View
1
Download
0
Category
Preview:
Citation preview
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 1/183
Elderly-friendly city services for active and healthy ageing
City4Age frailty and MCI risk model
Deliverable ID D2.11
Version 1.0
Contractual delivery date 30/11/2018
This version delivery date 28/11/2018
Status1 Final
Dissemination level2 PU
Leading partner MMED
Contributors UNIPV
This project has received funding from the European Union’s Horizon 2020 research and innovation
programme under grant agreement No 689731
1 ToC (v# = 0.0), Draft (v# < 1.0), Final (v# = 1.0), Improvement (v# > 1.0)
2 PU: Public, CO: Confidential, only for members of the consortium (including the Commission Services)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 2/183
History of changes
Version Date of issue Author(s) Description
0.1 04/11/2018 MMED Draft for peer review
1.0 28/11/2018 MMED Delivered version
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 3/183
Table of contents
1 Abstract ..................................................................................................................... 4
2 Executive summary .................................................................................................. 5
3 Introduction .............................................................................................................. 6
3.1 The case for City4Age unobtrusive behaviour monitoring ....................... 6
3.2 Measuring health......................................................................................... 9
4 City4Age risk modelling ........................................................................................ 14
4.1 Current methods to assess MCI and Frailty risk ...................................... 14
4.2 Assessing MCI and Frailty risk through unobtrusive technology ........... 18
5 Statistical assessment ............................................................................................. 30
5.1 Introduction ............................................................................................... 30
5.2 Correlation analysis .................................................................................. 30
5.3 Machine learning experiment ................................................................... 31
6 Lessons learned and recommendations.................................................................. 41
6.1 Lessons learned ......................................................................................... 41
6.2 Analysis of coverage from Pilots ............................................................. 44
6.3 Technical improvements........................................................................... 46
6.4 Towards a Machine Learning based approach......................................... 48
6.5 Towards market deployment .................................................................... 53
7 Conclusions ............................................................................................................ 55
8 Annex: Current Geriatric Instruments ................................................................... 56
8.1 MCI Instruments ....................................................................................... 56
8.2 Frailty Instruments .................................................................................... 73
9 Annex: Analysis of Items ....................................................................................... 87
10 Annex: Coverage of Geriatric Factors .......................................................... 128
11 Annex: Measures at City4Age Pilots ............................................................ 132
12 Annex: Correlation matrices for Pilots measures ......................................... 152
12.1 Athens Pilot ............................................................................................. 152
12.2 Birmingham Pilot .................................................................................... 154
12.3 Lecce Pilot............................................................................................... 155
12.4 Madrid Pilot ............................................................................................ 165
12.5 Montpellier Pilot ..................................................................................... 167
12.6 Singapore Pilot ........................................................................................ 169
13 Annex: FAI data from Pilots ......................................................................... 170
13.1 Athens Pilot ............................................................................................. 170
13.2 Birmingham Pilot .................................................................................... 171
13.3 Madrid Pilot ............................................................................................ 172
14 Annex: Data pre-processing SQL for the Athens Pilot experiment ............. 173
15 Annex: MultiSchemeAUC modifications ...................................................... 182
16 Annex: Computing *_time_per_visits features ............................................ 183
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 4/183
1 Abstract
The model represents a formalization of current geriatric knowledge which is relevant to MCI and
frailty risk detection, in the City4Age framework.
The main traits of the City4Age geriatric risk modelling are based upon the notions of “GEF” and
“GES” (Geriatric Factors and Sub-factors, i.e. conceptualizations of established geriatric determinants
of frailty/MCI) that can be synthesized from “Measures”, i.e. numbers generated by the various
technologies and methods for collecting data. Measures are analysed in various ways, generating
statistical indicators (called “NUI” in the project) that can be used to actually estimate GEFs and GESs.
Model’s NUIs can also be the basis (features) for building classifiers able to automatically screen
robustness in the elderly population. A proof-of-concept has been produced by training one such
classifier, based on real-life experimental data from the City4Age Athens Pilot site.
The deliverable includes proposals for further improvement, in terms of both (i) future data analytics
investigations that can reveal more about the inner structure of the risk model, and (ii) actions that
City4Age Partners interested in the risk model’s market deployment can put in place to achieve
industry-grade accuracy.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 5/183
2 Executive summary
A large number of instruments and methods for measuring the risk of MCI and frailty onset have been
proposed in the medical literature, and many of them have been clinically validated and are routinely
used in the geriatric practice.
However, such instruments are based on data collection methods (administration of questionnaires,
meter-based measurement, direct observation) that present several drawbacks in terms of high
deployment costs and reduced timeliness of risk detection.
City4Age aims at harnessing several important technology developments, which appeared on the scene
in the last decade, that offer new ways to help achieve systematic health monitoring and early health
risk detection, targeting elderly populations in smart-cities.
To achieve this goal, a first necessary step has been to accurately survey current geriatric knowledge,
with the aim of identifying health indicators potentially usable within the City4Age paradigm. Such
indicators must feature the following characteristics:
Be routinely applied in clinical settings to diagnose and/or predict the onset of MCI and
frailty conditions in individuals
Be linkable to behavioural traits that can be measured by sensors and other datasets readily
available in the smart-city environment.
Overall, 8 instruments for MCI and 11 instruments for frailty have been identified and analysed.
The analysis revealed that geriatric practice has largely agreed on the identification of the most
important functional domains that are implicated in the course of MCI and frailty onset – examples are:
Mobility, Physical activity, Basic and Instrumental Activities of Daily Living, etc.
Based on this knowledge, a comprehensive, “first-in-the-art” City4Age geriatric risk model has been
outlined.
The model consists of a number of Geriatric Factors (GEFs), possibly subdivided into Geriatric Sub-
factors (GESs), that parallel functional domains with relevant detection/predictive power. Factors can
be gauged through Measures, derived from the application of unobtrusive sensing technology or other
smart-city datasets – examples are WALK_DISTANCE, OUTDOOR_NUM, PUBLICTRANSPORT_TIME, etc.
Two complementary approaches are enabled by such model:
Build Individual Monitoring Dashboards (WP5) that exploit the GEF/GES/Measure
hierarchy to provide clinicians with a comprehensive, visually organized, multi-factorial
geriatric assessment, building on domains that medical professionals are already familiar
with and already know how to interpret
Use Measures to extract relevant features (called “numerical indicators” or NUIs, in the
Project) that can be combined to form “behavioural markers” for frailty/MCI, able to
automatically discriminate between robust and non-robust subjects
Based on the above, forms of effective semi-automated screening of the elderly population in smart
cities can be envisaged.
Proposals for future improvement are discussed, addressing in particular:
Further data analytics investigation avenues, able to capture the full richness and structure of
the model through a Bayesian Network approach (e.g. better identification of conditional
dependence relationships among GEFs, GESs and Measures)
Practical steps that Partners can conduct to achieve industry-grade model accuracy, by
extending model coverage with new unobtrusive data collection technologies and by
optimizing existing ones
With respect to its previous version (i.e. deliverable D2.06), this deliverable presents the following
differences:
Sections 1, 2, 6, and 7 have been fully revised
Section 5 and Annexes 12 to 16 have been added anew
Section 3 and 4 and Annexes 8 to 11 are mostly unchanged (with the exception of subsection
4.2.4.6, which has been partially revised)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 6/183
3 Introduction
The overarching objective of the City4Age project is to demonstrate that data collection in smart cities
can significantly improve the management of health risks in the elderly populations (in particular, MCI
and frailty) by harnessing the potential of new technologies for unobtrusive behavioural sensing, which
are more and more pervading urban life.
In order to achieve this objective, one of the important areas to be tackled relates to the City4Age early
risk detection subsystem.
In particular, as a starting point, it is important to:
At a first level of detail – state the overall case for applying unobtrusive behaviour
monitoring to address MCI and frailty risk detection; compare it with the methods currently
used in geriatrics for managing this task; and derive an overall City4Age risk detection
concept
At a second level of detail – examine how currently established geriatrics instruments for
measuring health risks are structured, and propose new City4Age approaches that can
effectively take advantage of new technologies for unobtrusive behaviour measurement
The City4Age risk modelling effort – conducted in task T2.1 Modelling risks and resilience profiles
and reported in the rest of this deliverable – aims at structuring the consequences from the above
mentioned thinking into a conceptual framework, to be used as a guidance in the technical design and
implementation of the risk detection subsystem. Together with behavioural modelling – elaborated in
task T2.2 Modelling MCI/frailty related behaviours and reported in deliverable D2.2 – it constitutes the
scientific base for the work planned in the technical work-packages, from WP3 to WP6.
This section briefly introduces the rationale behind these elements and paves the way for the following
Sections, that report on the modelling approach chosen for City4Age.
3.1 The case for City4Age unobtrusive behaviour monitoring
It is well known, especially in geriatrics, that the early detection of risks relating to a specific health
condition improves the chances of enacting appropriate interventions that can halt or at least delay the
condition itself, with beneficial effects on both patients’ quality of life and cost of treatment3.
For instance, in the case of MCI, Petersen et al.4 emphasize the need for the clinician to detect the
earliest signs of cognitive impairment and highlight the importance of this quest. In fact, as MCI is
recognized as a possible precursor of AD, its early detection and subsequent intervention can help to
delay the onset of the disease. Considering, for instance, that an estimated 5.3 million Americans had
Alzheimer's disease in 2015 and the overall cost of caring for these people is around $226 billion5 (i.e.
more than $40,000 per patient per year) detecting and treating MCI such that the progression to AD
can be delayed by even a single year, would imply very important savings. In addition, this analysis
only considers the economic facet of the issue, which is easier to measure, but even more important is
the impact on the quality of life of elderlies and of their carers.
However, under current conditions, MCI detection is a challenge.
Earliest signs of MCI may consist, for instance, in some degree of forgetfulness, beyond what is
justified by normal aging. The forgetfulness may be apparent to those closest to the person but not to
the casual observer6. In fact, relying, as current practice, on self-reported detection of MCI signs and
symptoms (by the patient or by those around her/him) has several drawbacks:
3 Onder et al., Measures of physical performance and risk for progressive and catastrophic disability: Results
from the women’s health and aging study, Journals of Gerontology: Medical Sciences, 2005
4 Petersen et al., Practice parameter: early detection of dementia: mild cognitive impairment (an evidence based
review), Neurology, 2001
5 http://www.alz.org/facts/
6 Petersen, Mild Cognitive Impairment, The New England Journal of Medicine, 2011
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 7/183
It may not be easy to discriminate signs and symptoms of MCI insurgence from normal
effects of aging
Patients and their carers may not be knowledgeable about exactly which signs or symptoms
to look for
Some patients, especially older ones, may be subject to forms of denial against worsening
conditions (e.g. minimizing symptoms) to avoid care7.
Similar considerations apply to frailty8. Frailty has been shown to be directly correlated with high risk
for adverse health outcomes, such as falls or hospitalization, that increase care costs and decrease
quality of life. On the other hand, when detected early, possibly at the pre-frail stage, frailty can
potentially be prevented or treated. For this reason, the need to recognize frailty in a timely manner is
noted as one of the top priorities in both gerontology, general practice, and public health.
Several important technology developments, which appeared on the scene in the last decade, may offer
new ways to address this type of issues, and help to achieve a systematic health monitoring and early
health risk detection approach.
A paradigmatic example of these technologies is the modern smartphone which – packed with sensors
that can capture many features, such as e.g. orientation, acceleration, location or voice, and equipped
with local computational power and an always-on network connection – can seamlessly interact with
other computer systems in order to conduct complex analysis on the generated data streams and infer
interesting aspects of human behaviour. The smartphone proliferation among users and its continuous
presence and usage, make it the ideal unobtrusive human activity data collection platform9.
In addition, smart cities environments offer extra opportunities, as ever more urban infrastructure is
deployed on the basis of technologies such as RFID cards (e.g. for check-in services), proximity
devices (e.g. BLE beacons for in-shop proximity detection), or intelligent meters of various kind (e.g.
for energy or water consumption)10
.
Prospects are also offered by “software sensors”, such as social network logging applications.
The endless possibilities arising from the combination of these technologies allow to measure, monitor
and analyse human behaviour in unprecedented ways.
With reference to the objectives of City4Age, and among many similar works that can be found in the
literature, Rantz et al. report on a simple yet paradigmatic case that shows what can be practically
achieved11
.
The authors installed an integrated sensors network in apartments of volunteer residents, hosted in an
“aging in Place” retirement community that allows residents to remain in their apartments even if their
health deteriorates.
Among others, the sensors network included several passive infrared (PIR) motion detectors installed
in various locations, to detect presence in different rooms and to consequently infer specific activities.
In particular, the objective of the sensors network is to detect changes in activities that could be linked
to respective changes in health status, and to consequently offer relevant clinical interventions to help
residents age in place.
Alerts are generated and sent to clinicians whenever sensor activity (within definite daytime frames)
deviates from normal (in the particular example, the cut-off point was established at 4 standard
deviations from the mean of the previous 14 days, and it was chosen conservatively, to err on the side
of generating too many alerts rather than possibly missing a crucial one).
7 Trull et al., Ambulatory assessment, Annual Review of Clinical Psychology. 2013
8 Lucas et al., Frailty in the older adult: will you recognize the signs?, Nurse practitioner, 2014
9 Lathia et al., Smartphones for Large-Scale Behavior Change Interventions, IEEE Pervasive Computing, 2013
10 Hancke et al., The Role of Advanced Sensing in Smart Cities, Sensors, 2013
11 Rantz et al. Using Sensor Networks to Detect Urinary Tract Infections in Older Adults, IEEE 13th International
Conference on e-Health Networking, Applications and Services, 2011
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 8/183
It is important to note that alerts are not diagnosis. In fact, whenever an alert is generated, the resident
is evaluated by a registered nurse, in order to check if the situation warrants deeper medical
investigation.
In this context, the authors report about several case studies related to the early detection of urinary
tract infections (UTI), a serious health threat for older women that, when treated too late, may lead to
kidney damage, system-wide infections or even death.
Since a common indicator of UTI observed by clinicians is urgency and frequent urination, particularly
at night, alerts based on increased nightly activity of the bathroom PIR motion detector have been used
to test residents for UTIs.
The authors show how, out of three case studies, two of them conducted to an effective early diagnosis
of UTI, which allowed a correspondingly early treatment and full recovery of the patient (the third one
being a false positive). If the diagnosis would have been delayed until the elderly person would have
self-reported relevant symptoms to her GP or caregivers, the situation would have been much worse,
with a less certain and more costly outcome.
Although relatively simple and limited to the indoor environment, this example is paradigmatic of the
state-of-the-art in the field, and clarifies how unobtrusive and continuous behavioural sensing, paired
with an appropriate data interpretation and alerting mechanism, can concretely contribute to the
improvement of health management for elderly people and to the reduction of related care costs.
On this basis, it is possible to highlight some essential characteristics that should also be replicated in
the City4Age risk detection approach (refer to Figure 1 below):
Suitable data streams, coming from sensors and other datasets available in the smart city
environment, shall be linked to valid and reliable health status indicators, established in the
medical practice and associated with specific behavioural patterns, and used to address
relevant health risks through (semi-)automated behaviour recognition and analysis
The ultimate objective of the risk detection subsystem is to interpret data and consequently
generate alerts, against which a conventional medical investigation will possibly lead to
diagnosis and intervention decisions (note that it is out of scope for the system to
automatically generate diagnosis). Alerts are accompanied with information, collected from
the data streams, which is presented to medical personnel through appropriate data
dashboards, in order to support clinical assessment and decision making
Figure 1. Risk detection subsystem in City4Age
A central element to consider in the design of the detection, alerting and data interpretation
mechanism is the achievement of the traditional maximization/balance among higher
City4Age Risk detection subsystem
Alerting mechanism Data dashboards
Visit (diagnosis, interventions)
Behaviour
Datasets
Health indicators
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 9/183
specificity (reduce false positives) and higher sensitivity (reduce false negatives). It is
important to note that this aspect has a clear and direct link with ROI estimation, as (i) true
positives represents actual savings and quality of life improvements made possible by the
enactment of early medical interventions; (ii) false positives represent costs sustained for an
additional, unnecessary medical investigation of a healthy person and (iii) false negatives
represent lost opportunities to provide better care and quality of life to subjects that would
have needed it (in City4Age, this analysis is to be carried out as part of Task T8.2
Exploitation plan and sustainability).
3.2 Measuring health
In order to implement the scheme illustrated in the previous section, a crucial element is the availability
of suitable health measurement indicators, that are:
valid and reliable
able to predict the onset of critical health conditions in individuals (MCI and frailty, in the
case of City4Age)
linked to behavioural traits that can be feasibly measured by sensors and other datasets
available in the city environment.
The effort to derive high quality health measurement indicators is a longstanding one, that will
probably always continue. In the last decades, many indicators of the health or well-being of
individuals have been developed, in order to address three ultimate aims12
:
Diagnosis of illness
Predict the need for care
Evaluate the outcomes of treatment
This is in agreement with the distinction made by Bombardier and Tugwell between the three purposes
of health measurement indicators: diagnostic, prognostic, and evaluative13
.
In particular, the second item – relating to the prognostic/predictive value of indicators – is the one of
interest to City4ge.
In this respect, it is also worth to note that several indicators have been validated for both concurrent
validity and predictive validity, depending on whether the focus is on current or future health status.
For instance, a questionnaire on hearing difficulties may be compared with the results coming from an
audiometric test to assess its concurrent validity, or the same questionnaire may be compared with
future patient’s health outcomes, to assess its predictive validity. It is worth to note that, in the context
of early detection envisaged by City4Age, concurrent validity is also important.
Given the above frame it is important to survey and analyse existing indicators, in order to identify how
they could meet City4Age requirements, and be used to constitute the foundation of the project’s risk
detection model. As a first step, we discuss in details the structure of existing indicators, in order to
base relevant City4Age proposals later on.
3.2.1 Structure of instruments
Defining health measurement indicators normally implies assembling a selection of elementary items,
that are intended to represent the specific clinical aspects that are of interest.
Underlying each indicator is a model, that takes item values as input and – according to relevant
algorithms derived through various techniques, such as domain expertise application, statistics
modelling, knowledge engineering, machine learning, etc. – produces a final outcome that is the
measure that was sought for.
12
McDowell, Measuring Health: A Guide to Rating Scales and Questionnaires, Oxford University Press, 2006
13 Bombardier et al., Methodological framework to develop and select indices for clinical trials: statistical and
judgmental approaches, Journal of Rheumatology, 1982
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 10/183
Examples of elementary items may be a measure on a specimen analysed in a laboratory, the flexion of
a limb observed by a physiotherapist, an estimate of working capacity assessed by a clinician, or a self-
reported behaviour obtained through a questionnaire response.
Items may in turn be grouped into specific categories or domains. For example, in the Mini-Mental
State Examination indicator, the domain “Orientation to time” includes five questions, from broadest to
most narrow, such “What year is this?”, “What season is this?”, “What month is this?”, “What is
today’s date?”, “What day of the week is this?”.
Sometimes, domains correspond to a single item, such as in the Lawton Instrumental Activities of
Daily Living (IADL) Scale, where domains such as “shopping”, “ability to use the telephone”, “mode
of transportation”, directly correspond to eponymous items to be measured.
Ultimately, each item is associated with (a set of) values, among which the item measure, or score, is
drawn. Item scores then contribute – through application of the underlying model, as above mentioned
– to the generation of the instrument outcome score.
In current medical literature there is no uniform terminology to describe these notions, and several
different terms – e.g. instruments, measures, measurement methods, indicators, scales, etc. – are used
more or less interchangeably, sometimes even indicating different concepts.
On the other hand, in City4Age modelling it is important to agree on some kind of formal definition of
the above ideas, upon which to base the subsequent design and implementation work.
For this reason, the specific schema illustrated in Figure 2 below is proposed.
Figure 2. Health measurement instruments strcture
In the figure, the following definitions apply:
Instrument: the overall measuring method, aimed at providing a final, comprehensive score
on a certain health entity
Category: a group of Items (see below) that contribute to the appraisal of a certain facet of
the measured health entity
Items: the actual, basic elements that will be measured to produce scores (see below).
Values: set of values associated to an Item. The Item’s measurement produces a specific
score, drawn from this set.
Model: the mathematical entity (algorithms) that allows to compute the Instrument outcome
score on the basis of the scores measured for single Items.
Examples of the application of these definitions to two established Instruments are illustrated in the
following table.
Category
Item Values
Instrument
1..*
1..* 1
Model 1
1..*
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 11/183
Instrument The Nottingham Extended Activities
of Daily Living
Fried Frailty Index
Example of
Category
Transportation Motility
Example of
Item
Question “Did you travel on public
transport?”
Test “time to walk 15 feet”
Example of
Score
“Not at all”, “With help”, “On your
own with help”, “On your own”
Number of seconds
Example of
(part of) Model
Count 1 if answer is “On your own
with help” or “On your own”, 0
otherwise
Count 1 if threshold is exceeded (Note: threshold depends on sex and
height – not reported here to save space)
Table 1. Example of Instrument definitions
The above terminology has been devised with the ultimate aim of facilitating communication across
the interdisciplinary knowledge areas represented in the project, including geriatricians, behavioural
scientists, information technology engineers, researchers, and data scientists. For this reason, some
trade-offs have been made, e.g. by using the term “Category” instead of the term “Domain”, often used
in geriatrics to denote groups of functionally related items, because the term “Domain” has a strong but
unrelated connotation in the information technology field.
An important observation regards the meaning of the term “score”, as used in the above definitions.
While in principle the outcome of measuring an Item (or the outcome of the overall Instrument) can
always be reduced to a numerical score, the following different types apply:
Nominal or categorical score: numbers are used just as labels for categories (e.g. 0 for false
and 1 for true, for a classifier outcome)
Ordinal score: in addition to the previous type, order is meaningful and represents growing
quantities (e.g. 1 = mild, 2 = moderate, 3 = strong)
Interval score: in addition to the previous type, difference among numbers is meaningful, i.e.
there exists the concept of unit of measure (e.g. temperature in oC)
Ratio score: in addition to the previous type, there is a meaningful zero point, so ratios
among scores are also meaningful (e.g. number of seconds to complete an action)
3.2.2 Instruments for City4Age
While in the 1970s the medical field has been characterized by an intense proliferation of new
Instruments and related models, that were often created relatively hastily and sometimes with
insufficient substantiation, the more recent trend is to focus on a less wide number of high quality
Instruments, making validation and reliability assessment the priority.
Following this trend, the best strategy for City4Age is to also concentrate on a well-chosen set of
existing Instruments, selected on the basis of the following characteristics, as suggested in the opening
to this subsection:
Predictive value for the conditions addressed by the project (MCI and frailty) and/or ability
of timely detect the onset of these condition
Measurability in behavioural terms (i.e. the model underlying the Instrument is linked to the
subject’s behaviour)
Feasibility of measurement through sensors/datasets readily available in a smart city, in the
timeframe expected for the project exploitation (as determined in WP3 and in Task T8.2
Exploitation plan and sustainability)
These points deserve additional discussion.
Concerning measurability in behavioural terms, three approaches are possible (refer to
Figure 3):
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 12/183
Use existing indicators, that rely on models directly linked to behaviours (part a in the
Figure), such as for instance the Lawton Instrumental Activities of Daily Living Scale,
which includes Item values such as “Takes care of all shopping needs independently”. In this
case, the City4Age behaviour reconstruction function shall detect the current behaviour (e.g.
the above “Takes care of all shopping needs independently”) while the risk detection
function should calculate a relevant score and possibly generate an alert if, according to the
model rules, a change is detected (e.g. from “Takes care of all shopping needs
independently” to “Shops independently for small purchases”).
Use existing indicators, that rely on models that, although not directly linked to behaviours,
can be feasibly correlated to them (part b in the Figure). For instance, the Fried Frailty Index
includes the test “time to walk 15 feet”. Although this is not directly a behaviour
(measurement is taken manually, by a physician with a chronometer), the relevant input to
the underlying model (i.e. walking speed) can be correlated to behaviour, for example by
monitoring daily walking speed through smartphone sensors.
Devise new geriatrics Instruments, based on new models – for example, based on machine
learning techniques – that directly connect the sensors and datasets information, gathered by
the City4Age data collection subsystem, to a relevant outcome score (part c in the Figure).
At a first level, the first and second approaches will be the preferred ones in City4Age.
In fact, by relying on existing Instruments for the detection of MCI and frailty onset, which are already
established and accepted by geriatrics researchers and clinicians, the City4Age risk detection
subsystem can count on their proven validity and reliability characteristics, as derived from decades
of geriatrics research and clinical practice, as well as on a higher level of trust from clinicians.
Figure 3. Determining City4Age model
On the other hand, the third approach would be a relatively innovative one. In fact, it is worth to note
that established Instruments have been conceived, validated and published at a time when the very
mechanisms for personal data collection envisaged by City4Age were not generally available to the
medical community. Today, such availability brings in the possibility that correspondingly new
indicators, that directly relate sensor readings to relevant outcome scores, could be developed and
Reconstruct behaviour
Apply existing
Instrument
Sensors and
datasets
Score
Reconstruct behaviour
Apply existing
Instrument
Sensors and
datasets
Score
Apply new Instrument
Sensors and
datasets
Score
Infer input for existing Instrument
(a)
(b)
(c)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 13/183
tested. In fact, research in using unobtrusive sensing for direct health measurement is in its infancy, and
few studies in this direction have started to appear in the recent literature14
.
Although this is a very stimulating approach, it is also more challenging for two reasons:
Models that directly link data features to MCI or frailty outcomes would be opaque with
respect to the underlying behaviours, and might not sufficiently clarify the link between
specific changes in behaviour and consequences on MCI and frailty; this makes it
problematic for geriatricians to accept such outcomes (this is a general and well-known issue
in the application of automatically, machine-learned models to clinical practice, which is not
specific to City4Age alone)
Relevant machine learning algorithms currently proposed in the literature are generally at the
very early stages of development and, to the best of the City4Age team’s knowledge,
although related research has been conducted in labs or smart-home settings, at the time of
writing no significant effort is yet available for measuring frailty and MCI on the basis of
unobtrusive technologies and datasets deployed in city-wide contexts, for monitoring city-
wide behaviours.
These challenges notwithstanding, the project has still made an effort to preliminary investigate this
third approach in order to at least generate educated recommendations for researchers who are willing
to undertake the full endeavour in the near future. Moreover, several technical achievements in
City4Age (e.g. data collection subsystem, the human activity reconstruction subsystem, the data
analytics platform for risk detection, etc.) represent important enablers for furthering this line of
investigation. In fact, in its last phases the Project has even gone someway further, by proposing a
kind of “hybrid” approach, in which a reasonably performant machine learning component, rather than
being used as “stand-alone”, is applied as a first step in a multi-level screening scheme, that could be
actually implemented in practice. Details are discussed in Sections 5 and 6.
A final note regards technical feasibility. It has been hinted that, when selecting Items to consider for
the City4Age risk detection subsystem, it is also necessary to take into account how the inputs to the
underlying models can be reliably obtained from sensors and other smart-city datasets. For instance
(with reference to previous examples), it may be relatively easy to link an indicator like the number of
nightly visits to the bathroom to a dataset coming from a PIR motion detector, installed in the
bathroom itself (like in the UTI early detection example), but it is quite a different story to understand
in a reliable way if a person is taking care of all her/his shopping needs or if she/he started to shop only
for small items (as in the case of the Lawton IADL scale application). Although seeking technical
feasibility is a crucial task of City4Age work-packages WP3-WP5, relevant issues in this area will still
be consistently considered in this deliverable, in order to ensure that the modelling approach is
conducive to reasonable technical implementations.
14
Ben-Zeev et al., Next-Generation Psychiatric Assessment: Using Smartphone Sensors to Monitor Behavior and
Mental Health, Psychiatric Rehabilitation Journal, 2015
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 14/183
4 City4Age risk modelling
To concretely apply the concepts introduced in the previous Section to the management of health risks
in the elderly populations, it is important to establish how such health risks are to be precisely
characterized.
The project has chosen to concentrate on MCI and frailty, two typical conditions of aging that
influence multiple determinants of health, including both physical and cognitive ones.
Accordingly, in this Section we present how City4Age models MCI and frailty risks in a way that can
be used to support the development of an effective, technology-based early detection system.
In particular, the Section will first illustrate how MCI and frailty are measured today in clinical
practice. On that basis, the challenge of achieving the same measurement through unobtrusive sensing
technology is then addressed and a model to underpin such technology-based assessment – i.e. the
City4Age Geriatric Risk Model – is outlined and discussed.
4.1 Current methods to assess MCI and Frailty risk
Although there is no agreed gold standard to measure MCI or frailty, geriatric medicine has
nevertheless conceived, proposed and clinically validated several Instruments that are routinely
used since many decades in clinical practice for detecting (and in some cases predicting) the onset of
these syndromes.
As such Instruments represent well the currently available medical knowledge on the issue, they are the
natural starting point to formulate a computational risk model to be used with the City4Age approach.
For this reason, in this subsection, after providing a brief introduction to both MCI and frailty, we will
illustrate a survey and analysis of currently established Instruments that has been conducted in the
frame of Task T2.1 (Modelling risks and resilience profiles) and present relevant results, on which the
subsequent discussion is based.
4.1.1 Brief introduction to MCI
Mild cognitive impairment (MCI) represents an intermediate state of cognitive function between the
changes seen in aging and those fulfilling the criteria for dementia and often Alzheimer’s disease.
MCI is classified into two categories:
amnestic MCI, which is characterized by memory impairments that still do not meet the
criteria for dementia, while other areas (e.g. attention, language, executive function,
visuospatial skills) are not affected
non-amnestic (anamnestic) MCI, which impacts non-memory related areas
Amnestic MCI is more common than anamnestic MCI and it has been related to the onset of AD.
Non-amnestic MCI has been related to other kinds of dementia, not linked to AD, such as
frontotemporal lobar degeneration or dementia with Lewy bodies.
Prevalence of mild cognitive impairment ranges from 10 to 20% in persons older than 65 years of age.
For example, the Mayo Clinic Study of Aging, a prospective, population-based study of persons
without dementia who were between 70 and 89 years of age at enrolment, found a prevalence of
amnestic mild cognitive impairment of 11.1% and of non-amnestic mild cognitive impairment of 4.9%.
Given its role as a possible precursor of more severe forms dementia, MCI has recently received a lot
of attention in clinical practice as well as in research settings.
In fact, on one side, its detection can help to enact effective disease-delaying lifestyle interventions,
and, on the other side, being the earliest manifestation of cognitive disorders, it may also be important
in formulating research hypotheses.
Discriminating between MCI and normal aging can be a challenge. Slight forgetfulness, e.g.
misplacing objects and having difficulty recalling words, can be a normal effect of aging.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 15/183
Amnestic MCI is characterized by a more prominent memory impairment. For example, patients may
forget important information that they used to remember easily, such as appointments, telephone
conversations, or recent events of interest to them (e.g., sporting events), while all other aspects of
function are normally preserved.
The American Academy of Neurology recommends the following criteria for an MCI diagnosis:
self-reported memory problems, preferably confirmed by another person
greater-than-normal memory impairment, measured with standard tests
normal thinking and reasoning skills
no impairments in ADLs
Currently, there is no approved pharmacological intervention for MCI.
There is evidence of benefits from cognitive rehabilitation interventions (e.g. use of mnemonics,
association strategies, computer assisted training programs) for amnestic MCI patients15
.
Other recommended interventions include aerobic exercise, involvement in intellectually stimulating
activities, and participation in social activities, given that these – even if not fully confirmed – might be
beneficial and pose little risk.
4.1.2 Brief introduction to frailty
Frailty is a geriatric syndrome of decreased reserve and resistance to stressors, resulting from
cumulative declines across multiple physiologic systems, causing increased vulnerability to adverse
outcomes16
.
People living with frailty are more at risk of dramatic health outcomes after apparent minor events, like
for instance an infection or a new medication. These outcomes include falls, disability,
institutionalization, hospitalization or even mortality.
On the other hand, frail people may be relatively low users of health care services, and be little known
to their GP, until they undergo a major health decline, possibly as a result of a marginal episode.
There is evidence that older people may not recognize themselves as living with frailty and do not
accept to be considered as ‘frail’, a term that is heavily associated with vulnerability and dependence17
.
There are two broad models of frailty:
the Phenotype model, based on a group of patient characteristics (unintentional weight loss,
reduced muscle strength, reduced gait speed, self-reported exhaustion and low energy
expenditure) which, if present, can predict poorer outcomes
the Cumulative Deficit model, that assumes an accumulation of deficits (ranging from
symptoms e.g. loss of hearing or low mood, through signs such as tremor, through to various
diseases such as dementia) which can occur with ageing and which combine to increase the
‘frailty index’ which in turn will increase the risk of an adverse outcome
Usually, three ordered frailty levels are identified: frail, pre-frail and robust (non-frail).
In City4Age, the ‘pre-frail’ level is an important one, as it has been referred to in the literature as a state
of less ‘inevitability’ that may be more amenable to interventions than the frail state18
. In fact, frailty
may be preventable and early detection and interventions can minimize transitions from the pre-frail
(and pre-disabled) state to the frail state, reduce the chance of adverse outcomes, and reduce healthcare
costs19
.
15
Jean et al., Cognitive intervention programs for individuals with mild cognitive impairment: systematic review
of the literature, The American journal of geriatric psychiatry, 2010
16 Fried et al., Frailty in Older Adults: Evidence for a Phenotype, Journal of Gerontology, 2001
17 British Geriatrics Society, Fit for frailty (Report), 2014
18 Gill et al., Transitions between frailty states among community living older persons, Archives of internal
medicine, 2006
19 Morley et al., Frailty consensus: a call to action, Journal of the American Medical Directors Association, 2013
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 16/183
The prevalence of frailty in community-dwelling older Europeans (65 years and older) varies between
5.8% and 27.3%. In addition, between 34.6% and 50.9% are classified as ‘pre-frail’20
.
Current recommendations to detect frailty include:
periodic social service assessment
review after referral for community intervention
primary care review when interacting with older people
assessment from home carers
assessment by ambulance crew, when called e.g. for a fall or other urgent matter
In terms of prevention, unhealthy behaviours that are implicated include:
insufficient physical activity, particularly resistance and aerobic exercise, which is beneficial
in preventing and treating the physical performance component of frailty
poor diet, particularly in terms of suboptimal protein/total calorie intake and vitamin D
insufficiency
In relation to this, the Action Group A3 of the European Innovation Partnership on Active and Healthy
Ageing has specific areas related to Food and Nutrition and to Physical Activity that investigate how
these aspects can be addressed in order to promote frailty prevention and to enact related multi-modal
interventions.
4.1.3 Analysis of Instruments to measure MCI and Frailty risk
To conduct the analysis of currently used methods for detecting and predicting MCI and frailty, a total
of 19 Instruments have been selected and surveyed: 8 for MCI and 11 for frailty.
The selection has been conducted on the basis of the requirements expressed in Section 3, including in
particular:
capability of the Instrument to detect or predict the onset of MCI or frailty, respectively, as
documented in the literature
possibility to (at least partially) implement the Instrument through behavioural monitoring
prospected technical feasibility of unobtrusive data collection and/or activity recognition
It is important to note that, with reference to the last two points, the selection presented here is rather
liberal, preferring to include Instruments that may eventually reveal as unmeasurable through
behavioural monitoring or technically unfeasible, rather than incurring the risk of leaving out good
ones.
The full results of the analysis work are reported in Annex, in Section 8.
For each Instrument listed in Section 8, the following elements have been compiled:
overall description, including the Instrument’s motivation, fundamental characteristics, and
why it is worth considering in City4Age
reference to literature, pointing to full information about the Instrument as well as to studies
that have addressed its diagnostic or predictive validity, as relevant
a synoptic view of the structure of the Instrument, along the definitions given in Section 3.
In particular, for each Instrument, this view includes:
o the list of Items
o for Items whose values correspond to potentially trackable behaviours, a full list of
those values
o the Category (i.e. geriatric domain) to which each Item belongs
o Moreover, in order to allow reference and tracking, each Item is given a unique ID,
which also identifies the Category and the Instrument to which it belongs.
20
Santos-Eggiman et al., Prevalence of frailty in middle-aged and older community-dwelling Europeans living in
10 countries, The journals of gerontology. Series A, Biological sciences and medical sciences, 2009
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 17/183
It is important to observe that the underlying geriatric domains probed by the surveyed Instruments (as
denoted by Items’ Categories) are generally shared from a relatively well recognizable and common
pool of few dozens of them.
This observation is a meaningful hint that – broadly speaking – geriatric practice has roughly agreed
on the identification of the most important domains implicated in MCI and frailty onset.
Code Category Remarks
Mt Motility21
Slowness, ambulation, balance
Ac Activity Physical activity level, energy expenditure
Ad ADLs Basic activities of daily living in general
Ia IADLs Instrumental activities of daily living in general
Fo Food Meal and/or food preparation
Ho Housekeeping
Ln Laundry
Co Communication Telephone and other types of communication
Sh Shopping
Tr Transportation Transportation beyond walking distance
Fi Finances Ability to manage own finances
Me Medication Ability to manage own medication
So Socialization
Cu Culture Cultural or entertainment activities
En Environment
Dp Dependence Dependence/disabilities and social support
He Health General health status, illnesses, sensory issues
We Weight
Wk Weakness
Ex Exhaustion Exhaustion, fatigue, energy reserve level
Ab Abstraction
At Attention Attention and calculation
Mr Memory Recall (short-term verbal memory)
Mo Mood Mood or depression issues
Ti Time Orientation to time
Sp Space Orientation to space
Vi Visuospatial Visuospatial/constructional praxis
La Language
De Demographics
Table 2. Categories for diagnosis/prediction of MCI/frailty in current Instruments
Given the crucial importance of this fact for the project endeavour, an effort has been made to extract
and list such domains, resulting in the array presented in Table 2 above.
21
with hindsight, “mobility” would be the correct term for this domain; however, to maintain consistency with
other project’s deliverables, we stick to “motility” asking the reader to keep this note in mind
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 18/183
It has to be noted that the above classification has not been always easy, and in some cases Items may
justifiably address more than one Category. In these cases, an effort has been made to identify the main
one and assign the Item to that Category.
This information – reported in full details in Section 8, to which the reader is referred – provides the
project’s common understanding about how geriatric risk is currently measured and assessed in clinical
practice, and it is used in the following as the foundation for City4Age risk modelling.
4.2 Assessing MCI and Frailty risk through unobtrusive technology
The next logical step to undertake, is to harness existing geriatric knowledge – as derivable from the
Instruments discussed in the previous sub-section – and use it to build a model able to organize data
collected through unobtrusive technology and present it to geriatricians in a form that matches such
existing knowledge, so that they can formulate relevant interpretations, according to the current norms
and practices they have been trained to use.
The result of this effort is presented in this sub-section.
Work has been based on a comprehensive, multifarious, joint effort including:
The survey of existing Instruments, presented above (as completed at M4, i.e. March 2016)
Feedback received from work-package WP3 (Unobtrusive acquisition of personal data), that
performed a revision of the proposed Instruments and Items to be used in City4Age and
provided an initial assessment of which technologies could be feasibly deployed to detect
which measures (as made available in several WP3 deliverables at M12, i.e. November
2016)
Feedback from the six City4Age Pilot testbeds (in Athens, Birmingham, Lecce, Madrid,
Montpellier, Singapore), that revised Instruments and Items and provided an assessment of
which measures they were ready to address through the deployment of which technologies
(as reported in several deliverables and subsequent mapping work conducted in the frame of
work-package WP7, from M5 to M18, i.e. from April 2016 to June 2017)
4.2.1 Two case studies
As mentioned in sub-section 3.2.2 (Figure 3, part a), one possibility for City4Age modelling would be
to re-use unmodified, existing instruments and directly measure their Items through unobtrusive
technology, substituting questions addressed to care recipients or informants, direct observation or
obtrusive measurement, with automatic data collection.
However, this straightforward solution proves to be very difficult to achieve in practice, with reduced
chances of success.
In order to better illustrate this point and understand what are the underlying challenges, an analysis of
two case studies – for the landmark instruments Fried Frailty Index (frailty) and Lawton scale of
Instrumental Activities of Daily Living (MCI) – is proposed below.
Fried Frailty Index
The Fried Frailty Index (refer for details in sub-section 8.2.1) is composed of six Items, covering five
Categories, associated to corresponding geriatric domains:
Item We.Fi.01 measures Category “Weight” through the detection of unintended weight loss
of at least 10 pounds in last year. The detection of this Item is performed by directly asking
the person
Item Ex.Fi.02 measures Category “Exhaustion” by asking the person how frequently she/he
felt that everything she/he did was an effort in last week, and classifying the response along
a 4-point Likert scale, with cut-offs 1, 3 and 5 days
Item Ex.Fi.03 also measures Category “Exhaustion” by asking the person how frequently
she/he felt that she/he could not get going in last week, and classifying a response along a
4-point Likert scale, with cut-offs 1, 3 and 5 days
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 19/183
Item Mt.Fi.04 measures Category “Motility” by gauging walking speed along a 15-feet
stretch with a stopwatch and assessing if it is greater than specific cut-off thresholds,
depending on gender and height
Item Ac.Fi.05 measures Category “Activity” by assessing if the number of kCal expenditure
per week is above 270. The assessment is performed by asking the person which of 18
moderately strenuous household chores she/he performed in the last two weeks and applying
a formula that includes activity-specific MET22
, body weight, number of sessions, session
duration, number of months per year the activity was done. Input parameters are obtained by
directly asking the person
Item Wk.Fi.06 measures Category “Weakness” by comparing the person’s grip strength,
gauged through a Jamar hand dynamometer, to specific cut-off thresholds depending on
gender and BMI23
.
The Fried Frailty Index is a classifier that categorizes people into three different stages: not frail, pre-
frail, and frail. The classification is performed as follows: a total score is computed by adding up 1 for
every Category that results as positive after measurement (the two Items for Exhaustion are composed
with or disjunction); then, a score of 0 corresponds to the not frail stage, a score of 1 or 2 corresponds
to the pre-frail stage, and a score of 3, 4 or 5 corresponds to the frail stage.
As it can be seen, all Items are measured obtrusively: four of them by posing direct questions to the
person under examination (Items We.Fi.01, Ex.Fi.02, Ex.Fi.03 and Ac.Fi.05) and two of them through
the application of a relevant measurement meter (Item Mt.Fi.04 with a stopwatch, and Item Wk.Fi.06
with a Jamar dynamometer).
