Multimed Tools ApplDOI 10.1007/s11042-015-3190-4

Facebook: a new tool for collecting health data?

Maria Claudia Buzzi1 ·Marina Buzzi1 ·Daniele Franchi2 ·Davide Gazze1 ·Giorgio Iervasi2 ·Andrea Marchetti1 ·Alessandro Pingitore2 ·Maurizio Tesconi1

Received: 4 August 2015 / Revised: 4 December 2015 / Accepted: 18 December 2015© Springer Science+Business Media New York 2016

Abstract This study investigates the use of social networks as a scientific tool for gatheringmedical data from young subjects while promoting healthier habits. Our first hypothesis isthat social networks can facilitate epidemiological studies, reducing time and cost. The sec-ond question is whether social networks can enable the collection of data from young andhealthy subjects who are otherwise beyond the reach of conventional social health polls. AFacebook application was created to collect data concerning adherence to the Mediterraneandiet, considering the significant risk of cardiovascular and neurological degenerative dis-eases in subjects with poor adherence to a healthy diet. More than 1400 users were recruitedin a short time without any promotional action. Collected data indicate that adherence to

� Marina Buzzimarina.buzzi@iit.cnr.it

Maria Claudia Buzziclaudia.buzzi@iit.cnr.it

Daniele Franchidaniele.franchi@ifc.cnr.it

Davide Gazzedavide.gazze@iit.cnr.it

Giorgio Iervasigiorgio.iervasi@ifc.cnr.it

Andrea Marchettiandrea.marchetti@iit.cnr.it

Alessandro Pingitorealessandro.pingitore@ifc.cnr.it

Maurizio Tesconimaurizio.tesconi@iit.cnr.it

1 Institute of Informatics and Telematics (IIT), CNR, Pisa, Italy

2 Institute of Clinical Physiology (IFC), CNR, Pisa, Italy

Multimed Tools Appl

the Mediterranean diet is in general greater in older users vs young (p <0.01) and Italian vsOther Countries (mainly participants from the US) (p <0.03), while no statistical differenceswere found concerning gender. Results show that the proposed approach offers advantagesin terms of reduced cost, faster data gathering and processing, and improved efficiency com-pared to a form-based epidemiology campaign. However, the initial network may influencethe sample constitution in age and geographical location, especially if the spread does notbecome viral and autonomous. Based on the case study, we provide designers of Facebookapps with some simple guideline suggestions aimed at maximizing the heterogeneity of thesample, in order to collect significant data. The proposed scenario, suitable for collectinghealth data, can easily be extended to other fields.

Keywords Social networks · Facebook · Mediterranean diet · Health · e-health

1 Introduction

ICT (Information and Communication Technology) is increasingly influencing medicalresearch and practice, improving efficiency in process, therapy, and practice while reducingerrors and costs.

In recent years, online social networks have revolutionized social interaction by connect-ing millions of people of different ages, educational levels and cultures, all over the world.Social networks are mainly used for personal or group communication, disseminating infor-mation, videos, pictures or games. Easy access and free service favor the spread of thiscommunication tool. Social networks fuel the sharing and development of real and virtualfriendships and facilitate meeting other people with common interests. However, by gath-ering millions of users, social networks also have great potential to be studied for scientificpurposes. One of them could be gathering epidemiological data, including those regardingyoung and healthy subjects.

In this study, we investigate whether Facebook, one of the most popular online socialnetworks in the world with 1 billion active users per month, is a tool suitable for carry-ing out an epidemiological study. Specifically, two hypotheses are provided and verified,implementing a small case study:

H1: Can social networks facilitate studies on healthy habits? Is it feasible and convenientto use Facebook as a cheap tool for collecting large amounts of data?

H2: Can social networks enable the collection of data from young and healthy sub-jects who are difficult to reach in conventional social health polls? Are these datasignificant?

To verify these hypotheses, a case study was developed. A Facebook application wasdesigned to collect data on fundamental aspects of quality of life: diet, exercise, sleepand stress. To verify the feasibility of this proposal, we have developed the first ques-tionnaire concerning adherence to the Mediterranean diet, a basic element in a goodquality of life. A sample of more than 1,400 users was electronically enrolled, and datawere collected and analyzed to verify the feasibility and the reliability of the proposedapproach.

The paper is organized into seven sections. After the related work introduction, the theMediterranean Diet app is introduced. Next, results of the survey involving 1,414 partici-

Multimed Tools Appl

pants are illustrated regarding their medical significance. Lastly, based on the case study,lessons learnt are discussed, together with the two hypotheses, to highlight pros and consof the proposed approach. Guidelines for using social networks as a research tool areintroduced, as well as future work.

2 Related work

2.1 ICT and health

ICT can empower health in a multitude of ways, ranging from research to clinical proce-dures, diagnostic techniques, telemedicine, decision-supporting systems (DSS), monitoringand tele-rehabilitation, Ambient Assisted Living, Personalized Care and so on, by meansof disparate techniques: computer vision, machine learning, artificial intelligence, intel-ligent data mining, sentiment analysis, search engines and question answering systems,persuasive technology, recommendation systems, big data and analytics, to mention onlya few.

Nowadays the Internet and touchscreen mobile devices offer a simple interaction envi-ronment for all, enabling natural interaction including touch and voice, enabling anyone toaccess eHealth services anytime, anyplace, on any device. Wearable sensors, bands, watchesand smart apps enable continuous self-monitoring of a variety of vital parameters (heartrate, sleep, diet, movement, etc.) while cloud architectures permit storing and sharing hugeamounts of data effectively and easily. Considering the aging of industrialized countries, it isimportant to design usable software for older people, i.e., accessible for special-needs users,easy to use, minimizing cognitive effort, appealing and challenging enough to engage theuser for long periods (for instance, occupational therapy for maintaining residual abilitiesover time). Recent studies highlight trends and challenges for pervasive Health Technologiesin the Internet of Things [4, 40]. However, shifting from objects to humans raises varioussecurity issues. Health big data could be a precious source for researchers, but storage andprocessing of sensitive personal data requires the application of robust security algorithmsand dealing with privacy and ethical issues [34].

The challenge is to identify new application scenarios applying efficient algorithms andtechniques derived by computer science branches, such as computer multimedia tools andapplications. The potential of multimedia tools in eHealth is a universe in continuous evo-lution. Combining images, videos, text and user feedback for inclusive continuous patientmonitoring is one of several opportunities. For instance, Sartori et al. investigated emotionsconveyed by paintings, applying a multimodal approach that combines computer visiontechniques and sentiment analysis to learn statistical patterns, showing that textual featuresimprove accuracy more than a visual-only analysis does [41] while Liu et al. defined amulti-task learning algorithm able to discriminate between style- and artist-patterns, thatshowed high accuracy in retrieval and recognition of painting style [30]. Zhang [50] pre-sented an interesting Machine Learning system for classifying aerial images that exploitsboth geometric and color/texture features extracted from each image. An aerial image suchas a medical image comprises millions of pixels. If we consider each pixel as a local fea-ture, the problem of classifying images becomes computationally intractable. This papershows an interesting feature extraction algorithm that allows a drastic reduction in the prob-lem of image categorization. This algorithm extracts graphlets [49], which are small-sizedgraphs constructed by joining basic components (nodes), each one linked with the otheradjacent nodes. It would be extremely valuable to translate these sophisticated multime-

Multimed Tools Appl

dia techniques, algorithms and protocols to advance the eHealth sector, for instance toimprove diagnosis supported by technology, improve evaluation tests in neuropsychologyor build usable and cheap tools for actively monitoring patient mood (e.g., early detectionof depression).

