Upload
ma0511
View
29
Download
1
Embed Size (px)
Citation preview
Hospital Heterogeneity: What Drives Quality of Care?
University of Manchester
May 29, 2015
Introduction
1 The need to improve the quality of healthcare in UK and in other developedcountries is well documented.
2 The conundrum of rising healthcare costs and disappointing quality exists in everydeveloped country including the NHS (Palmer, 2012)
3 Healthcare organizations are increasingly scrutinized by external agencies such asHealth Care Commission in England.
4 Such agencies are increasingly concern themselves with the quality of care.
5 Health systems around the world are placing increasing demands on health careorganisations to deliver improvements in the performance and quality of the ser-vices.
6 A central observation of the healthcare systems is the existence of substantialheterogeneity or variations in healthcare quality across hospitals which are majorproviders of healthcare services, treat the most critically ill patients and accruesa substantial amount of health expenditure.
(University of Manchester) May 29, 2015 2 / 39
Hetrogeneity across NHS trustsQuality of Care for Stroke Services
0
5
10
20 40 60 80score
coun
t
2004
0
5
10
40 60 80score
coun
t
2006
0
5
10
15
40 60 80 100score
coun
t
2008
05
1015
50 60 70 80 90 100score
coun
t
2010
(University of Manchester) May 29, 2015 3 / 39
IntroductionPossible Explanations
7 Differences in the quality of healthcare suggest opportunities to improve qualityexists throughout the healthcare system.
8 These in turn have generated huge academic and policy interest in understandingthe causes of these variations and what could be done to increase quality at underperforming hospitals.
9 Possible Explanations:
1 Resources
Physical CapitalHuman Capital
2 Optimal use of resources
Organizational factorsManagement practicesIncentive system (for instance clinical Best practice tariffs in 2010, Payment byresults, readmission penalties etc.)
3 Structural Characteristics4 External/Environmental factors
Regional factors—median wage, prevalence, mortality etc.Competition and other market–oriented policies
(University of Manchester) May 29, 2015 4 / 39
IntroductionPossible Explanations
7 Differences in the quality of healthcare suggest opportunities to improve qualityexists throughout the healthcare system.
8 These in turn have generated huge academic and policy interest in understandingthe causes of these variations and what could be done to increase quality at underperforming hospitals.
9 Possible Explanations:
1 Resources
Physical CapitalHuman Capital
2 Optimal use of resources
Organizational factorsManagement practicesIncentive system (for instance clinical Best practice tariffs in 2010, Payment byresults, readmission penalties etc.)
3 Structural Characteristics4 External/Environmental factors
Regional factors—median wage, prevalence, mortality etc.Competition and other market–oriented policies
(University of Manchester) May 29, 2015 4 / 39
IntroductionManagerial and Organisational factors
1 An emerging trend of literature focuses on managerial and organisational as-pects of healthcare
2 A number of hospitals have turned to management practices for example,“Lean” methodology that was originally developed by Toyota which focusedcontinuous improvement and team work (Bloom et al., 2015; McConnell etal., 2013 and 2014)
3 Problems of variations in and lower healthcare quality can be mitigated if clin-icians and hospitals have help in identifying the best strategies in evaluatingand treating patients, and if they work in teams characterized by good struc-tural organisation, management leadership, patient centred and coordination& communication tools which are understood by all involved and that helppatients avoid hospitalisations and stay as healthy as possible.
4 Challenges are not clinical, but organisational (Lee and Mongan, 2009; Ra-manujam and Rousseau, 2006; West 2001)
(University of Manchester) May 29, 2015 5 / 39
IntroductionManagerial and Organisational factors
1 An emerging trend of literature focuses on managerial and organisational as-pects of healthcare
2 A number of hospitals have turned to management practices for example,“Lean” methodology that was originally developed by Toyota which focusedcontinuous improvement and team work (Bloom et al., 2015; McConnell etal., 2013 and 2014)
3 Problems of variations in and lower healthcare quality can be mitigated if clin-icians and hospitals have help in identifying the best strategies in evaluatingand treating patients, and if they work in teams characterized by good struc-tural organisation, management leadership, patient centred and coordination& communication tools which are understood by all involved and that helppatients avoid hospitalisations and stay as healthy as possible.
4 Challenges are not clinical, but organisational (Lee and Mongan, 2009; Ra-manujam and Rousseau, 2006; West 2001)
(University of Manchester) May 29, 2015 5 / 39
IntroductionResearch Aims and Question
Research Question: What are the factors that determine the quality ofcare as measured by process of care?