The fundamental question here is if and how these Items can be measured through unobtrusive
technologies instead.
For two of them the answer is positive:
For Item Mt.Fi.04, a measure of walking speed can be derived by combining GPS and
accelerometer readings in a smartphone or wristband worn by the person. Some issues have
to be solved – such as for instance how to identify the right time windows when speed is
better measured – but literature shows that this can be feasibly done (see for example papers
mentioned in 8.1.8.2)
For Item Ac.Fi.05, accelerometer readings can estimate the kCal expenditure of a person by
recognizing several physical activities, including some mentioned in Fried which can be
detected by devices appearing on the market (see for instance the Withings [formerly Nokia
Health] smart-watches, that are trained to recognize more than 10 such activities24
)
For one Item the answer is a qualified yes:
Item We.Fi.01 can be measured semi-unobtrusively by fitting the home of a person under
examination with a smart scale, able to record periodic weight measures performed by the
person herself. While an explicit action form the person is required (she/he has to decide to
use the scale), this can be considered a normal, daily activity, that in most cases she would
have done anyway, while the rest of the burden (keeping track of historical measures) is
taken care of by the system
The remaining two Items are clearly infeasible, given the current state-of-the-art:
Items Ex.Fi.02 and Ex.Fi.03 require obtaining an assessment of feelings from the person
(although the area of emotional computing is studying how such feelings could be possibly
derived from available measurements, this is still an open and very challenging research
question that falls outside the scope of City4Age)
Item Wk.Fi.06 requires the application of a dynamometer, which is a necessarily obtrusive
action. To the knowledge of the project’s WP3 team, that surveyed existing technologies for
22
https://en.wikipedia.org/wiki/Metabolic_equivalent
23 https://en.wikipedia.org/wiki/Body_mass_index
24 https://www.withings.com/eu/en/watches
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 20/183
personal data collection, no feasible technology exist to derive muscular strength
unobtrusively.
In conclusion, the analysis above shows that less than half of the Fried Frailty Index Items can actually
be measured fully unobtrusively, making the idea of devising a direct “City4Age-enhanced” version of
such Instrument difficult to achieve in practice.
Lawton scale of Instrumental Activities of Daily Living
The Lawton scale of Instrumental Activities of Daily Living, described in details in sub-section 8.1.1,
is composed by eight Items, covering eight Categories, associated to corresponding geriatric domains:
Item Co.Li.01 measures Category “Communication” by assessing the person’s capability to
correctly use a telephone. The assessment is obtained by directly asking the person or an
informant how well the telephone is used (discriminating among behaviours such as:
“operates telephone on own initiative, looks up and dials numbers”; “dials a few well-known
numbers”; “answers telephone, but does not dial”; “does not use telephone at all”)
Item Sh.Li.02 measures Category “Shopping” by assessing the person’s capability to
independently shop for all her/his needs. The assessment is obtained by directly asking the
person or an informant how well shopping is performed (discriminating among behaviours
such as: “takes care of all shopping needs independently”; “shops independently for small
purchases”; “needs to be accompanied on any shopping trip”; “completely unable to shop”)
Item Fo.Li.03 measures Category “Food” by assessing the person’s capability to
independently prepare her/his own meals. The assessment is obtained by directly asking the
person or an informant how well meals are prepared (discriminating among behaviours such
as: “plans, prepares, and serves adequate meals independently”; “prepares adequate meals if
supplied with ingredients”; “heats and serves prepared meals or prepares meals but does not
maintain adequate diet”; “needs to have meals prepared and served”)
Item Ho.Li.04 measures Category “Housekeeping” by assessing the person’s capability to
keep her/his house clean and in order. The assessment is obtained by directly asking the
person or an informant how well housekeeping is performed (discriminating among
behaviours such as: “maintains house alone with occasional assistance (heavy work)”;
“performs light daily tasks such as dishwashing, bed making”; “performs light daily tasks,
but cannot maintain acceptable level of cleanliness”; “needs help with all home maintenance
tasks”; “does not participate in any housekeeping tasks”)
Item Ln.Li.05 measures Category “Laundry” by assessing the person’s capability to
independently take care of her/his laundry. The assessment is obtained by directly asking the
person or an informant how well laundry is performed (discriminating among behaviours
such as: “does personal laundry completely”; “launders small items, rinses socks, stockings,
etc.”; “all laundry must be done by others”)
Item Tr.Li.06 measures Category “Transportation” by assessing the person’s capability to
independently use transportation means, in order to travel to places beyond walking distance.
The assessment is obtained by directly asking the person or an informant how well
transportation means are used (discriminating among behaviours such as: “travels
independently on public transportation or drives own car”; “arranges own travel via taxi, but
does not otherwise use public transportation”; “travels on public transportation when
assisted or accompanied by another”; “travel limited to taxi or automobile with assistance of
another”; “does not travel at all”)
Item Me.Li.07 measures Category “Medication” by assessing the person’s capability to
independently take care of her/his own medications. The assessment is obtained by directly
asking the person or an informant how well medication responsibility is addressed
(discriminating among behaviours such as: “is responsible for taking medication in correct
dosages at correct time”; “takes responsibility if medication is prepared in advance in
separate dosages”; “is not capable of dispensing own medication”)
Item Fi.Li.08 measures Category “Finances” by assessing the person’s capability to
independently manage her/his own finances. The assessment is obtained by directly asking
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 21/183
the person or an informant how well finances are managed (discriminating among
behaviours such as: “manages financial matters independently – budgets, writes checks, pays
rent and bills, goes to bank”; “collects and keeps track of income”; “manages day-to-day
purchases, but needs help with banking, major purchases, etc.”; “incapable of handling
money”)
The Lawton scale assigns a score of 0 or 1 to every Item, depending on the behaviour that most closely
resembles the person’s highest functional level. For example, for Item Tr.Li.06, a score of 1 is assigned
to the first three behaviours (“travels independently on public transportation or drives own car”;
“arranges own travel via taxi, but does not otherwise use public transportation”; “travels on public
transportation when assisted or accompanied by another”) and a score of 0 to the remaining two
(“travel limited to taxi or automobile with assistance of another”; “does not travel at all”). A summary
score is then computed by summing up the individual Items’ scores, obtaining a result varying from 0
(low function) to 8 (high function).
As in the case of the Fried Frailty Index, all Items are measured obtrusively, by posing direct questions
to the person under examination or to an informant.
The challenge presented by the Lawton scale is somewhat different with respect to the Fried index. The
Items in the Lawton scale are based on complex activities that that are supposed to require a mix of
both physical and cognitive abilities, and are effective to detect the initial stages of MCI or frailty,
when deterioration is still mild and only affects more involved behaviours. Consequently, a possible
way to address automatic detection of this type of Items is through activity recognition.
In some cases the activities can be recognized directly or through observation of relevant proxies, that
can be easily measured with specific unobtrusive sensing technology. For example:
Item Co.Li.01 can be measured by directly observing telephone usage patterns in the
person’s smartphone
Item Sh.Li.02 can be measured by detecting shop entrances, as provided (for example) by
shop keepers that have installed appropriate technologies such as BLE beacons (e.g. to
support their business intelligence tasks – several City4Age Pilots are deploying such
solution)
Item Tr.Li.06 can be measured by detecting bus entrances from appropriate datasets
provided by the bus company (one City4Age Pilot is deploying such solution)
Other cases require more advanced human activity recognition methods, as those researched in the
project’s work-packageWP5, where activities are reconstructed from elementary actions, detected by
sensors. For example:
Item Fo.Li.03 can be measured by reconstructing cooking and meal preparation activities
from elementary actions such as opening/closing furniture elements in the kitchen, switching
on/off the oven, the cookers or other kitchen appliances, etc.
Item Ho.Li.04 can be measured by reconstructing the activity from elementary actions such
as switching on/off housekeeping appliances, tracking movement across different rooms in
the house (if local indoor positioning technologies are available, e.g. as in an appropriately
equipped smart home), etc.
Item Ln.Li.05 can be measured by tracking usage of the washing machine and, possibly,
movement across rooms
The above Items present different degrees of implementation difficulty, depending on how “dense” the
sensing environment is. For examples, activities that requires the usage of many electric appliances
(like Fo.Li.03 and, to a certain degree, Ho.Li.04) may be easier to recognize, as on/off switching can be
detected by deploying cheap, market available technologies, like for instance smart plugs. Others, like
Ln.Li.05 may be more difficult.
At least two Items – namely, Item Me.Li.07 and Fi.Li.08 – seem to be outside the reach of current
technologies, as they require detecting inner thought processes of the person that result in many
elementary actions that cannot be easily correlated with one another (neither in space nor in time) and
thus fall outside the capabilities of the City4Age Human Activity Recognition System described in
deliverable D5.3 The activity recognition system for smart cities.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 22/183
4.2.2 Deriving a computational model for geriatric risk assessment
The challenges illustrated above with reference to the cases of the Fried Frailty Index and the Lawton
scale of Instrumental Activity of Daily Living, are representative of difficulties that are found with
respect to any of the current geriatric Instruments studied in sub-section 4.1.3.
The fundamental conclusion to be derived is that, in general, only a limited number of an Instrument’s
Items can be unobtrusively measured through technology and, consequently, the creation of
technology-enhanced “unobtrusive doubles” of existing Instruments is difficult to achieve.
A strategy to get around this limitation can be based on the following three observations:
Although individual Instruments rely on few Categories (corresponding to relevant geriatric
domains), collectively the geriatric practice has identified a wider set of relevant
Categories, used across different Instruments. For example, the Fried Frailty Index and the
Lawton scale of Instrumental Activity of Daily Living include 5 and 8 Categories
respectively, while the survey presented in sub-section 4.1.3 identified a total of 29
Categories. It is worth to note that geriatricians already implicitly recognize the need to
cover more Categories than those addressed by a single Instrument, when they propose
assessment protocols based on test batteries that include multiple Instruments, collectively
covering a more comprehensive set of domains25
. A first investigation line thus consists of
aggregating (at least potentially) all Categories identified by clinicians as having relevant
diagnostic and/or predictive power, in order to increase the dimensions along which
available, technology-derived information can be harnessed
Although some individual Instruments rely on few Items per Category, collectively, the
geriatric practice has identified and experimented with a variety of different Items for
each Category, in different Instruments. For example, using again the Fried Frailty Index
and the Lawton scale of Instrumental Activity of Daily Living examples, the former
Instrument has a maximum of two Items per Category (in particular, for Exhaustion) and a
latter has only one Item per Category. However the survey presented in sub-section 4.1.3
identified 226 Items for 29 Categories, that is an average of almost 8 Items per Category. A
second investigation line thus consists in considering all Items that have been associated to a
certain Category (in one or more established Instruments) and assessing the feasibility of
implementing such Items through technology.
Finally, it is also possible to consider several additional proxy measures for each Item, that
can be more easily collected through sensing technology, with respect to corresponding
manually collected versions. In other words, when a certain Item, as formulated in a certain
current Instrument, cannot be directly collected through unobtrusive technology, perhaps a
strongly correlated proxy can be identified and validly used in its place. This is a line that
has been successfully applied in the literature, including to frailty measurement26
. To this
respect, it is also important to note that even several established Items are actually proxy
themselves of a deeper determinant that clinicians would like to measure. For example, the
hand grip strength measurement in Item Wk.Fi.06, in Fried Frailty Index, is a proxy of
sarcopenia, the actual determinant to be assessed.
All three lines ultimately rely on the basic observation that the current limited set of Categories and
Items addressed by established Instruments is not dictated by considerations of diminishing returns
when collecting more data, but rather by the increasing costs associated to the collection process (both in terms of time and money spent). In fact, the major advantage of an automated, technology
based approach to data collection – as advocated by City4Age – is that it does not have such limits, and
thus collection and usage of more data about more Categories and more Items – underlying
multiple and varied geriatric determinants – becomes both possible and logical.
25
See for example the effort to define a Comprehensive Geriatric Assessment diagnostic process for the frail
elderly population: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4164377/
26 Palmer K. et al., Frailty, prefrailty and employment outcomes in Health and Employment After Fifty (HEAF)
Study, Occupational and Environmental Medicine journal, 2017
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 23/183
For this reason, the strategy chosen for the City4Age geriatric risk modelling is to identify and
integrate in the model the largest possible number of Categories and Items, including proxies (as hinted
in subsection 3.2.2, Figure 3, part b), that can be extracted from established Instruments and feasibly
measured through the project data collection platform (feasibility is assessed not only according to
findings by the project’s WP3 work, but also by considering the actual technology proposed by Pilot
Partners for the testbed experiments).
To proceed in this direction a thorough and accurate reclassification work, addressing all of the Items,
from all Instruments surveyed in sub-section 4.1.3 above, has been conducted. In particular, the
following activities have been performed:
Identifying and grouping together all Items that refer to the same Category, even if
belonging to different Instruments
Make explicit the way each Item is measured in current practice
Label the diagnostic and/or predictive power of the Instrument to which each Item belongs
List the way(s) each Item is proposed for measurement through unobtrusive data collection
technologies at City4Age Pilots, in the frame of respective testbed experiments
The result of this work is presented in Table 9, in Section 9.
Such Table, to which the reader is referred, represents the fundamental basis for the derivation of the
City4Age model.
Before proceeding, some supplementary considerations are worth making.
Regarding the current data collection methods (fourth column in Table 9, heading Current
measurement), the Table confirms that such methods can be subdivided into three classes, illustrated in
Table 3 below.
Method Description Example
Report A report (in the form of a response
to a question) from a person – either
the person under examination or an
informant, such as a caregiver or a
family member
Mr.Ti.09
Question “Do you have problems
with your memory?”
Possible responses “Yes /
Sometimes / No”
Observation Direct observation of the conduction
of an assigned task by the health
professional conducting the
assessment
Fi.Da.06
Task: “The person is invited to
identify four different coins and
three notes”
Observation: “Correct / Incorrect
performance for each object”
Meter Gauging a physical quantity through
a relevant meter
Mt.Fi.04
Meter: “Stopwatch”
Measure: “Time to walk 15 feet
(4.57 meters)”
Table 3. Data collection methods in current Instruments
Table 3 clarifies, almost “at a glance”, how costly it is to measure Items with current methods and
confirms the rationale of the City4Age endeavour, that aims at substitute them with unobtrusive
technology.
Regarding the predictive vs. diagnostic value (fifth column in Table 9, heading Geriatric validity), it is
worth to note the good coverage of the former, that ensures that early detection of conditions’ onset is
possible.
Regarding Items measurement in City4Age, we need to recall that work-package WP3 (Unobtrusive
acquisition of personal data) identified the concept of Measure, that is a periodic number (mostly
daily, but also weekly or monthly) obtained from the City4Age data collection platform that, according
to previous discussion, represents either:
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 24/183
a direct quantification of the respective Item or
a proxy quantification of the respective Item
The identifiers reported in the last column of Table 9 (with heading City4Age measurement at Pilots)
represents the names of such Measures, as they are collected at City4Age Pilots, in the experiment
deployment. The full list of Measures that Pilot Partners proposed to collect in the experiment trials, is
presented in the Annex reported in Section 11.
Based on the above work, the following subsection illustrates how the City4Age risk modelling has
been formulated.
4.2.3 A model of Geriatric Factors
The work presented in the previous subsections conducted to the formulation of the City4Age
geriatric risk model summarized in Table 4 below.
The model consists of a number of Geriatric Factors (GEFs), some of which subdivided into Geriatric
Sub-factors (GESs), that parallel Categories – i.e. the diagnostic/predictive geriatric functional domains
discussed in Section 3.2 and identified in Section 4.1.3 above.
With respect to Categories, GEFs and GESs have been partially restructured, in order to address
several requirements aimed at obtaining the best possible match between what has been shown by
medical research to correlate well with the risk of MCI and frailty and what can be feasibly obtained
through unobtrusive sensing technologies. In particular:
GEFs with large sets of Items have been subdivided into composing GESs, able to better
capture the specifics of different groupings of Items, that share extra functional
cohesiveness. This has been done for the following GEFs: Motility, Basic ADLs (split along
the standard ADL list, identified in the literature), Instrumental ADLs (split along the
standard IADL list, identified in the literature), Environment, Health – Physical
GEFs targeted by a large number of Measures, derivable from unobtrusive sensing
technology, have been subdivided into composing GESs, able to isolate several specific
elements that such Measures allow to address. This has been done for the following GEFs:
Socialization, Cultural Engagement
Addition of GESs that, although not directly addressed in current Instruments, are
considered by many in the geriatrics practice to be a useful supplement in the overall
assessment of MCI/frailty risk and for which feasible detection technologies are available.
This includes in particular the addition of GESs Still/Moving, Moving across rooms (both of
them addressing the mobility domain) and New media communication (addressing the
Instrumental Activity of Daily Living domain)
Addition of the Gait balance GES, to represent a group of balance related Items, which are
assessed in several current Instruments (e.g. enter/exit car, enter/exit bed, TUG test, etc.),
with the expectation that such GES could be measured through smartphone/wristband
accelerometer readings (see also sub-section 6.2)
Explicit addition of the GES Going out to the Basic Activity of Daily Living GEF, to
complete the assessment of Basic ADLs with an activity frequently reported in literature for
this GEF and relatively easy to get from unobtrusive sensing
Four GESs have been added to the Health – Physical GEF, in order to complete physical
health assessment with data that geriatricians normally collect, beyond those mandated by
risk assessment Instruments (i.e. Falls, Appetite, Pain, Quality of sleep)
The Health – Cognitive GEF has been added to group the related Abstraction, Attention,
Memory and Mood functional domains and to parallel the Health – Physical GEF
The model also makes a distinction among two different types of GEFs, that is important to note with
reference to the City4Age approach:
Behavioural factors: i.e. the factors that depend on the actions and activities performed by
the person under monitoring; they are essentially the factors that should undergo a
continuous unobtrusive measurement, as advocated by the City4Age paradigm
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 25/183
Context and status factors: these are factors that are related to the person’s health profile
(both physical and cognitive) and her/his socio-economic status (environment quality and
dependence). Most of them are typically intended to be measured once, at the beginning of
the monitoring, and then updated with a periodicity much lower than for the behavioural
factors (e.g. once a year)
A number of Categories related to cognitive health, which are used in several existing Indicators (i.e.
Time, Space, Visuospatial, Language) have been dropped because they have been deemed redundant
with respect to those that have been included in the model (i.e. Abstraction, Attention, Memory, Mood).
The Demographic category has not been included, as these data are part of the Health Record that
geriatricians store for each monitored person and have always available at their hand, when performing
the assessment.
Geariatric factors Corresponding
Category
Remarks
Behavioural factors
Motility Mt Subdivided into sub-factors, to capture
sub-domains
Walking Mt Specific Items from Mt
Climbing stairs Mt Specific Items from Mt
Still/Moving n/a Proposed to enrich the assessment of the
mobility functional domain
Moving across rooms n/a Proposed to enrich the assessment of
mobility functional domain
Gait balance Mt Specific Items from Mt
Physical activity Ac
Basic Activities of Daily Living Ad Subdivided into sub-factors, to capture
sub-domains
Bathing and showering Ad Specific Items from Ad
Dressing Ad Specific Items from Ad
Self-feeding Ad Specific Items from Ad
Personal hygiene and
grooming Ad Specific Items from Ad
Toilet hygiene Ad Specific Items from Ad
Going out Ad Specific Items from Ad
Instrumental Activities of Daily
Living
Ia Subdivided into sub-factors, to capture
sub-domains
Ability to cook food Fo
Housekeeping Ho
Laundry Ln
Phone usage Co Specific Items relating to phone usage
New media
communication n/a Proposed to enrich the assessment of the
communication functional domain
Shopping Sh
Transportation Tr
Finance Management Fi
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 26/183
Geariatric factors Corresponding
Category
Remarks
Medication Me
Socialization So Subdivided into sub-factors, to test
potential sub-domains
Visits So
Attending senior centers So
Attending other social
places So
Restaurant So
Cultural engagement Cu Subdivided into sub-factors, to test
potential sub-domains
Visit entertainment /
culture places Cu
Watching TV Cu
Reading newspapers Cu
Reading books Cu
Context and status factors
Environment En Subdivided into sub-factors, to capture
sub-domains
Quality of housing En Specific items from En
Quality of neighbourhood En Specific items from En
Dependence Dp
Health – Physical He Subdivided into sub-factors, to capture
sub-domains
Falls n/a Proposed to complete physical health
assessment
Weight We
Weakness Wk
Exhaustion Ex
Pain n/a Proposed to complete physical health
assessment
Appetite n/a Proposed to complete physical health
assessment
Quality of sleep n/a Proposed to complete physical health
assessment
Visit to doctors He Introduced as a proxy of general health
(Items He.Fx.17, He.Ed.03, He.Ti.01,
He.Gr.05)
Visit to health related
places He Introduced as a proxy of general health
(Items He.Ed.06, He.Gr.09, He.Sb.02)
Health – Cognitive n/a Introduced to group 4 factors relating to
cognitive health
Abstraction Ab
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 27/183
Geariatric factors Corresponding
Category
Remarks
Attention At
Memory Mr
Mood Mo
Table 4. City4Age Geriatric Risk Model
Each GES or GEF with no GES, represented in Table 4, is in then gauged by one or more Measures
(as defined in subsection 4.2.2 above), that represent particular embodiments of geriatric Items
associated to the relevant GEF/GES. The mapping among Measures and GEFs/GESs is illustrated in
Annex, in Section 10, while the definition of Measures collected at City4Age Pilot sites is reported in
Section 11.
It has to be noted that the model does not mandate which specific set of Measures, among those
available to assess a certain GEF/GES, should be used in practice: this choice has been left to Pilots, on
the basis of considerations regarding technical feasibility, cost of deployment, local needs, etc.
The ultimate objective of the above described effort is to obtain a reliable model that balances two
(somehow conflicting – see subsection 6.1) requirements:
Provide geriatricians and health professional with a tool capable of presenting them with a
comprehensive geriatric assessment along categories of factors (GEFs and GESs) that they
are already knowledgeable with and already know how to interpret, because they are
rooted in established, validated Instruments. Such requirement allows, for instance, the
immediate usage of the City4Age geriatric risk model in the Individual Monitoring
Dashboards (IMD), specified in Deliverable D2.13 Requirements, user scenarios and data
visualization mock-ups for apps/dashboards
Enact a truly “data driven geriatrics” where all feasibly and unobtrusively collectable
datasets – embodied in Measures associated to GEFs and GESs – are potentially
harnessed and made available to form the geriatric assessment, compared to a
conventional “Instrument driven geriatrics”, where only a restricted amount of data is used,
as limited by relevant cost and time constraints
As a final note, we highlight that the GEFs/GESs hierarchy presented in Table 4 also represents a first
tentative causal structure for a probabilistic model that can be used to support the application of an
alternative Machine Learning approach to risk assessment, anticipated in subsection 3.2.2 and further
discussed in subsection 6.4 below.
4.2.4 How to compute Geriatric Factors from Measures
Although the implementation of appropriate risk detection algorithms that, using the model illustrated
in sub-section 4.2.3 above, actually computes a relevant quantitative assessment of risk is a task to be
addressed in work-package WP5 (Real time data analytics and health risks detection), we propose in
this sub-section a first exploratory approach aimed at completing the previous discussion and showing
the model “in action”. Part of this approach will also be used to conduct a quantitative assessment of
the effectiveness of the geriatric risk model, as reported later in Section 5.
4.2.4.1 Measures
As previously mentioned, data is provided by the City4Age data collection platform in the form of
Measures, i.e. numbers produced periodically (in this discussion we assume that the period is once per
day, although other periods are possible), that correspond to unobtrusively measurable Items,
associated to the geriatric model’s GEFs and GESs.
Measures can be used to assign a synthetic value to a GES (or to a GEF with no GESs) through the
process illustrated below.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 28/183
As an example, suppose that a Pilot installation has chosen to measure the GES Phone usage with daily
measures PHONECALLS_PLACED, PHONECALLS_RECEIVED and PHONECALLS_MISSED
(refer to Section 11 for the definition of these Measures).
4.2.4.2 Compute Numerical Indicators (NUIs)
NUIs represent specific features extracted from Measures. We assume that NUIs are assessed monthly.
Given a certain Measure, like for example PHONECALLS_PLACED, examples of NUIs are:
AVG(PHONECALLS_PLACED): the average number of daily phone calls placed, across
the month
CV(PHONECALLS_PLACED): coefficient of variation of daily phone calls placed, across
the month
BEST(PHONECALLS_PLACED): best quartile divided by average of daily phone calls
placed, across the month
DELTA(PHONECALLS_PLACED): difference among best quartile and average divided by
average of daily phone calls placed, across the month
4.2.4.3 Compute monthly values for a Measure
This value is computed for any Measure, every month, on the basis of that Measure’s NUIs.
It is a real number between 1 (worst) and 5 (best).
It expresses the variation of the current month Measure with respect to comparison with a reference
month called M0.
It can be computed in the following way.
Let’s call all NUIs for a given Measure at the current month as Ni (e.g. N1 = AVG, N2 = CV, etc.).
Let’s call all NUIs for a given Measure at M0 as Ni M0.
For each NUI a Diff function is computed as follows: Diff(Ni) = (Ni – Ni M0) / Ni M0.
(Note: some NUI may be negative, i.e. best geriatric values correspond to lower numbers; in that case
the formula to use is Diff(Ni) = – (Ni – Ni M0) / Ni M0; for example, this may be true for the CV NUI, if
geriatricians consider that higher variability for the specific Measure is a bad geriatric sign)
An individual value for the NUI, T(Ni), is computed according to the following rule:
1, if Diff(Ni) < –0,25
2, if –0,25 < Diff(Ni) < –0,10
3, if –0,10 < Diff(Ni) < +0,10
4, if +0,10 < Diff(Ni) < +0,25
5, if Diff(Ni) > +0,25
Eventually, the monthly value for the Measure is computed as a linear combination of T(Ni):
V(Measure) = i wvi*T(Ni)
The weights wvi are normalized (i.e. i wvi = 1)
4.2.4.4 Compute a monthly score for a simple Factor (GES or GEF with no GESs)
This score is computed for the factor, every month, on the basis of that factor’s Measures monthly
values.
It is a real number between 1 (worst) and 5 (best).
It expresses the variation of the current month factor with respect to a comparison with a reference
month called M0.
It can be computed in the following way.
Let’s call all Measures’ monthly values for a given factor at the current month as Vi (e.g. V1 = value
for PHONECALLS_PLACED, V2 = value for PHONECALLS_RECEIVED and V3 = value for
PHONECALLS_MISSED).
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 29/183
Let’s call all Measures’ monthly values for a given factor at M0 as Vi M0.
For each Measure monthly value a Diff function is computed as follows: Diff(Vi) = (Vi – Vi M0) / ViM0.
(Note: some Measures are negative, i.e. best geriatric values correspond to lower numbers; in that case
the formula to use is Diff(Vi) = – (Vi – Vi M0) / Vi M0; for example, this is true for the Measure
PHONECALLS_MISSED)
An individual score for the Measure, T(Vi), is computed according to the following rule:
1, if Diff(Vi) < –0,25
2, if –0,25 < Diff(Vi) < –0,10
3, if –0,10 < Diff(Vi) < +0,10
4, if +0,10 < Diff(Vi) < +0,25
5, if Diff(Vi) > +0,25
Eventually, the monthly score for the factor is computed as a linear combination of T(Vi):
S(Factor) = i wsi*T(Vi)
The weights wsi are normalized (i.e. i wsi = 1)
4.2.4.5 Monthly value for a composite Factor (i.e. GEF with composing GESs)
It is computed as a linear combination of values of composing GES:
S(GEF) = i wfi*S(GES)
The weights wfi are normalized (i.e. i wfi = 1)
4.2.4.6 Conclusions
It is important to reiterate that the above formulation has been provided as a first reasonable hypothesis
to support work on WP5, particularly for the implementations of IMDs, as specified in Deliverable
D2.13 Requirements, user scenarios and data visualization mock-ups for apps/dashboards.
To progress on it, additional work has been completed in WP5 and WP7, for example to
experimentally assess the values of the different weights in the proposed linear combinations,
according to information provided by Pilot testbeds and accompanying geriatricians judgments.
However, the above notwithstanding, by adopting the proposed vision we are finally able to derive a
fully applied instance of the City4Age Geriatric Risk Model, that builds on four different
abstraction levels:
GEFs, which represent higher level geriatric domains, and are evaluated on the basis of27
:
GESs, which represent lower level geriatric domains, are evaluated on the basis of:
NUIs, which represent features periodically extracted from:
Measures, which represent datasets provided by the City4Age unobtrusive data collection
platform (WP3)
Such a model flexibly supports different functions, such as:
Mapping measures of health status to GEFs/GESs, synthesizing them in a way that makes it
easier for clinicians to understand and to trust them
Using NUI features to train, validate and test machine-leaned models for fully automated
detection of MCI/frailty onset
Using the GEFs/GESs/Measures structure to infer causal networks among different
determinants of MCI/frailty
The first line has been implemented in WP5, embodied in City4Age IMDs. The second line has been
implemented in an experiment that proves the model’s validity, presented in Section 5 below. The third
line has been proposed as a future research agenda, based on Bayesian Network analysis, in subsection
6.4 below.
27
As previously illustrated, some GEFs are evaluated directly on the basis of NUIs, skipping the GES level
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 30/183
5 Statistical assessment
5.1 Introduction
Previous Sections have illustrated how the City4Age Geriatric Risk Model has been formulated, based
on a comprehensive investigation of current clinical knowledge, accompanied by in-depth discussions
with the Project WP3 team and the Pilot testbed teams.
In this section we attempt to provide a statistical assessment of the Model, examining on how well it
can perform in practice, based on experimental data collected at Pilot sites.
The endeavour is approached as a classification problem: i.e. to verify if it is possible to build a
classifier that (i) receives as input Pilot Measures that have been collected and organized consistently
with the City4Age risk model (see 4.2.3 and 4.2.4), and (ii) produce as output a binary classification of
elderly participants into robust and non-robust subjects, which is statistically well correlated with
ground truth evidence obtained through a clinically validated indicator of functional robustness.
In the subsections that follow, we will first examine the datasets collected at each Pilot site, in order to
assess their quality in relation to the above mentioned task. Then, we select a Pilot on which to conduct
the experiment, illustrate in details how the classification problem was setup and run, present the
quantitative results that were obtained, and finally discuss their meaning.
5.2 Correlation analysis
As a preliminary statistical assessment, we look at the correlation matrices that illustrate the
dependencies among the City4Age risk model’s Measures, collected at each Pilot site.
Data for this analysis has been gathered in June 2018, based on the Measures collected by Pilot sites
and uploaded to the City4Age Shared Repository up to that date, which amounted to several months-
worth of information for each of them (see Table 5 below).
Pilot Start date Assessment date Months
covered
Subjects
covered
Athens April 2017 June 2018 15 40
Birmingham January 2017 June 2018 18 35
Lecce January 2018 June 2018 6 24
Madrid September 2017 June 2018 10 15
Montpellier November 2016 June 2018 20 18
Singapore September 2016 June 2018 22 15
Table 5. Pilot data from City4Age Shared Repository used for correlation analysis
All correlation matrices are included in Annex, in Section 12.
As it could be expected, several Measures are relatively highly correlated with one another, across
several Pilots, as for example walk_distance and walk_steps (they both represents the same distance) or
seniorcenter_visits and seniorcenter_time (more visits are associated to more time), etc.
However, more interesting patterns can be observed through careful analysis:
Data coming from fitness wristbands are in general among the most strongly correlated. This
can be seen more clearly in the matrix from the Birmingham Pilot (subsection 12.2) the
Measures of which are based on the commercially available Withings wristband
(https://www.withings.com/, formerly Nokia Health). As hinted, walk_distance and
walk_steps are almost deterministically linked, with a coefficient of 0,96. In addition, these
Measures are in turn correlated with other three Measures related to physical activity:
physicalactivity_soft_time, physicalactivity_moderate_time and physicalactivity_calories,
with coefficients ranging from 0,44 to 0,82. These latter Measures are also somewhat
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 31/183
mutually correlated, with coefficients greater than 0,3. Finally, the sleep quality Measures
sleep_awake_time and sleep_wakeup_num are strongly mutually correlated with a
coefficient of 0,76. A general conclusion to be derived is that market available fitness
trackers – although very convenient for their low cost and wider availability – provide an
informative content that is more limited than initially apparent from the number of collected
features that they advertise. In a sense, this limit has to be expected, if we consider that all of
the mentioned features are basically derived from just one type sensor, i.e. the three-axis
accelerometer integrated in the wristband.
A similar conclusion can be reached about smart-homes systems based on presence detectors
in rooms, as visible in the Singapore Pilot matrix, which mostly uses this type of sensing
equipment. Apart from the obvious correlations among visits to rooms and time spent in the
same rooms, data shows that strong correlations also exist among Measures relating to
different rooms: for instance kitchen_visits correlates strongly with livingroom_time
(coefficient 0,48), livingroom_visits (coefficient 0,67) and restroom_visits (coefficient 0,65).
In turn restroom_visits is strongly linked to livingroom_time (coeffient 0,52) and
livingroom_visits (coefficient 0,64). Again, this can be explained by the fact that what
presence detectors actually measure is the pattern of movements across the home
In comparison with the above, the correlation matrices of Pilots that have addressed a larger
number of “city-scale” activities – such as Athens, Lecce, Madrid and Montpellier – are
relatively more “white”, signalling that such Measures actually bring in greater variety,
associated to more diverse and informative content that can be used to distinguish
comparatively weaker signals of incipient health decay.
In conclusion, the correlation analysis supports the City4Age rationale, i.e. that behaviour detection in
cities is likely to increase the chances of implementing better systems for the early detection of aging
conditions, as compared to existing, market available solutions.
5.3 Machine learning experiment
In order to further assess if the model’s Measures discussed in previous subsections have sufficient
statistical power to actually detect health decays, we relied on a supervised machine learning
experiment.
The underlying idea is to demonstrate that the Measures of the City4Age risk model, when collected
and fed to an appropriately trained classifier, are capable to discriminate robust from non-robust
people, as determined by a clinically validated scale, with an accuracy sufficient to prove the concept.
To proceed along this direction, the following tasks had to be completed:
Select an appropriate machine learning scheme to be used
Define and collect ground truth labels
Chose a Pilot experiment to obtain training, validation and test datasets
Conduct feature extraction (i.e. synthesis and selection) and produce the base dataset
Run the experiment and collect results
Analyse and discuss results
These aspects are respectively presented in the next subsections.
5.3.1 Choice of a machine learning scheme
The choice of an appropriate learning scheme to be used in the experiment is dictated by the following
set of requirements:
The learner should be robust against missing data, as missing data is a relevant issue,
frequently found in real-life, experimental datasets, including City4Age Pilots’ datasets (e.g.
when participants forget to switch their smartphone on, or inadvertently deactivate the
City4Age app).
Given the number of available data-points at City4Age Pilot sites (several tens of
participants) the learner should not fractionate the dataset (as typical of some popular
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 32/183
schemes like decision trees or rule learners) as fractions would contain an insufficient
number of data-points for the scheme to work properly
The learner should be robust against irrelevant features as – given the investigation
purpose of this work – it cannot be established in advance what is the best combination of
relevant features. In fact, such combination should be discovered by trial and error, based on
the generation of many candidate features and the support of dimensionality reduction
algorithms, as common in many data mining efforts28
.
In consideration of the above, the Naïve Bayes scheme was selected as the algorithm of choice, as it
satisfies all of the three mentioned requirements.
The price that has to be paid in exchange for the above benefits is that Naïve Bayes assumes that
features are conditionally independent of one another, which is not true in many datasets. However,
two considerations suggest that this downside can be manageable:
the Naïve Bayes algorithm has shown to perform surprisingly well in practice, even in cases
in which the conditional independence assumption among features is clearly violated29
most “city-scale” Measures collected at City4Age Pilots are in fact mutually uncorrelated, as
confirmed by the analysis presented in the previous subsection
5.3.2 Definition and collection of ground truth
To apply a supervised machine learning approach it is necessary to select and enact a method for
measuring ground truth labels. The method should be based on a relevant geriatric instrument for
measuring robustness in elderly people, addressing the following three requirements:
The instrument must be able to detect incipient functional decline in elderly citizens, in
community settings
Its administration must not be burdensome for elderly participants at Pilot sites, and it should
be efficient for Pilot Partners to administer
It must have been clinically validated
These requirements are well addressed by the Functional Ability Index (FAI) 30
, an instrument defined
in the frame of the longitudinal cohort study LUCAS [Note: the City4Age Consortium came into
contact with LUCAS researchers during the face-to-face meeting of the EIP on AHA’s Action Group
D4 on Age Friendly Environments, held in Hamburg on 16-17 November 2017, confirming the
invaluable networking potential of the EIP on AHA31
. Usage of FAI in City4Age was subsequently
discussed in an email exchange with instrument’s authors, in the second half of November 2017].
The index consists of a questionnaire with 11 questions, that can be easily administered (or even self-
administered) without the application of complex equipment that require appropriate management (e.g.
Jamar hand dynamometers, chronometers, etc.).
The FAI instrument excludes disabilities and targets items that are related to the early signs of decay,
including – similarly to the City4Age risk model – outdoor mobility and social interaction.
Differently from other indicators, FAI considers two different types of factors: (i) functional risk
factors, addressed by 6 questionnaire items, and (ii) resources, also addressed by 6 items (one of the 11
questions has double weight, being considered for both types of factors, with inverted scores).
Combinations of fitness/non-fitness classes along these two dimensions produce 4 different
possibilities:
28
Witten et al. Data Mining: Practical Machine Learning Tools and Techniques, Morgan Kaufman, 2017
29 Domingos et al. On the optimality of the simple Bayesian classifier under zero-one loss, Machine Learning,
1997
30 Dapp et al. Long-term prediction of changes in health status, frailty, nursing care and mortality in community-
dwelling senior citizens – results from the longitudinal urban cohort ageing study (LUCAS), BMC Geriatrics,
2014
31 https://ec.europa.eu/eip/ageing/home_en
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 33/183
People high on resources and low on health risks (Robust)
People high on resources and high on health risk (postRobust)
People low on resources and low on health risk (preFrail)
People low on resources and high on health risk (Frail)
According to conventional approaches (e.g. Fried Frailty Index, see subsection 8.2.1), this introduces a
new class, postRobust, that, according to authors, is intended to distinguishing between fully robust
persons and those not fully robust, but better off than pre-frail, i.e. very early on functional decline.
The predictive power of FAI has been validated in the frame of the LUCAS longitudinal study, with
data spanning over a period of 93 months.
Before usage in City4Age, the FAI questionnaire text has been slightly modified, as administration
rounds where closer in City4Age (3 months, see below) with respect to LUCAS (several years).
Consequently, reference periods of 6 months and 12 months in the original FAI formulation had to be
respectively shortened.
Three City4Age Pilots were available to collect FAI data for their elderly participants. In particular:
The Athens Pilot collected FAI data for 40 participants, in two rounds: in January/February
2018 and April 2018, generating 80 ground truth labels
The Birmingham Pilot collected data for 15 participants in one round, in March 2018,
generating 15 ground truth labels
The Madrid Pilot collected FAI data for 15 participants in January 2018 (1 participant), in
February 2018 (11 participants, including a reassessment of the participant measured in
January), in March 2018 (3 participants) and in June 2018 (1 participant), generating 15
ground truth labels (the measurement obtained in January 2018 was discarded, as it was too
close to the reassessment of the same participant in February 2018, as confirmed by no
change in the classification)
[Note: what above reported is the status at the time of writing: more rounds of FAI measurement are
foreseen at the mentioned Pilot sites, during the last month of the Project work-plan].
A summary of collected ground truth information is presented in Annex, in Section 13.
5.3.3 Selection of a Pilot experiment to obtain training and test datasets
In order to ensure a higher level of homogeneity in collected data and in the procedures for ground
truth measurement, it was decided to conduct the analysis on a Pilot per Pilot basis.
Looking at available ground truth data for the three Pilots mentioned above (see Section 13) it becomes
evident that the best Pilot site to conduct the machine learning assessment is the Athens one. In fact:
The Athens Pilot offers the highest potential number of data-points, having collected ground
truth labels for 40 participants in two rounds (totalling 80 labels). This guarantees a more
solid base for the machine learning algorithm to pick up signals in the data.
The Athens Pilot offers a more balanced number of labels from different robustness classes,
while participants in Birmingham and Madrid are predominantly Robust people, with only
one in fifteen being postRobust, in both cases. This class imbalance would make the
application of a supervised machine learning approach more difficult.
In addition, looking at the correlation matrix for the Athens Pilot, reported in subsection 12.1
and discussed in subsection 5.2 above, it can be seen that this testbed has the further
advantage of including a larger set of mutually independent Measures, that is advantageous
when applying the Naïve Bayes scheme, as postulated in subsection 5.3.1.
5.3.4 Feature extraction and dataset preparation
Having selected the Pilot experiment to work on, the next step is to decide about the features that will
be provided as input to the machine learning algorithm.
Features need to correspond to the City4Age risk model Measures (as collected at the Athens Pilot, in
this case) since the purpose of the experiment is the statistical assessment of the model’s detecting
power.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 34/183
The application of a supervised learning approach imposes the following additional considerations:
Feature synthesis: the Measures collected at the Athens Pilot belong to the City4Age
Measures dictionary, reported in Section 11. As illustrated in that Section, these Measures
are collected periodically, with different time granularities, down to daily frequency. A
method should be devised in order to synthesize the Athens Measures into relevant features,
that need to be associated with ground truth labels which, as mentioned in subsection 5.3.2,
have been gathered in two rounds, distanced approximately three months from one another
Feature selection: not all Measures are worth including in the experiment’s dataset. In
particular, having chosen to rely on a Naïve Bayes learner, it is beneficial to exclude all
those features that are believed to be obviously correlated with other ones, based on pre-
existing domain knowledge
Data cleaning: differently from other City4Age data analytics components developed on top
of the geriatric risk model, such as for instance the Individual Monitoring Dashboard, where
data interpretation is filtered by a domain expert that can apply her/his knowledge and
judgment to get around unclean data (missing data, outliers, etc.), in a data mining approach
an appropriate data cleaning process needs be run in advance, before feeding the data to the
learning algorithm
To illustrate in detail how these considerations have been applied in the present work, it is convenient
to examine the SQL query that has been used to extract the Athens Pilot’s dataset, in the first step of
the learning workflow. The text of the query is reported in Annex (Section 14), while the following
paragraphs illustrate the pre-processing operations that the query implements.