Searching for information is another critical point for many aspects, including infor-mation accuracy and completeness, different languages and knowledge of eHealth seekersand clinicians, result visibility and navigability. The vocabulary gap between health pro-fessionals and users impedes the efficient retrieval of medical data on the Internet. Nieet al. propose a medical terminology assignment scheme that combines local mining withglobal learning, showing promising performance and scalability potential [36]. Anotherapproach is applied by Bhaskar, who propose a graph-based Question Answering sys-tem that offers Natural Language interaction and restricts output to a small set of focusedresults [8].

Dynamic Clinical Data Mining promises to transform the onerous daily data entries ofclinicians into an effective tool for clinical care, making it possible to perform decisionsbased on collective experience. Celi et al. propose the creation of a system that supportsclinicians in making decisions, detecting similar prior cases in a global anonymous clinicaldatabase in order to suggest interventions and prognosis, based on prior outcomes, fuel-ing the potential of underused data of the electronic medical records [9]. More automaticdisease inference from community-based health services could support users for early self-diagnosis. Recently Nie et al. proposed a deep learning scheme to infer possible diseasesfrom user questions collected by community-based health sites, which showed good perfor-mance in experiments involving dataset labeled by on-line doctors [35]. Furthermore Nieinvestigated the potential of patients health records containing text and multimedia data(digital images, videos, scans, digital measurements). This is especially useful for moni-toring the progression of chronic diseases over time. The proposed adaptive multimodalmulti-task learning model unifies the modality agreement, temporal progression and modal-ity confidences. Experiments on an Alzheimer’s disease dataset confirm the effectivenessof the proposed model [36].

Data sharing, communication and cooperation in the design of usable tools and applica-tions benefits all of society. In this process the work of a multidisciplinary team that canapproach a problem from multiple perspectives is crucial. Interdisciplinary problem-solvingand interoperability are the basis of the application of ICT in the health field, fosteringnew avenues for innovative studies. Overall, implementing sustainable, usable multi-serviceeHealth programs is a complex challenge (technical, financial, organizational) that requiresa critical mass of skilled personnel, time to create and monitor the electronic services, anda concerted effort to build a single shared vision among all stakeholders (patients, clini-cians, patient organizations, caregivers, etc.) [33]. Moreover, HCI helps ensure efficient,high-quality services and favors a high degree of acceptance, meeting patient satisfaction[40]; thus, expected benefits would more than compensate for the initial effort of buildingeHealth programs.

Recent eHealth advances, social media and social networks all play fundamental rolessince the potential of user-generated content and self-care activities are still not fullyexploited. In this study we investigate the use of social media as a tool for supportingresearch in the medical field. We selected Facebook since it is a multimedia tool (comparedto textual microblogs such as Twitter), to reach a young and healthy population and to takeadvantage of the challenge mechanism that may be elicited by social-network friends, asdescribed in the following (score comparison).

Multimed Tools Appl

2.2 Social networks and health

Since the 1970s medical literature has investigated the use of social networks and socialsupport to promote health [5, 7, 12, 17]. Recent studies have investigated the role of onlinesocial networks in the healthcare system [16, 20, 23]. Online health communities enableinformation sharing on issues such as pathologies and treatments, user experiences pro-viding information, suggestions, emotional support and mutual motivation. Social contactsmediated by online social networks may even influence – positively or negatively – an indi-vidual’s health [11, 20, 21, 25]. Social networks promise to be even more useful as aneffective component of weight management programs [29]. Epidemiological studies ofteninvolve a large sample of subjects who fill out self-administered surveys. Currently, thesurvey is on paper or in electronic format to be filled out by means of computers andmobile devices. The traditional paper-based approach is very time-consuming and costly,especially with regard to personnel involved in the recruitment, preparation and submissionof material by post, and collection and storage of data that need to be converted to elec-tronic format. The use of social networks has been recently investigated as a tool for subjectrecruitment in health research, due to its diffusion among young users and its pervasivenessand ubiquity, thanks to the spread of touchscreen mobile devices [1, 29, 32]. Specifically,advantages of using online social media have been addressed by Lohse who compared thecost of a classic recruitment vs an approach based on social networks, finding consider-able cost savings when using the social-network enhanced approach [29]. However, to thebest of authors’ knowledge, a complete case study and guidelines for the design of socialnetwork apps for recruiting large samples of healthy and young population is still lack-ing. The enormous amount of data available on the Internet permits extracting informationfrom electronic sources in different ways, such as: 1) implicitly, discovering information onthe ”social web” by crawling existing data, i.e., forums, data-focused user groups, socialnetworks, and so on, or 2) explicitly, asking user cooperation for completing tasks (suchas Amazon Mechanical Turk, which mediates requests of rewarded small tasks) or requir-ing users’ data (such as web surveys using Google Docs). User consensus is required forthe use of aggregated data for scientific purposes. ”Crowdsourcing is a novel method ofresearch data collection that has the potential to collect real-world experience on treatmenteffectiveness” [3]. The most popular type of crowdsourcing in healthcare research col-lects patients’ symptoms and responses to treatments. Crowdsourcing could be a tool forcollecting huge amounts of data from heterogeneous populations (various ages, countries,cultures, etc). To standardize the gathered data it is important to adopt validated measur-able outcomes. To involve healthcare professionals in the design of a crowdsourcing siteimproves the accuracy of symptom and treatment categories. Data standardization is cru-cial for providing valuable insight into real-world treatment effectiveness [2]. We decided toimplement a Facebook application (henceforth app) to be proposed in the form of a friendly”competition”, instead of crowdsourcing for two reasons: to reach young and healthy users(our main target) since this social network accounts for millions of people who are notmotivated to participate in epidemiological studies, and to respect European laws on dataprivacy and obtain the explicit collaboration of users for providing data (collecting theirexplicit written consent data treatment). Furthermore, concerning user authentication, oneadvantage provided by social networks is the mechanism of federated authentication thatallows an app to use authentication with a social network (such as Facebook and Google+),avoiding the need for authentication in each environment and simplifying the user inter-action. In this way participants need not to change their usual way of interacting and any

Multimed Tools Appl

information of their demographics may be accessed directly by the app (asking for the user’spermission).