Syverson (2011)—internal and external drivers of firm productivity
Stroke process of care data from National Sentinel Stroke Audit 2004 to 2010from Royal College of Physicians
Process of measure includes:
1 Brain scan within 24 hours of stroke2 Screening for swallowing within 24 hours of admission3 Aspirin within 24 hours4 Physio assessment within 72 hours5 Patients weighed atleast once6 Mood assessed7 Rehabilitation goals agreed within discharge
Process measures summarized by a numerical score ranging from 0 to 100 usingthe indicator average method
(University of Manchester) May 29, 2015 6 / 39
IntroductionThree Dimensions of Quality
Under the health services research framework, healthcare quality can be definedor categorized by structure, process and outcomes.
Donabedian (1980) hypothesized that structure influences process which inturn influences outcomes.
Process measures illuminate the complicated process of delivering healthcareand describe the specific actions associated with healthcare delivery.
Structural measures focus on the characteristics of resources of the healthcaresystem, including institutional capacity (e.g. hospital size), system resources(e.g. stroke units, foundation trust status, QI participation)
Outcome measures focus on the end result of care or the effect of the careprocesses on the health and well–being of patients and population.
(University of Manchester) May 29, 2015 7 / 39
IntroductionProcess of Care Measures
Using process measures as quality indicators has several advantages (Rubin etal. 2001; Lilford et al. 2007)
Can more easily be used to provide feedback for quality improvement initiatives
Sensitivity and responsiveness to intervention
Generally requires less adjustment for patient severity and case–mix than most
outcome measures (Ukawa et al. 2014)
There is a lack of evidence to support a direct association between the qualityof care in process measures and improved performance in outcomes (Rubin etal. 2001)
Indicators to evaluate improvements of healthcare quality and standardization,as well as for public disclosure for accountability (Ukawa et al. 2014)
(University of Manchester) May 29, 2015 8 / 39
IntroductionShortcomings and Research Contributions
Importance of organizational and managerial factors in determining hospitalproductivity or quality
Process measures of quality over outcomes in health care
Interactions between the variables — Machine Learning algorithms
Predictive accuracy and out–of–sample predictions — Statistical Modelling: TwoCultures by Brieman (2001)
Longitudinal Data
Problematic causal inference and inappropriate policy recommendations
Time trend analyses
Geographic variations
(University of Manchester) May 29, 2015 9 / 39
IntroductionStroke Care
Focus on the quality of care for stroke care, also known as cerebral infarction
Stroke is one of the major causes of non–accidental deaths worldwide (Feiginet al. 2013).
In western countries, stroke is currently second biggest cause of death rankingafter heart disease and before cancer.
Stroke is a major cause of mortality and morbidity in the UK. In England 2011,there has been 110,000 incidences of stroke (Bray et al. 2013).
(University of Manchester) May 29, 2015 10 / 39
IntroductionDeaths in 2010 from circulatory diseases, England & Wales. Source: The Guardian, 2011
(University of Manchester) May 29, 2015 11 / 39
Theoretical frameworkRole and Importance of Organisational Quality
1 Theory of Complementarity in Organisations
(i) Brynjolfsson and Milgrom (2013); Milgrom and Roberts, (1995)
(ii) The interaction of two or agents or forces to produce an effect than the sum of
their individual effects.
(iii) The sets of factors or inputs when used together in the production process can be
mutually reinforcing in their effects of quality and performance
(iv) Dranove et al. (2014)—hospital technology adoption and costs
DefinitionTwo organisational practices for example x1 and x2 are complementary to each otherif their second derivative with respect to hospital’s objective function that is∂2f /∂x1∂x2 is ≥ 0. for all values of (x1, x2) with strict inequality for at least onevalue of the organisational practices.
(University of Manchester) May 29, 2015 12 / 39
Theoretical frameworkRole and Importance of Organisational Quality
1 Theory of Complementarity in Organisations
(i) Brynjolfsson and Milgrom (2013); Milgrom and Roberts, (1995)
(ii) The interaction of two or agents or forces to produce an effect than the sum of
their individual effects.