5.3.4.1 Features synthesis
Having held two rounds of ground truth collection based on FAI measurements with a distance of
approximately three months from one another (see subsection 5.3.2), it has been decided to associate
each FAI label with features that synthesise the City4Age risk model’s Measures collected at the Pilot
site in the trimester preceding the FAI assessment [Note: FAI scores in Athens were collected on
slightly different dates for different subjects: they have been considered as related to a given trimester
if measured within a temporal distance of less than 15days from the trimester’s end]. Such set of
synthesised features with the associated FAI labels, constitute the individual data-points to be used for
training and testing the machine learning algorithms.
Consistently with the approach suggested in subsection 4.2.4, the synthesis has been obtained by
computing the following values for each model Measure:
Average of Measures’ values collected across the trimester: this parallels the “AVG” NUI in
subsection 4.2.4
Standard deviation of Measures’ values collected across the trimester: this parallels the
“CV” NUI in subsection 4.2.4
Best quartile of Measures’ values collected across the trimester: this parallels the “BEST”
NUI in subsection 4.2.4
The “DELTA” NUI, which is also considered in subsection 4.2.4, has not been included here because,
being deterministically dependent on two of the above listed NUIs, it would have risked to confuse the
Naïve Bayes learner that, as mentioned, is based on the conditional independence assumption.
The above computations produce three features for each Measure, as discernible in the SQL query
reported in Section 14. For example, in the case of the cinema_time Measure, the query extracts the
three features, cinema_time_avg, cinema_time_sd, and cinema_time_best in the following way:
avg(cinema_time) as cinema_time_avg,
stddev(cinema_time) as cinema_time_sd,
percentile_cont(0.75) within group ( order by cinema_time ) as cinema_time_best
[Note: for most Measures, the best quartile is the 75th percentile; however, for home_time_best,
pharmacy_time_best and pharmacy_visits_best the best quartile is the 25th percentile, as these
Measures are negatively correlated with geriatric outcomes in the City4Age risk model].
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 35/183
The grouping by trimesters can be discerned in the query in Section 14, in particular (a) in the selection
of the time span covered by the query:
date(time_interval.interval_start) >= '2017-11-01' and
date(time_interval.interval_start) <= '2018-04-30'
(b) in the assignment of the temporary field trimester:
case when time_interval.interval_start <= '2018-01-31' then 1 else 2 end as trimester
and (c) in the grouping over trimester:
group by ct.cr_m, ct.trimester
5.3.4.2 Feature selection
As anticipated, some Measures have been excluded from the dataset because of their deterministic
correlation with other ones. In particular, among the 24 Measures collected at the Athens Pilot (listed in
subsection 12.1) the following 6 have been discarded:
cinema_visits_month and pharmacy_visits_month have been discarded because they report
the same information as, respectively, cinema_visits and pharmacy_visits (just differently
grouped along time)
supermarket_time and supermarket_visits have been discarded because they contain the
same values as, respectively, shops_time and shops_visits
walk_distance_outdoor_fast_perc has been discarded because it reports information which is
already conveyed by the best quartile feature of walk_distance_outdoor
walk_distance_outdoor_slow_perc has been excluded because it is deterministically
computed from walk_distance_outdoor_fast_perc (namely, 100% less
walk_distance_outdoor_fast_perc)
As a result of the selection, a total of 18 Measures remained, which generated a total of 54 features,
which are visible in the main select clause of the query in Section 14.
5.3.4.3 Data cleaning
As common in data mining workflows, especially those based on real-life experimental data, a data
cleaning step must be applied to remove values that are evidently the result of measurement errors.
In particular, the following data errors are targeted in the query in Section 14:
Missing data: this is a classical issue, present in all non-trivial data mining efforts. In the
case of City4Age experiments, the main cause of missing data is participants forgetting to
switch on or bring with them the smartphone or deactivating the City4Age georeferenced
app that enables the unobtrusive collection of model’s Measures. In the current experiment,
point-wise missing data were not considered problematic, as the Naïve Bayes scheme is
quite robust against this problem. However, care recipients that did not activate the
smartphone and/or the Ciy4Age georeferenced app for the whole extension of one trimester
where excluded from that trimester. This is done by checking the following condition over
the trimester group, assuming that being apparently at home for the whole trimester duration
would imply, with probability close to 1, that the City4Age app was simply never used:
having avg(home_time) != 86400. As a consequence, out of the 80 potential data-points
labelled in Athens, 34 had to be discarded, leaving a dataset with 46 valid data-points
Duplicated Measures: this issue is specific to the Athens Pilot and regards the Measure
restaurants_visits_week, which is reported twice with different values for few days. Analysis
revealed that the doubly reported values actually represent the cumulated sum of previous
weeks’ values, reported monthly. The problem was easily solved by grouping double
Measures (clause group by time_interval.interval_start, care_recipient.id,
variable.detection_variable_name in the query) and selecting the minimum value (field
min(measure.measure_value) in the query, the larger value being the incorrectly reported
monthly sum)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 36/183
Outliers, another typical issue in data mining, are also likely to be present in the City4Age datasets,
including in the Athens Pilot dataset. However, in the current experiment they were not corrected as:
They generally represent a relatively low share of the total number of samples (few
percentage points, on a rough estimation)
Although few values are clearly outside reasonable ranges, it is still not easy to determine a
clear-cut threshold, aimed to discard some values while keeping others (see considerations in
subsection 6.3)
It is expected that the probabilistic Naïve Bayes learner would help in mitigating outliers’
effects on the overall results.
5.3.5 Running the machine learning workflow
The machine learning experiment was run within the Weka data mining environment32
.
An additional pre-processing step was conducted with the Weka MergeManyValues filter, in order to
combine the postRobust, preFrail and Frail labels into a comprehensive nonRobust class, to be
separated by the classifier from the Robust class. This has been done for two reasons:
The limited size of the dataset makes it difficult to train a classifier with several class values
The rationale of the City4Age project is to discriminate robust persons from those that are
affected from early health decay, which is consistent with a binary classification problem
After execution of this pre-processing step, the final Athens Pilot dataset was produced, including 46
data-points, with 54 features and a binary class attribute each.
A first glance at the distribution of class values across individual features seems to confirm the
presence of informative content. For example, the following Figure 4 illustrates how the
seniorcenter_visits_avg feature can reasonably differentiate between Robust elderly citizens (who visit
the Athens Friendship Clubs, on average, up to every other day) and nonRobust elderly citizens (who
visit the Friendship Clubs every 5 days on average, at most). Similarly encouraging patterns can also
be observed in many of the other features.
Figure 4. Class distribution for the seniorcenter_visits_avg feature
32
https://www.cs.waikato.ac.nz/ml/weka/
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 37/183
On that basis, a learning experiment was thus set up and run with the following characteristics:
Execution of a preliminary dimensionality reduction step, through the Weka
PrincipalComponents filter, that operates Principal Components Analysis (PCA) before the
features are fed to the learner. This step is useful to further reduce redundancy among
features, and make them more suited to Naïve Bayes learning. Different values for the
varianceCovered parameter (the algorithm selects the smaller possible set of components
that covers at least this variance) have been tried in order to determine the most performant
setting. In particular, the following values have been tried: 0.75, 0.8, 0.85, 0.9, 0.95 plus no
PCA at all.
Running a Naïve Bayes scheme instance, for each different value of variance covered, as
above illustrated, and selection of the one with the best performance. The result of this step
is illustrated in Table 6. Performance has been measured in terms of AUC33
. Performance
assessment in each run was conducted through 10-fold cross-validation, to prevent
overfitting. Significance is assessed by using the corrected resampled paired t-tester34
, to
mitigate the issue of cross-validation folds insisting on the same underlying dataset. The
table shows that PCA is a useful step that improves performance, and that the best value for
the varianceCovered parameter is 0.8, corresponding to the selection of 7 components (see
second row in the table, shaded)
Running an outer, 10-times, 10-fold cross-validation testing experiment with the same
model, to obtain an unbiased quantitative measure of the performance for the above
described workflow, using different datasets for validation and testing to avoid “peeking at
data”. In particular, the experiment was conducted by using a modified version of the Weka
MultiScheme meta-learner, that uses AUC rather than percent correct to compare schemes
(renamed for this reason as MultiSchemeAUC, see Section 15).
The resulting unbiased performance estimate, obtained from the last step listed above, yields an
AUC = 0.710 different from the null hypothesis (random classification) with significance < 0.05,
which represents a convincing proof-of-concept for the City4Age Geriatric Risk Model
effectiveness.
Variance covered by
PCA step
(at least)
Number of
components selected
by PCA step
AUC Significantly different
from classification by
chance?
(significance level = .05)
0.75 6 0.71 yes
0.80 7 0.72 yes
0.85 7 0.71 yes
0.90 9 0.70 no
0.95 12 0.66 no
No PCA n/a 0.69 no
Table 6. Performance versus different values of variance covered
Starting from this persuasive result, seven additional experiments have been conducted, in order to
further investigate the validity of the City4Age risk modelling assumptions, by tweaking the feature
extraction process and observing resulting changes in performance.
In particular, the following modifications have been tried:
33
https://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_the_curve
34 Nadeau et al. Inference for the generalization error, Machine Learning, 2003
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 38/183
1. Remove standard deviation features: this correspond to removing the “CV” NUI from the
model
2. Remove best quartile features: this correspond to removing the “BEST” NUI from the model
3. Substitute average features with corresponding median features (i.e. 50th percentile): this is
to try a different way to measure central tendency (“AVG” NUI in the model)
4. Remove all “city-bound” features, i.e. only home-bound (home_time) and fitness tracker
provided (walk_*) features where left in: this is to check the relative importance of market
available unobtrusive data collection with respect to the full City4Age risk model
5. Only leave “city-bound” features (the complement of the previous selection): this is to check
the relative importance of smart-city-only unobtrusive data collection with respect to the full
City4Age risk model
6. Merge together all socializing Measures (other_social_*, restaurant_*, seniorcenter_*) in a
single Measure: this is to check if more dense but coarser features are preferable to the
sparser (i.e. with “many zeros”) but more granular ones currently included in the City4Age
risk model
7. Combine paired *_time and *_visits Measures in order to produce 6 additional
*_time_per_visit, features: this is to assess, with an example, the opportunity to further tune
model’s Measures, in order to improve machine learning algorithms performance in
extracting meaning from them (the query modification needed to add these new features is
reported in Annex, in Section 16)
The results from these additional experiments are summarized in Table 7 below. Row 0 in the Table
refers to the original experiment, for sake of comparison. We remind that all experiments presented in
the table have been conducted according to the previously described workflow, summarized at the
bottom of the table.
# Experiment description AUC Significance
0 Original model, with all 54 features
included
0.710 0.05
1 Removing standard deviation
features
0.674 0.1
2 Removing best quartile features 0.689 0.1
3 Substitute average features with
median (50th percentile) features
0.671 0.1
4 Remove city-bound features 0.668 0.1
5 Leave only city-bound features 0.648 0.2
6 Merge socialization features 0.616 0.3
7 Add *_time_per_visit features 0.715 0.05
All experiments are based on:
PCA with best coveredVariance parameter, selected in the range {0.75, 0.80, 0.85, .90,
0.95, no PCA}
Naïve Bayes learning with:
o inner 10-fold cross-validation for parameter selection, based on best AUC
o outer 10-times, 10-fold cross-validation for unbiased performance assessment
Significance calculated with corrected resampled paired t-tester
Table 7. Experiments to assess the City4Age Geriatric Risk Model
In the following subsection, the above reported results are discussed and general conclusions on the
validity of the City4Age Geriatric Risk Model are offered.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 39/183
5.3.6 Discussion
The first conclusion to be derived from the quantitative results presented in the previous subsection is
the confirmation of the validity of the City4Age Geriatric Risk Model.
Row 0 in Table 7 clearly shows that Measures established by the City4Age risk model are able to
significantly discriminate among Robust and nonRobust subjects, and thus support the enactment of
the early detection paradigm that is central to the Project’s rationale.
The following Figure 5, that illustrates a Receiver Operating Characteristic (ROC) curve from one
experiment run, visually shows how the City4Age risk model behaves differently from the null
hypothesis (i.e. random classification – red line in the Figure, corresponding to AUC = 0.5).
Obviously, the resulting AUC performance, although very good, does not yet achieve the levels
required for a market deployable system. However, it clearly represents a convincing proof-of-concept
that confirms the success of the City4Age modelling endeavour and lay the basis for its industrial
scale-up (see next Section).
Figure 5. Example of ROC curve from one experiment run
Other notable conclusions can be derived from the results of the additional experiments, reported in
rows 1-7 of Table 7:
Removing standard deviation or best quartile features or substituting average features with
modal ones (rows 1-3 in Table 7) decreases both AUC performance and statistical
significance, confirming that the NUIs proposed in subsection 4.2.4.2 are a sound base
hypothesis
Removing either city-bound or non-city-bound features (rows 4-5 in Table 7) decreases both
AUC performance and statistical significance, confirming the assumption of the City4Age
model that both types of features should be used in order to accurately detect health decay
Merging socialization features (that contain several daily zero values) in a single
consolidated feature to reduce sparseness (row 6 in Table 7) decreases both AUC
performance and (dramatically) statistical significance. This effect vindicates the City4Age
risk model assumption that it is better to collect more granular Measures, even at the cost of
more sparseness, as differences in their expression can be meaningful for the detection of
decaying conditions
Although the signal is relatively weak (a 0.05 increase in AUC with no increase in
significance), row 7 in Table 7 also deserves attention, as it highlights that the City4Age risk
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 40/183
model may still be augmented with additional interesting Measures, which have escaped the
first drafting phase of the model and can contribute to improve its accuracy
A last interesting consideration that can be derived from experimental data, regards putting into
perspective the number of Measures that should be included in a model for screening frailty/MCI,
depending on (i) the applied data collection method (obtrusive vs. unobtrusive) and (ii) the size of the
datasets used to train it. The landmark, non-technology-based Fried Frailty Index includes 6 Items,
collected through self-reported questions and meter-based measurements covering 5 domains (see
subsection 8.2.1.3). The Index was derived from a cohort of more than 5,000 subjects (see subsection
8.2.1.2). On the other hand, for example, the study discussed in subsection 8.1.8 shows the possibility
to predict MCI on the basis of a single Measure (gait speed) with an SVM learner that achieves an
AUC = 0.81 when trained on 984 data-points. However, although such measure was collected through
automatic sensing equipment, it still required a specific “lab-like” installation in the subject’s home,
that is more obtrusive than allowed by City4Age requirements. Finally, the work discussed in this
Section yields AUC = 0.71 with a Naïve Bayes learner trained on 46 data-points based on a set of 18
Measures, which cover 9 geriatric factors. All Measures are collected in a fully automated,
unobtrusive way and all of them are necessary to achieve good performance, as shown by experiments
#4 and #5 in the previous subsection.
Thus, it seems that when committing to an automated, unobtrusive data collection approach, the
number of Measures to be included in the model needs to be relatively increased. This is
reasonable, as unobtrusive measurement is likely to provide less information content “per Measure”,
with respect to manual or technically more burdensome, but more precise methods; and this
disadvantage must be counterbalanced by a corresponding increase in the number of Measures used.
From a cost point of view, this is not an issue as City4Age demonstrates that modern technologies do
allow to collect larger numbers of Measures at much lower cost (for instance, the 18 Measures
collected at the Athens Pilot were all obtained from just a single, georeferenced smartphone app).
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 41/183
6 Lessons learned and recommendations
Critical analysis of the model presented in Section 4 above and, in particular, its application to the
Pilots’ testbed cases, allowed to point out several areas that may be addressed to further advance and
improve the City4Age risk modelling strategy.
Results of such analysis are presented in this Section.
In particular, subsection 6.1 provides an overview of the lessons learned during the development of the
City4Age Geriatric Risk Model, that provide the basis for the recommendations presented in the rest of
the section.
Such recommendations are subsequently articulated as follows:
subsection 6.2 presents a detailed study of the coverage of the geriatric risk model in relation
to the data collection technologies deployed at the Pilot testbeds and discusses additional
research that can be implemented to improve the situation
subsection 6.3 presents possible technical improvements, relating to the way current
measures are collected, that can have positive impact on the model’s accuracy
as anticipated in subsection 3.2.2 and building upon the encouraging results presented in
Section 5, subsection 6.4 proposes the investigation of a new approach to geriatric risk
modelling, based on machine learning methods
subsection 6.5 proposes a sequence of steps that Partners interested in the commercial
exploitation of the City4Age risk model may want to apply in order to achieve industry-
grade accuracy, and move beyond the proof-of-concept milestone that was the main
objective of the Research and Innovation Action conducted within the City4Age Project
6.1 Lessons learned
In this section, we review the whole process that has been followed to address health risk detection,
from the initial conception of the geriatric risk model, to its further review based on additional insights
obtained from the implementation of Pilot sites’ data collection platforms, to the final assessment based
on experimental data generated by Pilots. This review allows to identify difficulties that have been
encountered and ways to overcome them.
During the course of the Project, work has proceeded along the following three phases:
In a first phase, as illustrated in Section 4, the effort has been focussed in identifying the
largest possible number of domains to be included in the model, by looking at established
geriatric knowledge, crystallized in existing instruments for detecting MCI and Frailty,
widely used in clinical practice. This decision has been driven by two observations: (i) that
“automating” the measurement of individual, established geriatric instruments is difficult,
since they are normally characterized by few domains, of which even fewer are addressable
with unobtrusive data collection technologies, and (ii) that the parsimoniousness of existing
instruments is not due to diminishing returns when collecting more data, but to the need to
contain the relatively high costs associated with conventional, “manual” data collection. The
approach proposed for City4Age has a double advantage:
o as it still relies on valid geriatric knowledge expressed in several established
instruments, it is clinically strong and minimizes the risks of moving along
medically unsound directions
o it follows a consistent strategy, common to many data analytics efforts, based on
initially widening the number hypothesised features, so as to maximize the
probability that good ones are not left out of consideration
In particular, the result of the first phase led to the identification of 45 geriatric factors, that
were then reorganized around 10 high level factors, 8 of which subdivided in 43 sub-factors (2
higher level factors do not have sub-factors). Each sub-factor (or factor without sub-factors) is
in turn linkable to several ways to measure it through unobtrusive technology. It is worth to
note that such reorganization also represents a first attempt at eliciting a causal model of
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 42/183
robustness in elderly citizens, conducive to more sophisticated data analysis strategies, e.g.
based on Bayesian Networks (sub-section 6.4).
In a second phase, the risk model obtained as above described was discussed with Pilots
Partners, in particular to verify the possibility to implement it on top of the specific data
collection platforms planned at the 6 different City4Age testbed sites. This work led to
several important results:
o The City4Age risk model was completed with the addition of a dictionary of 107
candidate measures – linked to factors and/or sub-factors, as previously mentioned –
to be potentially implemented at testbeds
o It turned out that Pilots were able or willing to collect relatively small subsets of the
above measures, targeting a correspondently restricted subset of geriatric factors
o It also turned out that geriatricians, although recognizing the vast potential of
counting on bigger datasets, had difficulty in considering a large number of
individual factors and measures, having been trained to mostly assess synthesized
scores, like those produced by currently used instruments (e.g. those reported in
Section 8)
These issues led to the elaboration of Pilot specific configurations of the risk model, to be used
in Individual Monitoring Dashboards (IMDs – developed within WP5 and discussed in that
work-package’s deliverables) that could be flexibly adapted to both (i) the factors and
measures actually collected at each Pilot site and (ii) the specific needs expressed by local
clinical and/or geriatric staff. An example of such configuration can be seen in Table 8 below,
for the Athens Pilot case discussed in Section 5. The underlying assumption is that a well-
crafted IMD configuration would make it possible to synthesise the right information needed
by medical professionals at each specific site, for efficiently assessing the robustness level of a
subject and to decide about appropriate interventions.
Two main difficulties had to be dealt with, while applying this approach:
o devising the right IMD configuration is a challenging task, since it implies deciding
on a large number of parameters (weights relating measures with sub-factors, and
sub-factors with factors, as illustrated in subsection 4.2.4) which represent a
relatively new and unfamiliar terrain for geriatricians
o the fact that – although IMD assessment represents a clear step forward in the
direction of a much earlier and efficient detection for frailty or MCI, when compared
with the existing situation – the careful analysis of the IMD for a certain subject still
represents a complex action to be manually carried out by a medical professional
(e.g. the GP), which is relatively hard to scale up without more automated support
In a third phase, when experimental Pilot data have become available for analysis, a
reassessment of the above described scenario has been possible, that led to the following
conclusions regarding the application of the City4Age risk modelling in practice:
o As discussed in Section 5, it has been proved that even a relatively limited set of
measures and factors (18 measures covering 9 sub-factors, in the case of the Athens
Pilot) are in principle capable to act as a kind of “behavioural marker” that, while
not sufficient for a full diagnosis of frailty or MCI, can still be validly used to
“screen out” people who are robust enough to exclude, with reasonable confidence,
that they need further medical attention. If this approach is properly scaled up (e.g.
by applying the recommendations suggested in the next subsections) a new, data
driven approach to screening for frailty could be developed, that overcomes one of
the main obstacles that have so far led geriatric societies to recommend against such
screening – i.e. the expensiveness vs low specificity of currently available
instruments35
.
35
It is worth to note that geriatric societies invite to reconsider the recommendation against screening for frailty if
a “suitably validated electronic frailty index” could be devised, i.e. the very concept proved in Section 5; see for
example BGS conclusion in https://www.bgs.org.uk/resources/introduction-to-frailty
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 43/183
o At a second level, only those subjects for which the above mentioned “screen out”
procedure yielded a positive result, would undergo a more in-depth analysis,
conducted through IMDs at the GP practice, as above illustrated. Such IMD-based
examination would still be a preliminary assessment (with respect e.g. to a full
CGA), but it can be conducted quickly and relatively inexpensively, as (i) data has
already been collected and synthesised (respectively) by the City4Age data
collection platform and the City4Age IMD, and (ii) only the time needed for the GP
to review IMD information regarding the few “screened in” cases has to be
accounted for. In particular, in this scenario, the purpose of the IMD would be to (a)
provide a potentially larger “quantity” of information to support patient
examination, as collected in smart environments; while at the same time (b) provide
clinicians with an organized synthesis of such information, based on the
GEF/GES/Measure hierarchy and a specific “local” configuration, that make it
easier for them to navigate such more complex knowledge asset
o Only the small number of subjects that are “red-flagged” by the IMD assessment
would then be referred to face-to-face visits at the geriatrician’s or neurologist’s
practice, for further assessment and, possibly, for a CGA.
This two-level screening procedure can contribute to increase overall specificity while still
significantly containing costs, since false positives that pass through the automated “screen
out” procedure will most likely be filtered out by subsequent IMD analysis.
Geriatric factors Measures NUIs
Motility
AVG
SD
BEST_QUARTILE
for each Measure
Walking WALK_DISTANCE_OUTDOOR
WALK_TIME_OUTDOOR
WALK_SPEED_OUTDOOR
Basic Activities of Daily Living
Going out –HOME_TIME36
Instrumental Activities of Daily Living
Shopping SHOPS_VISITS
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
Transportation TRANSPORT_TIME
Socialization
Attending senior centers SENIORCENTER_VISITS
SENIORCENTER_VISITS_WEEK
Attending other social places OTHERSOCIAL_VISITS
OTHERSOCIAL _TIME
Restaurant RESTAURANTS_VISITS_WEEK
RESTAURANTS_TIME
Cultural engagement
Visit entertainment / culture
places CINEMA_VISITS
CINEMA_TIME
Health – Physical
Visit to health related places –PHARMACY_VISITS
–PHARMACY _TIME
Table 8. City4Age Geriatric Risk Model configuration for the Athens Pilot case
36
A minus sign indicates that the Measure is negatively correlated with geriatric outcomes
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 44/183
The claim resulting from the above described work is that the City4Age Geriatric Risk Modelling is a
flexible tool, with proven validity, that can be used in different ways, adapted to specific application
cases and implementing multi-level approaches to screening for robustness in elderly populations, that
can optimize healthcare management.
We also claim that the City4Age Geriatric Risk Modelling is an innovative, first-in-the-art
contribution to the field, as:
(1) to the best of our knowledge, at the time of writing no other effort exists that targets the early
detection of age related health decay through city-scale behaviour analysis (e.g. visiting shops,
using transportation means, socializing, engaging in cultural activities, etc.). In fact, current
efforts are mostly focussed on “home-scale” behaviours (e.g. ADLs) detected through smart
home kits, or “body-scale” behaviours (e.g. walking, physical activity, sleep quality, etc.)
detected through fitness tracking apps
(2) our experiments have shown that city-scale behaviour analysis can achieve superior
classification performance, with respect to “home-scale” and “body-scale” only approaches.
On the other side, it is acknowledged that the above results have been achieved in the form of a “proof-
of-concept” scheme, which is satisfactory for a low-TRL, Research and Innovation effort like
City4Age, but that needs to be improved in several directions to achieve the higher TRLs that are
needed for large scale market deployment. Many opportunities to proceed along this path have been
discovered during the Project research work, and are discussed in details in the following subsections.
6.2 Analysis of coverage from Pilots
Table 10 in Section 10, lists the geriatric factors included in the City4Age risk model and, for each of
them, reports which Measures – obtained from the data collection platforms deployed at Pilots – can be
used to assess them (e.g. according to the guidelines provided in sub-section 4.2.4 above).
For each factor, the union of all the Measures collected at each of the six project Pilots is considered,
arguing that if one Measure could be feasibly deployed at one site, it could be similarly deployed at
another one, and thus the union of all of them realistically represents the potential offered by the
City4Age project to any new City wishing to install the system – for example, during the exploitation
phase, after the end of the Project.
By looking at Table 10, we can infer a good general coverage for the model, through both direct and
proxy Measures, and, in the case of some factors, even an overabundant one.
On the other hand, there are still areas where coverage is absent or poor, relying only on very few or on
only tentative Measures.
In the following paragraphs, we review the most important of these areas and propose interventions
that, if applied, would correct the situation.
Gait balance
GES Gait balance is not addressed by any Pilot, although this is an important factor, that summarizes a
considerable number of Items, as discussed in sub-section 4.2.3 above.
The current ways to assess such factor are varied: through reports from the monitored persons (e.g.
Items Mt.Ne.03, Mt.Ne.04 or Mt.Ne.05), through direct observation (e.g. Mt.So.02) or through a
meter-based measurement (e.g. Mt.Ed.11).
However, since the advent of affordable accelerometers (such as those integrated in smartphones or
fitness wristbands), several authors have researched the usage of such unobtrusive gauges in order to
assess gait balance, for instance to address the needs of Parkinsonian patients’ treatment.
A relevant survey in this direction, conducted by Hubble and co-authors, has been published on PLoS
One in 2015 (Wearable Sensor Use for Assessing Standing Balance and Walking Stability in People
with Parkinson’s Disease: A Systematic Review37
). Authors have found that, out of 26 papers surveyed,
addressing gait balance measurement through wearable sensors, 31% were of low methodological
37
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4403989/
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 45/183
quality, 58% were of moderate methodological quality and 11% were of high methodological quality.
81% of the studies reported differences in sensor-based measures of standing balance or walking
stability between different groups of Parkinson’s disease patients and/or healthy controls.
Yang and co-authors have studied the application of the principle of accelerometry measurement to a
wide variety of balance and postural problems, as those addressed by the factors included in the
City4Age risk model (not limited to Gait balance) including posture and movement classification,
estimation of energy expenditure, fall detection and balance control evaluation. Their paper, published
in Sensors in 2015 (A Review of Accelerometry-Based Wearable Motion Detectors for Physical
Activity Monitoring38
) also mentions the possibility to automatically obtain TUG test timings.
Clearly, the above mentioned papers report about research endeavours that certainly need more work to
be transferred into practice; however, they still point to stimulating directions that could be worth
investigating in further research efforts on geriatric risk modelling.
Basic ADLs
Basic ADLs are still poorly covered: out of six Basic ADLs identified in the geriatric practice only one
(Going out) is reliably measured in City4Age testbeds. Some Pilots have proposed to address three
more (Bathing and showering, Personal hygiene and grooming and Toilet hygiene) with tentative
proxies; however their reliability is limited. Two remaining ADLs are not addressed by any Pilot
(Dressing and Self-feeding).
The general problem with Basic ADLs is that they are composed by sets of many granular actions that
are difficult to reliably detect through sensors (as an example, consider the description for Item
Ad.Da.13: “The person is taken to a bathroom and asked to take the cap off a tube of toothpaste, put
toothpaste on a toothbrush, turn on the tap, brush teeth, dampen washcloth, put soap on washcloth,
wash the face and turn off the tap”).
In fact, this level of behaviour requires a “dense sensing environment” in order to be effectively
captured, i.e. an environment where sensors are attached to almost all objects that are related to the
activities under monitoring.
City4Age’s component HARS (Human Activity Recognition System), developed within work-package
WP5, has been shown to be very successful at detecting activities in a dense sensing environment,
including recognizing many geriatrically meaningful ones, such as “Use toilet” (addressing GES Toilet
Hygiene), “Take shower” (addressing GES Bathing and showering), “Get dressed” (addressing GES
Dressing) “Brush teeth” (addressing GES “Personal hygiene and grooming”)39
.
However, few City4Age Pilots are currently equipped with the dense sensing infrastructure needed to
recognize such activities, as this is typically found in highly advanced AAL dwellings, such as some
nursing homes, while it is more difficult to implement in conventional homes or in City
neighbourhoods.
Given the above difficulties, a possible innovative way to be investigated in order to recognize Basic
ADLs without requiring dense sensing, is to appropriately filter accelerometer data, detected (as
previously mentioned) by wearables such as fitness wristband or smartwatches.
For example, the already mentioned Withings smartwatch trackers have been trained to recognize more
than 10 sports activities in this way (e.g. run, swim, ping-pong, volleyball, dancing, etc.40
). A similar
direction could be investigated for recognizing Basic ADLs.
38
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3231187/
39 Relevant details are reported in Deliverable D5.3 The activity recognition system for smart cities, v3
40 It is worth to note that one of the City4Age Pilots is using this capability of Withings trackers in order to
measure GEF Physical Activity
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 46/183
Instrumental ADLs
Only four Instrumental ADLs are convincingly covered, while one is addressed through a tentative
proxy (Laundry) and four are not addressed by any Pilot (Housekeeping, New media communication,
Financial management, Medication).
Some of the Instrumental IADLs can be analysed with the HARS system. For example, one particular
good candidate could be GES Housekeeping, given the involvement of several electrical appliances,
the usage of which can be detected for instance through easy-to-deploy smart-plugs and the
involvement of patterns of rooms changes, which are detected at several City4Age Pilots. Such
elements might make available the right kind of “densely sensed” elementary actions that are needed to
effectively drive HARS.
GES New media communication could be relatively easily implemented, given the fact that this activity
actually “happens online”.
Context and status factors
GEFs in the Context and status factors class (Environment, Dependence, Health-Physical and Health-
Cognitive) are also less covered, with many GESs not addressed by any Pilot and several of them
addressed through obtrusive measurement. However this issue is less essential since – as mentioned in
sub-section 4.2.3 – these factors are mostly intended to be measured with very low frequency (typically
yearly) and thus are not a central part of the continuous unobtrusive monitoring paradigm of City4Age.
6.3 Technical improvements
In addition to improving model coverage, benefits can be extracted also by making factors detection
more accurate through technical improvement in the derivation of underlying measures.
Here we mainly refer to some possible developments that are suggested by the experiment conducted
on the Athens Pilot site, illustrated in Section 5, but the approach can be generalised to all Pilot sites, in
order to benefit potential new instances of the City4Age platform.
As already mentioned, outliers present in experimental data can reveal areas where sensor
improvement is possible.
For instance, Figure 6 below shows the distribution of measure restaurant_visits_week across all care
receivers in the Athens Pilot41
. As it can be seen, one particular care receiver seems to have visited a
restaurant 42 times in a single week (third from left).
This value may be an outlier (it means 6 restaurant visits per day) that may signal a sensor
malfunctioning (e.g. the person is living close to a restaurant and the system misfires
POI_ENTER(location_type: restaurant) LEAs when the persons passes by). However, things are not so
clear as it may seem at a first glance: it may be that the person’s family owns a restaurant and the
system is misclassifying visits to family as visits to restaurants; in this case the sensing equipment
would be OK, but the geriatric value of the measure becomes questionable. Other care receivers in the
Figure are difficult to assess, too. For example, the second more frequent visitor to restaurants accounts
for a week with 13 visits: is this an outlier, signalling sensing issues? Or is she simply a person that has
the habit to have daily meals outside home? Would this be still meaningful from a geriatric point of
view?
Answering these questions would require a deeper analysis of the outliers cases, with a joint technical
and geriatric assessment, to make sound decisions. Such decisions could also lead to modifications in
the measures’ dictionary reported in Section 11 (for example, restaurant_visits_week may be deemed a
geriatrically valid measure up to 14 visits, while after that threshold it should be considered as
missing).
Other issues are more straightforward and may just require better filtering. For instance, Figure 7
shows that the distribution of the walk_speed_outdoor measure (again, in the Athens Pilot) has two
modes, one towards lower speed values, which is likely to be the actual walking speed, and another
41
City4AgeID pseudonyms on the x-axis have been removed for anonymization purposes
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 47/183
around 11 km/h that is probably biking speed, rather than walking speed. Moreover, there are outlier
values up to 130 km/h which certainly represent car speeds.
Figure 6. Value of measure restaurant_visits_week across care receivers
This type of issues can be addressed technically, by better filtering algorithms (in this case it should not
be difficult to establish reasonable cut-off values) or by jointly consider other measurements that allow
to better discriminate the situation (e.g. looking at accelerometer signal patterns, to discriminate
walking from biking from car riding).
Figure 7. Distribution of measure walk_speed_outdoor values (in m/s)
An interesting observation is that the walk_speed_outdoor measure was in fact not filtered in the
experiment illustrated in Section 5: it was included as it is, as represented in Figure 7. Nevertheless, the
experiment gave good results. This is far from being unreasonable: as it is clear from the discussion
in previous sections of this deliverable, biking is a physical activity which has relevant positive impact
on frailty status assessment (see e.g. the Fried Frailty Index in subsection 8.2.1, which explicitly
mention biking). Even outliers that show usage of a motorized transportation vehicle like a car – a
IADL factor – can be meaningful. Thus, after all, the current, unfiltered formulation of
walk_speed_outdoor can be totally acceptable in the model. On the other side, as experiment #6 in
subsection 5.3.5 shows, it seems to be better to keep different features separated, rather than merged in
a single “macro-feature” that summarizes them; thus, even if the current formulation of the
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 48/183
walk_speed_outdoor measure is a good indicator, separating it into different walking, biking and car
usage measures may lead to better model accuracy.
This multi-faceted issue has been mentioned as a paradigmatic example, to underline that there are no
simple cases and each issue should be analysed individually and in depth, if valid indications are to be
derived.
A last opportunity for technical improvement regards the very choice of measures, to be included in the
model. The current dictionary reported in Section 11 has been drafted based on extensive discussion
with Project Pilot representatives, on the backdrop of the first version of the geriatric factors’
modelling.
However, as showed by experiment #7 in subsection 5.3.5, new, more informative measures could still
be devised. In the case of experiment #7, for instance, measures regarding Point of Interest (POIs)
visits and time spent at POIs have been combined to derive a new measure, reporting the average time
spent on a visit for a give type of POI. The addition of these new features seems to have slightly
improved the accuracy of the model.
Other opportunities may arise if more attributes are collected for existing features. E.g., if walking
speed measurements are associated with a time stamp that defines when a certain sample was collected,
a, e.g., walk_speed_outdoor_morning measure can be added that, according to some authors, could be
more meaningful from a geriatric point of view (see relevant paper mentioned in 8.1.8.2).
6.4 Towards a Machine Learning based approach
In the previous sections of this deliverable, and in particular in Section 4, we have derived the
City4Age risk model by basing it as much as possible on current geriatric knowledge, as represented by
established Instruments and Items that are routinely used in the clinical practice today.
The big advantage of this approach is that geriatricians and medical professionals are presented with
information that they already know how to interpret.
However, we have also seen that, when making the effort to substitute current measurement methods
(based on reports, observations or meter application, see Table 3 in sub-section 4.2.2) with more
continuous and unobtrusive methods, based on technology, several challenges have to be tackled.
It seems that these challenges stem from the following basic fact: given the underlying factors that are
affected by MCI and frailty conditions, as identified in geriatric research and practice (which are
represented by GEFs and GESs in the City4Age model) the external variables influenced by such
factors that can be observed by current methods (on one side), and those observable with
unobtrusive sensing technologies (on the other side), are generally different.
Take for instance the GES Walking.
Established Instruments assess this factor with Items like the following:
Question “Did you walk around outside without help?” (e.g. Mt.Ne.01, observed through
self-report)
Question “Are you able to walk 100m?” (e.g. Mt.Sh.04, observed through self-report)
Walking speed (e.g. Mt.Fi.04, observed through meter-based measurement)
Of these, the second and last Items can also be observed by unobtrusive sensing (Measures
WALK_DISTANCE and WALK_SPEED_OUTDOOR) while the first one cannot, because sensing
technologies cannot easily detect the presence or absence of help.
On the other side, sensing technologies may provide additional Measures that are clearly related to
GES Walking, but cannot be readily measured with conventional methods. Examples are:
Walk steps
Walk time
More sophisticated features, such as walk distance (or speed, steps, time) in specific periods
of the day – e.g. in the morning, afternoon, evening or night
This is partly an effect of the fact that Instruments used in today’s geriatrics practice have been
developed and clinically validated several decades ago, when new technologies were not yet
extensively available and thus could not be considered.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 49/183
Given this context, an alternative strategy for risk assessment is to relax the goal of addressing as many
existing Items as possible, and – while still being guided by the underlying causal structure of the MCI
and frailty phenotypes, as reflected in the City4Age model – aim instead at capturing the full
richness of datasets made available by unobtrusive sensing, that are expected to correlate well
with the GEFs and GESs in the model.
In other words, in addition to populating Individual Monitoring Dashboards to support interpretation
by geriatricians, the model can also be used to train appropriate Machine Learning algorithms and
to derive new MCI/frailty classifiers, able to detect and predict the conditions, similarly to (or
possibly better than) individual current Instruments (see sub-section 3.2.2, Figure 3, part c).
This opportunity is further reinforced by the encouraging results obtained from the experiment
described in Section 5 that, although initially conducted for the sake of model’s statistical assessment,
definitely represents a first step in this direction.
In this sub-section, we clarify the link among the City4Age risk modelling and the possible application
of a general Machine Learning-based approach to modelling, in order to support possible future
research work.
6.4.1 Conceptual framework
As already mentioned, the City4Age geriatric risk model outlines the structure of the causal
connections leading from the MCI and Frailty status of the person to the evidence represented by
relevant behavioural datasets, obtained through available unobtrusive sensing technologies.
In this subsection, we will delineate more precisely this assertion.
For the sake of explanation, we will assume that the risk of MCI/Frailty is modelled by the following
probabilistic variable:
MCIFrailtyRisk = <care recipient will become MCI or frail within 1 year>
This definition can certainly be changed in many ways, as for instance:
Consider classification along more states (instead of Boolean one); for instance, the Fried
Frailty Index classifies persons along three states: not frail, pre-frail and frail; the Lawton
scale of Instrumental ADLs classifies persons on an integer scale from 0 to 8
Creating two different classifiers, one for MCI and one for frailty
Establishing a different prospective period for the onset of the condition
Etc.
However for the purpose of this section and for sake of simplicity, the above mentioned Boolean
classifier is sufficient.
On the other side, we have a number of Measures, i.e. probabilistic variables that represent an
elaboration over the readings of the various sensors deployed at a certain City4Age instance. Some
examples of these variables are:
WALK_SPEED_OUTDOOR = <average outdoor walking speed in meters/seconds in the day>
SHOPS_VISITS = <number of visits to monitored shops in the day>
PUBLICTRANSPORT_RIDES_MONTH = <number of times the user gets on the bus in a month>
(for brevity, we will refer to the above variables as WSO, SV, PRM, in the subsequent text).
Each of the variables has an associated periodicity (e.g. daily or monthly, in the case of the examples
above).
In order to keep the illustration simple, in the rest of this section we assume that each of these variables
is represented through a uniquely established monthly feature (e.g. the Measure’s value itself for
monthly Measures and the average over the month for daily Measures – details on other feature
extraction approaches are discussed in subsections 4.2.4 and 5.3.4 above, and subsection 6.4.4 below).
Framed in this way, the fundamental problem of risk detection in City4Age can be stated as follows:
Each month, we would like to infer the distribution of the query variable MCIFrailtyRisk
given the evidence variables WSO, SV, PRM, etc.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 50/183
In other words, we want to compute:
P(MCIFrailtyRisk | WSO’, SV’, PRM’, …)
where WSO’, SV’, PRM’, … are specific values for variables WSO, SV, PRM, … respectively, that
have been measured for a given care receiver in a given month.
This means in turn that we need to know the full joint probability distribution table (FJD table) for the
query variable and all the evidence variables:
P(MCIFrailtyRisk, WSO, SV, PRM, …) (1)
Representing such FJD table is clearly infeasible, since evidence variables are too many. At present,
107 of them have already been defined (as illustrated in Section 11), and more of them could be added
in future. This means that, even in the very simplified case of transforming those variables into
discrete, Boolean form (see below) the FJD table in (1) would still have 2108
entries.
We need a different, more tractable way to represent the FJD table in (1) that can be used to reason
about the model. In the following sub-sections we propose a Bayesian Network representation for
achieving this objective, exploiting conditional independence assumptions among GEF/GES variables.
6.4.2 Bayesian network description of the geriatric modelling
The geriatric model illustrated in Section 4 naturally lends itself to a representation as a Bayesian
Network.