2.3 Mediterranean diet and heart

Cardiovascular disease, including heart disease and stroke, is the leading cause of deathand disability in women and men in Western as well in developing countries [31, 47, 48].Cardiovascular disease leads to high direct and indirect medical care costs and is thus acritical medical and social issue. In this context, intensive lifestyle interventions have beenshown to exert beneficial effects on prevention and progression of cardiovascular diseases[6, 10, 15, 22, 28]. It is noteworthy that information on lifestyle is largely lacking, espe-cially for the age range of 15-40 years, since younger people rarely go for regular checkups.This study investigates whether social network, specifically Facebook, could be used as aweb platform suitable for collecting critical research data on lifestyle habits. Potential ben-efits of using this kind of tool include the rapid, cheap enrollment of young and healthysubjects, widely present on social networks and otherwise harder to reach via conventionalsocial health polls, even from different countries, and the storing of data in electronic format(database and content management systems). These are important advantages for epidemi-ological studies involving large and multiethnic population samples. To test this hypothesis,a multidisciplinary team including doctors and computer science researchers was created.The hypothesis is to utilize the Facebook platform for creating and disseminating four ques-tionnaires, each related to fundamental aspects of quality of life: diet, exercise, sleep andstress. To verify the feasibility of this proposal, we have developed the first questionnaireconcerning adherence to the Mediterranean diet, a basic element of a good quality of life,considering the significant risk of cardiovascular and neurological degenerative diseases insubjects with poor adherence to such a diet [15, 19, 24]. The test consists of questions aboutparticipants’ eating habits, taking into account the main foods in the Mediterranean diet,their frequency of consumption, and their proportions, as reported in the ”MediterraneanDiet Pyramid” [43].

3 Methods

Usability of a system, artifact, or software (SW) consists of its ease of use and learnability; itis the dominant factor in offering a fully satisfying experience for anyone [37, 38]. The ISO9241 defines usability as the effectiveness, efficiency and satisfaction with which specifiedusers achieve specified goals in particular environments [27]. In order to develop a moreusable application, the questionnaire was designed applying participatory design principles.Participatory design aims to fill the gap between designers and the users of the application,in order to create a product that closely adheres to the users’ needs [42]. One of the mainproblems of software designers using the classic SW design approach is defining a productaccording his/her own conceptual model which is usually different from those conceivedby users [39]. User-centered design puts user needs and abilities at the center of design,focusing on their expectations and removing extraneous or difficult elements. For thesereasons, we involved young people with a Facebook account in the design and test phases. Inall, ten individuals were involved, eight of them in the age range 18-40 years. Specifically,two males and two females under 40 years of age actively cooperated with us in the designphase. The feedback received helped us to make the user interfaces of the questionnairesimpler and more intuitive.

Multimed Tools Appl

3.1 The application requirements

Social networks were investigated to be used as a tool for easily and rapidly gathering dataof healthy users, at low cost. The proposed approach, based on a popular social network,involves a dynamic population, i.e., new subjects join the study over time while some maybe lost in follow-up. The overall system is composed of four applications, each one relatedto an important factor influencing quality of life:

– Mediterranean diet– Exercise– Sleep quality– Stress.

The four questionnaires will be launched at different times, possibly one every 6 monthsor 1 year. After completing the first questionnaire, the user will be contacted to invitehim/her to fill out the next ones. At this time, he/she will also be invited to repeat the filling-out of the first questionnaire. This will be repeated for each new questionnaire, collectingdata for the follow-ups, as shown in Fig. 1. After the four questionnaires are completed, thewhole cycle could be repeated in the future, in a way that appeals to the user, for instancechanging the user interface and modifying user feedback. At this time, only the first piece ofthe ’life-style puzzle’ – the Mediterranean Diet – has been implemented. We have focusedon the strategy to encourage subjects’ enrollment:

– capturing individuals’ attention– asking users to fill-in an online questionnaire– providing useful feedback in order to gain user trust and loyalty– asking users to share the questionnaire result on their wall (page where users’ content

are displayed), to publicize and promote the application.

The questionnaire is available in six languages: English, French, German, Italian,Portuguese and Spanish.

3.2 Mediterranean diet score

A score from 0 to 10 was assigned to the results. Adherence to the diet was defined as low(L = 0 − 5); medium (I = 6 − 8), and high (H ≥ 8) derived from the work of Sofi [43] andnormalized between 0-10 for easy comprehension of the user. After analyzing the results,we provided participants with information on their adherence to a Mediterranean diet andoffered suggestions for improving their alimentary habits. The feedback that we provided

Fig. 1 a) Factors affecting lifestyle; b) Temporal diagram of the questionnaire releases

Multimed Tools Appl

could be published on the user’s wall, promoting the use of the questionnaire by his/herfriends.

4 User interfaces

An important point concerns the user interface, and specifically the visual design of allgraphic components of the applications, which were developed to guarantee an easy, pleas-ant and effective interaction environment. ”The user interface should make it easy for usersto contribute. This is highly non-trivial” [14]. Simple and concise language was adopted toclearly formulate the questions and avoid any ambiguity. A pilot test with six individuals(three male and three female, four of them under 40 years old) not involved in the designphase was carried out to assess the usability of the questionnaire and the preferred layout.Two versions of the user interfaces were created and tested. The first version was based onimages, while the second was only textual, organized in a tabular format. Data from thispilot test has shown that young participants prefer the visually based version, with photosof elements related to the questions, but two older users (one between 50 and 60 years,and one over 60 years) experienced some disorientation upon opening the picture to selectthe answers. Thus, we decided to adopt the simple tabular model to avoid confusion andimprove usability for all users. We plan to re-propose the graphic model of the question-naire in future follow-ups (planned t2, t3 and t4 in Fig. 1), when the user is more familiarwith the questions and elements of ambiguity are reduced.

The app was designed to enable rapid filling out the questionnaire, so questions andanswers are organized to minimize the need to scroll down the user interface. In order tomake the app usable on all devices including those with small screens, interaction via tabletand smartphone has been taken into account for optimizing weight and size of the images,however implementing only one version of the application (simplifying the maintenance).When using handheld devices, image loading may degrade the apps performance, increasingthe time it takes to visualize the page.

The logo of the app was elaborated in a heart shape, recalling sea and sun, key elementsof the Mediterranean region. For each of the four lifestyle components the logo has beencustomized, adding an indicative name on the upper right corner of the heart.

After the user fills out the Mediterranean diet questionnaire (Fig. 2), the results regard-ing their adherence to the diet are provided. In order to promote healthy foods that arescarce or absent in the user’s diet, the user feedback is enriched with graphical elementsrelated to the missing food. The app implementing the questionnaire, promotes healthymissing elements (except alcohol although a small quantity of it is planned in the diet) in acustomized way.

4.1 Social virality

In order to collect epidemiological data, we need to reach as many Facebook usersas possible. We would like to make our application viral to be able to recruit manyusers but without spending money. We tried to take advantage of the social networkmechanisms: social networks are designed to share relations and activities betweenfriends or peers; thus the Mediterranean diet app proposes publicizing the results ofthe questionnaire (the score obtained and some suggestions) on one’s Facebook Wall,sharing and promoting it to friends, also stimulating competition through the resultscomparison.

Multimed Tools Appl

Fig. 2 Main user interface: ten questions on food habits and one on alcohol consumption

5 The facebook application

The MedDiet app is a web page delivering the questionnaire. The Facebook authenticationis necessary for reaching the survey. Figure 2 shows the main app user interface. There arefour logical sections:

– The aim of the research, informing the user about the questionnaire’s goal andrequesting their collaboration. A link allows one to reach a web page with more details

– Average weekly frequency of assumption– Elements of the Mediterranean diet– User answers.