(iii) The sets of factors or inputs when used together in the production process can be
mutually reinforcing in their effects of quality and performance
(iv) Dranove et al. (2014)—hospital technology adoption and costs
DefinitionTwo organisational practices for example x1 and x2 are complementary to each otherif their second derivative with respect to hospital’s objective function that is∂2f /∂x1∂x2 is ≥ 0. for all values of (x1, x2) with strict inequality for at least onevalue of the organisational practices.
(University of Manchester) May 29, 2015 12 / 39
DataOrganisational quality and other variables
Royal College of Physicians National Sentinel Stroke Audit
Total organisational score for each trust ranging from 0 to 100 calculated asan average from 8 domains:
1 Acute Stroke Care Organisation
2 Organisation of Care
3 Specialist Roles
4 Inter Disciplinary Services (for sites with a stroke unit)
5 TIA/Neurovascular Service
6 Quality Improvement and Research
7 Team Working–Team Meetings
8 Communication with Patients and Carer
Total no of neurologists, teaching status, no. of hospital beds (size), regionaletc.
(University of Manchester) May 29, 2015 13 / 39
Exploratory Analysis
Table 1: Descriptive Statistics for Clinical Process Score
Years
2004 2006 2008 2010
Minimum 25.00 31.00 40.00 52.00
First Quartile 51.75 58.25 63.00 73.00
Median 61.00 67.00 71.00 79.00
Mean 60.48 66.13 69.74 78.75
Third Quartile 68.25 75.75 77.00 85.00
Maximum 93.00 93.00 96.00 97.00
Standard Deviation 12.36 13.33 11.32 8.57
(University of Manchester) May 29, 2015 14 / 39
Exploratory Analaysis
Table 2: Descriptive Statistics for Organisational Score
Years
2006 2008 2010
Minimum 23.00 32.00 47.00
First Quartile 57.25 63.75 62.00
Median 64.00 71.00 69.00
Mean 63.74 70.59 69.65
Third Quartile 72.00 79.00 76.75
Maximum 89.00 95.00 96.00
Standard Deviation 11.96 11.37 10.20
(University of Manchester) May 29, 2015 15 / 39
Exploratory Analysis
Figure 1: Score against the year
40
60
80
100
2003 2005 2007 2009 2011Year
scor
e
Year
2004
2006
2008
2010
(University of Manchester) May 29, 2015 16 / 39
Exploratory Analysis
Figure 2: Organisational Performance and Process of Score Quality
40
60
80
100
lower quartile middle half upper quartileOrganisational Position
scor
e
orgposition
lower quartile
middle half
upper quartile
(University of Manchester) May 29, 2015 17 / 39
Exploratory Analysis
Figure 3: Regional Variations
40
60
80
100
scor
e
Regions
East
East Midlands
London
North East
North West
South Central
South East Coast
South West
West Midlands
Yorkshire & the Humber
(University of Manchester) May 29, 2015 18 / 39
MethodologyMachine Learning for Longitudinal Data
1 (Unbiased) Regression Trees for longitudinal data—Siminoff and Wu (2014)
2 Mixed effects model for longitudinal data with regression tree based estimationmethods—RE–EM tree (Random Effects–Expectation Maximization); Sela andSiminoff (2012)
3 Linear Mixed Effects Model (LME)
4 Model performance/assessment:
(i) In–sample fit/predictions—AIC & BIC
(ii) Out–sample predictions: Leave–one out and k–fold cross validation
5 Endogeneity Issues
6 Importance of making predictions—Friedman (1953) and Chetty (2015)
(University of Manchester) May 29, 2015 19 / 39
Models and Hypotheses
Table 3: Models
Models Variables
Model 1: Physical Capital
Beds
Operating theatres
Day Case theatres
Model 2: Human Capital
Total Clinical (medically qualified) staff
Total non–medical staff
Professionally qualified clinical staff
Nurses
Healthcare Scientists
Allied health professionals