In fact, the model has been built based on the established, current geriatric knowledge (see subsection
4.1 and Annex in Section 8), including in particular:
1. The definition of relevant geriatric functional domains (Categories) involved in MCI and
Frailty, that have been hypothesized to be implicated in the conditions’ development
2. The definition of related evidence variables (Items) that have been hypothesized to correlate
well with the above domains
3. A number of clinical trials, that have consistently validated the hypotheses mentioned in 1)
and 2) above, making it possible to build Instruments for detecting and/or predicting
MCI/Frailty onset, which are routinely used in the clinical practice
This medically sound basis allow us to drive out the causal relationships that are needed to build an
initial Bayesian Network model.
To streamline the discussion, before proceeding we will make one more simplifying step: we will
convert each evidence variable into a discrete, Boolean form, by establishing a relevant cut-off
threshold, and assigning true or false labels to all values laying, respectively, on one or the other side of
the threshold.
This approach is in fact common in many widely used scales. For instance, the Fried Frailty Index (see
8.2.1) defines the (Boolean) evidence variable WALK_SPEED for men taller than 173cm as follows:
WALK_SPEED = if “time to walk 15 feet < 7 secs” then true else false
In a similar way, we can also consider GEFs and GESs illustrated in sub-section 4.2.3 as Boolean
variables, ending up with the conditions needed to draw a discrete Boolean Bayesian Network, an
excerpt of which is represented in Figure 8 below.
For space reasons, the Figure only represents the MCIFrailtyRisk variable, the first level of GEFs and
the expansion of (just) the Motility GEF into its composing GESs. Of these GESs, only the Walking
GES is furtherly expanded in its composing Measures.
It should be appreciated that, when fully expanded, the network is much bigger than what appears in
the Figure: in fact, it may include (depending on Pilot installations) up to 10 GEFs, 43 GESs and 107
Measures, for a total of 161 nodes.
In general, in the network presented in Figure 8, MCIFrailtyRisk represents the query variable,
Measures represent evidence variables and GEFs and GESs represents hidden variables.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 51/183
Approximations notwithstanding, the network of Figure 8 conveys a solid, basilar semantics for the
City4Age geriatric risk model, that can be used to ground an appropriate probabilistic risk detection
system.
In particular, the network causally models:
how the person’s MCI/Frailty status, represented by the MCIFrailtyRisk node, influences a
number of functional “macro-domains”, represented by the GEFs nodes. For instance: the
capability of the person to perform coordinated movement (variable Motility), her ability to
carry out basic ADLs (variable Basic ADLs), or more demanding Instrumental ADLs
(variable IADLs), or her interest in cultural activities (variable Cultural engagement) or
socialization activities (variable Socialization), etc.
Figure 8. Excerpt of a Bayesian Network representing the City4Age geriatric risk model
how macro-domains may in turn influence a number of further geriatric functional domains
represented by GESs. This is the level normally addressed in established geriatric
Instruments, and it includes for example, ability to walk (variable Walking), shopping habits
(variable Shopping), ability to prepare own meals (variable Ability to cook food), etc.
how geriatric domains in turn influence the evidence variables, i.e. the Measures which
relate to the datasets on care receivers’ behaviour, as actually collected by the City4Age
sensing technologies developed in work-package WP3. These include, for example, walking
speed (variable WALK_SPEED_OUTDOOR), number of visits to shops (variable
SHOP_VISITS), number of cooking sessions carried out (variable MEALS_NUM), etc.
Fixing our attention on the Bayesian Network so derived, we are able to highlight an initial set of
important challenges to be tackled in exploring this research avenue, as discussed in the following
subsections.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 52/183
6.4.3 Validity of conditional independence assumptions
The initial Bayesian Network depicted in the previous subsection makes some conditional
independence assumptions that may require further checking.
For example, according to the network, GEFs and GESs do not interact with each other, and the same
is assumed for Measures.
However, a more focussed scrutiny may suggest additional potential conditional dependencies among
such variables.
For instance, the Motility GEF would probably have partial influence on GEFs Socialization and
Cultural engagement, or on some GESs, like for instance Shopping, as a declining movement
capability can in turn reduce the possibility to conduct the activities measured by such variables, other
conditions remaining unchanged.
On the other hand, such extra conditional dependencies are expected to be reasonably limited.
Moreover, with the availability of larger datasets it would be possible to learn better network structures
directly from data and to define and describe the most meaningful of them. Dependencies that reveal to
be rather tenuous would be discarded, while others may be added to the network.
6.4.4 Feature extraction
Measures are available at the City4Age testbed experiments as discussed in sub-section 4.2.2 and listed
in Section 11.
Extracting features from such Measures is not an easy task, as in the literature there are not many
endeavours that, like City4Age, aim at predicting MCI/frailty on the basis of a large set of attributes,
addressing simultaneously many different functional domains. This leaves us with limited cases that
can be used as a starting point.
Among the works that address a single factor, an interesting example that parallels the City4AAge
attempt in many aspects, is Akl et al., Autonomous Unobtrusive Detection of Mild Cognitive
Impairment in Older Adults, published in the IEEE Transactions on Biomedical Engineering in 201542
and mentioned in subsection 8.1.8.2.
In particular, the paper addresses the GES Walking, through the measurement of (indoor) walking
speed.
In addition to statistical pre-processing similar to what has been illustrated in sub-section 4.2.4, the
authors have investigated different feature types, as follows:
Average of measures, where features were extracted by taking the average of the individual
measures across a temporal window of appropriate size
Probability densities of measures, where features were extracted by estimating the
probability densities of the individual measures in the temporal window
Trajectories of measures, where features were extracted by concatenating the trajectories of
the individual Measures across the temporal window into one vector
The first of these strategies has been applied in City4Age too, in the experiment illustrated in Section 5
above, with good results.
However, with the availability of more and larger datasets, it could be worth to investigate the other
two approaches, particularly in view of the fact that the authors eventually found that third one
produced the best accuracy.
It is also important to underline that the feature extraction process need not be limited to the measures
listed in Section 11. In fact, all City4Age Pilots also store in the Project’s repository Low-level
Elementary Actions (LEAs), i.e. basic events denoting actions performed by the monitored person,
that can be used to compute – through appropriate pre-processing steps – new Measures that may lead
to more interesting features.
42
https://www.ncbi.nlm.nih.gov/pubmed/25585407
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 53/183
It is out of the scope of this deliverable to discuss LEAs (they are the subject of work-package WP3’s
deliverables), however we provide here a brief example to show what is possible:
In many City4Age Pilots the SHOPS_VISIT Measure (i.e. number of shops visited in the
day) is computed on the basis of a couple of LEAs – for example detected by BLE Beacon
technology – called POI_ENTER(location_type: Shop, location_id: shop_ID) and
POI_EXIT(location_type: Shop, location_id: shop_ID), that detect the entrance to and the
exit from a given shop
Such LEAs carry with them additional attributes, such as a timestamp and, in some cases, a
geographical position
Exploiting the above attributes, several extra Measures, more sophisticated than
SHOPS_VISIT, can be derived, such as for instance SHOPS_VISITS_MORNING,
SHOPS_VISITS_FAR, SHOPS_VISITS_LONG, etc.
As hinted in subsection 5.3.6, when discussing experiment #7, the addition of some of these measures
can contribute to accuracy improvement.
6.4.5 Further developments
In the above sub-sections we have proposed to formalize the City4Age geriatric model as a Bayesian
Network structure, and have hinted at some of the issues that need to be solved in order to derive a
classifier based on such network, able to predict the posterior probability of MCI/Frailty given the
evidence represented by Measures obtained by the City4Age data collection platform.
The purpose of the preceding discussion is to clarify several preliminary aspects that may impact the
design and learning of such classifier. However, to proceed in this direction, fundamental aspects have
to be additionally investigated, such as:
Selection of the type of Machine Learning algorithms to be used. For example, the work
from Akl et al. mentioned in the previous sub-section has investigated support vector
machines and random forests. Other algorithms may be more appropriate, given the structure
of the City4Age classification problem and the fact that, as mentioned, it simultaneously
considers many geriatric determinants
Opportunity to relax the discretization hypothesis, possibly considering more general models
that involve both discrete and continuous variables
An additional issue is to consider if it makes sense to use temporal models, such as for
instance Dynamic Bayesian Networks or Long-Short Term Memory networks as hinted in
the project’s DoA. The current interpretation of the City4Age geriatric risk model is that it is
used monthly to periodically assess the MCI/Frailty risks of a care recipient on the basis of
behaviour changes detected by Measures during that same month. In principle this would not
require temporal modelling, given the low temporal variability of robustness conditions,
such as MCI or frailty. On the other hand, to make a well grounded decision it would be
necessary to more precisely assess if the additional dependencies among the GEFs/GESs
variables in a given month and those in previous months, introduced by temporal modelling,
would be actually reflected in an increased predictive accuracy, that it is worth the additional
effort.
6.5 Towards market deployment
Counting on the soundness of the results presented in this deliverable, City4Age Partners, and in
particularly business Partners, may want to apply the lessons learned and the recommendations that
have been presented in this Section to enact a practical and feasible plan that leads to improvement of
the City4Age risk model effectiveness, and achieves industry-grade accuracy. Building on the results
from the Athens Pilot study, reported in Section 5, such plan may be articulated in the following steps:
First, consider larger sets of candidate model’s Measures, covering a wider range of geriatric
factors and sub-factors. In fact, while the Athens Pilot collects 24 measures that covers 9
factors on a total of 45 offered by the City4Age risk model, other Pilots show that many
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 54/183
more can be feasibly addressed by unobtrusive technologies. By just by merging all
technologies experimented in-the-field in all 6 City4Age Pilots, a total of up to 18 different
factors can be covered, addressing up to 70 different measures, more than doubling the
“informative stock” experimented at the Athens Pilot and correspondingly expanding the
opportunities to look for optimal subsets of features, that can provide superior accuracy
Second, as previously mentioned, current implementation of existing measures can be
improved. For example, we discussed the presence of outliers and missing data, that we have
in part left within the dataset used for the statistical assessment conducted on the Athens
Pilot model. It is a relatively straightforward engineering task – and a common one for
commercial partners – to go through each of the measures and try to mitigate or outright
eliminate such problems, boosting in the process the model performance.
Third, the market is continuously producing dramatic innovations in personal data collection
equipment, that can be easily integrated in the City4Age risk model (see recommendations in
subsection 6.2). In particular, since the start of the Ciy4Age Project, new, affordable, and
reliable technologies have appeared that can provide additional, interesting measures to
address several of the City4Age model’s geriatric factors, or even new factors (e.g.
wristband that can measure HR/HRV, recognise a range of sporting or other physical
activities, open APIs e.g. from Google, that offer travelling patterns and activity recognition
on Android platforms, etc.)
Fourth, by conducting large scale experiments (i.e. with thousands of subjects – rather than
10s or 100s as typical of Research and Innovation Actions like City4Age) more
sophisticated, highly parametrized learning schemes (such as Bayesian Networks, as
discussed in subsection 6.4, but also SVMs or even DNNs) can be used in place of the
simple Naïve Bayes model used for the assessment reported in Section 5 – which, although
convenient for investigating smaller datasets, it is a parametric, fixed-size model, that is
known to saturate performance when increasing data sizes, well before achieving the AUC
values that are needed for commercial deployment43
By proceeding in a coordinated way along this path, it should be feasible for City4Age Partners to
achieve industry-grade performance levels – e.g. by identifying more efficient “behavioural markers”
for automatically screening robustness in elderly citizens, and/or by better tuning IMD configurations
to support geriatric assessment by health care professionals – that would ultimately allow to bring the
City4Age early risk modelling and detection strategy to the market.
A final remark, linked to market exploitation, regards the extensibility of the model to other types of
conditions and diseases, that may also benefit from the City4Age approach to unobtrusive, smart-city
based data collection, beyond health in aging populations.
For instance, these may include “behaviour-linked” conditions, such as for instance:
Obesity (monitor physical activity behaviours)
Mental wellbeing at the workplace (monitor cognitive behaviours)
Quality of life after aggressive medical treatments, like for instance in oncology (monitor
ADLs and IADLs behaviours)
Etc.
Most of the City4Age features would still be valid in cases like these, while new, domain-specific
measures would need to be added, by following the same risk modelling processes that have been
illustrated in details in this deliverable.
43
Domingos, A Few Useful Things to Know About Machine Learning, Communications of the ACM, 2012
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 55/183
7 Conclusions
This deliverable has illustrated the work performed to develop and assess a consistent geriatric risk
detection model, appropriate for City4Age’s needs.
A thorough review of established geriatric Instruments for detecting and predicting the onset of MCI
and frailty, allowed to identify the functional domains implicated in these phenotypes and to derive a
set of Factors (Geriatric Factors – GEFs, and Geriatric Sub-factors – GESs) and Measures (collected by
unobtrusive technology) that constitute the basis for deriving the City4Age Geriatric Risk Model.
A tentative proposal to compute a quantitative assessment of GEFs and GESs on the basis statistical
features (or Numerical Indicators – NUIs) extracted from Measures is presented. These computational
rules enable the immediate application of the model for driving the Individual Monitoring Dashboards,
that are specified in deliverable D2.13 Requirements, user scenarios and data visualization mock-ups
for apps/dashboards, v3. In particular, IMDs based on the City4Age model present geriatricians with a
synthesized view of multiple determinants involved in MCI and frailty detection and prediction,
that they wish to see and are trained to interpret.
Measures and NUIs can also be used to train machine learning classifiers, based on the City4Age risk
model. A statistical assessment of the potential accuracy of the City4Age model has been conducted by
training one such classifier on real-life experimental data collected at the Athens Pilot site. A Naïve
Bayes model running on top of a PCA with covered variance of about 0.8 achieved an AUC = 0.710,
statistically different from random classification with significance < 0.05. This represents a
convincing proof-of-concept, that vindicates the original City4Age rationale: that unobtrusive
data collection in smart-city contexts can lead to effective early detection of health decay in
elderly citizens.
The above results envisage the application of the City4Age Geriatric Risk model at two levels:
To build “behavioural markers” based on machine learning techniques, that can be used to
“screen out” robust elderly people, which do not need further medical attention
To provide domain experts (e.g. geriatricians, neurologists) with Individual Monitoring
Dashboards to quickly and efficiently assess the health conditions of those subjects that are
not screened out by the previous step, in order to more precisely decide about the subset that
actually warrant a further comprehensive geriatric assessment
This two-levels protocol underpins a virtual, validated “electronic frailty index” that can be used to
(semi-automatically) screen the city-dwelling elderly population for robustness, achieving at the
same time sufficient specificity (less false positives) and reasonable cost, as advocated by major
geriatric societies.
An account of invaluable lessons that have been learned during the conduction of the City4Age
modelling work and several directions for further research activities and for the Model’s industrial
scale-up have also been presented and discussed.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 56/183
8 Annex: Current Geriatric Instruments
This Annex lists the Instruments (respectively, for MCI and frailty) that have been selected and
surveyed to derive the City4Age risk modelling, as discussed in sub-section 4.1.3 above.
8.1 MCI Instruments
8.1.1 The Lawton Instrumental Activities of Daily Living (IADL) Scale
8.1.1.1 Overview
This Instrument has been devised in order to provide an assessment of the functional status in older
adults and deliver objective data to assist with targeting individualized care needs.
It guides the clinician to focus on the person’s baseline capabilities, facilitating early recognition of
changes, that may represent decline and/or need for additional medical check.
It is based on the assessment of a set of independent living skills, which are considered more complex
than the basic activities of daily living as measured by e.g. the Katz Index of ADLs.
For this reason, it can detect milder forms of cognitive impairment, with respect to basic ADLs. In
particular, studies have addressed its relationship with MCI assessment and its value in predicting
cognitive decline.
It is based on 8 domains of function. Persons are scored according to their highest level of functioning
in each Category. A summary score ranges from 0 (low function, dependent) to 8 (high function,
independent).
The questionnaire has to be administered by a trained interviewer and collects self-reported
information.
The Instrument validity and reliability have been assessed in some studies, also by determining its
correlation with four other Instruments: Physical Classification (6-point rating of physical health),
Mental Status Questionnaire (10-point test of orientation and memory), Behaviour and Adjustment
rating scales (4-6-point measure of intellectual, person, behavioural and social adjustment), and the
PSMS (6-item ADLs).
8.1.1.2 References
A complete description of the Instrument can be found at the ConsultGeri website of the
Hartford Institute for Geriatric Nursing, URL: https://consultgeri.org/try-this/general-
assessment/issue-23
A study addressing the usage of Instrumental Activities of Daily Living for assessing MCI
and for predicting future cognitive decline is described in the paper referred by the following
Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/22053873
8.1.1.3 Structure
ID Category Item Values
Co.Li.01 Communication Ability to Use Telephone Operates telephone on own
initiative; looks up and dials
numbers /
Dials a few well-known
numbers /
Answers telephone, but does not
dial /
Does not use telephone at all
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 57/183
ID Category Item Values
Sh.Li.02 Shopping Shopping Takes care of all shopping needs
independently /
Shops independently for small
purchases /
Needs to be accompanied on any
shopping trip /
Completely unable to shop
Fo.Li.03 Food Food Preparation Plans, prepares, and serves
adequate meals independently /
Prepares adequate meals if
supplied with ingredients. /
Heats and serves prepared meals
or prepares meals but does not
maintain adequate diet /
Needs to have meals prepared
and served
Ho.Li.04 Housekeeping Housekeeping Maintains house alone with
occasion assistance (heavy
work) /
Performs light daily tasks such as
dishwashing, bed making /
Performs light daily tasks, but
cannot maintain acceptable level
of cleanliness /
Needs help with all home
maintenance tasks /
Does not participate in any
housekeeping tasks.
Ln.Li.05 Laundry Laundry Does personal laundry
completely /
Launders small items, rinses
socks, stockings, etc. /
All laundry must be done by
others
Tr.Li.06 Transportation Mode of Transportation Travels independently on public
transportation or drives own car /
Arranges own travel via taxi, but
does not otherwise use public
transportation /
Travels on public transportation
when assisted or accompanied by
another /
Travel limited to taxi or
automobile with assistance of
another /
Does not travel at all
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 58/183
ID Category Item Values
Me.Li.07 Medication Responsibility for Own
Medications
Is responsible for taking
medication in correct dosages at
correct time /
Takes responsibility if
medication is prepared in
advance in separate dosages /
Is not capable of dispensing own
medication
Fi.Li.08 Finances Ability to Handle Finances Manages financial matters
independently (budgets, writes
checks, pays rent and bills, goes
to bank); collects and keeps track
of income /
Manages day-to-day purchases,
but needs help with banking,
major purchases, etc. /
Incapable of handling money
8.1.2 The OARS Multidimensional Functional Assessment Questionnaire
8.1.2.1 Overview
The OARS Multidimensional Functional Assessment Questionnaire (OMFAQ) has been designed at
the Duke Center for the Study of Aging and Human Development, as a tool for assessing individual
functional status of older persons (including ability to carry out Activities of Daily Living), in order to
group together people of comparable functional status.
The questionnaire has to be administered by a trained interviewer and collects self-reported
information.
In this subsection we consider the part regarding the Instrumental Activities of Daily Living, which
includes questions from 56 to 62. This part has shown good validity and reliability, suggesting it is
superior to many other Instruments. (Even if the recent trend is to do so, it is to be noted that the OARS
team counsels against using the ADL and IADL parts separately from the other questionnaire
components).
The Instrument has also been used as a screening tool and has been studied in assessing the correlation
among MCI and IADLs, also in view of improving prediction of the outcomes of MCI.
8.1.2.2 References
The main reference about the Instrument can be found on the site of the Duke Center for the
Study of Aging and Human Development, at the following URL:
https://sites.duke.edu/centerforaging/services/older-americans-resources-and-services/
A study on the association of the Instrument with MCI is referred by the following Pubmed
URL: http://www.ncbi.nlm.nih.gov/pubmed/22337146
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 59/183
8.1.2.3 Structure
ID Category Item Values
Co.Oa.01 Communication Can you use the telephone? Without help, including looking
up numbers and dialling /
With some help (can answer
phone or dial operator in an
emergency, but need a special
phone or help in getting the
number or dialling) /
Completely unable to use the
telephone
Tr.Oa.02 Transportation Can you get to places out of
walking distance?
Without help (drive your own
car, or travel alone on buses, or
taxis) /
With some help (need someone
to help you or go with you when
traveling) /
Unable to travel unless
emergency arrangements are
made for a specialized vehicle
like an ambulance
Sh.Oa.03 Shopping Can you go shopping for
groceries or clothes?
Without help (taking care of all
shopping needs yourself,
assuming you had
transportation) /
With some help (need someone
to go with you on all shopping
trips) /
Completely unable to do any
shopping
Fo.Oa.04 Food Can you prepare your own
meals?
Without help (plan and cook full
meals yourself) /
With some help (can prepare
some things but unable to cook
full meals yourself) /
Completely unable to prepare
any meals
Ho.Oa.05 Housekeeping Can you do your housework? Without help (can clean floors,
etc.) /
With some help (can do light
housework but need help with
heavy work) /
Completely unable to do any
housework
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 60/183
ID Category Item Values
Me.Oa.06 Medication Can you take your own
medicine?
Without help (in the right doses
at the right time) /
With some help (able to take
medicine if someone prepares it
for you and/or reminds you to
take it) /
Completely unable to take your
medicines
Fi.Oa.07 Finances Can you handle your own
money?
Without help (write checks, pay
bills, etc.) /
With some help (manage day-to-
day buying but need help with
managing your check-book and
paying your bills) /
Completely unable to handle
money
8.1.3 The Nottingham Extended Activities of Daily Living
8.1.3.1 Overview
The Nottingham Extended Activities of Daily Living (NEADL) is a ranked assessment of daily living
scale that has been developed to assess activities which may be important to stroke patients who have
been discharged home.
However, its role has also been investigated in the transition from no cognitive impairment to mild
cognitive impairment (MCI) and dementia, in comparison to neuropsychological tests.
The questionnaire incorporates 22 ADL Items in four sections. Some of them regards basic ADLs, of
less use to City4Age, while others regard IADLs, that are more interesting. Subsection 8.1.3.3 below
lists all the Items, including ADLs.
Answers should reflect what has actually been done in the last few weeks. The questionnaire should be
intended as a record of activity rather than capability.
The questionnaire can be administered by an interviewer (including via telephone) or it can be self-
administered (e.g. via mail, as it has been initially designed as a postal questionnaire).
8.1.3.2 References
The paper that describes the Instrument can be found at the following URL:
http://cre.sagepub.com/content/1/4/301.short
A study finding a powerful predictive value for the Instrument, comparable to that of
neuropsychological tests such as MMSE, MoCA, Addenbrooke cognitive assessment and
frontal assessment battery, can be found at the URL:
http://www.neurores.org/index.php/neurores/article/view/316
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 61/183
8.1.3.3 Structure
ID Category Item Values
Mt.Ne.01 Motility Did you walk around outside? Not at all /
With help /
On your own with help /
On your own
Mt.Ne.02 Motility Did you climb stairs? Not at all /
With help /
On your own with help /
On your own
Mt.Ne.03 Motility Did you get in and out of a car? Not at all /
With help /
On your own with help /
On your own
Mt.Ne.04 Motility Did you walk over uneven
ground?
Not at all /
With help /
On your own with help /
On your own
Mt.Ne.05 Motility Did you cross roads? Not at all /
With help /
On your own with help /
On your own
Tr.Ne.06 Transportation Did you travel on public
transport?
Not at all /
With help /
On your own with help /
On your own
Ad.Ne.07 Food Did you manage to feed
yourself?
Not at all /
With help /
On your own with help /
On your own
Fo.Ne.08 Food Did you manage to make
yourself a hot drink?
Not at all /
With help /
On your own with help /
On your own
Fo.Ne.09 Food Did you take hot drinks from one
room to another?
Not at all /
With help /
On your own with help /
On your own
Ho.Ne.10 Housekeeping Did you do the washing up? Not at all /
With help /
On your own with help /
On your own
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 62/183
ID Category Item Values
Fo.Ne.11 Food Did you make yourself a hot
snack?
Not at all /
With help /
On your own with help /
On your own
Fi.Ne.12 Finances Did you manage your own
money when out?
Not at all /
With help /
On your own with help /
On your own
Ln.Ne.13 Laundry Did you wash small items of
clothing?
Not at all /
With help /
On your own with help /
On your own
Ho.Ne.14 Housekeeping Did you do your own
housework?
Not at all /
With help /
On your own with help /
On your own
Sh.Ne.15 Shopping Did you do your own shopping? Not at all /
With help /
On your own with help /
On your own
Ln.Ne.16 Laundry Did you do a full clothes wash? Not at all /
With help /
On your own with help /
On your own
Cu.Ne.17 Culture Did you read newspapers or
books?
Not at all /
With help /
On your own with help /
On your own
Co.Ne.18 Communication Did you use the telephone? Not at all /
With help /
On your own with help /
On your own
Cu.Ne.19 Culture Did you write letters? Not at all /
With help /
On your own with help /
On your own
So.Ne.20 Socialization Did you go out socially? Not at all /
With help /
On your own with help /
On your own
Ac.Ne.21 Activity Did you manage your own
garden?
Not at all /
With help /
On your own with help /
On your own
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 63/183
ID Category Item Values
Tr.Ne.22 Transportation Did you drive a car? Not at all /
With help /
On your own with help /
On your own
8.1.4 The Direct Assessment of Functional Status
8.1.4.1 Overview
The Direct Assessment of Functional Status is a performance-based measure for evaluating a broad
spectrum of behaviours related to Instrumental Activities of Daily Living.
Although it has been designed and validated as tool for the assessment of the functional competencies
of patients with dementia, rather than predicting cognitive degradation, the Instrument is included here
as a paradigmatic example of a performance-based method (i.e. method relying on direct observation
of subject’s behaviour in controlled settings) a practice that bear some resemblance to the City4Age
unobtrusive monitoring approach.
Items’ Categories are similar to other IADL related Instruments.
The Instrument needs to be administered by an expert team in appropriate settings.
It is worth to note that performance-based measures in general, and DAFS in particular, have been
found to be better in detecting cognitive deficits, than self- or collateral report questionnaires, although
their broad application is limited by higher costs, as they are more time consuming, require space,
specialized equipment and expert examiners.
8.1.4.2 References
A paper describing the Instrument is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/2738312
A paper that shows DAFS higher sensitivity in detecting MCI, compared with self- or
collateral report questionnaire, is referred by the following Pubmed URL:
http://jgp.sagepub.com/content/27/4/253.short
A paper addressing the validity of the Instrument and also mentioning the current economic
and organizational limitations to performance-based methods (that can possibly be overcome
by the City4Age paradigm) is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/12675100
8.1.4.3 Structure
ID Category Item Values
Ti.Da.01 Time The person is shown four
different times (0300 h, 0800 h,
1030 h and 1215 h) using a large
model of a clock and is asked to
tell the time
Correct / Incorrect answer for
each item
Ti.Da.02 Time The person is asked to state the
date, the day, the month and the
year.
Correct / Incorrect answer for
each item
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 64/183
ID Category Item Values
Co.Da.03 Communication The person is invited to dial the
operator, dial from a list of
telephone numbers, dial from
oral presentation and dial from
written presentation.
Correct / Incorrect performance
for each item
Co.Da.04 Communication The person is observed on
picking up the receiver, dialling,
hanging up and operating the
telephone in the correct
sequence.
Correct / Incorrect performance
for each item
Co.Da.05 Communication The person is invited to fold a
letter in half, put it in an
envelope, seal the envelope, put
on a stamp, address it (from a
presented stimulus card) and add
a return address (the person’s
own current address, without a
postal code).
Correct / Incorrect performance
for each item
Fi.Da.06 Finance The person is invited to identify
four different coins and three
notes
Correct / Incorrect performance
for each item
Fi.Da.07 Finance The person is invited to count
four amounts of money Correct / Incorrect performance
for each item
Sh.Da.08 Shopping Before the preparation of the
letter (Co.Da.05), the examiner
instructs the person that in 10
min she/he will be going to a
grocery store to select six items:
orange juice, spaghetti, cherry
jam, tuna fish, rice and
tomatoes. After 1 min to recall
as many grocery items as
possible
Number of recalled items
Sh.Da.09 Shopping The person is taken to a
simulated grocery store to pick
out the items from a total of 25
Number of recalled items
Sh.Da.10 Shopping The examiner then gives the
person a written grocery list
(milk, crackers, eggs and
laundry detergent) and asks
them to select the four items and
to hand them to the examiner
Correctly / Incorrectly picked
items
Fi.Da.11 Finance Given a note to pay for the
items, the person is invited to
make change
Change correctly / incorrectly
identified
Tr.Da.12 Transportation The examiner asks the person to
identify a driver's correct
response to 13 road signals
Correctly / Incorrectly identified
signals
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 65/183
ID Category Item Values
Ad.Da.13 ADLs The person is taken to a
bathroom and asked to take the
cap off a tube of toothpaste, put
toothpaste on a toothbrush, turn
on the tap, brush teeth, dampen
washcloth, put soap on
washcloth, wash the face and
turn off the tap
Correct / Incorrect performance
for each item
Ad.Da.14 ADLs The person is invited to use a
hairbrush, put on a coat, button a
coat (three buttons), fasten a zip
and tie shoelaces
Correct / Incorrect performance
for each item
Ad.Da.15 ADLs The person, sitting at a table,
shows how she/he would cut a
steak, take a bite of it, eat soup
and pour water into a glass and
drink it
Correct / Incorrect performance
for each item
8.1.5 The Mini–Mental state Examination
8.1.5.1 Overview
The mini–mental state examination (MMSE) is a 30-point questionnaire that is used extensively in
clinical and research settings to measure cognitive impairment. It was designed to give a practical
clinical assessment of change in cognitive status in geriatric inpatients.
It is also commonly used to screen for dementia, to estimate the severity and progression of cognitive
impairment and to follow the course of cognitive changes in an individual over time.
The Instrument has to be administered by a trained examiner.
Although one of the most frequently noted disadvantages of the MMSE relates to its lack of sensitivity
to MCI, we report it here because of its wider adoption and acceptance. Another disadvantage is that it
is affected by demographic factors, age and education being the most important.
Performance of specific MMSE domains as predictors of subsequent overall cognitive decline has been
object of study.
8.1.5.2 References
MMSE has been copyrighted by Psychological Assessment Resources (PAR). However,
their site (www.parinc.com) was down at the time of writing. A version of the MMSE
questionnaire can be found on the British Columbia Ministry of Health website at the URL:
http://www2.gov.bc.ca/gov/content/health/practitioner-professional-resources/bc-
guidelines/cognitive-impairment#resources
A review of information regarding the psychometric properties and utility of the Mini-
Mental State Examination can be found in the paper referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/1512391
Studies on the value of single MMSE domains in predicting decline over time can be found
at the following Pubmed URLs: http://www.ncbi.nlm.nih.gov/pubmed/19382130 and
http://www.ncbi.nlm.nih.gov/pubmed/19196632
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 66/183
8.1.5.3 Structure
ID Category Item Values
Ti.Mm.01 Time The person is asked to state the
year, season, month, date and
day of the week.
Correct / Incorrect answer for
each element
Sp.Mm.02 Space The person is asked to state the
country, province, city/town,
address (home: street address, in-
facility: building name) and
location (home: room name, in-
facility: floor)
Correct / Incorrect answer for
each element
Mr.Mm.03 Memory The examiner says “I am going
to name three objects. When I am
finished, I want you to repeat
them. Remember what they are
because I am going to ask you to
name them again in a few
minutes.” The examiner says the
following words slowly at 1‐second intervals ‐ ball/ car/ man
Correct / Incorrect repetition for
each word
At.Mm.04 Attention The examiner asks the person to
spell the word WORLD. Now
spell it backwards. (in
alternative, the interviewer
counting down from one hundred
by sevens)
Correct / Incorrect answer
(position of the first error)
Mr.Mm.05 Memory The examiner asks the person
what were the three objects
she/he asked to remember?
Correctly / Incorrectly recalled
words
La.Mm.06 Language The examiner show a wristwatch
and asks the person “What is this
called?”. Repeats with a pencil
Correctly / Incorrectly identified
items
La.Mm.07 Language The examiner asks the person to
repeat this phrase “No ifs, ands
or buts.”
Correct / Incorrect repetition
La.Mm.08 Language The examiner asks the person to
read the words on a page and
then do what it says. Then hand
the person the page with
“CLOSE YOUR EYES” on it. If
the subject reads and does not
close their eyes, repeat up to
three times. Score only if subject
closes eyes
Closing / Not closing eyes
La.Mm.09 Language The examiner hands the person a
pencil and paper, then says
“Write any complete sentence on
that piece of paper”
Writing a sentence that make /
does not make sense (ignore
spelling errors)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 67/183
ID Category Item Values
Vi.Mm.10 Visuospatial The examiner places a drawing
(interlocking pentagons, i.e. two
five-sided figures intersecting to
make a four-sided figure), eraser
and pencil in front of the person
and asks “Copy this drawing
please.” (Allow multiple tries;
wait until person is finished and
hands it back.)
Drawing correctly / incorrectly
copied
La.Mm.11 Language The examiner asks the person if
she/he is right or left‐handed,
takes a piece of paper and hold it
up in front of the person. Then
says “Take this paper in your
right/left hand [note: whichever
is non‐dominant], fold the paper
in half once with both hands and
put the paper down on the floor”.
Correct / incorrect performance
for each item
8.1.6 The Short Test of Mental Status
8.1.6.1 Overview
The Short Test of Mental Status (STMS) is a screening measure of cognition specifically developed for
use in dementia assessment and was intended to be more sensitive to problems of learning and mental
agility that may be seen in mild cognitive impairment (MCI).
In particular, STMS has been found to be better able than MMSE for predicting MCI.
The Instrument is to be administered by a trained examiner.
8.1.6.2 References
The Instrument is described in the paper referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/3561043
A study that compares the STMS and MMSE in detecting or predicting MCI is referred by
the following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/14676056
A study that identified STMS as the strongest predictor of MCI risk in NC subjects, among
several other considered demographic and clinical variables, is referred by the following
Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/25788555
8.1.6.3 Structure
ID Category Item Values
Sp.Sm.01 Space The person is asked to state
name, address, current location
(building), city state.
Correct / Incorrect answer for
each item
Ti.Sm.02 Time The person is asked to state date
(day), month, year. Correct / Incorrect answer for
each item
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 68/183
ID Category Item Values
At.Sm.03 Attention The following digit spans of
increasing length are presented in
sequence to the person, asking
for repetition: 2-9-6-8-3, 5-7-1-9-
4-6, 2-1-5-9-3-6-2.
Length of correctly repeated
spans
Mr.Sm.04 Memory The examiner tell four unrelated
words: “apple”, “Mr. Johnson”,
“charity”, “tunnel”. The person is
requested to repeat all words.
Number of trial to repeat all four
words
At.Sm.05 Attention The examiner proposes 4
arithmetic operations to be
computed: 5x13, 65-7, 58/2,
29+11
Correctly / Incorrectly computed
operations
Ab.Sm.06 Abstraction The examiner proposes three pair
of words and asks the person to
state the abstract interpretation:
orange/banana, dog/horse,
table/bookcase (e.g. dog/horse =
animal)
Correctly / Incorrectly abstracted
interpretations
Vi.Sm.07 Visuospatial The examiner asks the person to
draw a clock face showing 11:15.
Clock face correctly / incorrectly
drawn
Vi.Sm.08 Visuospatial The examiner shows the person
the drawing of a cube and asks to
copy it
Drawing correctly / incorrectly
copied
La.Sm.09 Language The examiner asks the person to
provide three information
element: “first president”,
“define an island”, “number of
weeks per year”
Correct/incorrect information
provided for each item
Mr.Sm.10 Memory The examiner asks the person to
recall the words from Mr.Sm.04
Correctly/incorrectly recalled
words
8.1.7 The Montreal Cognitive Assessment
8.1.7.1 Overview
The Montreal Cognitive Assessment (MoCA) is a 10-minute cognitive screening tool to assist first-line
physicians in detection of MCI as a clinical state that often progresses to dementia.
The MoCA has been assessed as a predictor of worsening cognitive conditions, e.g. in conversion from
MCI to Alzheimer’s disease.
The Instrument has to be administered by a trained examiner.
8.1.7.2 References
The Instrument is described in the paper found referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/15817019
A study that assesses the usefulness of the Instrument in predicting conversion to AD is
referred by the following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/24635004
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 69/183
8.1.7.3 Structure
ID Category Item Values
Vi.Mo.01 Visuospatial The examiner instructs the
subject: "Please draw a line,
going from a number to a letter
in ascending order. Begin here
[point to (1)] and draw a line
from 1 then to A then to 2 and
so on. End here [point to (E)].".
Correct / Incorrect answer for
each item
Vi.Mo.02 Visuospatial The examiner gives the
following instructions, pointing
to the cube: “Copy this drawing
as accurately as you can, in the
space below”.
Drawing correctly / incorrectly
copied
Vi.Mo.03 Visuospatial The examiner give the following
instructions: “Draw a clock. Put
in all the numbers and set the
time to 10 past 11”.
Clock face correctly / incorrectly
drawn
La.Mo.04 Language The examiner points to three
animal figures (lion, rhino,
camel) in turn and says: “Tell
me the name of this animal”.
Names correctly / incorrectly
repeated.
Mr.Mo.05 Memory The examiner reads a list of 5
words (face, velvet, church,
daisy, red) at a rate of one per
second, giving the following
instructions: “This is a memory
test. I am going to read a list of
words that you will have to
remember now and later on.
Listen carefully. When I am
through, tell me as many words
as you can remember. It doesn’t
matter in what order you say
them”. When the subject
indicates that she/he has finished
(has recalled all words), or can
recall no more words, the
examiner reads the list a second
time with the following
instructions: “I am going to read
the same list for a second time.
Try to remember and tell me as
many words as you can,
including words you said the
first time.”. At the end of the
second trial, the examiner
informs the person that she/he
will be asked to recall these
words again”
Correct / Incorrect repetition for
each word
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 70/183
ID Category Item Values
At.Mo.06 Attention The examiner gives the
following instruction: “I am
going to say some numbers and
when I am through, repeat them
to me exactly as I said them”.
Then she/he reads the five
number sequence at a rate of one
digit per second: 2-1-8-5-4.
Correctly / Incorrectly repeated
sequence
At.Mo.07 Attention The examiner gives the
following instruction: “Now I
am going to say some more
numbers, but when I am through
you must repeat them to me in
the backwards order.” Read the
three number sequence at a rate
of one digit per second: 7-4-2
Correctly / Incorrectly repeated
sequence
At.Mo.08 Attention The examiner reads a list of
letters (F B A C M N A A J K L
B A F A K D E A A A J A M O
F A A B) at a rate of one per
second, after giving the
following instruction: “I am
going to read a sequence of
letters. Every time I say the
letter A, tap your hand once. If I
say a different letter, do not tap
your hand”.
Correctly / Incorrectly tapped As
At.Mo.09 Attention The examiner gives the
following instruction: “Now, I
will ask you to count by
subtracting seven from 100, and
then, keep subtracting seven
from your answer until I tell you
to stop.”
Correct / Incorrect subtractions
La.Mo.10 Language The examiner gives the
following instructions: “I am
going to read you a sentence.
Repeat it after me, exactly as I
say it [pause]: I only know that
John is the one to help today.”
Following the response, say:
“Now I am going to read you
another sentence. Repeat it after
me, exactly as I say it [pause]:
The cat always hid under the
couch when dogs were in the
room.”
Correct / Incorrect repetition
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 71/183
ID Category Item Values
La.Mo.11 Language The examiner gives the
following instruction: “Tell me
as many words as you can think
of that begin with a certain letter
of the alphabet that I will tell
you in a moment. You can say
any kind of word you want,
except for proper nouns (like
Bob or Boston), numbers, or
words that begin with the same
sound but have a different
suffix, for example, love, lover,
loving. I will tell you to stop
after one minute. Are you
ready? [Pause] Now, tell me as
many words as you can think of
that begin with the letter F.
[time for 60 sec]. Stop.”
Words generated
Ab.Mo.12 Abstraction The examiner asks the subject to
explain what each pair of words
has in common, starting with the
example: “Tell me how an
orange and a banana are alike”.
If the subject answers in a
concrete manner, then say only
one additional time: “Tell me
another way in which those
items are alike”. If the subject
does not give the appropriate
response (fruit), say, “Yes, and
they are also both fruit.” Do not
give any additional instructions
or clarification. After the
practice trial, say: “Now, tell me
how a train and a bicycle are
alike”. Following the response,
administer the second trial,
saying: “Now tell me how a
ruler and a watch are alike” Do
not give any additional
instructions or prompts.
Correctly / Incorrectly abstracted
interpretations
Mr.Mo.13 Memory The examiner gives the
following instruction: “I read
some words to you earlier,
which I asked you to remember.
Tell me as many of those words
as you can remember.”
Correctly / Incorrectly recalled
words
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 72/183
ID Category Item Values
Ti.Mo.14 Time The examiner gives the
following instructions: “Tell me
the date today”. If the subject
does not give a complete
answer, then prompt accordingly
by saying: “Tell me the [year,
month, exact date, and day of
the week].”
Correct / Incorrect answer for
each element
Sp.Mo.15 Space Then the examiner says: “Now,
tell me the name of this place,
and which city it is in.”
Correct / Incorrect answer for
each element
8.1.8 Predicting MCI onset from gait speed analysis
8.1.8.1 Overview
Although not actually an established Instrument used in research or clinical settings, we would like to
present here a significant result that is representative of new directions that can be potentially pursued.
We refer to the study of Buracchio et al. (see References subsection below) that assesses how gait
speed can be used as predictor of MCI onset.
Subjects for the study were 204 healthy seniors (58% women) from the Oregon Brain Aging Study,
who have been evaluated for up to 20 years with annual neurologic, neuropsychological and motor
examinations.
The authors found that rates of change with aging of gait speed were significantly different between
MCI converters and non-converters (p<0.001).
Change points occurred approximately 14 years prior to MCI onset in men and approximately 6 years
for women.
As gait speed can be measured in a relatively easy and unobtrusive way through a smartphone, this
result is significant for City4Age. In fact, research work aimed at devising unobtrusive classifier
systems based on this model have recently appeared in the literature (see References subsection below).
Authors have also analysed tapping speed and found it significantly different between MCI converters
and non-converters both in the dominant hand (p<0.003) and non-dominant hand (p<0.001). However,
in this case, change points occurred after the onset of MCI, and thus the feature could not be used for
prediction.