The graphic form of the questionnaire is compact in order to facilitate data insertion.User choices are highlighted with a dark tone of color. Additional personal info is requiredin the bottom part of the main user interface of the questionnaire (Fig. 3). These data areuser country, age, weight and height, useful for better analyzing the data collected. The firstfew lines explain the need for this data and specify their anonymous treatment, according tothe Italian law.

Each Facebook app has its own page that can be personalized by adding a cover photo.Our page provides information on the study and a link to the questionnaire; the people who”like” the App number more than 800.

Multimed Tools Appl

Fig. 3 Personal information required by the MedDiet app for calculating user bmi and dealing with privacy

5.1 Loyalty

The pages of the results present the score obtained from the answers given to the question-naire, the range of adherence to the Mediterranean diet and some suggestions, as shown inFig. 4. Furthermore, in the last user interface, there is a ”thank you” to the user and theannouncement of the future questionnaires, to familiarize the participant with the wholeresearch program (exercise, sleep, and stress surveys).

5.2 Programming environment

Facebook offers developers Application Programming Interfaces (APIs) for enrichingwebsites (with features such as ’I like’) and for creating apps with Facebook’s native func-tionalities such as Requests and Bookmarks. Facebook Login makes it easy to connect withusers on apps or websites. It is possible to use several methods in JavaScript, PHP or mobileSDKs to speed up the registration process and quickly build a functional system. The GraphAPI for interacting with the Facebook website offers a simple view of the Facebook socialgraph, uniformly representing objects in the graph and the connections between them. Face-book Query Language (FQL) enables one to use an SQL-style interface to query the dataexposed by the Graph API. In the near future we plan to substitute the FQL language withGraph API due to Facebook policy changes. One drawback of using native APIs offered byFacebook is the maintenance of the app, since the APIs change over time, making it neces-sary to update the code of the application. Last, the Facebook native Multilanguage supportfacilitates the spread of the app in different languages. Currently the Mediterranean diet

Multimed Tools Appl

Fig. 4 Mediterranean Diet app: an example of results. Since the adherence is low, some missing foods aresuggested

app is available in six languages: Italian, French, English, Portuguese, Spanish and German.However, we have not yet carried out a plan for diffusion outside Italy.

6 Results

Preliminary results of this study have shown the feasibility of the proposal, i.e., using asocial network to gather health data for scientific purposes (second hypothesis). A total of1,414 subjects (976 female; age range 13-72, mean score 5.96; 438 male; age range 12-77,mean score 5.77) filled out the questionnaire.

To favor participation in the Mediterranean diet app at the beginning of the design phase,we chose to minimize the data entry, extracting data from the Facebook user profile (birth-day, city, country, etc.) and browser info (language). Actually, after analyzing collected datawe verified that not all participants shared their age in the Facebook profile, so the analysisof score related to participant’s age is restricted to a smaller data set. To collect a completedata set, we added the age field in the user interface, as shown in Fig. 5. The smaller data setincluded 994 users (688 female; age range 13-72 years, 326 male, age range 12-77 years).

Female prominence in nutrition research programs is in accord with [32]. Men are a hard-to-reach population [46]. According to the age clusters (under 35, 35-50, and over 50), it isnoteworthy that our population was predominantly younger than 50 (over 80 % of the totalpopulation) as opposed to the population enrolled in the larger cohort study of the Mediter-ranean diet in which 57 % of subjects were younger than 55 years [44]. This result fulfillsone aim of this study, which is the enrollment of younger subjects (second hypothesis).This is relevant in the light of primordial and primary prevention, where primordial pre-vention is defined as prevention of the development of risk factors, and primary preventionis designed to modify adverse levels of risk factors once present, with the goal of prevent-ing cardiovascular events [47]. Furthermore, the importance of primordial prevention arises

Multimed Tools Appl

Fig. 5 Adherence to the Mediterranean diet: Italian (blue vertical pattern) vs other participants (red parallelpattern). From left: Low, Medium and High adherence

from evidence that chronic diseases such as atherosclerosis and other degenerative diseasesaffect people at an earlier age [13]. In fact, it is accepted that the risk of many chronicdiseases is not just determined by risk factors in mid-adult life, but begins in childhoodor adolescence [48]. In this context, our approach also provides straightforward promo-tion of lifestyle education and information through a direct feedback mechanism, takinginto account that adherence to modifiable healthy lifestyle factors can substantially reducepremature mortality in women and men [45].

Regarding nationality, Italians were the majority (88 % of the total sample, correspond-ing to 84 % of the female population, and 91 % of the male population), although peoplefrom all over the world filled out the questionnaire (12 % of the total sample, correspond-ing to 16 % of total female population, and 9 % of male population). The opportunity toenroll multiethnic subjects, allowing population comparison studies on lifestyle habits, isa potential of Facebook that deserves to be further investigated. This is relevant in lightof the different incidence of cardiac events across populations, e.g., South Asians have ahigher prevalence of coronary heart disease and cardiovascular mortality than Europeans[18]. The global mean score and the score in the three age clusters were similar betweenfemales and males, although females showed a trend toward higher scores, particularly ata younger age. These results are consistent with data present in the literature [43], thusproving the feasibility of the proposed approach. Concerning the accuracy of gathered data,the level of trust of the answers is analogous to computer- or paper-based surveys carriedout by users, without supervision by clinical personnel. However, the influence of absenceof direct contact with the doctor should be further investigated, since it impacts on dataaccuracy.

Concerning the difference between countries, an initial difference emerges since themean score of the Italians was 6.03, which fits in as a medium adherence to the Mediter-ranean diet, compared to a score of 5.28 for the other participants, falling into the top of the

Multimed Tools Appl

Table 1 Chi-square test for gender and score. N = 1414, (p>0.1) shows no differences in Mediterraneandiet scores related to gender

Cases

Valid Missing Total

N Percent N Percent N Percent

Gender ∗ Score 1414 100.0 % 0 0.0 % 1414 100.0 %

Value df Asymp.

Sig. (2-sided)

Pearson chi-square 3.286 2 0.193

Likelihood ratio 3.274 2 0.195

N of valid cases 1414

low adherence. Figure 5 details the degree of adherence to the Mediterranean diet of Italianscompared to other participants.

6.1 Data validity

A Chi-Square test was performed to test the significance of the sample scores and verifywhether differences exist in adherence to the Mediterranean diet related to gender, age andgeographic location of participants. A significant difference between the expected and theobserved frequencies in males and females of the overall sample (N = 1414) was not con-firmed (χ2 =3.286; d.f.=2 p < 0.19), as shown in Table 1. Moreover, the overall sample (N= 1414) showed a significant difference related to the Mediterranean population vs othercountries i.e., Italian (representing the Mediterranean area) vs other countries (χ2 =6.851;d.f.=2, p < 0.05), as shown in Table 2.