General Medicine Group
Neurologists
Neurophysiology
Neurosurgeons
Model 3: Hospital Characteristics and Organisational
quality
Teaching status
Foundation Trust Status
Organisational performance
Model 4: Regional Health Variables
Standardised Stroke Mortality 30–day rate
Emergency stroke admissions
all SMR (area)
Model 5: Socio–Economic Variables
Median wage
Inequality
%regional population with no qualifications
(University of Manchester) May 29, 2015 20 / 39
Regression Tree Results
Figure 4: Unbiased REEM–Tree for Model 3
orgscorep < 0.001
1
≤ 67 > 67
Node 2 (n = 155)
40
50
60
70
80
90
100
gmgStaffp < 0.001
3
≤ 0.041 > 0.041
Node 4 (n = 144)
40
50
60
70
80
90
100
teachingp = 0.053
5
≤ 0 > 0
Node 6 (n = 9)
40
50
60
70
80
90
100Node 7 (n = 20)
40
50
60
70
80
90
100
(University of Manchester) May 29, 2015 21 / 39
Regression Tree Results
Figure 5: Unbiased REEM–Tree for Model 4
orgscorep < 0.001
1
≤ 67 > 67
Node 2 (n = 155)
40
50
60
70
80
90
100
gmgStaffp < 0.001
3
≤ 0.041 > 0.041
Node 4 (n = 144)
40
50
60
70
80
90
100
all_smrp = 0.045
5
≤ 109.62 > 109.62
Node 6 (n = 21)
40
50
60
70
80
90
100Node 7 (n = 8)
40
50
60
70
80
90
100
(University of Manchester) May 29, 2015 22 / 39
Regression Tree Results
Figure 6: Unbiased REEM–Tree for Model 4
orgscorep < 0.001
1
≤ 67 > 67
Node 2 (n = 155)
40
50
60
70
80
90
100
gmgStaffp < 0.001
3
≤ 0.041 > 0.041
Node 4 (n = 144)
40
50
60
70
80
90
100
all_smrp = 0.045
5
≤ 109.62 > 109.62
Node 6 (n = 21)
40
50
60
70
80
90
100Node 7 (n = 8)
40
50
60
70
80
90
100
(University of Manchester) May 29, 2015 23 / 39
Regression Tree Results
Figure 7: Unbiased REEM–Tree for Model 4
orgscorep < 0.001
1
≤ 67 > 67
median_wagep < 0.001
2
≤ 486.5 > 486.5
Node 3 (n = 111)
40
50
60
70
80
90
100Node 4 (n = 44)
40
50
60
70
80
90
100
median_wagep < 0.001
5
≤ 524.8 > 524.8
median_wagep = 0.02
6
≤ 420.5 > 420.5
Node 7 (n = 47)
40
50
60
70
80
90
100Node 8 (n = 103)
40
50
60
70
80
90
100
profqStaffp = 0.031
9
≤ 0.506 > 0.506
Node 10 (n = 11)
40
50
60
70
80
90
100Node 11 (n = 12)
40
50
60
70
80
90
100
(University of Manchester) May 29, 2015 24 / 39
Model Performance and AssessmentIn Sample Fits for models 1 to 5
Table 4: In Sample fits for models without year and ratios
Unbiased REEM–Tree REEM–Tree LME
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Log Likelihood AIC BIC Log Likelihood AIC BIC Log Likelihood AIC BIC
Models
Model 1 -1234.497 2474.99 2430.04 -1234.497 2474.99 2486.36 -1238.517 2489.03 2511.72
Model 2 -1197.678 2407.36 2457.86 -1192.093 2396.19 2418.87 -1164.889 2359.78 2416.07
Model 3 -1187.768 2387.54 2410.22 -1158.047 2332.10 2362.29 -1133.805 2303.61 2370.98
Model 4 -1187.121 2386.24 2408.93 -1179.302 2370.60 2393.85 -1139.121 2320.24 2398.64
Model 5 -1162.276 2340.55 2370.75 -1156.683 2329.37 2359.56 -1117.162 2282.32 2371.69
(University of Manchester) May 29, 2015 25 / 39
Regression Tree ResultsOut–sample–fits for models 1 to 5
Table 5: Out of Sample Predictions without year and ratios
LOOCV k–FOLD
Unbiased REEM–Tree REEM–Tree LME Unbiased REEM–Tree REEM–Tree LME
(1) (2) (3) (4) (5) (6)
Models
Model 1 9.220 9.264 9.222 9.314 9.601 9.319
Model 2 9.166 8.902 8.999 9.542 9.050 9.050
Model 3 8.800 9.057 8.348 8.857 8.991 8.330
Model 4 8.763 9.118 8.178 8.727 9.259 8.150
Model 5 7.982 8.210 7.634 8.133 8.498 7.682
(University of Manchester) May 29, 2015 26 / 39
Robustness AnalysisYear Robustness
Figure 8: Yearly Analysis
60
70
80
2004 2006 2008 2010Years
Val
ue
variable
score
overall
(University of Manchester) May 29, 2015 27 / 39
Regression Tree ResultsYear Robustness