8.1.8.2 References
The relevant paper from Buracchio er al. is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/20697049
A study that applies the Buracchio et al.s’ model to build an appropriate classifier, based on
datastreams coming from motion sensors placed in a smart home, is referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/25585407
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 73/183
8.1.8.3 Structure
ID Category Item Values
Mt.Gm.01 Motility Time in seconds to walk from a
starting point to a marker 15 feet
away, turn, and back at a normal
casual gait for a total of 30 feet
(9.14 meters).
Decrease of 0.013m/s/yr in all
subjects /
Decrease of 0.023m/s/yr in men
who are MCI converters, at 14.2
yr prior to MCI diagnosis /
Decrease of 0.025m/s/yr in
women who are MCI converters,
at 6.0 yr prior to MCI diagnosis
8.2 Frailty Instruments
8.2.1 Fried Frailty Index
8.2.1.1 Overview
The Fried Frailty Index (FFI) is a landmark Instrument that has been conceived as a phenotype
framework for defining frailty and it is one of the most popular and widespread.
It encompasses the assessment of five dimensions that are hypothesized to reflect systems the impaired
regulation of which underlies the syndrome: unintentional weight loss, exhaustion, muscle weakness,
slowness while walking, and low levels of activity.
Corresponding to these dimensions are five specific Items indicating adverse functioning, which are
implemented using a combination of self-reported and performance-based measures.
The instrument classifies people as
‘frail’, when three or more Items are found positive
‘pre-frail’, when one or two Items are found positive
‘non-frail’, when no Items are found positive
8.2.1.2 References
The seminal study of Fried et al. that defined FFI is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/11253156; the same study also offers concurrent and
predictive validity for the Instrument, based on data from the Cardiovascular Health Study
A paper that confirmed FFI validity with data from the Women's Health and Aging Studies
is referred by the following Pubmed URL: https://www.ncbi.nlm.nih.gov/pubmed/16567375
8.2.1.3 Structure
ID Category Item Values
We.Fi.01 Weight The person lost >10 pounds
unintentionally last year True / False
Ex.Fi.02 Exhaustion The person felt that everything
she/he did was an effort in last
week
Rarely or none of the time (<1
day) /
Some or little of the time (1 to 2
days) /
Moderate amount of the time (3
to 4 days) /
Most of the time
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 74/183
ID Category Item Values
Ex.Fi.03 Exhaustion The person felt that she/he could
not get going in last week
Rarely or none of the time (<1
day) /
Some or little of the time (1 to 2
days) /
Moderate amount of the time (3
to 4 days) /
Most of the time
Mt.Fi.04 Motility Time to walk 15 feet (4.57
meters) Number of seconds beyond/not
beyond thresholds:
For men with height <= 173 cm:
7 secs /
For men with height > 173 cm: 6
secs /
For men with height <= 173cm:
7 secs /
For men with height > 173cm: 6
secs
Ac.Fi.05 Activity Physical expenditure on activity
scale per week on 18 items
(Walking for exercise,
moderately strenuous household
chores, mowing or raking the
lawn, gardening, hiking,
jogging, biking, exercise cycle,
dancing, aerobics, bowling, golf,
singles or doubles tennis,
racquetball, calisthenics,
swimming. To compute kcals
expended per week, use the
formula: kcal/week = [activity-
specific MET (kcal/kg × hour) ]
× [duration per session (min) /
60 min] × [body weight (kg)] ×
[number of sessions in the last 2
wk / 2] × [number of months per
year activity was done])
kCal beyond / not beyond
threshold of 270 kCal
Wk.Fi.06 Weakness Grip strength (average of 3
trials, dominant hand, measured
with Jamar hand dynamometer)
Kg beyond/not beyond
thresholds:
For men, BMI <= 24: 29 kg /
For men, BMI 24.1-26: 30 kg /
For men, BMI 26.1-28: 30 kg /
For men, BMI > 28: 32 kg /
For women, BMI <= 23: 17kg /
For women, BMI 23.1-26: 17.3
kg /
For women, BMI 26.1-29: 18 kg
/
For women, BMI > 29: 21 kg
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 75/183
8.2.2 Study of Osteoporotic Fractures index
8.2.2.1 Overview
The Study of Osteoporotic Fractures index (SOF index) is an attempt to propose a simpler alternative
to the FFI Instrument, easier to use in the clinical practice.
It has been developed in the frame of the Study of Osteoporotic Fractures, within a cohort of 6701
women 69 years or older, and its predictive validity has been compared with that of the FFI Instrument.
It uses 3 components: weight loss, the subject's inability to rise from a chair 5 times without using her
arms, and reduced energy level.
Other studies have subsequently extended validation to additional cohorts, including men.
SOF index can classify people as ‘robust’ (no Item is found positive), ‘pre-frail’ (one Item is found
positive), or ‘frail’ (two or more Items are found positive)
8.2.2.2 References
The paper that describes the development of the SOF index Instrument and the comparison
of its predictive validity with that of the FFI Instrument, is referred by the following Pubmed
URL: http://www.ncbi.nlm.nih.gov/pubmed/18299493
A paper that further validates the SOF index and compares it with FFI in a diverse elderly
community-dwelling sample of men and women is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/19682112
A study aiming at determining the ability of the SOF index criteria to predict adverse health
outcomes at a one-year follow-up in a sample of older outpatients in Italy is referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/21871675
8.2.2.3 Structure
ID Category Item Values
We.So.01 Weight Weight loss (irrespective of
intent to lose weight) of 5% or
more in a 2 years period
True / False
Mt.So.02 Motility Subject's inability to rise from a
chair 5 times without using her
arms
True / False
Ex.So.03 Exhaustion Do you feel full of energy? Yes / No
8.2.3 SHARE-FI
8.2.3.1 Overview
SHARE-FI (SHARE Frailty Instrument) is an Instrument based in the on the Survey of Health, Ageing
and Retirement in Europe (SHARE).
SHARE-FI represents an attempt to operationalise the FFI Instrument in a very large European
population-based sample, offering an alternative to FFI in the European context.
To develop the Instrument, SHARE-FI authors selected the five SHARE variables that, in their view,
were the closest to FFI’s Items. On the other hand, their selection was not without significant
departures from FFI and thus SHARE-FI has been validated on its own.
The ultimate goal of the Instrument is to provide European community practitioners with a simple and
valid tool that addresses people over the age of 50.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 76/183
The SHARE-FI Instrument is publicly available as an HTML/Javascript downloadable calculator44
.
8.2.3.2 References
The main study that defines SHARE-FI can be found at the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/20731877. The paper also assesses the validity of the
Instrument in predicting mortality
A paper that provides further prospective validation of SHARE-FI, with a focus on
disability, is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/22186172
A study that shows that SHARE-FI predicts mortality similarly to a more complex frailty
Instruments based on Comprehensive Geriatrics Assessment (CGA) is referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/22994136
8.2.3.3 Structure
ID Category Item Values
Ex.Sh.01 Exhaustion In the last month, have you had
too little energy to do the things
you wanted to do?
Yes / No
We.Sh.02 Weight What has your appetite been
like? Diminution in desire for food
and/or eating less than usual /
No change in desire for food
and/or eating the same as usual /
Increase in desire for food and/or
eating more than usual.
Wk.Sh.03 Weakness Grip strength (highest among
four measures, two for each
hand, taken with a
dynamometer)
Kg (continuous measure)
Mt.Sh.04 Motility Because of a health or physical
problem, do you have any
difficulty walking 100 metres?
(Exclude any difficulties that
you expect to last less than three
months)
Yes / No
Mt.Sh.05 Motility Because of a health or physical
problem, do you have any
difficulty climbing one flight of
stairs without resting? (Exclude
any difficulties that you expect
to last less than three months)
Yes / No
Ac.Sh.06 Activity How often do you engage in
activities that require a low or
moderate level of energy such as
gardening, cleaning the car, or
doing a walk?
Hardly ever, or never /
One to three times a month /
Once a week /
More than once a week.
44
http://bmcgeriatr.biomedcentral.com/articles/10.1186/1471-2318-10-57
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 77/183
8.2.4 FRAIL scale
This 5-Items Instrument has been demonstrated as an excellent screening test for clinicians that need to
identify frail persons at risk of developing disability, declining in health functioning and mortality.
The FRAIL scale has been developed with the intent of providing an Instrument that does not requires
face-to-face examination, and could thus result in more efficient identification of the syndrome, that
could be accomplished by telephone or through self-administered forms. These features are aimed at
earlier recognition and treatment by practitioners.
8.2.4.1 References
The International Academy of Nutrition, Health, and Aging proposed the FRAIL scale in the
following papers: http://www.ncbi.nlm.nih.gov/pubmed/18165842 and
http://www.ncbi.nlm.nih.gov/pubmed/18261696
Another study that shows that the Instrument can predict future disability before the person
becomes disabled, is referred by the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/22836700
8.2.4.2 Structure
ID Category Item Values
Ex.Fr.01 Exhaustion How much of the time during
the past 4 weeks did you feel
tired?
Rarely or none of the time /
Some or little of the time /
Moderate amount of the time /
Most of the time
Mt.Fr.02 Motility By yourself and not using aids,
do you have any difficulty
walking up 10 steps without
resting?
Yes / No
Mt.Fr.03 Motility By yourself and not using aids,
do you have any difficulty
walking several hundred yards?
Yes / No
He.Fr..04 Health Did a doctor ever tell you that
you have [illness]? [where
illness is: hypertension,
diabetes, cancer (other than a
minor skin cancer), chronic lung
disease, heart attack, congestive
heart failure, angina, asthma,
arthritis, stroke, and kidney
disease
Number < 5
Number >= 5
We.Fr.05 Weight How much do you weigh with
your clothes on but without
shoes? One year ago in (MO,
YR), how much did you weigh
without your shoes and with
your clothes on?
Decrease > 5% /
Decrease < 5%
8.2.5 PRISMA-7
The PRISMA-7 is a seven-Item, self-completion questionnaire Instrument.
It is intended to be used as a postal questionnaire, or for people who are too unwell to undertake the 4
metre walking speed test.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 78/183
One point is scored for each of its seven questions and a score of 3 points or more is considered to
identify frailty.
As other similar endeavours, it has been conceived as a cost-effective tool to assess large number of
people, with good sensitivity and specificity. It aims to preview moderate to severe disabilities and may
be less efficient in detecting milder ones. In fact, it is used more as a case finding Instrument (current
state: prevalent cases) rather than a screening one (predicting incident cases).
8.2.5.1 References
A paper that illustrates how PRISMA-7 was conceived and validated is referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/17723247
A paper that compares PRISMA-7 properties with those of four other Instruments is referred
by the following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/23108163
8.2.5.2 Structure
ID Category Item Values
De.Pr.01 Demographics Are you more than 85 years? Yes / No
De.Pr.02 Demographics Are you male? Yes / No
He.Pr.03 Health In general, do you have any
health problems that require you
to limit your activities?
Yes / No
Dp.Pr.04 Dependence Do you need someone to help
you on a regular basis? Yes / No
He.Pr.05 Health In general, do you have any
health problems that require you
to stay at home?
Yes / No
Dp.Pr.06 Dependence In case of need, can you count
on someone close to you? Yes / No
Mt.Pr.07 Motility Do you regularly use a stick,
walker or wheelchair to get
about?
Yes / No
8.2.6 Edmonton Frail Scale
8.2.6.1 Overview
The Edmonton Frail Scale (EFS) is intended as a simple valid measure of frailty, covering multiple
important Categories, with scores ranging from 0 (not frail) to 17 (very frail).
The EFS does not depend on formal medical training to administer, requires less than 5 minutes of the
patient’s time, and it has been shown to be a valid measure of frailty compared with the clinical
impression of geriatric specialists after their more comprehensive assessment.
8.2.6.2 References
The paper that defines and validates the EFS Instrument is referred by the following Pubmed
URL http://www.ncbi.nlm.nih.gov/pubmed/16757522
A study that further validates EFS demonstrating that it is associated with increasing
comorbidity, hospital lengths of stay, lower use of invasive procedures, and increased
mortality in a known high-risk population (elderly patients with ACS) is referred by Pubmed
URL http://www.ncbi.nlm.nih.gov/pubmed/24183299
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 79/183
8.2.6.3 Structure
ID Category Item Values
Vi.Ed.01 Visuospatial Clock diagram: Place the
numbers in the correct positions
then place the hands to indicate
a time of “10 after 11”
No errors /
Minor spacing errors /
Other errors.
He.Ed.02 Health Hospital admissions in past year 0 /
1-2. /
>=2.
He.Ed.03 Health General health description Excellent, very good, or good /
Fair /
Poor.
Ia.Ed.04 IADLs Requires assistance with
activities such as meal
preparation, shopping,
transportation, dialling
telephone, housekeeping,
laundry, managing money,
taking medications
0-1 /
2-4 /
5-8.
Dp.Ed.05 Dependence Availability of individuals who
are willing and able to support
patient needs
Always /
Sometimes /
Never.
He.Ed.06 Health Five or more different
prescription medications on a
regular basis
Yes / No
Me.Ed.07 Medication Forgetfulness about taking
prescription medications Yes / No
We.Ed.08 Weight Weight loss Yes / No
Mo.Ed.09 Mood Reported feelings of sadness or
depression Yes / No
He.Ed.10 Health Unexpected urinary
incontinence Yes / No
Mt.Ed.11 Motility Patient begins by sitting in a
chair with back and arms
resting, then stands up and
walks approximately 3 m, and
returns to the chair and sits
down
0-10s /
11s-20s /
>= 20s
8.2.7 Tilburg Frailty Indicator
8.2.7.1 Overview
As most frailty assessments Instruments are dominated by biomedical indicators, the Tilburg Frailty
Indicator (TFI) was developed in order to extend this view, and to consider frailty from a live course
perspective, expressing relationships between life-course determinants, diseases, frailty and adverse
outcomes.
Based on this model, TFI addresses the measurement of frailty in community-dwelling older persons,
along three domains: physical, psychological and social.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 80/183
TFI has been shown to have good predictive validity for quality of life and adverse outcomes such as
disability and receiving personal care, nursing, and informal care.
8.2.7.2 References
The TFI Instrument can be found on the website of the Tilburg University, at the URL
https://www.tilburguniversity.edu/upload/ac3c1079-6188-4bea-b4af-
8f552c07a1d2_tfieng.pdf
A study that assesses the reliability and validity of the TFI Instrument is referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/20511102
8.2.7.3 Structure
ID Category Item Values
He.Ti.01 Health Do you feel physically healthy? Yes / No
We.Ti.02 Weight Have you lost a lot of weight
recently without wishing to do
so? (‘a lot’ is: 6 kg or more
during the last six months, or 3
kg or more during the last
month)
Yes / No
Mt.Ti.03 Motility Do you experience problems in
your daily life due to difficulty
in walking?
Yes / No
Mt.Ti.04 Motility Do you experience problems in
your daily life due to difficulty
maintaining your balance?
Yes / No
He.Ti.05 Health Do you experience problems in
your daily life due to poor
hearing?
Yes / No
He.Ti.06 Health Do you experience problems in
your daily life due to poor
vision?
Yes / No
Wk.Ti.07 Weakness Do you experience problems in
your daily life due to lack of
strength in your hands?
Yes / No
Ex.Ti.08 Exhaustion Do you experience problems in
your daily life due to physical
tiredness?
Yes / No
Mr.Ti.09 Memory Do you have problems with
your memory? Yes / Sometimes / No
Mo.Ti.10 Mood Have you felt down during the
last month? Yes / Sometimes / No
Mo.Ti.11 Mood Have you felt nervous or
anxious during the last month? Yes / Sometimes / No
Mo.Ti.12 Mood Are you able to cope with
problems well? Yes / No
Dp.Ti.13 Dependence Do you live alone? Yes / No
Dp.Ti.14 Dependence Do you sometimes miss having
people around you? Yes / Sometimes / No
Dp.Ti.15 Dependence Do you receive enough support
from other people? Yes / No
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 81/183
8.2.8 Comprehensive Frailty Assessment Instrument
8.2.8.1 Overview
The Comprehensive Frailty Assessment Instrument (CFAI) aims to broaden the body of knowledge
regarding the concept of frailty by introducing a multidimensional, self-administrated instrument
capturing 4 domains of frailty: physical, psychological, social and environmental.
In particular, CFAI was the first Instrument to include the environmental Category in the assessment of
frailty. It was developed on the basis of the TFI Instrument, that already added psychological and social
measurement to the traditional physical measurement.
The relationship of an aging individual with her/his spatial context is assumed to be essential and to
contribute to an aging individual’s quality of life. For instance, as authors report, evidence suggests that
the proximity of amenities and services may promote health either directly or indirectly through the
possibilities they provide for people to live healthy lives.
8.2.8.2 References
The paper that describes the development and initial assessment of the CFAI Instrument is
referred by the following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/23608069
A study that assesses the validity of the CFAI Instrument in comparison with the TFI
instrument is referred by the following URL:
http://www.europeangeriaticmedicine.com/article/S1878-7649(13)00045-4/abstract
8.2.8.3 Structure
ID Category Item Values
Ac.Cf.01 Activity Indicate how long you have
been hampered by your health
status in performing less
demanding activities like
carrying shopping bags
Not at all \
3 months or less \
More than 3 months
Mt.Cf.02 Motility Indicate how long you have
been hampered by your health
status in walking up a hill/stairs
Not at all \
3 months or less \
More than 3 months
Mt.Cf.03 Motility Indicate how long you have
been hampered by your health
status in bending or lifting
Not at all \
3 months or less \
More than 3 months
Mt.Cf.04 Motility Indicate how long you have
been hampered by your health
status in going for a walk
Not at all \
3 months or less \
More than 3 months
Mo.Cf.05 Mood To what extent do you agree
with the statement “Feeling
unhappy”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 82/183
ID Category Item Values
Mo.Cf.06 Mood To what extent do you agree
with the statement “Losing self-
confidence”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.07 Mood To what extent do you agree
with the statement “Unable to
cope with problems”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.08 Mood To what extent do you agree
with the statement “Feeling
pressure”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.09 Mood To what extent do you agree
with the statement “Feeling
worth nothing anymore”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.10 Mood To what extent do you agree
with the statement “I experience
a general sense of emptiness”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.11 Mood To what extent do you agree
with the statement “I miss
having people around me”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Mo.Cf.12 Mood To what extent do you agree
with the statement “I often feel
rejected”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Dp.Cf.13 Dependence There are plenty of people I can
lean on when I have problems Yes / No
Dp.Cf.14 Dependence There are many people I can
trust completely Yes / No
Dp.Cf.15 Dependence There are enough people I feel
close to Yes / No
Dp.Cf.16 Dependence How many persons can you rely
on among partner, son and
daughter-in-law? (Social support
network 1)
Number of persons
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 83/183
ID Category Item Values
Dp.Cf.17 Dependence How many persons can you rely
on among daughter, son-in-law
and grandchildren? (Social
support network 2)
Number of persons
Dp.Cf.18 Dependence How many persons can you rely
on among brother or sister (-in-
law), family, neighbours and
friends? (Social support network
3)
Number of persons
En.Cf.19 Environment My house is in a bad condition Yes / No
En.Cf.20 Environment My house is not comfortable Yes / No
En.Cf.21 Environment It is difficult to heath my house Yes / No
En.Cf.22 Environment There is insufficient comfort in
my house Yes / No
En.Cf.23 Environment I do not like the neighbourhood Yes / No
8.2.9 Groningen Frailty Indicator
8.2.9.1 Overview
The Groningen Frailty Indicator (GFI) has been developed to identify frailty of home-dwelling as well
as institutionalized elderly people. It comprises both a professional and a self-assessed version.
The GFI is widely used in clinical practice (i.e., geriatric centres, nursing homes, emergency
departments, traumatology, pulmonology, rheumatology, and surgical medicine), in outpatient settings,
and in clinical studies.
To obtain the self-reported version (presented here), the professional version of the GFI was modified
from a patient-orientated questionnaire (with Items such as “Has the patient recently felt downhearted
or sad?”) to an individual-oriented questionnaire (with Items such as “Have you recently felt
downhearted or sad?”) and, as a consequence, the formulations of all Items were adapted.
8.2.9.2 Reference
The paper that illustrates the development and initial test of the GFI Instrument can be found
at the University of Groningen website:
http://www.rug.nl/research/portal/publications/measuring-frailty(f91ecfcc-18e0-481f-84a7-
1230d62e032c).html
A couple of studies that address the predictive validity of GFI (and conclude that, for the
self-reported version, more work is needed) can be found at the following Pubmed URLs:
http://www.ncbi.nlm.nih.gov/pubmed/20353611 and
http://www.ncbi.nlm.nih.gov/pubmed/22579590
8.2.9.3 Structure
ID Category Item Values
Sh.Gr.01 Shopping Are you able to carry out
shopping single-handedly and
without any help?
Yes / No
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 84/183
ID Category Item Values
Mt.Gr.02 Motility Are you able to carry out
walking around outside (around
the house or to the neighbours)
single-handedly and without any
help?
Yes / No
Ad.Gr.03 ADLs Are you able to carry out
dressing and undressing single-
handedly and without any help?
Yes / No
Ad.Gr.04 ADLs Are you able to carry out going
to the toilet single-handedly and
without any help?
Yes / No
He.Gr.05 Health What mark do you give yourself
for physical fitness? Yes / No
He.Gr.06 Health Do you experience problems in
daily life due to poor vision? Yes / No
He.Gr.07 Health Do you experience problems in
daily life due to being hard of
hearing?
Yes / No
We.Gr.08 Weight During the last 6 months have
you lost a lot of weight
unwillingly? (3 kg in 1 month or
6 kg in 2 months)
Yes / No
He.Gr.09 Health Do you take 4 or more different
types of medicine? Yes / No
Mr.Gr.10 Memory Do you have any complaints
about your memory? Yes / No
Mo.Gr.11 Mood Do you sometimes experience
emptiness around yourself? Yes / No
Mo.Gr.12 Mood Do you sometimes miss people
around yourself? Yes / No
Mo.Gr.13 Mood Do you sometimes feel
abandoned? Yes / No
Mo.Gr.14 Mood Have you recently felt
downhearted or sad? Yes / No
Mo.Gr.15 Mood Have you recently felt nervous
or anxious? Yes / No
8.2.10 The Sherbrooke Postal Questionnaire
8.2.10.1 Overview
The Sherbrooke Postal Questionnaire is a simple six-Items Instruments, suitable for postal
administration, that has been developed in order to enact effective programmes of assessment and
surveillance in a context of secondary prevention. The Items’ Categories cover the physical (four
Items), social support (one Item), and cognitive (one Item) domains of functioning.
The authors found it valid for screening elderly individuals at risk for functional decline, although
another study, comparing it with the GFI and TFI Instruments, suggests that further research is needed.
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 85/183
8.2.10.2 References
The paper that illustrates the development of the SPQ Instruments, also assessing its
predictive validity, can be found at the following Pubmed URL:
http://www.ncbi.nlm.nih.gov/pubmed/8670547
The SPQ Instrument has been compared to the GFI and TFI instruments in the study referred
by the following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/20353611
8.2.10.3 Structure
ID Category Item Values
Dp.Sb.01 Dependence Do you live alone? Yes / No
He.Sb.02 Health Do you take more than three
different medications every day? Yes / No
Mt.Sb.03 Motility Do you regularly use a cane, a
walker or a wheelchair to move
about?
Yes / No
He.Sb.04 Health Do you see well? Yes / No
He.Sb.05 Health Do you hear well? Yes / No
Mr.Sb.06 Memory Do you have problems with
your memory? Yes / No
8.2.11 Frailty Index
8.2.11.1 Overview
The Frailty Index is presented here as an example of the Cumulative Deficit model.
We present the version discussed by Mitnitski et al. in the paper referred to in the Reference subsection
below.
It is composed by 40 Items that represent corresponding relevant health deficits (e.g. symptoms, health
attitudes, illnesses, and impaired function), the accumulation of which is assumed to represent the
frailty of the subject.
8.2.11.2 References
The development of the Instrument here presented is illustrated in the paper referred by the
following Pubmed URL: http://www.ncbi.nlm.nih.gov/pubmed/15215283; the paper also
offer predictive validation of the Instrument, by demonstrating the association between it
and mortality
8.2.11.3 Structure
ID Category Item Values
He.Fx.01 Health Eyesight 5 levels Likert scale
He.Fx.02 Health Hearing 5 levels Likert scale
Ad.Fx.03 ADLs Help to eat 3 levels Likert scale
Ad.Fx.04 ADLs Help to dress and undress 3 levels Likert scale
Ad.Fx.05 ADLs Ability to take care of
appearance 3 levels Likert scale
Mt.Fx.06 Motility Help to walk 3 levels Likert scale
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 86/183
ID Category Item Values
Mt.Fx.07 Motility Help to get in and out of bed 3 levels Likert scale
Ad.Fx.08 ADLs Help to take a bath or shower 3 levels Likert scale
Ad.Fx.09 ADLs Help to go to the bathroom 3 levels Likert scale
Co.Fx.10 Communication Help to use the telephone 3 levels Likert scale
Tr.Fx..11 Transportation Help to get to place out of
walking distance 3 levels Likert scale
Sh.Fx.12 Shopping Help in shopping 3 levels Likert scale
Fo.Fx.13 Food Help to prepare own meals 3 levels Likert scale
Ho.Fx.14 Housekeeping Help to do housework 3 levels Likert scale
Me.Fx.15 Medication Ability to take medicine 3 levels Likert scale
Fi.Fx.16 Finances Ability to handle own money 3 levels Likert scale
He.Fx.17 Health Self-rating of health 5 levels Likert scale
He.Fx.18 Health Troubles prevent normal
activities 3 levels Likert scale
Dp.Fx.19 Dependence Living alone Yes / No
He.Fx.20 Health Having a cough Yes / No
Ex.Fx.21 Exhaustion Feeling tired Yes / No
He.Fx.22 Health Nose stuffed up or sneezing Yes / No
He.Fx.23 Health High blood pressure Yes / No
He.Fx.24 Health Heart and circulation problems Yes / No
He.Fx.25 Health Stroke or effects of stroke Yes / No
He.Fx.26 Health Arthritis or rheumatism Yes / No
He.Fx.27 Health Parkinson’s disease Yes / No
He.Fx.28 Health Eye trouble Yes / No
He.Fx.29 Health Ear trouble Yes / No
He.Fx.30 Health Dental problems Yes / No
He.Fx.31 Health Chest problems Yes / No
He.Fx.32 Health Trouble with stomach Yes / No
He.Fx.33 Health Kidney trouble Yes / No
He.Fx.34 Health Losing control of bladder Yes / No
He.Fx.35 Health Losing control of bowels Yes / No
He.Fx.36 Health Diabetes Yes / No
He.Fx.37 Health Trouble with feet or ankles Yes / No
He.Fx.38 Health Trouble with nerves Yes / No
He.Fx.39 Health Skin problems Yes / No
He.Fx.40 Health Fractures Yes / No
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 87/183
9 Annex: Analysis of Items
The following Table presents the result of the Analysis of Items, extracted from currently established Instruments, aimed to assess the risk of MCI and frailty and to
predict the onset of the conditions.
The Table contains a row for each Item surveyed (see sub-section 4.1.3).
The first column assigns to the Item a unique ID that identifies it, for easy reference.
The second column describes the Item, while the third column reports the set of possible values for it.
The fourth column specifies the measurement method used for gauging the Item, in the Instrument to which it belongs, with the following meaning (see sub-section
4.2.2 for a discussion of this aspect):
Report: self-report or report from an informant (e.g. caregiver, family member, etc.)
Observation: observation by medical personnel
Meter: reading of a measurement instrument
The fifth column reports about the geriatric validity of the Instrument to which the Item belongs, with the following meaning:
Diagnostic: the Instrument has been validated for diagnostic usage (i.e. detection of the condition)
Predictive: the Instrument has been validated as a predictor of the condition’s onset
Predictive (study): studies have shown predictive validity for the Instrument, but the need of more work is acknowledged
Research (diagnostic | predictive): the Instrument derives from a research effort, not currently part of routine clinical practice
The last column reports which Measures, specified by the City4Age data collection platform (WP3), have been planned for deployment at the Project Pilot sites in
order to assess the Item (this provides a preliminary assessment of technical feasibility). The different fonts have the following meaning:
BOLD: means that the Measure is deployed as a direct assessment for the Item
NORMAL: means that the Measure is deployed as a proxy for the Item
ITALICS: means that the Measure is deployed as a tentative proxy for the Item; however it is recognized that verification against actual experimental data
is needed and more work may be necessary (e.g. more sophisticated filtering, connection with the City4Age Human Activity Recognition System,
combining more measures and datasets, etc.)
UNDERLINE: the Measure is not unobtrusive, i.e. it is obtained by directly asking a question to the person or to an informant, or by requesting the
person to do some specific measurement actions, such as using a smart scale, etc. (see discussion in sub-section 6.2)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 88/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mt.Ne.01 Did you walk around
outside?
Not at all /
With help /
On your own with help /
On your own
Report Predictive WALK_DISTANCE_OUTDOOR
WALK_STEPS_OUTDOOR
WALK_TIME_OUTDOOR
Mt.Cf.04 Indicate how long you have
been hampered by your
health status in going for a
walk
Not at all \
3 months or less \
More than 3 months
Report Diagnostic WALK_DISTANCE_OUTDOOR
WALK_STEPS_OUTDOOR
WALK_TIME_OUTDOOR
Mt.Gr.02 Are you able to carry out
walking around outside
(around the house or to the
neighbours) single-handedly
and without any help?
Yes / No Report Predictive (study) WALK_DISTANCE_OUTDOOR
WALK_STEPS_OUTDOOR
WALK_TIME_OUTDOOR
Mt.Fx.06 Help to walk 3 levels Likert scale Report Predictive WALK_DISTANCE_OUTDOOR
WALK_STEPS_OUTDOOR
WALK_TIME_OUTDOOR
Mt.Sh.04 Because of a health or
physical problem, do you
have any difficulty walking
100 metres? (Exclude any
difficulties that you expect
to last less than three
months)
Yes / No Report Predictive WALK_DISTANCE
WALK_STEPS
Mt.Fr.03 By yourself and not using
aids, do you have any
difficulty walking several
hundred yards?
Yes / No Report Predictive WALK_DISTANCE
WALK_STEPS
Mt.Ti.03 Do you experience problems
in your daily life due to
difficulty in walking?
Yes / No Report Predictive WALK_DISTANCE
WALK_STEPS
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 89/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mt.Fi.04 Time to walk 15 feet (4.57
meters) Number of seconds
beyond/not beyond
thresholds:
For men with height <= 173
cm: 7 secs /
For men with height > 173
cm: 6 secs /
For men with height <=
173cm: 7 secs /
For men with height >
173cm: 6 secs
Meter Diagnostic,
Predictive WALK_SPEED_OUTDOOR
WALK_SPEED_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST_PERC
WALK_DISTANCE_OUTDOOR_SLOW_PERC
WALK_TIME_OUTDOOR_FAST
Mt.Gm.01 Time in seconds to walk
from a starting point to a
marker 15 feet away, turn,
and back at a normal casual
gait for a total of 30 feet
(9.14 meters).
Decrease of 0.013m/s/yr in
all subjects /
Decrease of 0.023m/s/yr in
men who are MCI
converters, at 14.2 yr prior to
MCI diagnosis /
Decrease of 0.025m/s/yr in
women who are MCI
converters, at 6.0 yr prior to
MCI diagnosis
Meter Research
(predictive) WALK_SPEED_OUTDOOR
WALK_SPEED_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST_PERC
WALK_DISTANCE_OUTDOOR_SLOW_PERC
WALK_TIME_OUTDOOR_FAST
Mt.Ne.02 Did you climb stairs? Not at all /
With help /
On your own with help /
On your own
Report Predictive STAIRS_FLOOR_CHANGES_UP
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 90/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mt.Sh.05 Because of a health or
physical problem, do you
have any difficulty climbing
one flight of stairs without
resting? (Exclude any
difficulties that you expect
to last less than three
months)
Yes / No Report Predictive STAIRS_FLOOR_CHANGES_UP
Mt.Fr.02 By yourself and not using
aids, do you have any
difficulty walking up 10
steps without resting?
Yes / No Report Predictive STAIRS_FLOOR_CHANGES_UP
Mt.Cf.02 Indicate how long you have
been hampered by your
health status in walking up a
hill/stairs
Not at all \
3 months or less \
More than 3 months
Report Diagnostic STAIRS_FLOOR_CHANGES_UP
Mt.Ne.03 Did you get in and out of a
car?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Mt.Ne.04 Did you walk over uneven
ground?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Mt.Ne.05 Did you cross roads? Not at all /
With help /
On your own with help /
On your own
Report Predictive
Mt.So.02 Subject's inability to rise
from a chair 5 times without
using her arms
True / False Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 91/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mt.Ed.11 Patient begins by sitting in a
chair with back and arms
resting, then stands up and
walks approximately 3 m,
and returns to the chair and
sits down
0-10s /
11s-20s /
>= 20s
Meter Diagnostic
Mt.Pr.07 Do you regularly use a stick,
walker or wheelchair to get
about?
Yes / No Report Diagnostic
Mt.Sb.03 Do you regularly use a cane,
a walker or a wheelchair to
move about?
Yes / No Report Predictive
Mt.Ti.04 Do you experience problems
in your daily life due to
difficulty maintaining your
balance?
Yes / No Report Predictive
Mt.Cf.03 Indicate how long you have
been hampered by your
health status in bending or
lifting
Not at all \
3 months or less \
More than 3 months
Report Diagnostic
Mt.Fx.07 Help to get in and out of bed 3 levels Likert scale Report Predictive
Ac.Ne.21 Did you manage your own
garden?
Not at all /
With help /
On your own with help /
On your own
Report Predictive PHYSICALACTIVITY_INTENSE_TIME
PHYSICALACTIVITY_MODERATE_TIME
PHYSICALACTIVITY_SOFT_TIME
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 92/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ac.Fi.05 Physical expenditure on
activity scale per week on 18
items (Walking for exercise,
moderately strenuous
household chores, mowing
or raking the lawn,
gardening, hiking, jogging,
biking, exercise cycle,
dancing, aerobics, bowling,
golf, singles or doubles
tennis, racquetball,
calisthenics, swimming. To
compute kcals expended per
week, use the formula:
kcal/week = [activity-
specific MET (kcal/kg ×
hour) ] × [duration per
session (min) / 60 min] ×
[body weight (kg)] ×
[number of sessions in the
last 2 wk / 2] × [number of
months per year activity was
done])
kCal beyond / not beyond
threshold of 270 kCal
Report Diagnostic,
Predictive PHYSICALACTIVITY_CALORIES
PHYSICALACTIVITY_NUM
Ac.Sh.06 How often do you engage in
activities that require a low
or moderate level of energy
such as gardening, cleaning
the car, or doing a walk?
Hardly ever, or never /
One to three times a month /
Once a week /
More than once a week.
Report Predictive PHYSICALACTIVITY_MODERATE_TIME
PHYSICALACTIVITY_SOFT_TIME
PHYSICALACTIVITY_NUM
Ac.Cf.01 Indicate how long you have
been hampered by your
health status in performing
less demanding activities
like carrying shopping bags
Not at all \
3 months or less \
More than 3 months
Report Diagnostic PHYSICALACTIVITY_SOFT_TIME
PHYSICALACTIVITY_NUM
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 93/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ad.Fx.08 Help to take a bath or
shower 3 levels Likert scale Report Predictive BATHROOMS_VISITS
BATHROOM_TIME
Ad.Gr.03 Are you able to carry out
dressing and undressing
single-handedly and without
any help?
Yes / No Report Predictive (study)
Ad.Fx.04 Help to dress and undress 3 levels Likert scale Report Predictive
Ad.Ne.07 Did you manage to feed
yourself?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Ad.Fx.03 Help to eat 3 levels Likert scale Report Predictive
Ad.Da.15 The person, sitting at a table,
shows how she/he would cut
a steak, take a bite of it, eat
soup and pour water into a
glass and drink it
Correct / Incorrect
performance for each item
Observation Diagnostic
Ad.Da.13 The person is taken to a
bathroom and asked to take
the cap off a tube of
toothpaste, put toothpaste on
a toothbrush, turn on the tap,
brush teeth, dampen
washcloth, put soap on
washcloth, wash the face
and turn off the tap
Correct / Incorrect
performance for each item
Observation Diagnostic BATHROOMS_VISITS
BATHROOM_TIME
Ad.Da.14 The person is invited to use a
hairbrush, put on a coat,
button a coat (three buttons),
fasten a zip and tie shoelaces
Correct / Incorrect
performance for each item
Observation Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 94/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ad.Fx.05 Ability to take care of
appearance 3 levels Likert scale Report Predictive
Ad.Gr.04 Are you able to carry out
going to the toilet single-
handedly and without any
help?
Yes / No Report Predictive (study) RESTROOM_VISITS
RESTROOM_TIME
Ad.Fx.09 Help to go to the bathroom 3 levels Likert scale Report Predictive RESTROOM_VISITS
RESTROOM_TIME
Ia.Ed.04 Requires assistance with
activities such as meal
preparation, shopping,
transportation, dialling
telephone, housekeeping,
laundry, managing money,
taking medications
0-1 /
2-4 /
5-8.
Report Diagnostic
Fo.Li.03 Food Preparation Plans, prepares, and serves
adequate meals
independently /
Prepares adequate meals if
supplied with ingredients. /
Heats and serves prepared
meals or prepares meals but
does not maintain adequate
diet /
Needs to have meals
prepared and served
Report Predictive MEALS_NUM
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 95/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Fo.Oa.04 Can you prepare your own
meals?
Without help (plan and cook
full meals yourself) /
With some help (can prepare
some things but unable to
cook full meals yourself) /
Completely unable to prepare
any meals
Report Diagnostic MEALS_NUM
Fo.Fx.13 Help to prepare own meals 3 levels Likert scale Report Predictive MEALS_NUM
Fo.Ne.08 Did you manage to make
yourself a hot drink?
Not at all /
With help /
On your own with help /
On your own
Report Predictive MEALS_NUM
Fo.Ne.09 Did you take hot drinks from
one room to another?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Fo.Ne.11 Did you make yourself a hot
snack?
Not at all /
With help /
On your own with help /
On your own
Report Predictive MEALS_NUM
Ho.Ne.14 Did you do your own
housework?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 96/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ho.Li.04 Housekeeping Maintains house alone with
occasion assistance (heavy
work) /
Performs light daily tasks
such as dishwashing, bed
making /
Performs light daily tasks,
but cannot maintain
acceptable level of
cleanliness /
Needs help with all home
maintenance tasks /
Does not participate in any
housekeeping tasks.
Report Predictive
Ho.Oa.05 Can you do your housework? Without help (can clean
floors, etc.) /
With some help (can do light
housework but need help
with heavy work) /
Completely unable to do any
housework
Report Diagnostic
Ho.Ne.10 Did you do the washing up? Not at all /
With help /
On your own with help /
On your own
Report Predictive
Ho.Fx.14 Help to do housework 3 levels Likert scale Report Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 97/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ln.Li.05 Laundry Does personal laundry
completely /
Launders small items, rinses
socks, stockings, etc. /
All laundry must be done by
others
Report Predictive WASHINGMACHINE_SESSIONS
Ln.Ne.13 Did you wash small items of
clothing?
Not at all /
With help /
On your own with help /
On your own
Report Predictive WASHINGMACHINE_SESSIONS
Ln.Ne.16 Did you do a full clothes
wash?
Not at all /
With help /
On your own with help /
On your own
Report Predictive WASHINGMACHINE_SESSIONS
Co.Li.01 Ability to Use Telephone Operates telephone on own
initiative; looks up and dials
numbers /
Dials a few well-known
numbers /
Answers telephone, but does
not dial /
Does not use telephone at all
Report Predictive PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 98/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Co.Oa.01 Can you use the telephone? Without help, including
looking up numbers and
dialling /
With some help (can answer
phone or dial operator in an
emergency, but need a
special phone or help in
getting the number or
dialling) /
Completely unable to use the
telephone
Report Diagnostic PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Co.Ne.18 Did you use the telephone? Not at all /
With help /
On your own with help /
On your own
Report Predictive PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Co.Da.03 The person is invited to dial
the operator, dial from a list
of telephone numbers, dial
from oral presentation and
dial from written
presentation.
Correct / Incorrect
performance for each item
Observation Diagnostic PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 99/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Co.Da.04 The person is observed on
picking up the receiver,
dialling, hanging up and
operating the telephone in
the correct sequence.
Correct / Incorrect
performance for each item
Observation Diagnostic PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Co.Fx.10 Help to use the telephone 3 levels Likert scale Report Predictive PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
Co.Da.05 The person is invited to fold
a letter in half, put it in an
envelope, seal the envelope,
put on a stamp, address it
(from a presented stimulus
card) and add a return
address (the person’s own
current address, without a
postal code).
Correct / Incorrect
performance for each item
Observation Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 100/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Sh.Li.02 Shopping Takes care of all shopping
needs independently /
Shops independently for
small purchases /
Needs to be accompanied on
any shopping trip /
Completely unable to shop
Report Predictive SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Sh.Oa.03 Can you go shopping for
groceries or clothes?
Without help (taking care of
all shopping needs yourself,
assuming you had
transportation) /
With some help (need
someone to go with you on
all shopping trips) /
Completely unable to do any
shopping
Report Diagnostic SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Sh.Ne.15 Did you do your own
shopping?
Not at all /
With help /
On your own with help /
On your own
Report Predictive SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 101/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Sh.Da.08 Before the preparation of the
letter (Co.Da.05), the
examiner instructs the
person that in 10 min she/he
will be going to a grocery
store to select six items:
orange juice, spaghetti,
cherry jam, tuna fish, rice
and tomatoes. After 1 min to
recall as many grocery items
as possible
Number of recalled items Observation Diagnostic SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Sh.Da.09 The person is taken to a
simulated grocery store to
pick out the items from a
total of 25
Number of recalled items Observation Diagnostic SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Sh.Da.10 The examiner then gives the
person a written grocery list
(milk, crackers, eggs and
laundry detergent) and asks
them to select the four items
and to hand them to the
examiner
Correctly / Incorrectly
picked items
Observation Diagnostic SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Sh.Gr.01 Are you able to carry out
shopping single-handedly
and without any help?