To understand whether any relation exists between participants’ age and their adherenceto the Mediterranean diet, another Chi-Square analysis was performed on the reduced dataset (age clusters: <35, 35-50, >50; score clusters: low, medium, high). The test confirmed

Table 2 Chi-square test for Italian participants and those from other countries. N = 1414, (p< 0.05) indi-cates significant differences in Mediterranean diet scores correlated to geographic areas (Mediterranean vsothers)

Cases

Valid Missing Total

N Percent N Percent N Percent

Local ∗ Score 1414 100.0 % 0 0,0 % 1414 100.0 %

Value df Asymp.

Sig. (2-sided)

Pearson chi-square 6.851 2 0.033

Likelihood ratio 6.778 2 0.034

N of valid cases 1414

Multimed Tools Appl

Table 3 Chi-square test for age and score. N = 994, (p < 0.01) indicates significant differences inMediterranean diet scores related to age clusters

Cases

Valid Missing Total

N Percent N Percent N Percent

Age ∗ Score 994 100.0 % 0 0.0 % 994 100.0 %

Value df Asymp.

Sig. (2-sided)

Pearson chi-square 17.564 4 0.002

Likelihood ratio 18.103 4 0.001

N of valid cases 994

that there is a real difference in healthy diet habits, related to participant age (χ2 =17.564;d.f.=4, P< 0.01) in the restricted sample (N = 994), as illustrated in Table 3.

6.2 Data accuracy

To fully verify the second hypothesis of our study, (i.e., the accuracy of the collecteddata) we have to compare our results with large accurate studies performed with classicapproaches such as expensive paper-based surveys performed in a clinical environment. Werestricted this analysis to Italian subjects who were the majority of participants. Six hundredand three Italian participants specified height and weight. Figure 6 shows the distribution ofItalian participants bmi (body mass index) in comparison with nutritional indexes.

The curves delimit respectively, from the bottom to the top, the five areas correspondingto Underweight, Normal Range, Overweight, Obese, and Severely Obese. We consideredthe following value ranges: Underweight <18.5; Normal range 18.5-24.9; Overweight 25.0-29.9; Obese 30-34.9; Severely Obese ≥35. These data show some accumulation points,possibly due to participants rounding off their height to the next value. This factor might

Fig. 6 The bmi distribution of participants (N = 603). From the button to the top, the five areas of Under-weight (under the blue curve), Normal range (from the blue to the green curves), Overweight (from the greento the yellow curves), Obese (from the yellow to the red curves), Severely Obese (up the red curve)

Multimed Tools Appl

Fig. 7 The bmi distribution of participants aggregated by low, medium and high scores of Mediterraneandiet adherence. From left: Underweight (blue parallel pattern), Normal range (green vertical pattern);Overweight (yellow diagonal pattern); Obese-Severely Obese (red diagonal pattern)

imply an underestimation of overweight and obese participants, as discussed in the nextsection.

Figure 7 aggregates data by the scores of adherence to the Mediterranean Diet. We canobserve that individuals with medium and high scores are in lower underweight, overweightand obese percentages than participants with a low Mediterranean Diet adherence score.As internationally acknowledged, this alimentary regime benefits users in several ways,including keeping in physical shape and favoring an optimum bmi index.

According to Epicentro (National Center of Epidemiology, Surveillance and Health Pro-motion, http://www.epicentro.iss.it/), in 2012 (participants n = 151,603) 35.6 % of Italianadults were overweight and 10.4 % were obese. In our experiment we found 25.7 % of par-ticipants to be overweight and 8.0 % obese. This result does not contradict our findingssince our sample includes a high percentage of young people (51 % under 35 years of age,34 % between 35 and 50 years of age and only 15 % over 50 years of age). Consideringthat the percentage of population overweight increases with age: in Italy, being overweightmoves from 15.8 % in the age group 18-24 years to 45.8 % among those age 65-74 years,while obesity increased from 2.8 % to 15.9 % for the same groups. Furthermore, our sampleincludes a high number of women (61.2 %), which further decreased the overweight per-centage (in Italy overweight and obese men are respectively 39 % and 11 % vs 24 % and 10% of women).

Overall, the preliminary results of our small, economical study seem to be coherent withthose of a large, expensive study. However, more structured studies are needed to fullyevaluate the potential of social networks as an epidemiological tool.

7 Discussion

The case in study allows us to understand the advantages and limits of the proposed solu-tion. Few limitations emerged from our experience. First of all, the initial network heavily

Multimed Tools Appl

influences the data collection, especially with respect to country and age, since friends aregenerally clustered in one or few countries, and friends are more common between peers.In the collected sample the Italian population dominates, and the average age is 35.7 years.Thus, the younger population that the authors had access to could be the result of theinitial network as well. First, after an initial peak, the self-diffusion of the questionnaire,was decreasing over time. This limitation could be overcome in two ways: 1) buying a spe-cific advertising program to sustain and increase participant recruitment, 2) repeating theinvitation to new friends and friends of friends or contacting categories such as schools,associations etc. To obtain multi-ethnic subjects, it would be appropriate to launch theapp in different countries at the same time. Since students represent the most conspicu-ous young population, to be reached via social networks educational cooperation programswith schools could be initiated. Furthermore, in contrast with [32] where authors adoptedan advertising program to recruit a high number of participants in a few days, we only usethe free available features of Facebook, such as publishing results on the wall and invitingfriends to play the MedDiet app. A more ”aggressive advertising” policy for the app dif-fusion is suitable for enrolling a larger sample of subjects. In our project, we attempted todisseminate the questionnaire cost-free. Another focal point concerns the validity of the col-lected data. Data accuracy depends on user precision in inserting data such as age, weightand height, so user attention has to be called to this point, as a reminder. Data such asweight and height, inserted by nurses or clinicians, might be more reliable than those spec-ified by users. Without a clear vision of the meaning and value of input data the user mightapproximate these values, increasing statistical error in the data analysis. Thus it is vitalto emphasize to users the accuracy of input data, by providing some kind of reinforce-ment, such as a graph of their bmi over time. Based on lessons learnt and experience someguidelines can be provided to designers of epidemiological studies, polls, and campaigns:

– Design the whole app with health professionals (participatory design)– Define standard data format and score for required data with health professionals– Make the page/form visually appealing and fully usable (provide simple interaction,

comprehensible content, minimize the number of clicks, avoid user errors, make themvisible and easily recoverable, etc.)

– Make the app robust and available (reliability is crucial so as not to lose users)– Inform the user about the aim of the research and emphasize the importance of accuracy

of input data (age, weight, height)– Reward the user: provide a reinforcement for making the app socially interesting, in

form of scores, visual feedback, info, bmi progress over time; etc.– Build user loyalty (if any follow-ups)– Manage security and privacy (set up secure connection, keep data safe, inform user on

data treatments, and user rights, etc.)– Get user consent for data collection, anonymize data and aim of the study– Perform pilot test before launching the app– Introduce into the app the automatic request for sending an invitation to friends– Promote the application through other social networks such as Twitter, LinkedIn,

Google+, Academia, and so on– Set up specific agreements with associations or Ministries for spreading the survey (for

instance, setting up specific agreements with the Education Ministry for diffusion inschools/universities).