Figure 9: Unbiased REEM–Tree for Model 5 with year (ratios)
yearp < 0.001
1
≤ 2008 > 2008
orgscorep < 0.001
2
≤ 67 > 67
Node 3 (n = 97)
50
60
70
80
90
gmgStaffp < 0.001
4
≤ 0.037 > 0.037
Node 5 (n = 55)
50
60
70
80
90
Node 6 (n = 12)
50
60
70
80
90
clinicalStaffp < 0.001
7
≤ 0.134 > 0.134
Node 8 (n = 149)
50
60
70
80
90
Node 9 (n = 15)
50
60
70
80
90
(University of Manchester) May 29, 2015 28 / 39
In sample fits
Figure 10: In–sample fit plot
40 50 60 70 80 90
6070
8090
Actual Score
Fitt
ed V
alue
s
(University of Manchester) May 29, 2015 29 / 39
Diagnostic Plots
Figure 11: Diagnostic Analysis
60 70 80 90
−15
−10
−5
05
10Residuals against Fitted values
Fitted Score
Res
idua
ls
−3 −1 0 1 2 3
−15
−10
−5
05
10
Normal Q−Q Plot
Theoretical Quantiles
Sam
ple
Qua
ntile
s
(University of Manchester) May 29, 2015 30 / 39
Robustness AnalysisFurther Robustness
Figure 12: Further Robustness
yearp < 0.001
1
≤ 2008 > 2008
orgscorep < 0.001
2
≤ 67 > 67
Node 3 (n = 97)
50
60
70
80
90
100
gmgStaffp < 0.001
4
≤ 0.037 > 0.037
Node 5 (n = 55)
50
60
70
80
90
100Node 6 (n = 12)
50
60
70
80
90
100
orgscorep = 0.003
7
≤ 77 > 77
Node 8 (n = 112)
50
60
70
80
90
100
gmgStaffp < 0.001
9
≤ 0.037 > 0.037
Node 10 (n = 39)
50
60
70
80
90
100Node 11 (n = 13)
50
60
70
80
90
100
(University of Manchester) May 29, 2015 31 / 39
Robustness AnalysisFurther Robustness: Cross Section
Figure 13: Further Robustness: 2010 Cross Section
orgscorep < 0.001
1
≤ 77 > 77
Node 2 (n = 112)
60
70
80
90
100
gmgStaffp = 0.003
3
≤ 0.042 > 0.042
Node 4 (n = 42)
60
70
80
90
100Node 5 (n = 10)
60
70
80
90
100
(University of Manchester) May 29, 2015 32 / 39
Model Assessment and PerformanceYear Robustness
Table 6: Year Robustness — In sample and out of sample performance (Ratios)
In–Sample Fits Out of Sample Predictions
Log Lik AIC BIC LOOCV k–FOLD
(1) (2) (3) (4) (5)
Unbiased REEM–Tree -1155.66 2325.32 2351.76 7.540 7.469
REEM–Tree -1141.89 2299.77 2329.97 7.728 7.519
LME -1113.739 2277.48 2370.487 7.527 7.568
(University of Manchester) May 29, 2015 33 / 39
Limitations & Extensions
1 Data
2 Randomised Experiments
(University of Manchester) May 29, 2015 34 / 39
Summary & Conclusions
The empirical findings are consistent with the productivity literature and cor-roborates with previous research on managerial and organisational determinantsof healthcare quality that have used different designs, data and methods and,they offer predictive support for the theory used in this study as well as in theeconomics literature on the role of institutions and productivity.
Robust
Regression Trees and machine learning methods
Healthcare exceptionalism—Chandra et al. (2013)
Can inform and improve the decision making process for healthcare qualityimprovement and also in general contributes to data driven decision making inhealthcare.
Limitations
Roadmap
(University of Manchester) May 29, 2015 35 / 39
Summary & Conclusions
The empirical findings are consistent with the productivity literature and cor-roborates with previous research on managerial and organisational determinantsof healthcare quality that have used different designs, data and methods and,they offer predictive support for the theory used in this study as well as in theeconomics literature on the role of institutions and productivity.
Robust
Regression Trees and machine learning methods
Healthcare exceptionalism—Chandra et al. (2013)
Can inform and improve the decision making process for healthcare qualityimprovement and also in general contributes to data driven decision making inhealthcare.