Yes / No Report Predictive (study)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 102/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Sh.Fx.12 Help in shopping 3 levels Likert scale Report Predictive
Tr.Li.06 Mode of Transportation Travels independently on
public transportation or
drives own car /
Arranges own travel via taxi,
but does not otherwise use
public transportation /
Travels on public
transportation when assisted
or accompanied by another /
Travel limited to taxi or
automobile with assistance of
another /
Does not travel at all
Report Predictive PUBLICTRANSPORT_RIDES_MONTH
PUBLICTRANSPORT_DISTANCE_MONTH
PUBLICTRANSPORT_TIME
Tr.Ne.06 Did you travel on public
transport?
Not at all /
With help /
On your own with help /
On your own
Report Predictive PUBLICTRANSPORT_RIDES_MONTH
PUBLICTRANSPORT_DISTANCE_MONTH
PUBLICTRANSPORT_TIME
Tr.Oa.02 Can you get to places out of
walking distance?
Without help (drive your own
car, or travel alone on buses,
or taxis) /
With some help (need
someone to help you or go
with you when traveling) /
Unable to travel unless
emergency arrangements are
made for a specialized
vehicle like an ambulance
Report Diagnostic TRANSPORT_TIME
PUBLICTRANSPORT_RIDES_MONTH
PUBLICTRANSPORT_DISTANCE_MONTH
PUBLICTRANSPORT_TIME
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 103/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Tr.Fx..11 Help to get to place out of
walking distance 3 levels Likert scale Report Predictive TRANSPORT_TIME
PUBLICTRANSPORT_RIDES_MONTH
PUBLICTRANSPORT_DISTANCE_MONTH
PUBLICTRANSPORT_TIME
Tr.Ne.22 Did you drive a car? Not at all /
With help /
On your own with help /
On your own
Report Predictive
Tr.Da.12 The examiner asks the
person to identify a driver's
correct response to 13 road
signals
Correctly / Incorrectly
identified signals
Observation Diagnostic
Fi.Li.08 Ability to Handle Finances Manages financial matters
independently (budgets,
writes checks, pays rent and
bills, goes to bank); collects
and keeps track of income /
Manages day-to-day
purchases, but needs help
with banking, major
purchases, etc. /
Incapable of handling money
Report Predictive
Fi.Oa.07 Can you handle your own
money?
Without help (write checks,
pay bills, etc.) /
With some help (manage
day-to-day buying but need
help with managing your
check-book and paying your
bills) /
Completely unable to handle
money
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 104/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Fi.Ne.12 Did you manage your own
money when out?
Not at all /
With help /
On your own with help /
On your own
Report Predictive
Fi.Da.06 The person is invited to
identify four different coins
and three notes
Correct / Incorrect
performance for each item
Observation Diagnostic
Fi.Da.07 The person is invited to
count four amounts of
money
Correct / Incorrect
performance for each item
Observation Diagnostic
Fi.Da.11 Given a note to pay for the
items, the person is invited
to make change
Change correctly / incorrectly
identified
Observation Diagnostic
Fi.Fx.16 Ability to handle own
money 3 levels Likert scale Report Predictive
Me.Li.07 Responsibility for Own
Medications
Is responsible for taking
medication in correct dosages
at correct time /
Takes responsibility if
medication is prepared in
advance in separate dosages /
Is not capable of dispensing
own medication
Report Predictive
Me.Oa.06 Can you take your own
medicine?
Without help (in the right
doses at the right time) /
With some help (able to take
medicine if someone
prepares it for you and/or
reminds you to take it) /
Completely unable to take
your medicines
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 105/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Me.Ed.07 Forgetfulness about taking
prescription medications Yes / No Report Diagnostic
Me.Fx.15 Ability to take medicine 3 levels Likert scale Report Predictive
So.Ne.20 Did you go out socially? Not at all /
With help /
On your own with help /
On your own
Report Predictive VISITS_PAYED_WEEK
VISITS_RECEIVED_WEEK
VISITORS_WEEK
SENIORCENTER_VISITS
SENIORCENTER_VISITS_WEEK
SENIORCENTER_VISITS_MONTH
SENIORCENTER_TIME
SENIORCENTER_TIME_OUT_PERC
SENIORCENTER_LONG_VISITS
FOODCOURT_VISITS_MONTH
FOODCOURT_VISITS_WEEK
FOODCOURT_TIME
PUBLICPARK_VISITS
PUBLICPARK _VISITS_MONTH
PUBLICPARK _TIME
OTHERSOCIAL_VISITS
OTHERSOCIAL _LONG_VISITS
OTHERSOCIAL _TIME
OTHERSOCIAL _TIME_OUT_PERC
RESTAURANTS_VISITS_WEEK
RESTAURANTS_VISITS_MONTH
RESTAURANTS_TIME
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 106/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Cu.Ne.17 Did you read newspapers or
books?
Not at all /
With help /
On your own with help /
On your own
Report Predictive CINEMA_VISITS
CINEMA_VISITS_MONTH
CINEMA_TIME
CULTUREPOI_VISITS_MONTH
CULTUREPOI_VISITS_TIME_PERC_MONTH
TVWATCHING_TIME
TVWATCHING_TIME_PERC
Cu.Ne.19 Did you write letters? Not at all /
With help /
On your own with help /
On your own
Report Predictive
En.Cf.19 My house is in a bad
condition Yes / No Report Diagnostic
En.Cf.20 My house is not comfortable Yes / No Report Diagnostic
En.Cf.21 It is difficult to heath my
house Yes / No Report Diagnostic
En.Cf.22 There is insufficient comfort
in my house Yes / No Report Diagnostic
En.Cf.23 I do not like the
neighbourhood Yes / No Report Diagnostic
Dp.Ti.13 Do you live alone? Yes / No Report Predictive
Dp.Sb.01 Do you live alone? Yes / No Report Predictive
Dp.Fx.19 Living alone Yes / No Report Predictive
Dp.Pr.06 In case of need, can you
count on someone close to
you?
Yes / No Report Diagnostic
Dp.Ed.05 Availability of individuals
who are willing and able to
support patient needs
Always /
Sometimes /
Never.
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 107/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Dp.Cf.17 How many persons can you
rely on among daughter,
son-in-law and
grandchildren? (Social
support network 2)
Number of persons Report Diagnostic
Dp.Cf.16 How many persons can you
rely on among partner, son
and daughter-in-law? (Social
support network 1)
Number of persons Report Diagnostic
Dp.Cf.18 How many persons can you
rely on among brother or
sister (-in-law), family,
neighbours and friends?
(Social support network 3)
Number of persons Report Diagnostic
Dp.Cf.13 There are plenty of people I
can lean on when I have
problems
Yes / No Report Diagnostic
Dp.Cf.14 There are many people I can
trust completely Yes / No Report Diagnostic
Dp.Cf.15 There are enough people I
feel close to Yes / No Report Diagnostic
Dp.Ti.15 Do you receive enough
support from other people? Yes / No Report Predictive
Dp.Pr.04 Do you need someone to
help you on a regular basis? Yes / No Report Diagnostic
Dp.Ti.14 Do you sometimes miss
having people around you? Yes / Sometimes / No Report Predictive
He.Fx.17 Self-rating of health 5 levels Likert scale Report Predictive GP_VISITS_MONTH
GP_TIME_MONTH
He.Ed.03 General health description Excellent, very good, or good
/ Fair /
Poor.
Report Diagnostic GP_VISITS_MONTH
GP_TIME_MONTH
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 108/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
He.Ti.01 Do you feel physically
healthy? Yes / No Report Predictive GP_VISITS_MONTH
GP_TIME_MONTH
He.Gr.05 What mark do you give
yourself for physical fitness? Yes / No Report Predictive (study) GP_VISITS_MONTH
GP_TIME_MONTH
He.Ed.06 Five or more different
prescription medications on
a regular basis
Yes / No Report Diagnostic PHARMACY_VISITS
PHARMACY _VISITS_WEEK
PHARMACY _VISITS_MONTH
PHARMACY _TIME
He.Gr.09 Do you take 4 or more
different types of medicine? Yes / No Report Predictive (study) PHARMACY_VISITS
PHARMACY _VISITS_WEEK
PHARMACY _VISITS_MONTH
PHARMACY _TIME
He.Sb.02 Do you take more than three
different medications every
day?
Yes / No Report Predictive PHARMACY_VISITS
PHARMACY _VISITS_WEEK
PHARMACY _VISITS_MONTH
PHARMACY _TIME
He.Fr..04 Did a doctor ever tell you
that you have [illness]?
[where illness is:
hypertension, diabetes,
cancer (other than a minor
skin cancer), chronic lung
disease, heart attack,
congestive heart failure,
angina, asthma, arthritis,
stroke, and kidney disease
Number < 5
Number >= 5
Report Predictive
He.Ed.02 Hospital admissions in past
year 0 /
1-2. /
>=2.
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 109/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
He.Pr.03 In general, do you have any
health problems that require
you to limit your activities?
Yes / No Report Diagnostic
He.Pr.05 In general, do you have any
health problems that require
you to stay at home?
Yes / No Report Diagnostic
He.Fx.02 Hearing 5 levels Likert scale Report Predictive
He.Fx.29 Ear trouble Yes / No Report Predictive
He.Sb.05 Do you hear well? Yes / No Report Predictive
He.Ti.05 Do you experience problems
in your daily life due to poor
hearing?
Yes / No Report Predictive
He.Gr.07 Do you experience problems
in daily life due to being
hard of hearing?
Yes / No Report Predictive (study)
He.Fx.01 Eyesight 5 levels Likert scale Report Predictive
He.Fx.28 Eye trouble Yes / No Report Predictive
He.Sb.04 Do you see well? Yes / No Report Predictive
He.Ti.06 Do you experience problems
in your daily life due to poor
vision?
Yes / No Report Predictive
He.Gr.06 Do you experience problems
in daily life due to poor
vision?
Yes / No Report Predictive (study)
He.Fx.18 Troubles prevent normal
activities 3 levels Likert scale Report Predictive
He.Fx.20 Having a cough Yes / No Report Predictive
He.Fx.22 Nose stuffed up or sneezing Yes / No Report Predictive
He.Ed.10 Unexpected urinary
incontinence Yes / No Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 110/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
He.Fx.23 High blood pressure Yes / No Report Predictive
He.Fx.24 Heart and circulation
problems Yes / No Report Predictive
He.Fx.25 Stroke or effects of stroke Yes / No Report Predictive
He.Fx.26 Arthritis or rheumatism Yes / No Report Predictive
He.Fx.27 Parkinson’s disease Yes / No Report Predictive
He.Fx.30 Dental problems Yes / No Report Predictive
He.Fx.31 Chest problems Yes / No Report Predictive
He.Fx.32 Trouble with stomach Yes / No Report Predictive
He.Fx.33 Kidney trouble Yes / No Report Predictive
He.Fx.34 Losing control of bladder Yes / No Report Predictive
He.Fx.35 Losing control of bowels Yes / No Report Predictive
He.Fx.36 Diabetes Yes / No Report Predictive
He.Fx.37 Trouble with feet or ankles Yes / No Report Predictive
He.Fx.38 Trouble with nerves Yes / No Report Predictive
He.Fx.39 Skin problems Yes / No Report Predictive
He.Fx.40 Fractures Yes / No Report Predictive
We.Fi.01 The person lost >10 pounds
unintentionally last year True / False Report Diagnostic,
Predictive
WEIGHT
We.So.01 Weight loss (irrespective of
intent to lose weight) of 5%
or more in a 2 years period
True / False Report Predictive WEIGHT
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 111/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
We.Ti.02 Have you lost a lot of weight
recently without wishing to
do so? (‘a lot’ is: 6 kg or
more during the last six
months, or 3 kg or more
during the last month)
Yes / No Report Predictive WEIGHT
We.Gr.08 During the last 6 months
have you lost a lot of weight
unwillingly? (3 kg in 1
month or 6 kg in 2 months)
Yes / No Report Predictive (study) WEIGHT
We.Ed.08 Weight loss Yes / No Report Diagnostic WEIGHT
We.Fr.05 How much do you weigh
with your clothes on but
without shoes? One year ago
in (MO, YR), how much did
you weigh without your
shoes and with your clothes
on?
Decrease > 5% /
Decrease < 5%
Report Predictive WEIGHT
We.Sh.02 What has your appetite been
like? Diminution in desire for food
and/or eating less than usual /
No change in desire for food
and/or eating the same as
usual /
Increase in desire for food
and/or eating more than
usual.
Report Predictive WEIGHT
Wk.Ti.07 Do you experience problems
in your daily life due to lack
of strength in your hands?
Yes / No Report Predictive WEAKNESS
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 112/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Wk.Fi.06 Grip strength (average of 3
trials, dominant hand,
measured with Jamar hand
dynamometer)
Kg beyond/not beyond
thresholds:
For men, BMI <= 24: 29 kg /
For men, BMI 24.1-26: 30 kg
/
For men, BMI 26.1-28: 30 kg
/
For men, BMI > 28: 32 kg /
For women, BMI <= 23:
17kg /
For women, BMI 23.1-26:
17.3 kg /
For women, BMI 26.1-29: 18
kg /
For women, BMI > 29: 21 kg
Meter Diagnostic,
Predictive
WEAKNESS
Wk.Sh.03 Grip strength (highest
among four measures, two
for each hand, taken with a
dynamometer)
Kg (continuous measure)
Meter Predictive WEAKNESS
Ex.Fi.02 The person felt that
everything she/he did was an
effort in last week
Rarely or none of the time
(<1 day) /
Some or little of the time (1
to 2 days) /
Moderate amount of the time
(3 to 4 days) /
Most of the time
Report Diagnostic,
Predictive
EXHAUSTION
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 113/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ex.Fi.03 The person felt that she/he
could not get going in last
week
Rarely or none of the time
(<1 day) /
Some or little of the time (1
to 2 days) /
Moderate amount of the time
(3 to 4 days) /
Most of the time
Report Diagnostic,
Predictive
EXHAUSTION
Ex.So.03 Do you feel full of energy? Yes / No Report Predictive EXHAUSTION
Ex.Sh.01 In the last month, have you
had too little energy to do
the things you wanted to do?
Yes / No Report Predictive EXHAUSTION
Ex.Fr.01 How much of the time
during the past 4 weeks did
you feel tired?
Rarely or none of the time /
Some or little of the time /
Moderate amount of the time
/
Most of the time
Report Predictive EXHAUSTION
Ex.Ti.08 Do you experience problems
in your daily life due to
physical tiredness?
Yes / No Report Predictive EXHAUSTION
Ex.Fx.21 Feeling tired Yes / No Report Predictive EXHAUSTION
Ab.Sm.06 The examiner proposes three
pair of words and asks the
person to state the abstract
interpretation:
orange/banana, dog/horse,
table/bookcase (e.g.
dog/horse = animal)
Correctly / Incorrectly
abstracted interpretations
Observation Predictive WORD_PAIRS_TEST
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 114/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ab.Mo.12 The examiner asks the
subject to explain what each
pair of words has in
common, starting with the
example: “Tell me how an
orange and a banana are
alike”. If the subject answers
in a concrete manner, then
say only one additional time:
“Tell me another way in
which those items are alike”.
If the subject does not give
the appropriate response
(fruit), say, “Yes, and they
are also both fruit.” Do not
give any additional
instructions or clarification.
After the practice trial, say:
“Now, tell me how a train
and a bicycle are alike”.
Following the response,
administer the second trial,
saying: “Now tell me how a
ruler and a watch are alike”
Do not give any additional
instructions or prompts.
Correctly / Incorrectly
abstracted interpretations
Observation Predictive WORD_PAIRS_TEST
At.Mo.09 The examiner gives the
following instruction: “Now,
I will ask you to count by
subtracting seven from 100,
and then, keep subtracting
seven from your answer
until I tell you to stop.”
Correct / Incorrect
subtractions
Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 115/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
At.Mm.04 The examiner asks the person
to spell the word WORLD.
Now spell it backwards. (in
alternative, the interviewer
counting down from one
hundred by sevens)
Correct / Incorrect answer
(position of the first error)
Observation Predictive (study)
At.Sm.03 The following digit spans of
increasing length are
presented in sequence to the
person, asking for repetition:
2-9-6-8-3, 5-7-1-9-4-6, 2-1-
5-9-3-6-2.
Length of correctly repeated
spans
Observation Predictive
At.Sm.05 The examiner proposes 4
arithmetic operations to be
computed: 5x13, 65-7, 58/2,
29+11
Correctly / Incorrectly
computed operations
Observation Predictive
At.Mo.06 The examiner gives the
following instruction: “I am
going to say some numbers
and when I am through,
repeat them to me exactly as
I said them”. Then she/he
reads the five number
sequence at a rate of one
digit per second: 2-1-8-5-4.
Correctly / Incorrectly
repeated sequence
Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 116/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
At.Mo.07 The examiner gives the
following instruction: “Now
I am going to say some more
numbers, but when I am
through you must repeat
them to me in the backwards
order.” Read the three
number sequence at a rate of
one digit per second: 7-4-2
Correctly / Incorrectly
repeated sequence
Observation Predictive
At.Mo.08 The examiner reads a list of
letters (F B A C M N A A J
K L B A F A K D E A A A J
A M O F A A B) at a rate of
one per second, after giving
the following instruction: “I
am going to read a sequence
of letters. Every time I say
the letter A, tap your hand
once. If I say a different
letter, do not tap your hand”.
Correctly / Incorrectly tapped
As
Observation Predictive
Mr.Gr.10 Do you have any complaints
about your memory? Yes / No Report Predictive (study) MEMORY
Mr.Sb.06 Do you have problems with
your memory? Yes / No Report Predictive MEMORY
Mr.Ti.09 Do you have problems with
your memory? Yes / Sometimes / No Report Predictive MEMORY
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 117/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mr.Mm.03 The examiner says “I am
going to name three objects.
When I am finished, I want
you to repeat them.
Remember what they are
because I am going to ask
you to name them again in a
few minutes.” The examiner
says the following words
slowly at 1‐second intervals ‐ ball/ car/ man
Correct / Incorrect repetition
for each word
Observation Predictive (study) MEMORY
Mr.Mm.05 The examiner asks the person
what were the three objects
she/he asked to remember?
Correctly / Incorrectly
recalled words
Observation Predictive (study) MEMORY
Mr.Sm.04 The examiner tell four
unrelated words: “apple”,
“Mr. Johnson”, “charity”,
“tunnel”. The person is
requested to repeat all words.
Number of trial to repeat all
four words
Observation Predictive MEMORY
Mr.Sm.10 The examiner asks the person
to recall the words from
Mr.Sm.04
Correctly/incorrectly recalled
words
Observation Predictive MEMORY
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 118/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mr.Mo.05 The examiner reads a list of
5 words (face, velvet,
church, daisy, red) at a rate
of one per second, giving the
following instructions: “This
is a memory test. I am going
to read a list of words that
you will have to remember
now and later on. Listen
carefully. When I am
through, tell me as many
words as you can remember.
It doesn’t matter in what
order you say them”. When
the subject indicates that
she/he has finished (has
recalled all words), or can
recall no more words, the
examiner reads the list a
second time with the
following instructions: “I am
going to read the same list
for a second time. Try to
remember and tell me as
many words as you can,
including words you said the
first time.”. At the end of the
second trial, the examiner
informs the person that
she/he will be asked to recall
these words again”
Correct / Incorrect repetition
for each word
Observation Predictive MEMORY
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 119/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mr.Mo.13 The examiner gives the
following instruction: “I
read some words to you
earlier, which I asked you to
remember. Tell me as many
of those words as you can
remember.”
Correctly / Incorrectly
recalled words
Observation Predictive MEMORY
Mo.Ed.09 Reported feelings of sadness
or depression Yes / No Report Diagnostic
Mo.Ti.10 Have you felt down during
the last month? Yes / Sometimes / No Report Predictive
Mo.Ti.11 Have you felt nervous or
anxious during the last
month?
Yes / Sometimes / No Report Predictive
Mo.Ti.12 Are you able to cope with
problems well? Yes / No Report Predictive
Mo.Cf.05 To what extent do you agree
with the statement “Feeling
unhappy”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Cf.06 To what extent do you agree
with the statement “Losing
self-confidence”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 120/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mo.Cf.07 To what extent do you agree
with the statement “Unable
to cope with problems”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Cf.08 To what extent do you agree
with the statement “Feeling
pressure”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Cf.09 To what extent do you agree
with the statement “Feeling
worth nothing anymore”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Cf.10 To what extent do you agree
with the statement “I
experience a general sense
of emptiness”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Cf.11 To what extent do you agree
with the statement “I miss
having people around me”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 121/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Mo.Cf.12 To what extent do you agree
with the statement “I often
feel rejected”
Completely disagree /
Disagree /
Neither agree nor disagree /
Agree /
Completely agree.
Report Diagnostic
Mo.Gr.11 Do you sometimes
experience emptiness around
yourself?
Yes / No Report Predictive (study)
Mo.Gr.12 Do you sometimes miss
people around yourself? Yes / No Report Predictive (study)
Mo.Gr.13 Do you sometimes feel
abandoned? Yes / No Report Predictive (study)
Mo.Gr.14 Have you recently felt
downhearted or sad? Yes / No Report Predictive (study)
Mo.Gr.15 Have you recently felt
nervous or anxious? Yes / No Report Predictive (study)
Ti.Da.01 The person is shown four
different times (0300 h, 0800
h, 1030 h and 1215 h) using a
large model of a clock and is
asked to tell the time
Correct / Incorrect answer
for each item
Observation Diagnostic
Ti.Da.02 The person is asked to state
the date, the day, the month
and the year.
Correct / Incorrect answer
for each item
Observation Diagnostic
Ti.Mm.01 The person is asked to state
the year, season, month, date
and day of the week.
Correct / Incorrect answer
for each element
Observation Predictive (study)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 122/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Ti.Mo.14 The examiner gives the
following instructions: “Tell
me the date today”. If the
subject does not give a
complete answer, then
prompt accordingly by
saying: “Tell me the [year,
month, exact date, and day
of the week].”
Correct / Incorrect answer
for each element
Observation Predictive
Ti.Sm.02 The person is asked to state
date (day), month, year. Correct / Incorrect answer
for each item
Observation Predictive
Sp.Mm.02 The person is asked to state
the country, province,
city/town, address (home:
street address, in-facility:
building name) and location
(home: room name, in-
facility: floor)
Correct / Incorrect answer
for each element
Observation Predictive (study)
Sp.Sm.01 The person is asked to state
name, address, current
location (building), city
state.
Correct / Incorrect answer
for each item
Observation Predictive
Sp.Mo.15 Then the examiner says:
“Now, tell me the name of
this place, and which city it
is in.”
Correct / Incorrect answer
for each element
Observation Predictive
Vi.Sm.07 The examiner asks the person
to draw a clock face showing
11:15.
Clock face correctly /
incorrectly drawn
Observation Predictive
Vi.Sm.08 The examiner shows the
person the drawing of a cube
and asks to copy it
Drawing correctly /
incorrectly copied
Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 123/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Vi.Mm.10 The examiner places a
drawing (interlocking
pentagons, i.e. two five-sided
figures intersecting to make a
four-sided figure), eraser and
pencil in front of the person
and asks “Copy this drawing
please.” (Allow multiple
tries; wait until person is
finished and hands it back.)
Drawing correctly /
incorrectly copied
Observation Predictive (study)
Vi.Mo.01 The examiner instructs the
subject: "Please draw a line,
going from a number to a
letter in ascending order.
Begin here [point to (1)] and
draw a line from 1 then to A
then to 2 and so on. End here
[point to (E)].".
Correct / Incorrect answer
for each item
Observation Predictive
Vi.Mo.02 The examiner gives the
following instructions,
pointing to the cube: “Copy
this drawing as accurately as
you can, in the space
below”.
Drawing correctly /
incorrectly copied
Observation Predictive
Vi.Mo.03 The examiner give the
following instructions:
“Draw a clock. Put in all the
numbers and set the time to
10 past 11”.
Clock face correctly /
incorrectly drawn
Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 124/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
Vi.Ed.01 Clock diagram: Place the
numbers in the correct
positions then place the
hands to indicate a time of
“10 after 11”
No errors /
Minor spacing errors /
Other errors.
Observation Diagnostic
La.Mm.06 The examiner show a
wristwatch and asks the
person “What is this called?”.
Repeats with a pencil
Correctly / Incorrectly
identified items
Observation Predictive (study)
La.Mm.07 The examiner asks the person
to repeat this phrase “No ifs,
ands or buts.”
Correct / Incorrect repetition Observation Predictive (study)
La.Mm.08 The examiner asks the person
to read the words on a page
and then do what it says.
Then hand the person the
page with “CLOSE YOUR
EYES” on it. If the subject
reads and does not close their
eyes, repeat up to three times.
Score only if subject closes
eyes
Closing / Not closing eyes Observation Predictive (study)
La.Mm.09 The examiner hands the
person a pencil and paper,
then says “Write any
complete sentence on that
piece of paper”
Writing a sentence that make
/ does not make sense (ignore
spelling errors)
Observation Predictive (study)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 125/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
La.Mm.11 The examiner asks the person
if she/he is right or left‐handed, takes a piece of
paper and hold it up in front
of the person. Then says
“Take this paper in your
right/left hand [note:
whichever is non‐dominant],
fold the paper in half once
with both hands and put the
paper down on the floor”.
Correct / incorrect
performance for each item
Observation Predictive (study)
La.Sm.09 The examiner asks the person
to provide three information
element: “first president”,
“define an island”, “number
of weeks per year”
Correct/incorrect information
provided for each item
Observation Predictive
La.Mo.04 The examiner points to three
animal figures (lion, rhino,
camel) in turn and says:
“Tell me the name of this
animal”.
Names correctly / incorrectly
repeated.
Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 126/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
La.Mo.10 The examiner gives the
following instructions: “I am
going to read you a sentence.
Repeat it after me, exactly as
I say it [pause]: I only know
that John is the one to help
today.” Following the
response, say: “Now I am
going to read you another
sentence. Repeat it after me,
exactly as I say it [pause]:
The cat always hid under the
couch when dogs were in the
room.”
Correct / Incorrect repetition Observation Predictive
La.Mo.11 The examiner gives the
following instruction: “Tell
me as many words as you
can think of that begin with
a certain letter of the
alphabet that I will tell you
in a moment. You can say
any kind of word you want,
except for proper nouns (like
Bob or Boston), numbers, or
words that begin with the
same sound but have a
different suffix, for example,
love, lover, loving. I will tell
you to stop after one minute.
Are you ready? [Pause]
Now, tell me as many words
as you can think of that
begin with the letter F. [time
for 60 sec]. Stop.”
Words generated Observation Predictive
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 127/183
ID Item Values Current
measurement
Geriatric validity City4Age measurement at Pilots
De.Pr.01 Are you more than 85 years? Yes / No Report Diagnostic
De.Pr.02 Are you male? Yes / No Report Diagnostic
Table 9. Analysis of Items
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 128/183
10 Annex: Coverage of Geriatric Factors
For the meaning of the different fonts in the third column of the following Table, see note in Section 9
(in particular, regarding the last column of Table 9).
Geriatric factors Corresponding
Category
Measures proposed by Pilots for testbed
experiments
Motility Mt
Walking Mt WALK_DISTANCE
WALK_STEPS
WALK_DISTANCE_OUTDOOR
WALK_STEPS_OUTDOOR
WALK_TIME_OUTDOOR
WALK_SPEED_OUTDOOR
WALK_SPEED_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST
WALK_DISTANCE_OUTDOOR_FAST_PERC
WALK_DISTANCE_OUTDOOR_SLOW_PERC
WALK_TIME_OUTDOOR_FAST
Climbing stairs Mt STAIRS_FLOOR_CHANGES_UP
Still/Moving n/a STILL_TIME
Moving across
rooms n/a ROOM_CHANGES
BEDROOM_VISITS
BEDROOM_TIME
BATHROOM_VISITS
BATHROOM_TIME
KITCHEN_VISITS
KITCHEN_TIME
LIVINGROOM_VISITS
LIVINGROOM _TIME
RESTROOM_VISITS
RESTROOM_TIME
Gait balance Mt
Physical activity Ac PHYSICALACTIVITY_CALORIES
PHYSICALACTIVITY_INTENSE_TIME
PHYSICALACTIVITY_MODERATE_TIME
PHYSICALACTIVITY_SOFT_TIME
PHYSICALACTIVITY_NUM
Basic Activities of Daily
Living
Ad
Bathing and
showering Ad BATHROOMS_VISITS
BATHROOM_TIME
Dressing Ad
Self-feeding Ad
Personal hygiene
and grooming Ad BATHROOMS_VISITS
BATHROOM_TIME
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 129/183
Geriatric factors Corresponding
Category
Measures proposed by Pilots for testbed
experiments
Toilet hygiene Ad RESTROOM_VISITS
RESTROOM_TIME
Going out Ad OUTDOOR_NUM
OUTDOOR_TIME
HOME_TIME
Instrumental Activities of
Daily Living
Ia
Ability to cook food Fo MEALS_NUM
Housekeeping Ho
Laundry Ln WASHINGMACHINE_SESSIONS
Phone usage Co PHONECALLS_PLACED
PHONECALLS_RECEIVED
PHONECALLS_LONG_PLACED_PERC
PHONECALLS_LONG_RECEIVED_PERC
PHONECALLS_MISSED
PHONECALLS_PLACED_PERC
PHONECALLS_RECEIVED_PERC
PHONECALLS_SHORT_PLACED_PERC
PHONECALLS_SHORT_RECEIVED_PERC
New media
communication n/a
Shopping Sh SHOPS_VISITS
SHOPS_VISITS_WEEK
SHOPS_TIME
SHOPS_OUTDOOR_TIME_PERC
SUPERMARKET_VISITS
SUPERMARKET_VISITS_WEEK
SUPERMARKET _TIME
SUPERMARKET _TIME_PERC
Transportation Tr TRANSPORT_TIME
PUBLICTRANSPORT_RIDES_MONTH
PUBLICTRANSPORT_DISTANCE_MONTH
PUBLICTRANSPORT_TIME
Finance
Management Fi
Medication Me
Socialization So
Visits So VISITS_PAYED_WEEK
VISITS_RECEIVED_WEEK
VISITORS_WEEK
Attending senior
centers So SENIORCENTER_VISITS
SENIORCENTER_VISITS_WEEK
SENIORCENTER_VISITS_MONTH
SENIORCENTER_TIME
SENIORCENTER_TIME_OUT_PERC
SENIORCENTER_LONG_VISITS
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 130/183
Geriatric factors Corresponding
Category
Measures proposed by Pilots for testbed
experiments
Attending other
social places So FOODCOURT_VISITS_MONTH
FOODCOURT_VISITS_WEEK
FOODCOURT_TIME
PUBLICPARK_VISITS
PUBLICPARK _VISITS_MONTH
PUBLICPARK _TIME
OTHERSOCIAL_VISITS
OTHERSOCIAL _LONG_VISITS
OTHERSOCIAL _TIME
OTHERSOCIAL _TIME_OUT_PERC
Restaurant So RESTAURANTS_VISITS_WEEK
RESTAURANTS_VISITS_MONTH
RESTAURANTS_TIME
Cultural engagement Cu
Visit entertainment /
culture places Cu CINEMA_VISITS
CINEMA_VISITS_MONTH
CINEMA_TIME
CULTUREPOI_VISITS_MONTH
CULTUREPOI_VISITS_TIME_PERC_MONTH
Watching TV Cu TVWATCHING_TIME
TVWATCHING_TIME_PERC
Reading newspapers Cu
Reading books Cu
Environment En
Quality of housing En
Quality of
neighbourhood En
Dependence Dp
Health – Physical He
Falls n/a FALLS_MONTH
Weight We WEIGHT
Weakness Wk WEAKNESS
Exhaustion Ex EXHAUSTION
Pain n/a PAIN
Appetite n/a APPETITE
Quality of sleep n/a SLEEP_AWAKE_TIME
SLEEP_DEEP_TIME
SLEEP_LIGHT_TIME
SLEEP_REM_TIME
SLEEP_TIME
SLEEP_TOSLEEP_TIME
SLEEP_WAKEUP_NUM
Visit to doctors He GP_VISITS_MONTH
GP_TIME_MONTH
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 131/183
Geriatric factors Corresponding
Category
Measures proposed by Pilots for testbed
experiments
Visit to health
related places He PHARMACY_VISITS
PHARMACY _VISITS_WEEK
PHARMACY _VISITS_MONTH
PHARMACY _TIME
Health – Cognitive n/a
Abstraction Ab WORD_PAIRS_TEST
Attention At
Memory Mr MEMORY
Mood Mo
Table 10. Coverage of City4Age geriatric factors from Pilots
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 132/183
11 Annex: Measures at City4Age Pilots
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
APPETITE Appetite reported in the day on a 1-5
scale
Daily integer Lecce (Health – physical,
Appetite)
BATHROOM_TIME Time in seconds spent in bathroom in
the day.
Daily integer Montpellier (Motility,
Moving across rooms; Basic
ADLs, Bathing and
showering), Singapore
(Motility, Moving across
rooms; Basic ADLs, Bathing
and showering)
BATHROOMS_VISITS Number of bathrooms entrances in
the day.
Daily integer Montpellier (Motility,
Moving across rooms; Basic
ADLs, Bathing and
showering), Singapore
(Motility, Moving across
rooms; Basic ADLs, Bathing
and showering)
BEDROOM_TIME Time in seconds spent in bedroom in
the day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms), Singapore (Motility,
Moving across rooms)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 133/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
BEDROOM_VISITS Number of bedroom entrances in the
day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms), Singapore (Motility,
Moving across rooms)
CINEMA_TIME Time in seconds spent in monitored
cinema/theatres in the day.
Daily integer Athens (Cultural
Engagement, Visiting
entertainment/culture places),
Montpellier (Cultural
Engagement, Visiting
entertainment/culture places)
CINEMA_VISITS Number of visits to monitored
cinema/theatres in the day.
Daily integer Athens (Cultural
Engagement, Visiting
entertainment/culture places),
Montpellier (Cultural
Engagement, Visiting
entertainment/culture places)
CINEMA_VISITS_MONTH Number of visits to monitored
cinema/theatres in the month.
Monthly integer Athens (Cultural
Engagement, Visiting
entertainment/culture places),
Montpellier (Cultural
Engagement, Visiting
entertainment/culture places)
CULTUREPOI_VISITS_MONTH Number of visits to monitored
cultural places in the month.
Monthly integer Madrid (Cultural
Engagement, Visiting
entertainment/culture places)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 134/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
CULTUREPOI_VISITS_TIME_PERC_MONTH Percentage of time spent in monitored
cultural places with respect to time
spent outside home in the month.
Monthly float Madrid (Cultural
Engagement, Visiting
entertainment/culture places)
EXHAUSTION Exhaustion measured along a 1-4
scale sampled in the month, according
to the following definition (from
Fried Frailty Index):
The person felt that everything
she/he did was an effort in last week
1. Rarely or none of the time (<1
day)
2. Some or little of the time (1 to 2
days)
3. Moderate amount of the time (3
to 4 days) /
4. Most of the time
Monthly float Madrid (Health – physical,
Weight)
FALLS_MONTH Number of falls detected in the
month.
Monthly integer Montpellier (Health –
physical, Fall), Singapore
(Health – physical, Fall)
FOODCOURT_TIME Time in seconds spent in FoodCourt
in the day.
Daily integer Singapore (Socialization,
Attending Other Social
Places)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 135/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
FOODCOURT_VISITS_MONTH Number of visits to FoodCourt in the
month.
Monthly integer Singapore (Socialization,
Attending Other Social
Places)
FOODCOURT_VISITS_WEEK Number of visits to FoodCourt in the
week.
Weekly integer Singapore (Socialization,
Attending Other Social
Places)
GP_TIME_MONTH Time spent at GP’s practice in the
month,
Monthly integer Lecce (Health – physical,
Visits to doctors)
GP_VISITS_MONTH Number of visits to GP in the month. Monthly integer Lecce (Health – physical,
Visits to doctors)
HEART_RATE Resting heart rate, measured in rpms
as well as the daily minutes in 5
ranges of heart rate in reference to the
maximum heart rate regime.
Daily float Madrid (Note: to be further
researched during the
experiment)
HOME_TIME Time in seconds spent at home in the
day.
Daily integer Athens (Basic ADLs, Going
out), Lecce (Basic ADLs,
Going out), Montpellier
(Basic ADLs, Going out),
Singapore (Basic ADLs,
Going out)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 136/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
KITCHEN_TIME Time in seconds spent in kitchen in
the day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms; Instrumental ADLs,
Ability to cook food),
Singapore (Motility, Moving
across rooms; Instrumental
ADLs, Ability to cook food)
KITCHEN_VISITS Number of kitchen entrances in the
day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms; Instrumental ADLs,
Ability to cook food),
Singapore (Motility, Moving
across rooms; Instrumental
ADLs, Ability to cook food)
LIVINGROOM _TIME Time in seconds spent in living room
in the day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms), Singapore (Motility,
Moving across rooms)
LIVINGROOM_VISITS Number of living room entrances in
the day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms), Singapore (Motility,
Moving across rooms)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 137/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
MEALS_NUM Meals prepared by the person in the
day.
Weekly integer Montpellier (Instrumental
ADLs, Ability to cook food),
Singapore (Instrumental
ADLs, Ability to cook food)
MEMORY Memory performance reported in the
day, as numerical indicator returned
from the CANTAB system
Monthly An object with a
float property for
each CANTAB
results test score
(the name of the
property is the
name of the
CANTAB test
code).
Lecce (Health – cognitive,
Memory)
OTHERSOCIAL _LONG_VISITS Number of long visits to
OtherSocialPlace in the day.
Daily integer Lecce (Socialization,
Attending Other Social
Places)
OTHERSOCIAL _TIME Time in seconds spent in
OtherSocialPlace in the day.
Daily integer Athens (Socialization,
Attending Other Social
Places), Lecce (Socialization,
Attending Other Social
Places)
OTHERSOCIAL _TIME_OUT_PERC Percentage of time spent in
OtherSocialPlace with respect to total
time spent outside home in the day.
Daily float Lecce (Socialization,
Attending Other Social
Places)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 138/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
OTHERSOCIAL_VISITS Number of visits to OtherSocialPlace
in the day.
Daily integer Athens (Socialization,
Attending Other Social
Places), Lecce (Socialization,
Attending Other Social
Places)
OUTDOOR_NUM Total number of exits from home in
the day.
Daily integer Montpellier (Basic ADLs,
Going out), Singapore (Basic
ADLs, Going out)
OUTDOOR_TIME Time in seconds spent outdoor in the
day.
Note: as outdoor and indoor states
are complementary (i.e. outdoor
indoor) INDOOR_TIME can be
computed by subtracting
OUTDOOR_TIME from 86.400
(seconds in a day).
Daily integer Lecce (Basic ADLs, Going
out), Montpellier (Basic
ADLs, Going out), Singapore
(Basic ADLs, Going out)
PAIN Pain level reported in the day on a 1-5
scale
Daily integer Lecce (Health – physical,
Pain)
PERCEIVED_TEMPERATURE Average measure of the thermal
comfort (thermal sensation) during
the day.
Note: This measure can be collected
by hours as minimum time gap. We
have to agree what is the more
convenient and meaningful.
Daily float Madrid (Note: to be further
researched during the
experiment)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 139/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
PHARMACY _TIME Time in seconds spent in monitored
pharmacies in the day.
Daily integer Athens (Instrumental ADLs,
Shopping; Health – physical,
Visits to health related places)
PHARMACY _VISITS_MONTH Number of visits to monitored
pharmacies in the month.
Monthly integer Athens (Instrumental ADLs,
Shopping; Health – physical,
Visits to health related
places), Lecce (Instrumental
ADLs, Shopping; Health –
physical, Visits to health
related places), Montpellier
(Instrumental ADLs,
Shopping; Health – physical,
Visits to health related places)
PHARMACY _VISITS_WEEK Number of visits to monitored
pharmacies in the week.
Weekly integer Athens (Instrumental ADLs,
Shopping; Health – physical,
Visits to health related
places), Montpellier
(Instrumental ADLs,
Shopping; Health – physical,
Visits to health related places)
PHARMACY_VISITS Number of visits to monitored
pharmacies in the day.
Daily integer Athens (Instrumental ADLs,
Shopping; Health – physical,
Visits to health related
places), Montpellier
(Instrumental ADLs,
Shopping; Health – physical,
Visits to health related places)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 140/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
PHONECALLS_LONG_PLACED_PERC Percentage of long phone calls placed
on total phone calls placed in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
PHONECALLS_LONG_RECEIVED_PERC Percentage of long phone calls
received on total phone calls received
in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
PHONECALLS_MISSED Number of phone calls missed in the
day.
Daily integer Lecce (Instrumental ADLs,
Phone usage)
PHONECALLS_PLACED Number of phone calls placed in the
day.
Daily integer Lecce (Instrumental ADLs,
Phone usage), Montpellier
(Instrumental ADLs, Phone
usage)
PHONECALLS_PLACED_PERC Percentage of phone calls placed on
total phone calls in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
PHONECALLS_RECEIVED Number of phone calls received in the
day.
Daily integer Lecce (Instrumental ADLs,
Phone usage), Montpellier
(Instrumental ADLs, Phone
usage)
PHONECALLS_RECEIVED_PERC Percentage of phone calls received on
total phone calls in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
PHONECALLS_SHORT_PLACED_PERC Percentage of short phone calls placed
on total phone calls placed in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 141/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
PHONECALLS_SHORT_RECEIVED_PERC Percentage of short phone calls
received on total phone calls received
in the day.
Daily float Lecce (Instrumental ADLs,
Phone usage)
PHYSICALACTIVITY_CALORIES Total calories in kcal burned in the
day.
Daily float Birmingham (Physical
Activity), Madrid (Physical
Activity)
PHYSICALACTIVITY_INTENSE_TIME Time in seconds spent in intense
activities in the day.
Daily integer Birmingham (Physical
Activity)
PHYSICALACTIVITY_MODERATE_TIME Time in seconds spent in moderate
activities in the day.
Daily integer Birmingham (Physical
Activity)
PHYSICALACTIVITY_NUM Number of physical activity sessions
attended in the day.
Daily integer
PHYSICALACTIVITY_SOFT_TIME Time in seconds spent in soft
activities in the day.
Daily integer Birmingham (Physical
Activity)
PUBLICPARK _TIME Time in seconds spent in PublicPark
in the day.
Daily integer Singapore (Socialization,
Attending Other Social
Places)
PUBLICPARK _VISITS_MONTH Number of visits to PublicPark in the
month.