Another concern could be the security and privacy of the acquired personal data. Regard-ing privacy, we manage personal data in an anonymized way. Concerning this project,

Multimed Tools Appl

all data collected have been processed anonymously for scientific purposes, and specificconsent from the users was required. Italian law (Legislative Decree n. 30 June 2003, n.196) in fact explicitly requires obtaining prior consent. The questionnaire provides the userwith this information and requires a check box to be clicked in order to authorize pro-cessing of this data. In order to analyze data results over time, each Facebook user ischaracterized by a unique identifier that links the user with their data profile. This iden-tifier allows choosing from multiple answers on the same questionnaire (if present in thedatabase), selecting the most recent (also in a fixed range of time) for statistical pro-cessing. Subsequent versions by a user of the same questionnaire can be compared forthe follow-up campaign (after 1 year or more). The European Community has defined acode of practice called Safer Social Networking Principles for the EU, signed on Febru-ary 10, 2009, adopted by twenty of the largest providers, including Facebook (EU, 2009).Furthermore, the revision of the European legal framework on data protection is alreadydefined. On January 25, 2011 the European Commission presented the new Regulationson the protection of personal data, for ensuring the practical application of key rulesand principles (replacing Directive 95/467CE) [26]. The new regulations on data protec-tion are composed of nearly 100 articles, binding the countries on all aspects of dataprocessing, from health to sensitive information, data banks, the right of access and theright to oblivion. Particular attention is devoted to data protection on the Internet with-out the specific consent of the user, if they are used to profile behavior on the net formarketing purposes.

8 Conclusion

Human-computer interaction is now central and pervasive in users’ lives. Social networksfavor user collaboration, fueling feedback and idea-sharing, and have the potential to serveas a base for building scientific applications and research. The collective intelligence hasthe potential to improve knowledge, processes and data. Especially in healthcare, the pos-sibility of distributing tasks to users, such as self-monitoring and self-care, is important forempowering users, increasing awareness and improving behaviors in order to prevent ordelete the onset of disease and overall personal and social well-being.

In this study we investigated the use of social networks as a tool with a double aim: tocollect data from healthy and young users and to model their behavior. Our first hypothesisis supported by the results of this preliminary study. Social networks can be used as a toolto enroll subjects and archive critical data, greatly reducing the time needed to enroll sub-jects, and automating the storage, processing and analysis of collected data. Social networkswould also enable the collection of data from young and healthy subjects, who are otherwisehard to reach with the use of conventional social health polls. In addition, sharing infor-mation on social networks can attract an increasing number of participants daily, with verylittle effort. However, limitations have been experienced concerning the second hypothesissince both age and geographic distribution of initial participants may influence the result.This is obvious since main social network relations reflect personal or friend links, follow-ing a ”leopard spots model”. This suggests that for gathering data from all over the worldand from young and healthy subjects it is important to distribute initial network points indifferent countries, possibly reaching schools/universities. The secondary and more ambi-tious objective is improving in lifestyle and habits of the sample, now based on feedbackprovided by the Facebook applications developed. For this, it would be necessary to intro-duce mechanisms for user loyalty, such as additional apps released at consecutive times.

Multimed Tools Appl

Monitoring users, especially in the case of low scores, may require asking them to repeatthe tests over time to check for any changes in lifestyle.

From this experience, several values emerge: the extensibility of the model to other areas,with its advantages of low cost, easy maintenance, and rapidity and wide representation ofthe sample. This merits further investigation. As future work, we plan to complete the fourapplications in order to verify other important aspects of lifestyle. Furthermore, a massivecampaign using different media such as press communications, agreement with the educa-tion minister, or a Facebook advertising program could be explored to boost the spread ofthe application and increase the sample by at least one order of magnitude.

Additional research is needed to exploit social networks to create an effective multidi-rectional communication channel, involving peer education and creating collaborative tools,in order to model young peoples’ behavior toward healthy attitudes. Our experience wasvaluable for understanding how to tune the data gathered from social networks. Prelimi-nary results suggest further exploration of the potential of social networks in the healthcaresector. As future research, we suggest developing this kind of study by increasing user feed-back, to enable self-monitoring of individual health status and increase wellbeing. For thisit is necessary to increase app functions and change the user interface, by taking advan-tage of several techniques including gamification, persuasive interfaces and principles ofapplied behavior analysis. A multidisciplinary approach including psychologists, clinicians,ICT scientists, and users to implement a participatory design approach is suggested to max-imize the app’s usability – making interaction simple, fast and easy to use, providing a fullysatisfying experience for anyone.

References

1. Amerson R (2011) Facebook: A tool for nursing education research. J Nurs Educ 50(7):414–416.doi:10.3928/01484834-20110331-01

2. Armstrong AW, Harskamp CT, Cheeney S, Schupp CW (2012) Crowdsourcing for research datacollection in rosacea. Dermatology Online Journal 18(3):15. http://www.ncbi.nlm.nih.gov/pubmed/22483526

3. Armstrong AW, Wu J, Harskamp CT, Cheeney S, Schupp CW (2013) Crowdsourcing for data collection:a pilot study comparing patient-reported experiences and clinical trial data for the treatment of seborrheicdermatitis. Skin research and technology : official journal of International Society for Bioengineeringand the Skin (ISBS) [and] International Society for Digital Imaging of Skin (ISDIS) [and] InternationalSociety for Skin Imaging (ISSI) 19(1):55–7. doi:10.1111/j.1600-0846.2012.00667.x. http://www.ncbi.nlm.nih.gov/pubmed/22891700

4. Arnrich B, Mayora O, Bardram J, Troster G (2009) Pervasive healthcare. Methods Inf Med.doi:10.3414/me09-02-0044

5. Barabsi AL (2007) Network medicine from obesity to the diseasome. N Engl J Med 357(4):404–407.doi:10.1056/NEJMe078114. PMID: 17652657

6. Behrens G, Fischer B, Kohler S, Park Y, Hollenbeck A, Leitzmann M (2013) Healthy lifestyle behaviorsand decreased risk of mortality in a large prospective study of u.s. women and men. Eur J Epidemiol28(5):361–372. doi:10.1007/s10654-013-9796-9

7. Berkman LF, Syme SL (1979) Social networks, host resistance, and mortality: a nine-year follow-up study of alameda county residents. Am J Epidemiol 109(2):186–204. http://aje.oxfordjournals.org/content/109/2/186.abstract

8. Bhaskar P, Buzzi M, Geraci F, Pellegrini M (2015) From literature to knowledge: Exploiting pubmed toanswer biomedical questions in natural language. In: Renda ME, Bursa M, Holzinger A, Khuri S (eds)Information Technology in Bio- and Medical Informatics, Lecture Notes in Computer Science, vol 9267.Springer International Publishing, pp 3–15. doi:10.1007/978-3-319-22741-2 1

9. Celi LA, Zimolzak AJ, Stone DJ (2014) Dynamic clinical data mining: Search engine-based decisionsupport. JMIR Medical Informatics 2(1):e13. doi:10.2196/medinform.3110