Limitations
Roadmap
(University of Manchester) May 29, 2015 35 / 39
References
1 Bloom, N., Propper, C., Seiler, S and Van Reenen J. (2015). The Impact of Competition
on Management Quality: Evidence from Public Hospitals. The Review of Economic
Studies, 0:1–33
2 Bloom, N. and Van Reenen, J. (2007). Measuring and explaining management practices
across rms and countries. The Quarterly Journal of Economics, 122(4):1351–1408.
3 Bray, D. B., Ayis, S., Campbell, J., Hoffman, A., Roughton, M., Tyrrell, P.J., Wolfe, C.
and Rudd, A. 2013. Associations between the organisation of stroke services, process
of care, and mortality in England: prospective cohort study. BMJ 346:f2827
4 Breiman, L. 2001. Statistical Modeling: The Two Cultures. Statistical Science. 16(3),
pp. 199–231 Brynjolfsson, E. and Milgrom, P. (2013). Complementarity In Orga-
nization. In Gibbons, R. and Roberts, J., editors, The Handbook of Organizational
Economics. Princeton University Press.
5 Chandra, A., Finkelstein, A., Sacarny, A., and Syverson, C. (2013). Healthcare ex-
ceptionalism? Productivity and allocation in the us healthcare sector. NBER Working
Paper, National Bureau of Economic Research.
6 Chetty, R. (2015). Behavioral economics and public policy: A pragmatic perspective.
7 Donabedian, A., 1980. The Definition of Quality and Approaches to Its Assessment.
Ann Arbor, MI: Health Administration Press.
(University of Manchester) May 29, 2015 36 / 39
References
8 Friedman, M. (1953). The methodology of positive economics. Essays in Positive
Economics, 3(3).
9 McConnell, K. J., Chang, A. M., Maddox, T. M., Wholey, D. R., and Lindrooth, R. C.
(2014). An Exploration of Management Practices in Hospitals. Healthcare, 2(2):121–
129.
10 McConnell, K. J., Lindrooth, R. C., Wholey, D. R., Maddox, T. M., and Bloom, N.
(2013). Management practices and the quality of care in cardiac units. JAMA Internal
Medicine, 173(8):684–692.
11 Milgrom, P. and Roberts, J. (1995). Complementarities and fit strategy, structure,
and organizational change in manufacturing. Journal of Accounting and Economics,
19(2):179–208.
12 Palmer, K. (2012). Stronger incentives for quality improvement needed in NHS in
England. Journal of Health Services Research & Policy, 17(2):65–67.
13 Ramanujam, R. and Rousseau, D. M. (2006). The challenges are organizational not
just clinical. Journal of Organizational Behavior, 27(7):811–827.
14 Dranove, D., Forman, C., Goldfrab, A. and Greenstein, S. (2014). The Trillion Dollar
Conundrum: Complementarities and Health Information Technology. American Eco-
nomic Journal: Economic Policy, 6(4): 239-70
(University of Manchester) May 29, 2015 37 / 39
References
15 Rubin, H. R., Pronovost, P., and Diette, G. B. (2001). The advantages and disadvan-
tages of process-based measures of health care quality. International Journal for Quality
in Health Care, 13(6):469–474.
16 Sela, R. J. and Simonoff, J. S. (2012). RE–EM trees: a data mining approach for
longitudinal and clustered data. Machine Learning, 86(2):169–207.
17 Simonoff, J. S. and Fu, W. (2014). Unbiased Regression Trees for Longitudinal Data.
SSRN Working Paper.
18 Syverson, C. 2011. What Determines Productivity? Journal of Economic Literature,
49(2), pp. 326–365
19 Ukawa, N., Ikai, H., and Imanaka, Y. (2014). Trends in hospital performance in acute
myocardial infarction care: a retrospective longitudinal study in japan. International
Journal for Quality in Health Care, 26(5):516–523.
20 West, E. (2001). Management matters: the link between hospital organisation and
quality of patient care. Quality in Health Care, 10(1):40–48.
(University of Manchester) May 29, 2015 38 / 39
AppendixData Sources
1 National Sentinel Stroke Audit 2004 to 2010, Royal College of Physicians
2 Hospital Estate and Facilities Data
3 NHS Workforce Statistics
4 Department of Health QMCO
5 Health and Social Care Information Centre (HSCIC)
6 Office for National Statistics; ASHE
7 NOMIS
8 QOF
(University of Manchester) May 29, 2015 39 / 39