Monthly integer Singapore (Socialization,
Attending Other Social
Places)
PUBLICPARK_VISITS Number of visits to PublicPark in the
day.
Daily integer Singapore (Socialization,
Attending Other Social
Places)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 142/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
PUBLICTRANSPORT_DISTANCE_MONTH Distance in km travelled on the bus in
a month.
Monthly float Madrid (Instrumental ADLs,
Transportation)
PUBLICTRANSPORT_RIDES_MONTH Number of times the user gets on the
bus in a month.
Monthly integer Madrid (Instrumental ADLs,
Transportation)
PUBLICTRANSPORT_TIME Time in seconds spent in public
transportation in the day.
Daily integer
RESTAURANTS_TIME Time in seconds spent in monitored
restaurants in the day.
Weekly integer Athens (Socialization,
Restaurant), Montpellier
(Socialization, Restaurant)
RESTAURANTS_VISITS_MONTH Number of visits to monitored
restaurants in the month.
Monthly integer Montpellier (Socialization,
Restaurant)
RESTAURANTS_VISITS_WEEK Number of visits to monitored
restaurants in the week.
Weekly integer Athens (Socialization,
Restaurant), Montpellier
(Socialization, Restaurant)
RESTROOM_TIME Time in seconds spent in restroom in
the day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms; Basic ADLs, Toilet
hygiene), Singapore (Motility,
Moving across rooms; Basic
ADLs, Toilet hygiene)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 143/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
RESTROOM_VISITS Number of restroom entrances in the
day.
Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms; Basic ADLs, Toilet
hygiene), Singapore (Motility,
Moving across rooms; Basic
ADLs, Toilet hygiene)
ROOM_CHANGES Number of room changes in the day. Daily integer Lecce (Motility, Moving
across rooms), Montpellier
(Motility, Moving across
rooms), Singapore (Motility,
Moving across rooms)
SENIORCENTER_LONG_VISITS Number of long visits to
SeniorCenter in the day.
Daily integer Lecce (Socialization,
Attending Senior Centers)
SENIORCENTER_TIME Time in seconds spent in
SeniorCenter in the day.
Daily integer Athens (Socialization,
Attending Senior Centers),
Lecce (Socialization,
Attending Senior Centers)
SENIORCENTER_TIME_OUT_PERC Percentage of time spent in
SeniorCenter with respect to total
time spent outside home in the day.
Daily float Lecce (Socialization,
Attending Senior Centers)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 144/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
SENIORCENTER_VISITS Number of visits to SeniorCenter in
the day.
Daily integer Athens (Socialization,
Attending Senior Centers),
Lecce (Socialization,
Attending Senior Centers),
Singapore (Socialization,
Attending Senior Centers)
SENIORCENTER_VISITS_MONTH Number of visits to SeniorCenter in
the month.
Monthly integer Singapore (Socialization,
Attending Senior Centers)
SENIORCENTER_VISITS_WEEK Number of visits to SeniorCenter in
the week.
Weekly integer Singapore (Socialization,
Attending Senior Centers)
SHOPS_OUTDOOR_TIME_PERC Percentage of time spent in monitored
shops with respect to time spent
outdoor in the day.
Daily float Athens (Instrumental ADLs,
Shopping), Lecce
(Instrumental ADLs,
Shopping)
SHOPS_TIME Time in seconds spent in monitored
shops in the day.
Daily integer Athens (Instrumental ADLs,
Shopping), Lecce
(Instrumental ADLs,
Shopping), Montpellier
(Instrumental ADLs,
Shopping), Singapore
(Instrumental ADLs,
Shopping)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 145/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
SHOPS_VISITS Number of visits to monitored shops
in the day.
Daily integer Athens (Instrumental ADLs,
Shopping), Lecce
(Instrumental ADLs,
Shopping), Montpellier
(Instrumental ADLs,
Shopping), Singapore
(Instrumental ADLs,
Shopping)
SHOPS_VISITS_WEEK Number of visits to monitored shops
in the week.
Weekly integer Montpellier (Instrumental
ADLs, Shopping), Singapore
(Instrumental ADLs,
Shopping)
SLEEP_AWAKE_TIME Total time in seconds spent awake
while at rest in the day.
Daily integer Birmingham (Health –
physical, Quality of sleep),
Madrid (Health – physical,
Quality of sleep)
SLEEP_DEEP_TIME Total time in seconds spent in deep
sleeping in the day.
Daily integer Birmingham (Health –
physical, Quality of sleep),
Madrid (Health – physical,
Quality of sleep)
SLEEP_LIGHT_TIME Total time in seconds spent in light
sleeping in the day.
Daily integer Birmingham (Health –
physical, Quality of sleep),
Madrid (Health – physical,
Quality of sleep)
SLEEP_REM_TIME Total time in seconds spent in REM
sleeping in the day.
Daily integer Madrid (Health – physical,
Quality of sleep)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 146/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
SLEEP_TIME Total time in seconds spent sleeping
in the day.
Day integer Madrid (Health – physical,
Quality of sleep),
Montpellier (Health –
physical, Quality of sleep),
Singapore (Health – physical,
Quality of sleep)
SLEEP_TOSLEEP_TIME Total time in seconds the user spent
falling asleep in the day.
Daily integer Birmingham (Health –
physical, Quality of sleep)
SLEEP_WAKEUP_NUM Number of times the user woke up in
the day.
Daily integer Birmingham (Health –
physical, Quality of sleep)
STAIRS_FLOOR_CHANGES_UP Number of floor changes performed
in the day by climbing stairs upwards.
Daily integer
Lecce (Motility, Climbing
stairs)
STILL_TIME Time in seconds spent in the still state
in the day.
Note: as still and moving states are
complementary (i.e. still
moving) MOVING_TIME can be
computed by subtracting
STILL_TIME from 86.400 (seconds
in a day).
Daily integer Lecce (Motility,
Still/Moving)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 147/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
SUPERMARKET _TIME Time in seconds spent in monitored
supermarkets in the day.
Daily integer Athens (Instrumental ADLs,
Shopping), Lecce
(Instrumental ADLs,
Shopping), Montpellier
(Instrumental ADLs,
Shopping)
SUPERMARKET _TIME_PERC Percentage of time spent in monitored
supermarkets on total time spent in
shops in the day.
Daily float Lecce (Instrumental ADLs,
Shopping)
SUPERMARKET_VISITS Number of visits to monitored
supermarkets in the day.
Daily integer Athens (Instrumental ADLs,
Shopping), Lecce
(Instrumental ADLs,
Shopping), Montpellier
(Instrumental ADLs,
Shopping)
SUPERMARKET_VISITS_WEEK Number of visits to monitored
supermarkets in the week.
Weekly integer Athens (Instrumental ADLs,
Shopping), Montpellier
(Instrumental ADLs,
Shopping)
TRANSPORT_TIME Time in seconds spent in
transportation (either public or
private) in the day.
Daily integer Athens (Instrumental ADLs,
Transportation)
TVWATCHING_TIME Time in seconds spent watching TV
in the day.
Daily integer Lecce (Cultural Engagement,
Watching TV)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 148/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
TVWATCHING_TIME_PERC Percentage of time spent watching
TV with respect to time spent inside
home in the day.
Daily float Lecce (Cultural Engagement,
Watching TV)
VISITORS_WEEK Number of visitors met in the week. Weekly integer Lecce (Socialization, Visits),
Singapore (Socialization,
Visits)
VISITS_PAYED_WEEK Number of visits payed in the week. Weekly integer Lecce (Socialization, Visits),
Montpellier (Socialization,
Visits)
VISITS_RECEIVED_WEEK Number of visits received in the
week.
Weekly integer Lecce (Socialization, Visits),
Montpellier (Socialization,
Visits), Singapore
(Socialization, Visits),
WALK_DISTANCE Total distance in meters walked in the
day.
Daily integer Birmingham (Motility,
Walking), Madrid (Motility,
Walking)
WALK_DISTANCE_OUTDOOR Total distance in meters walked
outdoor in the day.
Daily integer Athens (Motility, Walking),
Lecce (Motility, Walking),
Montpellier (Motility,
Walking), Singapore
(Motility, Walking)
WALK_DISTANCE_OUTDOOR_FAST Distance in meters walked outdoor at
fast speed in the day.
Daily float Lecce (Motility, Walking)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 149/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
WALK_DISTANCE_OUTDOOR_FAST_PERC Percentage of distance walked
outdoor at fast speed in the day on
total distance walked outdoor.
Daily float Athens (Motility, Walking),
Lecce (Motility, Walking)
WALK_DISTANCE_OUTDOOR_SLOW_PERC Percentage of distance walked
outdoor at slow speed in the day on
total distance walked outdoor.
Daily float Athens (Motility, Walking),
Lecce (Motility, Walking)
WALK_SPEED_OUTDOOR Average outdoor walking speed in
meters/seconds in the day.
Daily float Athens (Motility, Walking),
Lecce (Motility, Walking),
Madrid (Motility, Walking)
WALK_SPEED_OUTDOOR_FAST Average outdoor walking speed at
fast speed in meters/seconds in the
day.
Daily float Lecce (Motility, Walking)
WALK_STEPS Number of steps done in the day. Daily integer Birmingham (Motility,
Walking)
WALK_STEPS_OUTDOOR Number of steps done outdoor in the
day.
Daily integer
Montpellier (Motility,
Walking), Singapore
(Motility, Walking)
WALK_TIME_OUTDOOR Time in seconds spent walking
outdoor in the day.
Daily integer Athens (Motility, Walking),
Lecce (Motility, Walking),
Madrid (Motility, Walking),
Montpellier (Motility,
Walking), Singapore
(Motility, Walking)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 150/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
WALK_TIME_OUTDOOR_FAST Time in seconds spent walking at fast
speed outdoor in the day.
Daily integer Lecce (Motility, Walking)
WASHINGMACHINE_SESSIONS Washing machine usage sessions in
the day.
Daily integer Lecce (Instrumental ADLs,
Laundry)
WEAKNESS <yet to be defined; following
definitions are currently considered>
Grip strength in kg sampled in the
month.
Or
The person experiences problems in
daily life due to lack of strength in
hands.
Monthly float
or
boolean
Madrid (Health – physical,
Weakness)
WEIGHT User weight in kg sampled in the
month.
Monthly float Madrid (Health – physical,
Weight)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 151/183
Payload property name Meaning Reporting
frequency
Payload
property type
Needed by (for what)
WORD_PAIRS_TEST The result of the word pair test (from
the “Short Test of Mental Status”
scale) sampled in the month
(replicated below for reader’s
convenience).
Word Pair Test: The examiner
proposes three pair of words and asks
the person to state the abstract
interpretation: orange/banana,
dog/horse, table/bookcase (e.g.
dog/horse = animal).
Monthly float Madrid (Health – cognitive,
Abstraction)
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 152/183
12 Annex: Correlation matrices for Pilots measures
The correlation values reported in the following subsections have been assessed in June 2018, based on the status of the City4Age Shared Repository at that time.
12.1 Athens Pilot
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
1 1 0,6 0 -0,01 0,04 0,08 -0,01 0 0 0 0 -0,01 -0,02 -0,01 -0,01 0,04 -0,01 0,04 0,03 0,05 0,05 -0,05 0,03 0,01
2 0,6 1 0 -0,05 0,02 0,06 -0,02 0,01 0 -0,01 -0,01 -0,01 -0,01 -0,01 -0,01 0,06 -0,01 0,06 0,04 0,12 0,07 -0,07 0,04 0,06
3 0 0 1 -0,02 0,04 0,04 0 0,03 0,2 0 0 0,02 0,01 -0,01 -0,01 0,01 -0,01 0,01 0 0,01 0 0 -0,01 0,02
4 -0,01 -0,05 -0,02 1 -0,09 -0,1 -0,24 -0,31 -0,04 -0,04 0,01 -0,23 -0,24 0,04 -0,18 -0,25 -0,18 -0,25 -0,13 -0,37 -0,06 0,06 -0,01 -0,84
5 0,04 0,02 0,04 -0,09 1 0,77 0,03 0,3 0,06 -0,02 -0,02 -0,04 -0,01 -0,03 -0,02 0,09 -0,02 0,09 0 0,06 0 0 0,07 0,04
6 0,08 0,06 0,04 -0,1 0,77 1 0,05 0,47 0,06 -0,02 -0,03 -0,03 0 -0,02 -0,01 0,18 -0,01 0,18 0,01 0,12 0,01 -0,01 0,07 0,08
7 -0,01 -0,02 0 -0,24 0,03 0,05 1 0,63 0,06 -0,01 0 0,18 0,26 0,07 0,27 0,18 0,27 0,18 -0,02 0,09 -0,03 0,03 0 0,09
8 0 0,01 0,03 -0,31 0,3 0,47 0,63 1 0,07 -0,01 0 0,14 0,26 0,06 0,19 0,38 0,19 0,38 0,01 0,22 -0,01 0,01 0,05 0,21
9 0 0 0,2 -0,04 0,06 0,06 0,06 0,07 1 0 0 0,01 0 0 0 0,03 0 0,03 0 0,01 -0,02 0,02 0 0,04
10 0 -0,01 0 -0,04 -0,02 -0,02 -0,01 -0,01 0 1 0,17 0,03 0,07 -0,01 -0,01 0,03 -0,01 0,03 0,02 0,03 0,02 -0,02 0,02 0,02
11 0 -0,01 0 0,01 -0,02 -0,03 0 0 0 0,17 1 0,01 0,06 0,02 0 0,01 0 0,01 0,02 0,01 0,02 -0,02 0,02 0
12 -0,01 -0,01 0,02 -0,23 -0,04 -0,03 0,18 0,14 0,01 0,03 0,01 1 0,68 -0,01 0,09 0,07 0,09 0,07 -0,03 0,01 -0,02 0,02 0,05 0
13 -0,02 -0,01 0,01 -0,24 -0,01 0 0,26 0,26 0 0,07 0,06 0,68 1 0,03 0,16 0,19 0,16 0,19 -0,04 0,06 -0,03 0,03 0,07 0,08
14 -0,01 -0,01 -0,01 0,04 -0,03 -0,02 0,07 0,06 0 -0,01 0,02 -0,01 0,03 1 0,75 0,41 0,75 0,41 -0,03 -0,04 0 0 0,03 -0,11
15 -0,01 -0,01 -0,01 -0,18 -0,02 -0,01 0,27 0,19 0 -0,01 0 0,09 0,16 0,75 1 0,55 1 0,55 -0,02 0,05 -0,02 0,02 0,01 0,04
16 0,04 0,06 0,01 -0,25 0,09 0,18 0,18 0,38 0,03 0,03 0,01 0,07 0,19 0,41 0,55 1 0,55 1 0,01 0,27 0,02 -0,02 0,07 0,2
17 -0,01 -0,01 -0,01 -0,18 -0,02 -0,01 0,27 0,19 0 -0,01 0 0,09 0,16 0,75 1 0,55 1 0,55 -0,02 0,05 -0,02 0,02 0,01 0,04
18 0,04 0,06 0,01 -0,25 0,09 0,18 0,18 0,38 0,03 0,03 0,01 0,07 0,19 0,41 0,55 1 0,55 1 0,01 0,27 0,02 -0,02 0,07 0,2
19 0,03 0,04 0 -0,13 0 0,01 -0,02 0,01 0 0,02 0,02 -0,03 -0,04 -0,03 -0,02 0,01 -0,02 0,01 1 0,31 0,61 -0,61 0,33 0,15
20 0,05 0,12 0,01 -0,37 0,06 0,12 0,09 0,22 0,01 0,03 0,01 0,01 0,06 -0,04 0,05 0,27 0,05 0,27 0,31 1 0,25 -0,25 0,18 0,39
21 0,05 0,07 0 -0,06 0 0,01 -0,03 -0,01 -0,02 0,02 0,02 -0,02 -0,03 0 -0,02 0,02 -0,02 0,02 0,61 0,25 1 -1 0,63 0,11
22 -0,05 -0,07 0 0,06 0 -0,01 0,03 0,01 0,02 -0,02 -0,02 0,02 0,03 0 0,02 -0,02 0,02 -0,02 -0,61 -0,25 -1 1 -0,63 -0,11
23 0,03 0,04 -0,01 -0,01 0,07 0,07 0 0,05 0 0,02 0,02 0,05 0,07 0,03 0,01 0,07 0,01 0,07 0,33 0,18 0,63 -0,63 1 -0,03
24 0,01 0,06 0,02 -0,84 0,04 0,08 0,09 0,21 0,04 0,02 0 0 0,08 -0,11 0,04 0,2 0,04 0,2 0,15 0,39 0,11 -0,11 -0,03 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 153/183
Legenda
1 cinema_time
2 cinema_visits
3 cinema_visits_month
4 home_time
5 othersocial_time
6 othersocial_visits
7 pharmacy_time
8 pharmacy_visits
9 pharmacy_visits_month
10 restaurants_time
11 restaurants_visits_week
12 seniorcenter_time
13 seniorcenter_visits
14 shops_outdoor_time_perc
15 shops_time
16 shops_visits
17 supermarket_time
18 supermarket_visits
19 transport_time
20 walk_distance_outdoor
21 walk_distance_outdoor_fast_perc
22 walk_distance_outdoor_slow_perc
23 walk_speed_outdoor
24 walk_time_outdoor
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 154/183
12.2 Birmingham Pilot
1 2 3 4 5 6 7 8 9 10 11
1 1 0,12 0,31 0,34 0,07 -0,05 0,05 0 0,13 0,51 0,44
2 0,12 1 0,14 0,07 -0,02 -0,03 0,08 0,01 -0,02 0,2 0,12
3 0,31 0,14 1 0,27 -0,07 0 0,03 0 -0,12 0,82 0,75
4 0,34 0,07 0,27 1 -0,16 -0,14 -0,05 -0,02 -0,2 0,56 0,61
5 0,07 -0,02 -0,07 -0,16 1 0,25 0,13 0,29 0,76 -0,08 -0,09
6 -0,05 -0,03 0 -0,14 0,25 1 -0,09 0,03 0,24 -0,06 -0,02
7 0,05 0,08 0,03 -0,05 0,13 -0,09 1 0,06 0,19 0,05 0
8 0 0,01 0 -0,02 0,29 0,03 0,06 1 0,04 0,01 0,01
9 0,13 -0,02 -0,12 -0,2 0,76 0,24 0,19 0,04 1 -0,09 -0,09
10 0,51 0,2 0,82 0,56 -0,08 -0,06 0,05 0,01 -0,09 1 0,96
11 0,44 0,12 0,75 0,61 -0,09 -0,02 0 0,01 -0,09 0,96 1
Legenda
1 physicalactivity_calories
2 physicalactivity_intense_time
3 physicalactivity_moderate_time
4 physicalactivity_soft_time
5 sleep_awake_time
6 sleep_deep_time
7 sleep_light_time
8 sleep_tosleep_time
9 sleep_wakeup_num
10 walk_distance
11 walk_steps
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 155/183
12.3 Lecce Pilot
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
1 1 0,29 -0,02 0,32 0 0 0,37 0,41 0,25 0,13 -0,05 0,02 0,54 0 0 0 0 0,46 0,15 0,09 -0,02 0,01
2 0,29 1 -0,08 0,17 0 0 0,23 0,28 0,21 0,19 -0,09 0,04 0,22 0 0 0 0 0,28 0,12 -0,03 -0,03 -0,1
3 -0,02 -0,08 1 0,39 0 0 0,23 0,06 0,29 0,37 0,31 0 0,52 0 0 0,01 0 0,16 0,39 0,36 -0,02 0,09
4 0,32 0,17 0,39 1 0 0 0,66 0,69 0,52 0,59 0,4 0,03 0,6 0 0 0,01 0 0,57 0,42 0,54 -0,04 0,27
5 0 0 0 0 1 0,34 0 0 0 0 0 0,41 0,01 0,27 0,05 0,36 0,05 0,01 0,01 0 -0,11 0
6 0 0 0 0 0,34 1 0 0 0 0 0 0,16 0 0,12 0,12 0,21 0,28 0,01 0 0 -0,03 0
7 0,37 0,23 0,23 0,66 0 0 1 0,72 0,71 0,55 0,25 0,01 0,65 0 0 0,02 0 0,36 0,52 0,59 -0,01 0,28
8 0,41 0,28 0,06 0,69 0 0 0,72 1 0,61 0,54 0,2 0,02 0,59 0 0 0,01 0 0,51 0,48 0,46 -0,04 0,34
9 0,25 0,21 0,29 0,52 0 0 0,71 0,61 1 0,52 0,17 0,02 0,56 0 0 0,01 0 0,4 0,47 0,49 0,01 0,38
10 0,13 0,19 0,37 0,59 0 0 0,55 0,54 0,52 1 0,44 0,04 0,47 0 0 0,01 0 0,48 0,34 0,49 -0,02 0,3
11 -0,05 -0,09 0,31 0,4 0 0 0,25 0,2 0,17 0,44 1 0,01 0,26 0 0 -0,01 0 0,18 0,16 0,38 -0,02 0,18
12 0,02 0,04 0 0,03 0,41 0,16 0,01 0,02 0,02 0,04 0,01 1 0,01 0,56 0,09 0,53 0,09 0,02 0,01 0 -0,36 0,03
13 0,54 0,22 0,52 0,6 0,01 0 0,65 0,59 0,56 0,47 0,26 0,01 1 0,01 0 0,02 0 0,53 0,47 0,43 -0,04 0,2
14 0 0 0 0 0,27 0,12 0 0 0 0 0 0,56 0,01 1 0,22 0,32 0,17 0 0,01 0 -0,16 0
15 0 0 0 0 0,05 0,12 0 0 0 0 0 0,09 0 0,22 1 0,18 0,96 0 0 0 -0,02 0
16 0 0 0,01 0,01 0,36 0,21 0,02 0,01 0,01 0,01 -0,01 0,53 0,02 0,32 0,18 1 0,22 0,01 0,02 0 -0,14 0
17 0 0 0 0 0,05 0,28 0 0 0 0 0 0,09 0 0,17 0,96 0,22 1 0 0 0 -0,02 0
18 0,46 0,28 0,16 0,57 0,01 0,01 0,36 0,51 0,4 0,48 0,18 0,02 0,53 0 0 0,01 0 1 0,17 0,31 -0,02 0,21
19 0,15 0,12 0,39 0,42 0,01 0 0,52 0,48 0,47 0,34 0,16 0,01 0,47 0,01 0 0,02 0 0,17 1 0,42 -0,03 0,03
20 0,09 -0,03 0,36 0,54 0 0 0,59 0,46 0,49 0,49 0,38 0 0,43 0 0 0 0 0,31 0,42 1 0,03 0,38
21 -0,02 -0,03 -0,02 -0,04 -0,11 -0,03 -0,01 -0,04 0,01 -0,02 -0,02 -0,36 -0,04 -0,16 -0,02 -0,14 -0,02 -0,02 -0,03 0,03 1 -0,02
22 0,01 -0,1 0,09 0,27 0 0 0,28 0,34 0,38 0,3 0,18 0,03 0,2 0 0 0 0 0,21 0,03 0,38 -0,02 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 156/183
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
1 0,37 0 0 0 0,01 0 0 0 0,01 0 0,01 0,16 0 -0,23 0,01 0,01 0 0 0 0,01 0 0,01
2 0,31 0 0 0 0,02 0 0,01 -0,01 -0,02 0,01 -0,01 0,14 -0,01 0,01 0,01 0,01 0 0 0 0,03 0,01 0,02
3 0,06 0 0 0 -0,01 -0,01 -0,01 0 0,01 -0,02 0 0,26 0,1 -0,19 0,01 0,01 0 0 0 -0,01 0 0,01
4 0,38 0 0 0 0,01 0 0 0 0,01 0 0 0,44 0,26 0,06 0,01 0,01 0 0 0 0,03 0,01 0,01
5 0 0 0,01 -0,01 0,05 0,33 0,22 0,31 0,13 0,25 0,24 0,01 0 0 0,18 0,22 0,1 0,05 0,16 0,17 0,1 0,15
6 0 0 0 0 0,03 0,13 0,08 0,13 0,05 0,09 0,09 0 0 0 0,08 0,12 0,32 0,01 0,05 0,06 0,06 0,06
7 0,52 0 0 0 0 0 0 0 0,01 -0,01 0,01 0,48 0,23 0,14 0,01 0,01 0 0 0 -0,01 0 0,01
8 0,82 0 0 0 0,01 0 0 0 0,01 0 0 0,4 0,31 0,28 0,01 0,01 0 0 0 0,02 0,01 0,01
9 0,41 0 0 0 -0,01 0 -0,01 0 0,01 -0,01 0,01 0,61 -0,02 0,16 0,02 0,02 0 0 0 0 0 0,01
10 0,49 0 0 0 0,01 0 0 0 0,01 0,01 0,01 0,4 0,2 0,26 0,01 0,01 0 0 0 0,04 0,01 0,01
11 0,06 0 0 0 0,01 0,01 0,01 0,01 0,01 0,01 0,01 0,21 0,17 0,1 0 0 0 0 0 0,02 0,01 -0,01
12 0,02 0 0,01 -0,01 0,19 0,4 0,35 0,41 0,26 0,41 0,38 0 -0,01 -0,02 0,28 0,33 0,12 0,17 0,42 0,67 0,41 0,57
13 0,48 0 0 0 0,01 0 0 0,01 0,02 0 0,01 0,34 0,07 -0,14 0 0 0 0 0 -0,01 0 0,02
14 0 -0,01 0 -0,01 0,15 0,39 0,28 0,4 0,22 0,34 0,35 0 0 -0,01 0,32 0,45 0,22 0,05 0,23 0,24 0,17 0,2
15 0 -0,01 0 0 0,03 0,07 0,05 0,07 0,04 0,06 0,06 0 0 0 0,03 0,06 0,97 0,01 0,06 0,04 0,25 0,04
16 0,01 0 -0,01 -0,01 0,12 0,24 0,19 0,27 0,16 0,23 0,25 0,01 -0,02 -0,01 0,12 0,23 0,23 0,04 0,18 0,18 0,15 0,27
17 0 0 0 0 0,02 0,05 0,04 0,07 0,04 0,05 0,06 0 0 0 0,02 0,05 0,99 0,01 0,06 0,04 0,25 0,05
18 0,38 0 0 0 0,01 0 0 0,01 0,02 0 0,01 0,56 0,03 0,02 -0,01 0 0 0 0 0,01 0 0,02
19 0,33 0 0 0 -0,01 0 -0,01 0,01 0,02 -0,01 0,01 0,25 0,17 0,17 0,02 0,01 0 0 0 -0,01 0 0,02
20 0,25 0 0 0 0 0 0 0 0 0 0 0,29 0,19 0,23 0,02 0,02 0 0 0 -0,01 0 0
21 -0,03 0 -0,05 -0,03 -0,16 -0,22 -0,2 -0,17 -0,08 -0,23 -0,16 -0,01 -0,03 -0,01 -0,04 -0,01 -0,02 -0,02 -0,18 -0,29 -0,18 -0,25
22 0,31 0 0 0 0,02 0 0,01 0 0 0,01 0 0,24 0,26 0,31 0,02 0,01 0 0 0 0,03 0,01 0
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 157/183
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
1 0 0,02 0 0,01 0 0,02 0,56 0 0 0,23 0,41 0,32 0,02 0,02 -0,01 0,02 -0,01 0 0,01 0,02 0 0,16 0,04
2 0,01 0,04 0,01 0,01 0 0,05 0,3 0 0,02 0,46 -0,04 0,36 0,01 0,02 0,04 0,04 0,06 0,06 0 0,02 0 0,09 0,04
3 0 0,01 0 0 0 0,02 -0,1 -0,01 0,01 0,04 0,36 0,26 0 0 -0,01 -0,02 -0,02 -0,02 0,01 0 0 0,31 0,27
4 0 0,01 0,01 0 0 0,02 0,54 -0,01 0,01 0,34 0,62 0,54 0,01 0,01 0,04 0,04 0,04 0,05 0,01 0,01 0 0,28 0
5 0,11 0,12 0,22 0,01 -0,03 -0,02 0 0,22 0,1 0 0 0,01 0,18 0,17 0,17 0,17 0,17 0,2 0,15 0,17 -0,01 0,01 0
6 0,06 0,08 0,13 0,01 0 0,01 0 0,1 0,05 0 0 0 0,06 0,06 0,07 0,07 0,08 0,09 0,05 0,06 0,01 0 0
7 0,01 0,02 0 0,01 0 0,03 0,51 -0,01 0,02 0,3 0,38 0,36 0,01 0,01 0,04 0,05 0,04 0,05 0,01 0,01 0 0,42 -0,01
8 0 0,01 0,01 0 0 0,01 0,85 -0,01 0,01 0,33 0,4 0,52 0,01 0,01 0,05 0,04 0,05 0,05 0,01 0,01 0 0,17 -0,11
9 0,01 0,02 0 0,01 0 0,03 0,49 0 0,01 0,3 0,35 0,35 0,01 0,01 0,04 0,05 0,04 0,04 0,01 0,01 0 0,62 -0,05
10 0,01 0,02 0,01 0,01 0 0,03 0,39 -0,01 0,01 0,2 0,41 0,4 0,01 0,01 0,06 0,04 0,05 0,05 0,02 0,01 0 0,3 0,18
11 0 -0,02 0,01 -0,01 0 -0,03 0,03 -0,01 -0,02 0,16 0,21 0,21 0,01 0,01 0,02 0,01 0,01 0,01 0,01 0,01 0 0,08 0,07
12 0,41 0,48 0,38 0,29 0,01 0,29 0,01 0,42 0,2 0,03 0,02 0,02 0,14 0,15 0,4 0,35 0,43 0,45 0,03 0,15 0 0,02 0,02
13 0,01 0,03 0 0,01 0 0,04 0,43 -0,01 0,02 0,15 0,49 0,38 0,02 0,02 0,02 0,03 0,02 0,02 0,02 0,02 0 0,26 0,02
14 0,16 0,2 0,39 0,16 0,05 0,16 0 0,55 0,3 0 0,01 0,01 0,04 0,04 0,17 0,17 0,2 0,21 0 0,04 0,04 0,01 0
15 0,1 0,28 0,19 0,12 0,25 0,33 0 0,08 0,05 0 0 0 0,01 0,01 0,04 0,03 0,04 0,04 0 0,01 0,41 0 0
16 0,28 0,33 0,27 0,23 0,1 0,27 0,01 0,4 0,25 -0,01 0,01 0,01 0,09 0,09 0,19 0,18 0,19 0,22 0,05 0,09 0,03 0,02 0,01
17 0,11 0,29 0,19 0,12 0,24 0,34 0 0,07 0,04 0 0 0 0,02 0,02 0,04 0,03 0,04 0,04 0 0,02 0,41 0 0
18 0,01 0,03 0 0,01 0 0,05 0,54 -0,01 0 0,16 0,37 0,44 0,02 0,02 0,03 0,02 0,02 0,02 0,02 0,02 0 0,27 0,07
19 0,01 0,03 0 0,01 0 0,04 0,4 0 0,02 0,19 0,25 0,44 0,02 0,02 0,01 0 0,01 0,01 0,02 0,02 0 0,18 -0,01
20 0 0 0,01 0 0 0 0,3 -0,01 0 0,29 0,18 0,24 0 0 0,02 0,02 0,02 0,02 0 0 0 0,38 0,09
21 -0,18 -0,21 -0,09 -0,12 0 -0,13 -0,02 -0,1 0,05 -0,02 -0,04 -0,03 0,01 0 -0,02 -0,04 -0,05 -0,06 0,08 0 0 0,06 0,01
22 0 0,01 0,01 0 0 0,01 0,22 -0,01 -0,01 0,05 0,28 0,15 0,01 0,01 0,04 0 0,04 0,03 0,01 0,01 0 0,22 -0,13
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 158/183
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
23 0,37 0,31 0,06 0,38 0 0 0,52 0,82 0,41 0,49 0,06 0,02 0,48 0 0 0,01 0 0,38 0,33 0,25 -0,03 0,31
24 0 0 0 0 0 0 0 0 0 0 0 0 0 -0,01 -0,01 0 0 0 0 0 0 0
25 0 0 0 0 0,01 0 0 0 0 0 0 0,01 0 0 0 -0,01 0 0 0 0 -0,05 0
26 0 0 0 0 -0,01 0 0 0 0 0 0 -0,01 0 -0,01 0 -0,01 0 0 0 0 -0,03 0
27 0,01 0,02 -0,01 0,01 0,05 0,03 0 0,01 -0,01 0,01 0,01 0,19 0,01 0,15 0,03 0,12 0,02 0,01 -0,01 0 -0,16 0,02
28 0 0 -0,01 0 0,33 0,13 0 0 0 0 0,01 0,4 0 0,39 0,07 0,24 0,05 0 0 0 -0,22 0
29 0 0,01 -0,01 0 0,22 0,08 0 0 -0,01 0 0,01 0,35 0 0,28 0,05 0,19 0,04 0 -0,01 0 -0,2 0,01
30 0 -0,01 0 0 0,31 0,13 0 0 0 0 0,01 0,41 0,01 0,4 0,07 0,27 0,07 0,01 0,01 0 -0,17 0
31 0,01 -0,02 0,01 0,01 0,13 0,05 0,01 0,01 0,01 0,01 0,01 0,26 0,02 0,22 0,04 0,16 0,04 0,02 0,02 0 -0,08 0
32 0 0,01 -0,02 0 0,25 0,09 -0,01 0 -0,01 0,01 0,01 0,41 0 0,34 0,06 0,23 0,05 0 -0,01 0 -0,23 0,01
33 0,01 -0,01 0 0 0,24 0,09 0,01 0 0,01 0,01 0,01 0,38 0,01 0,35 0,06 0,25 0,06 0,01 0,01 0 -0,16 0
34 0,16 0,14 0,26 0,44 0,01 0 0,48 0,4 0,61 0,4 0,21 0 0,34 0 0 0,01 0 0,56 0,25 0,29 -0,01 0,24
35 0 -0,01 0,1 0,26 0 0 0,23 0,31 -0,02 0,2 0,17 -0,01 0,07 0 0 -0,02 0 0,03 0,17 0,19 -0,03 0,26
36 -0,23 0,01 -0,19 0,06 0 0 0,14 0,28 0,16 0,26 0,1 -0,02 -0,14 -0,01 0 -0,01 0 0,02 0,17 0,23 -0,01 0,31
37 0,01 0,01 0,01 0,01 0,18 0,08 0,01 0,01 0,02 0,01 0 0,28 0 0,32 0,03 0,12 0,02 -0,01 0,02 0,02 -0,04 0,02
38 0,01 0,01 0,01 0,01 0,22 0,12 0,01 0,01 0,02 0,01 0 0,33 0 0,45 0,06 0,23 0,05 0 0,01 0,02 -0,01 0,01
39 0 0 0 0 0,1 0,32 0 0 0 0 0 0,12 0 0,22 0,97 0,23 0,99 0 0 0 -0,02 0
40 0 0 0 0 0,05 0,01 0 0 0 0 0 0,17 0 0,05 0,01 0,04 0,01 0 0 0 -0,02 0
41 0 0 0 0 0,16 0,05 0 0 0 0 0 0,42 0 0,23 0,06 0,18 0,06 0 0 0 -0,18 0
42 0,01 0,03 -0,01 0,03 0,17 0,06 -0,01 0,02 0 0,04 0,02 0,67 -0,01 0,24 0,04 0,18 0,04 0,01 -0,01 -0,01 -0,29 0,03
43 0 0,01 0 0,01 0,1 0,06 0 0,01 0 0,01 0,01 0,41 0 0,17 0,25 0,15 0,25 0 0 0 -0,18 0,01
44 0,01 0,02 0,01 0,01 0,15 0,06 0,01 0,01 0,01 0,01 -0,01 0,57 0,02 0,2 0,04 0,27 0,05 0,02 0,02 0 -0,25 0
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 159/183
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
23 1 0 0 0 0,01 0 0 0 0 0 0 0,17 0,26 0,18 0 0 0 0 0 0,02 0,01 0,01
24 0 1 0 0 0 0 0 0 -0,01 0 -0,01 0 0 0 0 0 -0,01 0 0 0 -0,02 0
25 0 0 1 0 0 0,05 0,09 0,02 0 0,01 0,02 0 0 0 -0,01 0 0 0,08 0,08 0,03 0,02 0
26 0 0 0 1 -0,01 0,11 0,05 0,03 0,01 0,05 -0,01 0 0 0 -0,01 -0,01 0 0 0 -0,01 0 -0,01
27 0,01 0 0 -0,01 1 0,25 0,09 0,17 0,04 0,22 0,17 0 0 0 0,05 0,08 0,03 -0,03 0,05 0,1 0,05 0,09
28 0 0 0,05 0,11 0,25 1 0,68 0,58 0,17 0,74 0,52 0 0 0 0,19 0,22 0,08 -0,02 0,13 0,16 0,11 0,24
29 0 0 0,09 0,05 0,09 0,68 1 0,24 0,07 0,95 0,25 -0,01 0 0 0,16 0,16 0,06 0 0,13 0,18 0,15 0,21
30 0 0 0,02 0,03 0,17 0,58 0,24 1 0,71 0,44 0,87 0,01 -0,01 -0,01 0,21 0,28 0,09 0,01 0,13 0,15 0,1 0,25
31 0 -0,01 0 0,01 0,04 0,17 0,07 0,71 1 0,18 0,85 0,02 -0,02 -0,02 0,1 0,15 0,05 0,04 0,11 0,13 0,12 0,13
32 0 0 0,01 0,05 0,22 0,74 0,95 0,44 0,18 1 0,43 -0,01 0,01 0 0,19 0,21 0,07 0 0,15 0,2 0,17 0,25
33 0 -0,01 0,02 -0,01 0,17 0,52 0,25 0,87 0,85 0,43 1 0,01 -0,01 -0,01 0,16 0,23 0,08 0,04 0,15 0,17 0,13 0,23
34 0,17 0 0 0 0 0 -0,01 0,01 0,02 -0,01 0,01 1 0,15 0,18 0,03 0,02 0 0 0 -0,02 -0,01 0,02
35 0,26 0 0 0 0 0 0 -0,01 -0,02 0,01 -0,01 0,15 1 0,55 0,02 0,01 0 0 0 0 0 -0,02
36 0,18 0 0 0 0 0 0 -0,01 -0,02 0 -0,01 0,18 0,55 1 0,02 0,01 0 0 0 -0,01 0 -0,01
37 0 0 -0,01 -0,01 0,05 0,19 0,16 0,21 0,1 0,19 0,16 0,03 0,02 0,02 1 0,68 0,06 -0,01 0,06 0,11 0,04 0,05
38 0 0 0 -0,01 0,08 0,22 0,16 0,28 0,15 0,21 0,23 0,02 0,01 0,01 0,68 1 0,1 -0,05 0,08 0,09 0,03 0,1
39 0 -0,01 0 0 0,03 0,08 0,06 0,09 0,05 0,07 0,08 0 0 0 0,06 0,1 1 0,01 0,07 0,05 0,25 0,05
40 0 0 0,08 0 -0,03 -0,02 0 0,01 0,04 0 0,04 0 0 0 -0,01 -0,05 0,01 1 0,7 0,4 0,21 -0,07
41 0 0 0,08 0 0,05 0,13 0,13 0,13 0,11 0,15 0,15 0 0 0 0,06 0,08 0,07 0,7 1 0,67 0,53 0
42 0,02 0 0,03 -0,01 0,1 0,16 0,18 0,15 0,13 0,2 0,17 -0,02 0 -0,01 0,11 0,09 0,05 0,4 0,67 1 0,59 0,01
43 0,01 -0,02 0,02 0 0,05 0,11 0,15 0,1 0,12 0,17 0,13 -0,01 0 0 0,04 0,03 0,25 0,21 0,53 0,59 1 0,04
44 0,01 0 0 -0,01 0,09 0,24 0,21 0,25 0,13 0,25 0,23 0,02 -0,02 -0,01 0,05 0,1 0,05 -0,07 0 0,01 0,04 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 160/183
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
23 0 0,01 0 0 0 0,01 0,77 0 0,01 0,24 0,3 0,39 0,01 0,01 0,04 0,02 0,03 0,03 0,01 0,01 0 0,06 -0,1
24 -0,01 -0,01 -0,01 -0,01 -0,03 -0,01 0 0 0 0 0 0 0 0 0,01 0 0 0 0 0 -0,05 0 0
25 0,01 0 -0,01 0 0 -0,01 0 -0,01 -0,01 0 0 0 0,01 0,01 0 0,02 0,01 0,01 0 0,01 0 0 0
26 0 0 -0,01 0 0 0 0 -0,01 -0,01 0 0 0 0 0 0,03 0,02 0,03 0,03 0 0 0 0 0
27 0,16 0,11 0,08 0,09 -0,01 0,07 0 0,13 0,09 0,02 0 0,01 0,02 0,02 0,13 0,13 0,14 0,14 -0,02 0,02 0 -0,01 0,01
28 0,19 0,21 0,31 0,13 0,03 0,14 0 0,29 0,19 0 0 0 0,08 0,08 0,25 0,21 0,24 0,27 0 0,08 0 0 0
29 0,14 0,21 0,32 0,11 0,13 0,16 0 0,2 0,17 0 0 0 0,09 0,1 0,32 0,19 0,28 0,28 0,01 0,1 0 -0,01 0
30 0,19 0,21 0,28 0,14 0,02 0,15 0 0,29 0,16 0 0,01 0,01 0,1 0,1 0,23 0,16 0,2 0,24 0,01 0,1 0 0 0
31 0,1 0,13 0,2 0,09 0,07 0,1 0 0,16 0,11 -0,01 0,02 0,02 0,11 0,11 0,22 0,11 0,17 0,18 0,03 0,11 0,01 0,01 0,01
32 0,18 0,24 0,35 0,14 0,12 0,18 0 0,24 0,19 0,01 0 0 0,1 0,11 0,35 0,21 0,31 0,32 0,01 0,11 0 -0,01 0
33 0,17 0,2 0,29 0,13 0,05 0,15 0 0,26 0,16 -0,01 0,01 0,01 0,12 0,12 0,26 0,17 0,23 0,25 0,02 0,12 0 0,01 0
34 0,01 0,03 0 0,01 0 0,04 0,26 -0,01 -0,02 0,36 0,37 0,39 0,01 0,01 0,01 0,02 0 0,01 0,02 0,01 0 0,38 0,24
35 -0,01 -0,03 0,01 -0,01 0 -0,04 0,2 -0,01 -0,01 0,2 0,31 0,33 0,02 0,02 0,01 -0,04 0 0 0,03 0,02 0 -0,06 0,2
36 -0,01 -0,02 0,01 -0,01 0 -0,03 0,27 -0,01 -0,02 0,18 -0,14 0,03 0,01 0,01 0,01 -0,03 0 0 0,02 0,01 0 -0,01 -0,08