Multimed Tools Appl

10. Chainani-Wu N, Weidner G, Purnell DM, Frenda S, Merritt-Worden T, Pischke C, Campo R, KempC, Kersh ES, Ornish D (2011) Changes in emerging cardiac biomarkers after an intensive lifestyleintervention. Am J Cardiol 108(4):498–507. doi:10.1016/j.amjcard.2011.03.077

11. Chou WYS, Beckjord EB, Moser RP, Hesse BW (2009) Social media use in the united states:Implications for health communication. J Med Internet Res 11(4):e48. doi:10.2196/jmir.1249

12. Christakis NA (2004) Social networks and collateral health effects. BMJ 329(7459):184–185.doi:10.1136/bmj.329.7459.184

13. Darnton-Hill I, Nishida C, James W (2004) A life course approach to diet, nutrition and the preventionof chronic diseases. Public Health Nutr 7(1a). doi:10.1079/phn2003584

14. Doan A, Ramakrishnan R, Halevy AY (2011) Crowdsourcing systems on the world-wide web. CommunACM 54(4):86–96. doi:10.1145/1924421.1924442

15. Estruch R, Ros E, Salas-Salvado J, Covas MI, Corella D, Aros F, Gomez-Gracia E, Ruiz-GutierrezV, Fiol M, Lapetra J, Lamuela-Raventos RM, Serra-Majem L, Pinto X, Basora J, Munoz MA, SorlıJV, Martınez JA, Martınez-Gonzalez MA (2013) Primary prevention of cardiovascular disease with amediterranean diet. N Engl J Med 368(14):1279–1290. doi:10.1056/nejmoa1200303

16. Farmer AD, Bruckner Holt CE, Cook MJ, Hearing SD (2009) Social networking sites: a novel portal forcommunication. Postgrad Med J 85(1007):455–459. doi:10.1136/pgmj.2008.074674

17. Fleury J (1992) An exploration of the role of social networks in cardiovascular risk reduction. Heart &Lung: J Crit Care 22(2):134–144

18. Forouhi N, Sattar N, Tillin T, McKeigue P, Chaturvedi N (2006) Do known risk factorsexplain the higher coronary heart disease mortality in south asian compared with europeanmen? prospective follow-up of the southall and brent studies, uk. Diabetologia 49(11):2580–2588.doi:10.1007/s00125-006-0393-2

19. Gardener S, Gu Y, Rainey-Smith SR, Keogh JB, Clifton PM, Mathieson SL, Taddei K,Mondal A, Ward VK, Scarmeas N, Barnes M, Ellis KA, Head R, Masters CL, Ames D,Macaulay SL, Rowe CC, Szoeke C, Martins RN (2012) Translational Psychiatry 2(10):e164.doi:10.1038/tp.2012.91

20. Greene J, Choudhry N, Kilabuk E, Shrank W (2011) Online social networking by patients with dia-betes: A qualitative evaluation of communication with facebook. J Gen Intern Med 26(3):287–292.doi:10.1007/s11606-010-1526-3

21. Griffiths KM, Calear AL, Banfield M (2009) Systematic review on internet support groups(ISGs) and depression (1): Do ISGs reduce depressive symptoms? J Med Internet Res 11(3):e40.doi:10.2196/jmir.1270

22. Hamer M, Stamatakis E (2009) Physical activity and mortality in men and womenwith diagnosed cardiovascular disease. Eur J Cardiovasc Prev Rehabil 16(2):156–160.doi:10.1097/hjr.0b013e32831f1b77

23. Hawn C (2009) Take two aspirin and tweet me in the morning: How twitter, facebook, and other socialmedia are reshaping health care. Health Affairs 28(2):361–368. doi:10.1377/hlthaff.28.2.361

24. Hoevenaar-Blom MP, Nooyens ACJ, Kromhout D, Spijkerman AMW, Beulens JWJ, van derSchouw YT, de Mesquita BB, Verschuren WMM (2012) Mediterranean style diet and 12-year incidence of cardiovascular diseases: The EPIC-NL cohort study. PLoS ONE 7(9):e45,458.doi:10.1371/journal.pone.0045458

25. Hoey LM, Ieropoli SC, White VM, Jefford M (2008) Systematic review of peer-support programs forpeople with cancer. Patient Educ Couns 70(3):315–337. doi:10.1016/j.pec.2007.11.016

26. Implementation of the safer social networking principles for the eu. Accessed: 2015-12-0427. Iso 9241-11: Ergonomic requirements for office work with visual display terminals (vdts) - part 11 :

Guidance on usability (1998)28. Leon AS, Connett J, Jacobs DR, Rauramaa R (1987) Leisure-time physical activity levels and risk of

coronary heart disease and death: The multiple risk factor intervention trial. JAMA 258(17):2388–2395.doi:10.1001/jama.1987.03400170074026

29. Li JS, Barnett TA, Goodman E, Wasserman RC, Kemper AR (2012) Approaches to the preventionand management of childhood obesity: The role of social networks and the use of social media andrelated electronic technologies: A scientific statement from the american heart association. Circulation127(2):260–267. doi:10.1161/cir.0b013e3182756d8e

30. Liu G, Yan Y, Ricci E, Yang Y, Han Y, Winkler S, Sebe N (2015) Inferring painting style with multi-taskdictionary learning. In: 0001 QY, Wooldridge M (eds) Proceedings of the Twenty-Fourth InternationalJoint Conference on Artificial Intelligence, IJCAI 2015, Buenos Aires, Argentina, July 25-31, 2015, pp2162–2168. AAAI Press. http://ijcai.org/papers15/Abstracts/IJCAI15-306.html

31. Lloyd-Jones D, Adams RJ, Brown T. M, Carnethon M, Dai S, Simone G. D, Ferguson T. B,Ford E, Furie K, Gillespie C, Go A, Greenlund K, Haase N, Hailpern S, Ho PM, HowardV, Kissela B, Kittner S, Lackland D, Lisabeth L, Marelli A, McDermott MM, Meigs J,Mozaffarian D, Mussolino M, Nichol G, Roger VL, Rosamond W, Sacco R, Sorlie P, Stafford R,

Multimed Tools Appl

Thom T, Wasserthiel-Smoller S, Wong ND, Wylie-Rosett J (2009) Heart disease and stroke statis-tics – 2010 update: A report from the american heart association. Circulation 121(7):e46–e215.doi:10.1161/circulationaha.109.192667

32. Lohse B (2013) Facebook is an effective strategy to recruit low-income women to online nutritioneducation. J Nutr Educ Behav 45(1):69–76. doi:10.1016/j.jneb.2012.06.006

33. McNelly RA, Latifi R, Barker G (2008) Integrating telemedicine and telehealth: Putting it all together.Current principles and practices of telemedicine and e-Health 131:23

34. Meingast M, Roosta T, Sastry S (2006) Security and privacy issues with health careinformation technology. In: 2006 International Conference of the IEEE Engineering inMedicine and Biology Society. Institute of Electrical & Electronics Engineers (IEEE).doi:10.1109/iembs.2006.260060