37 0,07 0,05 0,2 0,04 -0,02 0,03 0,01 0,14 0,1 0,02 0,02 0,03 0,03 0,03 0,14 0,13 0,16 0,16 -0,02 0,03 -0,01 0,02 0,01
38 0,1 0,1 0,27 0,11 0 0,09 0,01 0,27 0,17 0,01 0,01 0,02 0,05 0,06 0,19 0,15 0,19 0,2 0,01 0,06 -0,02 0,02 0,01
39 0,11 0,29 0,21 0,12 0,24 0,33 0 0,1 0,06 0 0 0 0,02 0,02 0,05 0,05 0,05 0,06 0,01 0,02 0,4 0 0
40 0,02 0,01 0,03 0,07 0 0,04 0 -0,02 -0,03 0 0 0 0 0 0,09 0,09 0,1 0,1 0 0 0 0 0
41 0,22 0,2 0,19 0,25 0,03 0,19 0 0,1 0,02 0 0 0 0,02 0,02 0,19 0,17 0,2 0,21 -0,02 0,02 0 0 0
42 0,15 0,17 0,19 0,15 0 0,13 0,01 0,15 0,03 0,03 0,01 0,02 0,04 0,04 0,28 0,25 0,31 0,32 -0,02 0,04 0 0,01 0,02
43 0,21 0,31 0,2 0,22 0,19 0,3 0 0,1 0,02 0,01 0 0,01 0,03 0,03 0,24 0,14 0,21 0,2 -0,02 0,03 0,08 0 0,01
44 0,56 0,66 0,24 0,35 0,01 0,37 0 0,18 0,04 0,02 0,02 0 0,14 0,14 0,24 0,21 0,25 0,27 0,01 0,14 0 0,01 0,01
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 161/183
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
45 0 0,01 0 0 0,11 0,06 0,01 0 0,01 0,01 0 0,41 0,01 0,16 0,1 0,28 0,11 0,01 0,01 0 -0,18 0
46 0,02 0,04 0,01 0,01 0,12 0,08 0,02 0,01 0,02 0,02 -0,02 0,48 0,03 0,2 0,28 0,33 0,29 0,03 0,03 0 -0,21 0,01
47 0 0,01 0 0,01 0,22 0,13 0 0,01 0 0,01 0,01 0,38 0 0,39 0,19 0,27 0,19 0 0 0,01 -0,09 0,01
48 0,01 0,01 0 0 0,01 0,01 0,01 0 0,01 0,01 -0,01 0,29 0,01 0,16 0,12 0,23 0,12 0,01 0,01 0 -0,12 0
49 0 0 0 0 -0,03 0 0 0 0 0 0 0,01 0 0,05 0,25 0,1 0,24 0 0 0 0 0
50 0,02 0,05 0,02 0,02 -0,02 0,01 0,03 0,01 0,03 0,03 -0,03 0,29 0,04 0,16 0,33 0,27 0,34 0,05 0,04 0 -0,13 0,01
51 0,56 0,3 -0,1 0,54 0 0 0,51 0,85 0,49 0,39 0,03 0,01 0,43 0 0 0,01 0 0,54 0,4 0,3 -0,02 0,22
52 0 0 -0,01 -0,01 0,22 0,1 -0,01 -0,01 0 -0,01 -0,01 0,42 -0,01 0,55 0,08 0,4 0,07 -0,01 0 -0,01 -0,1 -0,01
53 0 0,02 0,01 0,01 0,1 0,05 0,02 0,01 0,01 0,01 -0,02 0,2 0,02 0,3 0,05 0,25 0,04 0 0,02 0 0,05 -0,01
54 0,23 0,46 0,04 0,34 0 0 0,3 0,33 0,3 0,2 0,16 0,03 0,15 0 0 -0,01 0 0,16 0,19 0,29 -0,02 0,05
55 0,41 -0,04 0,36 0,62 0 0 0,38 0,4 0,35 0,41 0,21 0,02 0,49 0,01 0 0,01 0 0,37 0,25 0,18 -0,04 0,28
56 0,32 0,36 0,26 0,54 0,01 0 0,36 0,52 0,35 0,4 0,21 0,02 0,38 0,01 0 0,01 0 0,44 0,44 0,24 -0,03 0,15
57 0,02 0,01 0 0,01 0,18 0,06 0,01 0,01 0,01 0,01 0,01 0,14 0,02 0,04 0,01 0,09 0,02 0,02 0,02 0 0,01 0,01
58 0,02 0,02 0 0,01 0,17 0,06 0,01 0,01 0,01 0,01 0,01 0,15 0,02 0,04 0,01 0,09 0,02 0,02 0,02 0 0 0,01
59 -0,01 0,04 -0,01 0,04 0,17 0,07 0,04 0,05 0,04 0,06 0,02 0,4 0,02 0,17 0,04 0,19 0,04 0,03 0,01 0,02 -0,02 0,04
60 0,02 0,04 -0,02 0,04 0,17 0,07 0,05 0,04 0,05 0,04 0,01 0,35 0,03 0,17 0,03 0,18 0,03 0,02 0 0,02 -0,04 0
61 -0,01 0,06 -0,02 0,04 0,17 0,08 0,04 0,05 0,04 0,05 0,01 0,43 0,02 0,2 0,04 0,19 0,04 0,02 0,01 0,02 -0,05 0,04
62 0 0,06 -0,02 0,05 0,2 0,09 0,05 0,05 0,04 0,05 0,01 0,45 0,02 0,21 0,04 0,22 0,04 0,02 0,01 0,02 -0,06 0,03
63 0,01 0 0,01 0,01 0,15 0,05 0,01 0,01 0,01 0,02 0,01 0,03 0,02 0 0 0,05 0 0,02 0,02 0 0,08 0,01
64 0,02 0,02 0 0,01 0,17 0,06 0,01 0,01 0,01 0,01 0,01 0,15 0,02 0,04 0,01 0,09 0,02 0,02 0,02 0 0 0,01
65 0 0 0 0 -0,01 0,01 0 0 0 0 0 0 0 0,04 0,41 0,03 0,41 0 0 0 0 0
66 0,16 0,09 0,31 0,28 0,01 0 0,42 0,17 0,62 0,3 0,08 0,02 0,26 0,01 0 0,02 0 0,27 0,18 0,38 0,06 0,22
67 0,04 0,04 0,27 0 0 0 -0,01 -0,11 -0,05 0,18 0,07 0,02 0,02 0 0 0,01 0 0,07 -0,01 0,09 0,01 -0,13
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 162/183
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
45 0 -0,01 0,01 0 0,16 0,19 0,14 0,19 0,1 0,18 0,17 0,01 -0,01 -0,01 0,07 0,1 0,11 0,02 0,22 0,15 0,21 0,56
46 0,01 -0,01 0 0 0,11 0,21 0,21 0,21 0,13 0,24 0,2 0,03 -0,03 -0,02 0,05 0,1 0,29 0,01 0,2 0,17 0,31 0,66
47 0 -0,01 -0,01 -0,01 0,08 0,31 0,32 0,28 0,2 0,35 0,29 0 0,01 0,01 0,2 0,27 0,21 0,03 0,19 0,19 0,2 0,24
48 0 -0,01 0 0 0,09 0,13 0,11 0,14 0,09 0,14 0,13 0,01 -0,01 -0,01 0,04 0,11 0,12 0,07 0,25 0,15 0,22 0,35
49 0 -0,03 0 0 -0,01 0,03 0,13 0,02 0,07 0,12 0,05 0 0 0 -0,02 0 0,24 0 0,03 0 0,19 0,01
50 0,01 -0,01 -0,01 0 0,07 0,14 0,16 0,15 0,1 0,18 0,15 0,04 -0,04 -0,03 0,03 0,09 0,33 0,04 0,19 0,13 0,3 0,37
51 0,77 0 0 0 0 0 0 0 0 0 0 0,26 0,2 0,27 0,01 0,01 0 0 0 0,01 0 0
52 0 0 -0,01 -0,01 0,13 0,29 0,2 0,29 0,16 0,24 0,26 -0,01 -0,01 -0,01 0,14 0,27 0,1 -0,02 0,1 0,15 0,1 0,18
53 0,01 0 -0,01 -0,01 0,09 0,19 0,17 0,16 0,11 0,19 0,16 -0,02 -0,01 -0,02 0,1 0,17 0,06 -0,03 0,02 0,03 0,02 0,04
54 0,24 0 0 0 0,02 0 0 0 -0,01 0,01 -0,01 0,36 0,2 0,18 0,02 0,01 0 0 0 0,03 0,01 0,02
55 0,3 0 0 0 0 0 0 0,01 0,02 0 0,01 0,37 0,31 -0,14 0,02 0,01 0 0 0 0,01 0 0,02
56 0,39 0 0 0 0,01 0 0 0,01 0,02 0 0,01 0,39 0,33 0,03 0,03 0,02 0 0 0 0,02 0,01 0
57 0,01 0 0,01 0 0,02 0,08 0,09 0,1 0,11 0,1 0,12 0,01 0,02 0,01 0,03 0,05 0,02 0 0,02 0,04 0,03 0,14
58 0,01 0 0,01 0 0,02 0,08 0,1 0,1 0,11 0,11 0,12 0,01 0,02 0,01 0,03 0,06 0,02 0 0,02 0,04 0,03 0,14
59 0,04 0,01 0 0,03 0,13 0,25 0,32 0,23 0,22 0,35 0,26 0,01 0,01 0,01 0,14 0,19 0,05 0,09 0,19 0,28 0,24 0,24
60 0,02 0 0,02 0,02 0,13 0,21 0,19 0,16 0,11 0,21 0,17 0,02 -0,04 -0,03 0,13 0,15 0,05 0,09 0,17 0,25 0,14 0,21
61 0,03 0 0,01 0,03 0,14 0,24 0,28 0,2 0,17 0,31 0,23 0 0 0 0,16 0,19 0,05 0,1 0,2 0,31 0,21 0,25
62 0,03 0 0,01 0,03 0,14 0,27 0,28 0,24 0,18 0,32 0,25 0,01 0 0 0,16 0,2 0,06 0,1 0,21 0,32 0,2 0,27
63 0,01 0 0 0 -0,02 0 0,01 0,01 0,03 0,01 0,02 0,02 0,03 0,02 -0,02 0,01 0,01 0 -0,02 -0,02 -0,02 0,01
64 0,01 0 0,01 0 0,02 0,08 0,1 0,1 0,11 0,11 0,12 0,01 0,02 0,01 0,03 0,06 0,02 0 0,02 0,04 0,03 0,14
65 0 -0,05 0 0 0 0 0 0 0,01 0 0 0 0 0 -0,01 -0,02 0,4 0 0 0 0,08 0
66 0,06 0 0 0 -0,01 0 -0,01 0 0,01 -0,01 0,01 0,38 -0,06 -0,01 0,02 0,02 0 0 0 0,01 0 0,01
67 -0,1 0 0 0 0,01 0 0 0 0,01 0 0 0,24 0,2 -0,08 0,01 0,01 0 0 0 0,02 0,01 0,01
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 163/183
45 1 0,72 0,22 0,76 0,1 0,55 0 0,11 0,02 0,01 0,01 0 0,09 0,09 0,19 0,15 0,19 0,2 0,01 0,09 0,02 0,01 0
46 0,72 1 0,3 0,57 0,34 0,76 0,01 0,16 0,04 0,03 0,04 0 0,11 0,11 0,26 0,17 0,23 0,24 0,01 0,11 0,08 0,02 0,01
47 0,22 0,3 1 0,18 0,2 0,24 0 0,28 0,2 0,01 0 0,01 0,07 0,08 0,38 0,18 0,31 0,29 -0,02 0,08 0,08 0 0
48 0,76 0,57 0,18 1 0,15 0,73 0 0,11 0,03 0,01 0,01 0 0,07 0,07 0,15 0,1 0,15 0,15 0,01 0,07 0,03 0,01 0
49 0,1 0,34 0,2 0,15 1 0,46 0 0,01 0 0 0 0 0 0 0,13 -0,06 0,02 0 0 0 0,06 0 0
50 0,55 0,76 0,24 0,73 0,46 1 0,01 0,1 0,03 0,04 0,05 0 0,06 0,06 0,18 0,09 0,15 0,14 0 0,06 0,1 0,03 0,01
51 0 0,01 0 0 0 0,01 1 0 0,01 0,29 0,34 0,46 0,01 0,01 0,03 0,03 0,03 0,03 0,01 0,01 0 0,19 -0,07
52 0,11 0,16 0,28 0,11 0,01 0,1 0 1 0,54 -0,02 -0,01 0 0,02 0,03 0,17 0,15 0,19 0,19 -0,02 0,03 0 0 0
53 0,02 0,04 0,2 0,03 0 0,03 0,01 0,54 1 -0,03 -0,01 0,02 0,05 0,05 0,19 0,21 0,23 0,22 0 0,05 0 0,02 0
54 0,01 0,03 0,01 0,01 0 0,04 0,29 -0,02 -0,03 1 0,14 0,29 0 0 0 0,01 0,01 0,01 -0,01 0 0 0,22 0,06
55 0,01 0,04 0 0,01 0 0,05 0,34 -0,01 -0,01 0,14 1 0,51 0,01 0,01 0 0,02 0 0,01 0,02 0,01 0 0,2 0,27
56 0 0 0,01 0 0 0 0,46 0 0,02 0,29 0,51 1 0,02 0,02 0,02 0 0,02 0,02 0,02 0,02 0 0,22 0,26
57 0,09 0,11 0,07 0,07 0 0,06 0,01 0,02 0,05 0 0,01 0,02 1 0,99 0,33 0,36 0,26 0,4 0,81 0,99 0 0,01 0,01
58 0,09 0,11 0,08 0,07 0 0,06 0,01 0,03 0,05 0 0,01 0,02 0,99 1 0,36 0,33 0,27 0,41 0,81 1 0 0,01 0,01
59 0,19 0,26 0,38 0,15 0,13 0,18 0,03 0,17 0,19 0 0 0,02 0,33 0,36 1 0,56 0,88 0,88 0,19 0,36 0,01 0,03 0,01
60 0,15 0,17 0,18 0,1 -0,06 0,09 0,03 0,15 0,21 0,01 0,02 0 0,36 0,33 0,56 1 0,72 0,79 0,21 0,33 0 0,04 0
61 0,19 0,23 0,31 0,15 0,02 0,15 0,03 0,19 0,23 0,01 0 0,02 0,26 0,27 0,88 0,72 1 0,95 0,02 0,27 0 0,03 0
62 0,2 0,24 0,29 0,15 0 0,14 0,03 0,19 0,22 0,01 0,01 0,02 0,4 0,41 0,88 0,79 0,95 1 0,23 0,41 0 0,03 0
63 0,01 0,01 -0,02 0,01 0 0 0,01 -0,02 0 -0,01 0,02 0,02 0,81 0,81 0,19 0,21 0,02 0,23 1 0,81 0 0,01 0,01
64 0,09 0,11 0,08 0,07 0 0,06 0,01 0,03 0,05 0 0,01 0,02 0,99 1 0,36 0,33 0,27 0,41 0,81 1 0 0,01 0,01
65 0,02 0,08 0,08 0,03 0,06 0,1 0 0 0 0 0 0 0 0 0,01 0 0 0 0 0 1 0 0
66 0,01 0,02 0 0,01 0 0,03 0,19 0 0,02 0,22 0,2 0,22 0,01 0,01 0,03 0,04 0,03 0,03 0,01 0,01 0 1 0,36
67 0 0,01 0 0 0 0,01 -0,07 0 0 0,06 0,27 0,26 0,01 0,01 0,01 0 0 0 0,01 0,01 0 0,36 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 164/183
Legenda
1 ability_to_cook_food 23 personal_hygiene_and_grooming 45 shops_time
2 abstraction 24 pharmacy_visits_month 46 shops_visits
3 appetite 25 phonecalls_long_placed_perc 47 still_time
4 bathing_and_showering 26 phonecalls_long_received_perc 48 supermarket_time
5 bedroom_time 27 phonecalls_missed 49 supermarket_time_perc
6 bedroom_visits 28 phonecalls_placed 50 supermarket_visits
7 dependence 29 phonecalls_placed_perc 51 toilet_hygiene
8 dressing 30 phonecalls_received 52 tvwatching_time
9 exhaustion 31 phonecalls_received_perc 53 tvwatching_time_perc
10 falls 32 phonecalls_short_placed_perc 54 visits
11 gait_balance 33 phonecalls_short_received_perc 55 visit_to_doctors
12 home_time 34 quality_of_sleep 56 visit_to_health_related_places
13 housekeeping 35 reading_books 57 walk_distance_outdoor
14 kitchen_time 36 reading_newspapers 58 walk_distance_outdoor_fast
15 kitchen_visits 37 restroom_time 59 walk_distance_outdoor_fast_perc
16 livingroom_time 38 restroom_visits 60 walk_distance_outdoor_slow_perc
17 livingroom_visits 39 room_changes 61 walk_speed_outdoor
18 medication 40 seniorcenter_long_visits 62 walk_speed_outdoor_fast
19 memory 41 seniorcenter_time 63 walk_time_outdoor
20 mood 42 seniorcenter_time_out_perc 64 walk_time_outdoor_fast
21 outdoor_time 43 seniorcenter_visits 65 washingmachine_sessions
22 pain 44 shops_outdoor_time_perc 66 weakness
67 weight
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 165/183
12.4 Madrid Pilot
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
1 1 -0,1 0,03 0 0 0 0 0,53 0,51 -0,02 0 0 0 0 -0,01 0,01 -0,01 -0,02
2 -0,1 1 -0,15 0,03 0,03 0,11 0 0,08 0,1 0,07 0,17 0,29 -0,19 -0,12 -0,09 0,03 -0,09 -0,05
3 0,03 -0,15 1 0,01 0 0,01 0 0,01 0 0 0,02 0,03 -0,05 0,06 -0,03 -0,07 -0,06 -0,06
4 0 0,03 0,01 1 0,36 0,22 0,05 0 0 0,04 0 0 -0,01 -0,01 -0,11 -0,09 -0,07 -0,06
5 0 0,03 0 0,36 1 0,05 -0,05 0,01 0,01 -0,08 0,05 -0,01 0,03 0,02 0,19 0,13 0,06 0,06
6 0 0,11 0,01 0,22 0,05 1 0,12 0 0 0,03 0,17 0,3 -0,01 -0,01 -0,01 -0,02 0 0,02
7 0 0 0 0,05 -0,05 0,12 1 0 0 0,06 -0,04 -0,03 -0,01 -0,01 0 -0,02 0,01 0,04
8 0,53 0,08 0,01 0 0,01 0 0 1 0,99 0,01 0 0 0 0 0 0,01 0 -0,01
9 0,51 0,1 0 0 0,01 0 0 0,99 1 0,01 0 0 0 0 0 0,01 0 -0,01
10 -0,02 0,07 0 0,04 -0,08 0,03 0,06 0,01 0,01 1 -0,03 0,03 0,04 0,02 -0,05 -0,11 0,07 0,19
11 0 0,17 0,02 0 0,05 0,17 -0,04 0 0 -0,03 1 0,63 0 0 -0,03 -0,02 0,03 -0,03
12 0 0,29 0,03 0 -0,01 0,3 -0,03 0 0 0,03 0,63 1 0 0 -0,03 -0,04 -0,01 -0,03
13 0 -0,19 -0,05 -0,01 0,03 -0,01 -0,01 0 0 0,04 0 0 1 0,33 0,06 -0,02 0,06 0,01
14 0 -0,12 0,06 -0,01 0,02 -0,01 -0,01 0 0 0,02 0 0 0,33 1 0,03 -0,02 0,15 -0,01
15 -0,01 -0,09 -0,03 -0,11 0,19 -0,01 0 0 0 -0,05 -0,03 -0,03 0,06 0,03 1 0,51 0,24 0,44
16 0,01 0,03 -0,07 -0,09 0,13 -0,02 -0,02 0,01 0,01 -0,11 -0,02 -0,04 -0,02 -0,02 0,51 1 -0,19 -0,04
17 -0,01 -0,09 -0,06 -0,07 0,06 0 0,01 0 0 0,07 0,03 -0,01 0,06 0,15 0,24 -0,19 1 0,31
18 -0,02 -0,05 -0,06 -0,06 0,06 0,02 0,04 -0,01 -0,01 0,19 -0,03 -0,03 0,01 -0,01 0,44 -0,04 0,31 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 166/183
Legenda
1 culturepoi_visits_month
2 othersocial_time
3 othersocial_visits
4 outdoor_num
5 outdoor_time
6 publicpark_time
7 publicpark_visits
8 publictransport_distance_month
9 publictransport_rides_month
10 publictransport_time
11 seniorcenter_time
12 seniorcenter_visits
13 supermarket_time
14 supermarket_visits
15 walk_distance
16 walk_speed_outdoor
17 walk_steps
18 walk_time_outdoor
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 167/183
12.5 Montpellier Pilot
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
1 1 0,27 -0,25 -0,07 -0,02 0 -0,06 -0,18 -0,12 -0,02 -0,03 0,14 0,02 0,06 -0,03 0 0 0 -0,02 -0,08 0 -0,01 -0,03 -0,04 -0,22 -0,01 0
2 0,27 1 -0,01 0,45 -0,12 -0,06 0,03 -0,22 -0,1 0,1 0,42 0,04 -0,01 -0,03 -0,08 -0,04 -0,02 0,13 0,02 -0,15 0,02 0,54 0,05 0,04 -0,22 0 0,03
3 -0,25 -0,01 1 0,5 -0,01 -0,02 -0,01 -0,24 -0,1 -0,61 -0,23 -0,24 -0,08 0,01 0,01 0 0,01 0,02 0,02 -0,3 -0,37 -0,03 -0,01 0,01 0,59 0,01 0,04
4 -0,07 0,45 0,5 1 -0,06 -0,04 0,06 -0,25 0,05 -0,21 0,35 -0,12 -0,07 -0,06 -0,04 -0,02 -0,01 0,09 0,02 -0,26 -0,25 0,57 0,05 0,04 0,14 0 0,07
5 -0,02 -0,12 -0,01 -0,06 1 0,01 -0,03 0,01 0,02 0,01 -0,06 0 0,02 0,03 -0,11 0 0,16 0,01 -0,14 0,03 0 -0,07 0,53 0,03 0,02 -0,07 0,03
6 0 -0,06 -0,02 -0,04 0,01 1 0 0 0 0,02 -0,04 0 0 0 0,13 0,7 0 -0,26 0 0 0,01 -0,05 0,01 0 0,01 0 0
7 -0,06 0,03 -0,01 0,06 -0,03 0 1 0,11 0,16 0,16 0,18 -0,04 -0,16 -1 0 0 0 0 0 0 0,12 0,18 -0,02 0 -0,01 0 0,01
8 -0,18 -0,22 -0,24 -0,25 0,01 0 0,11 1 0,5 -0,06 0,13 0,06 -0,08 -0,11 0,02 0 0,01 -0,01 0 0,03 0,26 0,13 -0,01 -0,01 -0,06 0,01 0,04
9 -0,12 -0,1 -0,1 0,05 0,02 0 0,16 0,5 1 0,09 0,65 0,11 -0,07 -0,15 0,02 0 0,01 -0,01 0 -0,05 0,36 0,67 -0,01 -0,01 -0,04 0,01 0,13
10 -0,02 0,1 -0,61 -0,21 0,01 0,02 0,16 -0,06 0,09 1 0,39 0,08 -0,04 -0,16 -0,04 -0,02 -0,05 0 -0,02 -0,02 0,3 0,19 0,04 0,03 -0,39 -0,01 0,01
11 -0,03 0,42 -0,23 0,35 -0,06 -0,04 0,18 0,13 0,65 0,39 1 0,09 -0,06 -0,18 -0,05 -0,03 -0,01 0,08 0,02 -0,09 0,33 0,92 0,02 0,04 -0,25 0 0,09
12 0,14 0,04 -0,24 -0,12 0 0 -0,04 0,06 0,11 0,08 0,09 1 0,1 0,04 0 0 0 0 0 0,22 0,22 0,08 0 0 -0,27 0 0,01
13 0,02 -0,01 -0,08 -0,07 0,02 0 -0,16 -0,08 -0,07 -0,04 -0,06 0,1 1 0,16 0,15 0 0,03 -0,01 -0,01 0,07 0,03 -0,07 -0,01 -0,01 -0,06 0 -0,03
14 0,06 -0,03 0,01 -0,06 0,03 0 -1 -0,11 -0,15 -0,16 -0,18 0,04 0,16 1 0 0 0 0 0 0 -0,12 -0,18 0,02 0 0,01 0 -0,01
15 -0,03 -0,08 0,01 -0,04 -0,11 0,13 0 0,02 0,02 -0,04 -0,05 0 0,15 0 1 0,18 -0,02 -0,02 0,01 0,02 -0,01 -0,05 -0,09 0,01 0 0,01 0
16 0 -0,04 0 -0,02 0 0,7 0 0 0 -0,02 -0,03 0 0 0 0,18 1 0 -0,19 0 0,01 -0,01 -0,03 0 0 0,01 0 0
17 0 -0,02 0,01 -0,01 0,16 0 0 0,01 0,01 -0,05 -0,01 0 0,03 0 -0,02 0 1 0 -0,2 0 0,01 -0,01 -0,27 -0,49 -0,01 -0,11 -0,06
18 0 0,13 0,02 0,09 0,01 -0,26 0 -0,01 -0,01 0 0,08 0 -0,01 0 -0,02 -0,19 0 1 0 -0,01 -0,02 0,09 0,01 0 -0,01 0 0
19 -0,02 0,02 0,02 0,02 -0,14 0 0 0 0 -0,02 0,02 0 -0,01 0 0,01 0 -0,2 0 1 0 0 0,02 -0,01 0,47 0 0,56 -0,06
20 -0,08 -0,15 -0,3 -0,26 0,03 0 0 0,03 -0,05 -0,02 -0,09 0,22 0,07 0 0,02 0,01 0 -0,01 0 1 0,33 -0,13 0 0,01 -0,13 0 -0,07
21 0 0,02 -0,37 -0,25 0 0,01 0,12 0,26 0,36 0,3 0,33 0,22 0,03 -0,12 -0,01 -0,01 0,01 -0,02 0 0,33 1 0,34 -0,04 0 -0,26 0 0,04
22 -0,01 0,54 -0,03 0,57 -0,07 -0,05 0,18 0,13 0,67 0,19 0,92 0,08 -0,07 -0,18 -0,05 -0,03 -0,01 0,09 0,02 -0,13 0,34 1 0,03 0,04 -0,17 0,01 0,12
23 -0,03 0,05 -0,01 0,05 0,53 0,01 -0,02 -0,01 -0,01 0,04 0,02 0 -0,01 0,02 -0,09 0 -0,27 0,01 -0,01 0 -0,04 0,03 1 0,39 0,01 -0,04 0,07
24 -0,04 0,04 0,01 0,04 0,03 0 0 -0,01 -0,01 0,03 0,04 0 -0,01 0 0,01 0 -0,49 0 0,47 0,01 0 0,04 0,39 1 0 0,11 0,08
25 -0,22 -0,22 0,59 0,14 0,02 0,01 -0,01 -0,06 -0,04 -0,39 -0,25 -0,27 -0,06 0,01 0 0,01 -0,01 -0,01 0 -0,13 -0,26 -0,17 0,01 0 1 0,02 0,01
26 -0,01 0 0,01 0 -0,07 0 0 0,01 0,01 -0,01 0 0 0 0 0,01 0 -0,11 0 0,56 0 0 0,01 -0,04 0,11 0,02 1 -0,08
27 0 0,03 0,04 0,07 0,03 0 0,01 0,04 0,13 0,01 0,09 0,01 -0,03 -0,01 0 0 -0,06 0 -0,06 -0,07 0,04 0,12 0,07 0,08 0,01 -0,08 1
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 168/183
Legenda
1 bathroom_time
2 bathroom_visits
3 bedroom_time
4 bedroom_visits
5 cinema_visits
6 cinema_visits_month
7 home_time
8 kitchen_time
9 kitchen_visits
10 livingroom_time
11 livingroom_visits
12 meals_num
13 outdoor_num
14 outdoor_time
15 pharmacy_visits
16 pharmacy_visits_month
17 pharmacy_visits_week
18 restaurants_visits_month
19 restaurants_visits_week
20 restroom_time
21 restroom_visits
22 room_changes
23 shops_visits
24 shops_visits_week
25 sleep_time
26 supermarket_visits_week
27 visits_received_week
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 169/183
12.6 Singapore Pilot
1 2 3 4 5 6 7 8 9 10 11
1 1 0,1 -0,01 0,01 0 -0,02 -0,03 -0,02 -0,03 -0,04 -0,02
2 0,1 1 0,01 0,08 0,04 0,04 -0,04 -0,03 -0,05 -0,06 0
3 -0,01 0,01 1 0,39 -0,11 -0,12 -0,17 -0,14 -0,18 -0,18 -0,1
4 0,01 0,08 0,39 1 -0,03 0,01 -0,11 -0,09 -0,13 -0,1 0,01
5 0 0,04 -0,11 -0,03 1 0,22 -0,08 -0,03 -0,14 0,15 0,13
6 -0,02 0,04 -0,12 0,01 0,22 1 0,48 0,67 -0,21 0,65 0,79
7 -0,03 -0,04 -0,17 -0,11 -0,08 0,48 1 0,74 -0,21 0,52 0,58
8 -0,02 -0,03 -0,14 -0,09 -0,03 0,67 0,74 1 -0,17 0,64 0,78
9 -0,03 -0,05 -0,18 -0,13 -0,14 -0,21 -0,21 -0,17 1 -0,1 -0,19
10 -0,04 -0,06 -0,18 -0,1 0,15 0,65 0,52 0,64 -0,1 1 0,71
11 -0,02 0 -0,1 0,01 0,13 0,79 0,58 0,78 -0,19 0,71 1
Legenda
1 bathroom_time
2 bathroom_visits
3 bedroom_time
4 bedroom_visits
5 kitchen_time
6 kitchen_visits
7 livingroom_time
8 livingroom_visits
9 restroom_time
10 restroom_visits
11 room_changes
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 170/183
13 Annex: FAI data from Pilots
13.1 Athens Pilot
Frailty score
Robustness score
Label Frailty score
Robustness score
Label Frailty score
Robustness score
Label Frailty score
Robustness score
Label
0 4 Robust
0 3 Robust
0 4 Robust
1 2 preFrail
1 3 Robust
0 3 Robust
1 2 preFrail
2 4 Robust
1 3 Robust
1 2 preFrail
3 3 postRrobust
0 3 Robust
0 4 Robust
1 3 Robust
2 3 Robust
1 3 Robust
0 4 Robust
0 3 Robust
0 4 Robust
0 3 Robust
0 3 Robust
2 2 preFrail
0 3 Robust
2 2 preFrail
1 2 preFrail
4 2 Frail
1 2 preFrail
3 3 postRrobust
0 4 Robust
0 3 Robust
0 4 Robust
1 4 Robust
0 3 Robust
3 3 postRrobust
0 3 Robust
3 4 postRrobust
2 3 Robust
0 4 Robust
1 2 preFrail
1 4 Robust
3 4 postRrobust
1 3 Robust
3 4 postRrobust
1 3 Robust
2 3 Robust
0 4 Robust
2 3 Robust
0 4 Robust
0 4 Robust
1 3 Robust
1 3 Robust
1 3 Robust
1 3 Robust
1 3 Robust
1 3 Robust
1 3 Robust
0 3 Robust
0 4 Robust
1 3 Robust
1 3 Robust
0 3 Robust
0 4 Robust
1 2 preFrail
0 4 Robust
1 4 Robust
5 0 Frail
3 2 Frail
4 0 Frail
2 2 preFrail
3 3 postRrobust
0 4 Robust
3 3 postRrobust
0 4 Robust
1 3 Robust
1 3 Robust
1 3 Robust
3 2 Frail
0 4 Robust
2 3 Robust
1 3 Robust
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 171/183
13.2 Birmingham Pilot
Frailty score
Robustness score Label
0 5 Robust
0 3 Robust
0 5 Robust
0 5 Robust
0 5 Robust
0 6 Robust
0 5 Robust
0 5 Robust
2 6 Robust
0 4 Robust
4 3 postRrobust
0 4 Robust
0 5 Robust
0 4 Robust
0 6 Robust
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 172/183
13.3 Madrid Pilot
Frailty score
Robustness score Label
1 5 Robust
1 5 Robust
0 5 Robust
0 5 Robust
0 4 Robust
0 4 Robust
0 5 Robust
0 4 Robust
2 3 Robust
3 3 postRrobust
0 4 Robust
0 6 Robust
0 4 Robust
0 4 Robust
1 5 Robust
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 173/183
14 Annex: Data pre-processing SQL for the Athens Pilot experiment
Note: for full data anonymization, the City4AgeID pseudonyms have been obscured in the query below
select
avg(cinema_time) as cinema_time_avg,
stddev(cinema_time) as cinema_time_sd,
percentile_cont(0.75) within group ( order by cinema_time ) as cinema_time_best,
avg(cinema_visits) as cinema_visits_avg,
stddev(cinema_visits) as cinema_visits_sd,
percentile_cont(0.75) within group ( order by cinema_visits ) as cinema_visits_best,
avg(home_time) as home_time_avg,
stddev(home_time) as home_time_sd,
percentile_cont(0.25) within group ( order by home_time ) as home_time_best,
avg(othersocial_time) as othersocial_time_avg,
stddev(othersocial_time) as othersocial_time_sd,
percentile_cont(0.75) within group ( order by othersocial_time ) as othersocial_time_best,
avg(othersocial_visits) as othersocial_visits_avg,
stddev(othersocial_visits) as othersocial_visits_sd,
percentile_cont(0.75) within group ( order by othersocial_visits ) as othersocial_visits_best,
avg(pharmacy_time) as pharmacy_time_avg,
stddev(pharmacy_time) as pharmacy_time_sd,
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 174/183
percentile_cont(0.25) within group ( order by pharmacy_time ) as pharmacy_time_best,
avg(pharmacy_visits) as pharmacy_visits_avg,
stddev(pharmacy_visits) as pharmacy_visits_sd,
percentile_cont(0.25) within group ( order by pharmacy_visits ) as pharmacy_visits_best,
avg(restaurants_time) as restaurants_time_avg,
stddev(restaurants_time) as restaurants_time_sd,
percentile_cont(0.75) within group ( order by restaurants_time ) as restaurants_time_best,
avg(restaurants_visits_week) as restaurants_visits_week_avg,
stddev(restaurants_visits_week) as restaurants_visits_week_sd,
percentile_cont(0.75) within group ( order by restaurants_visits_week ) as restaurants_visits_week_best,
avg(seniorcenter_time) as seniorcenter_time_avg,
stddev(seniorcenter_time) as seniorcenter_time_sd,
percentile_cont(0.75) within group ( order by seniorcenter_time ) as seniorcenter_time_best,
avg(seniorcenter_visits) as seniorcenter_visits_avg,
stddev(seniorcenter_visits) as seniorcenter_visits_sd,
percentile_cont(0.75) within group ( order by seniorcenter_visits ) as seniorcenter_visits_best,
avg(shops_outdoor_time_perc) as shops_outdoor_time_perc_avg,
stddev(shops_outdoor_time_perc) as shops_outdoor_time_perc_sd,
percentile_cont(0.75) within group ( order by shops_outdoor_time_perc ) as shops_outdoor_time_perc_best,
avg(shops_time) as shops_time_avg,
stddev(shops_time) as shops_time_sd,
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 175/183
percentile_cont(0.75) within group ( order by shops_time ) as shops_time_best,
avg(shops_visits) as shops_visits_avg,
stddev(shops_visits) as shops_visits_sd,
percentile_cont(0.75) within group ( order by shops_visits ) as shops_visits_best,
avg(transport_time) as transport_time_avg,
stddev(transport_time) as transport_time_sd,
percentile_cont(0.75) within group ( order by transport_time ) as transport_time_best,
avg(walk_distance_outdoor) as walk_distance_outdoor_avg,
stddev(walk_distance_outdoor) as walk_distance_outdoor_sd,
percentile_cont(0.75) within group ( order by walk_distance_outdoor ) as walk_distance_outdoor_best,
avg(walk_speed_outdoor) as walk_speed_outdoor_avg,
stddev(walk_speed_outdoor) as walk_speed_outdoor_sd,
percentile_cont(0.75) within group ( order by walk_speed_outdoor ) as walk_speed_outdoor_best,
avg(walk_time_outdoor) as walk_time_outdoor_avg,
stddev(walk_time_outdoor) as walk_time_outdoor_sd,
percentile_cont(0.75) within group ( order by walk_time_outdoor ) as walk_time_outdoor_best,
max(status) as status
from crosstab('
select
format(''%s%s'', date(time_interval.interval_start), care_recipient.id) as row_name,
case when time_interval.interval_start <= ''2018-01-31'' then 1 else 2 end as trimester,
date(time_interval.interval_start), care_recipient.id,
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 176/183
variable.detection_variable_name, min(measure.measure_value)
from
city4age_sr.variation_measure_value as measure inner join
city4age_sr.user_in_role as care_recipient on measure.user_in_role_id = care_recipient.id inner join
city4age_sr.pilot as pilot on care_recipient.pilot_code = pilot.pilot_code inner join
city4age_sr.cd_detection_variable as variable on measure.measure_type_id = variable.id inner join
city4age_sr.time_interval as time_interval on measure.time_interval_id = time_interval.id
where
pilot.pilot_code = ''ath'' and
date(time_interval.interval_start) >= ''2017-11-01'' and date(time_interval.interval_start) <= ''2018-04-30''
group by time_interval.interval_start, care_recipient.id, variable.detection_variable_name
order by 1
','
select cat
from (VALUES
(''cinema_time''),
(''cinema_visits''),
(''cinema_visits_month''),
(''home_time''),
(''othersocial_time''),
(''othersocial_visits''),
(''pharmacy_time''),
(''pharmacy_visits''),
(''pharmacy_visits_month''),
(''restaurants_time''),
(''restaurants_visits_week''),
(''seniorcenter_time''),
(''seniorcenter_visits''),
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 177/183
(''shops_outdoor_time_perc''),
(''shops_time''),
(''shops_visits''),
(''supermarket_time''),
(''supermarket_visits''),
(''transport_time''),
(''walk_distance_outdoor''),
(''walk_distance_outdoor_fast_perc''),
(''walk_distance_outdoor_slow_perc''),
(''walk_speed_outdoor''),
(''walk_time_outdoor'')) as v(cat)
order by 1
') as ct(
row_name text,
trimester numeric,
day_m date,
cr_m numeric,
cinema_time numeric,
cinema_visits numeric,
cinema_visits_month numeric,
home_time numeric,
othersocial_time numeric,
othersocial_visits numeric,
pharmacy_time numeric,
pharmacy_visits numeric,
pharmacy_visits_month numeric,
restaurants_time numeric,
restaurants_visits_week numeric,
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 178/183
seniorcenter_time numeric,
seniorcenter_visits numeric,
shops_outdoor_time_perc numeric,
shops_time numeric,
shops_visits numeric,
supermarket_time numeric,
supermarket_visits numeric,
transport_time numeric,
walk_distance_outdoor numeric,
walk_distance_outdoor_fast_perc numeric,
walk_distance_outdoor_slow_perc numeric,
walk_speed_outdoor numeric,
walk_time_outdoor numeric
) inner join (VALUES
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'postRobust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'postRobust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 179/183
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Frail'),
([City4AgeID obscured], 1, 'postRobust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'postRobust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'postRobust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'preFrail'),
([City4AgeID obscured], 1, 'Frail'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 1, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 180/183
([City4AgeID obscured], 2, 'preFrail'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'preFrail'),
([City4AgeID obscured], 2, 'Frail'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'postRobust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Frail'),
([City4AgeID obscured], 2, 'postRobust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'preFrail'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 181/183
([City4AgeID obscured], 2, 'postRobust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'preFrail'),
([City4AgeID obscured], 2, 'Robust'),
([City4AgeID obscured], 2, 'Frail')
) as labels(cr_m, trimester, status)
on ct.cr_m = labels.cr_m and ct.trimester = labels.trimester
group by ct.cr_m, ct.trimester
having avg(home_time) != 86400
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 182/183
15 Annex: MultiSchemeAUC modifications
Note: modifications that have been implemented to convert Weka MultiScheme meta-learner into City4Age MultiSchemeAUC are shaded
public void buildClassifier(Instances data) throws Exception {
...
double AUC = evaluation.weightedAreaUnderROC();
if (m_Debug) {
System.err.println("Error rate: " + Utils.doubleToString(AUC, 6, 4)
+ " for classifier "
+ currentClassifier.getClass().getName());
}
if ((i == 0) || (AUC > bestPerformance)) {
bestClassifier = currentClassifier;
bestPerformance = AUC;
bestIndex = i;
}
...
Project City4Age Grant Agreement #689731
D2.11 City4Age frailty and MCI risk model 183/183
16 Annex: Computing *_time_per_visits features
case when sum(cinema_visits) != 0 then sum(cinema_time) / sum(cinema_visits) else NULL end as cinema_time_per_visit,
case when sum(othersocial_visits) != 0 then sum(othersocial_time) / sum(othersocial_visits) else NULL end as othersocial_time_per_visit,
case when sum(pharmacy_visits) != 0 then sum(pharmacy_time) / sum(pharmacy_visits) else NULL end as pharmacy_time_per_visit,
case when sum(restaurants_visits_week) != 0 then sum(restaurants_time) / sum(restaurants_visits_week) else NULL end as restaurants_time_per_visit,
case when sum(seniorcenter_visits) != 0 then sum(seniorcenter_time) / sum(seniorcenter_visits) else NULL end as seniorcenter_time_per_visit,
case when sum(shops_visits) != 0 then sum(shops_time) / sum(shops_visits) else NULL end as shops_time_per_visit,
Recommended