35. Nie L, Wang M, Zhang L, Yan S, Zhang B, Chua T (2015) Disease inference from health-related questions via sparse deep learning. IEEE Trans Knowl Data Eng 27(8):2107–2119.doi:10.1109/TKDE.2015.2399298

36. Nie L, Zhang L, Yang Y, Wang M, Hong R, Chua T (2015) Beyond doctors: Future health predictionfrom multimedia and multimodal observations. In: Proceedings of the 23rd Annual ACM Confer-ence on Multimedia Conference, MM ’15, Brisbane, Australia, October 26 - 30, 2015, pp 591–600.doi:10.1145/2733373.2806217

37. Nielsen J (1993) Usability engineering. Morgan Kaufmann Publishers Inc., San Francisco38. Nielsen J Designing web usability: The practice of simplicity. dl.acm.org (1999). Query date: 2015-10-1639. Norman D. A. (2002) The design of everyday things. Basic Books, Inc., New York40. Rocker C, Ziefle M, Holzinger A (2014) From computer innovation to human integration: Current trends

and challenges for pervasive health technologies. In: Holzinger A, Ziefle M, Rocker C (eds) PervasiveHealth. Springer, London, pp 1–17. doi:10.1007/978-1-4471-6413-5 1

41. Sartori A, Yan Y, zbal G, Salah AAA, Salah AA, Sebe N (2015) Looking at mon-drian’s victory boogie-woogie: What do i feel?. In: 0001 QY, Wooldridge M (eds)Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence,IJCAI 2015, Buenos Aires, Argentina, July 25-31, 2015, pp. 2503–2509. AAAI Press.http://ijcai.org/papers15/Abstracts/IJCAI15-355.html

42. Schuler D, Namioka A (1993) Participatory design: perspectives on systems design. L. ErlbaumAssociates Inc.

43. Sofi F, Cesari F, Abbate R, Gensini GF, Casini A (2008) Adherence to mediterranean diet and healthstatus: meta-analysis. BMJ 337. doi:10.1136/bmj.a1344

44. Trichopoulou A, Costacou T, Bamia C, Trichopoulos D (2003) Adherence to a mediterranean diet andsurvival in a greek population. N Engl J Med 348(26):2599–2608. doi:10.1056/NEJMoa025039. PMID:12826634

45. van den Brandt PA (2011) The impact of a mediterranean diet and healthy lifestyle on prematuremortality in men and women. Am J Clin Nutr 94(3):913–920. doi:10.3945/ajcn.110.008250

46. Vandelanotte C, Caperchione CM, Ellison M, George ES, Maeder A, Kolt GS, Duncan MJ, KarunanithiM, Noakes M, Hooker C, Viljoen P, Mummery WK (2013) What kinds of website and mobile phone–delivered physical activity and nutrition interventions do middle-aged men want? J Health Commun18(9):1070–1083. doi:10.1080/10810730.2013.768731

47. Weintraub WS, Daniels SR, Burke LE, Franklin BA, Goff DC, Hayman LL, Lloyd-Jones D, PandeyDK, Sanchez EJ, Schram AP, Whitsel LP (2011) Value of primordial and primary prevention for cardio-vascular disease: A policy statement from the american heart association. Circulation 124(8):967–990.doi:10.1161/cir.0b013e3182285a81

48. Who J, Consultation FE (2003) Diet, nutrition and the prevention of chronic diseases. World HealthOrgan Tech Rep Ser 916(i-viii)

49. Zhang L, Gao Y, Xia Y, Lu K, Shen J, Ji R (2014) Representative discovery of structurecues for weakly-supervised image segmentation. IEEE Transactions on Multimedia 16(2):470–479.doi:10.1109/TMM.2013.2293424

50. Zhang L, Han Y, Yang Y, Song M, Yan S, Tian Q (2013) Discovering discriminativegraphlets for aerial image categories recognition. IEEE Trans Image Process 22(12):5071–5084.doi:10.1109/TMM.2013.2293424

Multimed Tools Appl

Maria Claudia Buzzi National Council Research (CNR), Institute of Informatics and Telematics (IIT).Research focus on Web, eHealth, eGovernment and user interface design. She has coordinated severalprojects, like the infrastructure of the Italian Health Electronic Records. Member of Program Committees ofmany International Conferences, she has published more than 50 scientific papers. She is also involved in themanagement of the Italian W3C Office. http://www.iit.cnr.it/claudia.buzzi.

Marina Buzzi National Council Research (CNR) Head of the Web Application for the Future Inter-net Lab, at Institute of Informatics and Telematics (IIT). Research activities on Web usability and IoT,eHealth. Coordination of technological and research projects. She follows activities of the W3C and IETFgroups. Professional member of ACM and SIGACCESS since 2004. Co-author of one hundred internationalpublications.

Multimed Tools Appl

Daniele Franchi His activities include Biological systems and interactive simulator of cardiopulmonarydistrict; Telemedicine software for biomedical signal monitoring; EKG transmission standard protocol andquality system for diagnostic instrumentation; European informative system for movement data analysis;HIS integrated Nurse informative system; Cardiovascular risk evaluation; Web software for outpatient visiton underwater medicine.

Davide Gazze PHD in Information Engineering. His research interest includes “Web of Data”, “Social andSemantic Web”, “Social Media Analysis”, and “eHealth”. He has participated to the FP7 European Projects“Collaborative information, Acquisition, Processing, Exploitation and Reporting for the prevention of organ-ised crime” and “Open Polarity Enhanced Named Entity Recognition”. Currently, he works in the EuropeanProject Cassandra (Computer Assisted Solutions for Studying the Availability aNd DistRibution of novelpsychoActive substances).

Multimed Tools Appl

Giorgio Iervasi National Council Research (C.N.R.) Director of the Institute of Clinical Physiology (IFC-CNR). Research activity has focused in the cardio-endocrine-metabolic area. Last years were dedicated onnew factors of cardiovascular risk, in particular the metabolic ones, and on the pathogenic role and clinicalimplications of mild thyroid dysfunction on cardiac disorders.

AndreaMarchetti Technologist- Responsible of the “Social and Semantic Web” Lab. Main research interest:web of data, digital humanities and social media analysis. Authors of several international publications, he isparticipating to several FP7 European Projects and cooperates into the activities of the W3C Italian Office,promoting the Linked Open Data technologies.

Multimed Tools Appl

Alessandro Pingitore Medicine and Surgery (1991), PhD in Cardiovascular Physiopathology (1997), postgraduate school in Cardiovascular Disease (2002). Research fields: assessment of cardiac function, thy-roid/heart relationship, role of social networks on public healthy and cardiovascular prevention. Author of119 articles published on international peer-reviewed journals, 2 books, and 10 book chapters.

Maurizio Tesconi Researcher in Computer Science of the Web Application for the Future Internet group atthe Institute of Informatics and Telematics of the National Research Council (CNR) of Pisa. His researchinterests include social web mining and social media analysis, visual analytics, and natural language pro-cessing. He is also part of the permanent team of the European Laboratory on Big Data Analytics and SocialMining, performing advanced research and analyses on the emerging challenges posed by Big Data.