Patients in England are more likely to choose general practices with higher clinical quality as measured by Quality and Outcomes Framework (QOF) points. A 1 standard deviation increase in QOF points is associated with a 20% increase in demand for a practice. While the effect of quality on an individual's choice is small, the large number of potential patients means quality has a large effect on total demand for a practice. Quality is the main driver of choice, more so than distance, practice characteristics, or patient attributes. This provides incentive for practices to improve quality to attract more patients.

1. Does quality affect patients’ choice
of doctor? Evidence from the UK
Rita Santos* Hugh Gravelle*
Carol Propper**
*Centre for Health Economics, University of York
**Imperial College, London; CMPO, University of Bristol
Competition and market mechanisms in health care.workshop. Nuffield Trust, 13 September 2013

2. Motivation
• Governments increasingly adopting policies to create/enhance
choice in health care (and other public services)
• Improve match of patients and providers
• Encourage providers to compete on quality in fixed price regime
• Competition will improve quality only if demand is responsive to
quality
– But
• Noisy quality measures
• Consumers may not value quality
• Does quality affect patients’ choice of provider?
2

3. Our Research Question
• Do English patients respond to differences in quality when
choosing a general practice?
• England is a good “test bed”
– Patients register with a single general practice
– GPs are gatekeepers for elective hospital care
– Care is tax funded and free at point of use, so choice should be driven
only by clinical quality, distance and other practice attributes
– Government actively promotes provision of information to patients eg
NHS Choices website
– Rich set of quality measures for general practice
3

4. What we do
• Data on the choices of 3.4M patients in 2874 small areas (LSOAs)
from amongst nearly 1000 general practices
• Focus on quality, distance, and practice attributes
• Consider appropriate measure of quality, patient heterogeneity,
endogeneity of quality
• We find
– Patients more likely to choose practices which have higher clinical quality
as measured by published data on performance (QOF points)
– Robust across age and gender, socio-economic status
– Quality has small effect on probability patient chooses particular practice
– But large number of potential patients for a practice mean that quality has
large effect on total demand
• 1 SD increase in QOF points would increase demand for a practice by 20%
4

5. Literature on health care quality, demand, and
competition
• Quality and demand
• Hospital sector – quality affects choice of hospital
– Extensive US literature
– Recent UK studies: Sivey 2011; Beckert et al 2012, Gaynor et al 2012
• General practice: few studies of impact of quality on choice because
of lack of good measures of quality
– UK: Dixon et al, 1997
– Norway: Biorn & Godager, 2010); Iversen and Luras 2011
• Competition and quality
• Hospital sector: competition improves quality in fixed price systems
• UK hospital sector: Cooper et al (2011); Gaynor et al (2013)
• UK general practice: Pike (2010) - practices with more rivals within
500m have higher patient satisfaction, smaller ACSC admission rate
5

6. Outline of talk
• Institutional setting
• Data
• Model
• Results
• Conclusions
6

7. Institutional setting: English NHS
• No charges to patients for hospital or primary care
• List system (patient registration) for general practices
• GPs as gatekeepers for elective hospital care
• 8300 practices with 4.2 GPs and 6600 patients
• Practice contracts with NHS
– GMS – mix of capitation, lump sum, quality related payments
– PMS - paid as if GMS plus uplift to reflect additional services
negotiated with local PCT
– Quality and Outcomes Framework – payment for achievement
of quality indicators
– Both contracts link total revenue closely to number of patients
7

8. Institutional background: policy
• Abolition of national body controlling entry (2002)
• Devolution of entry control to local health authorities
• Removal on restrictions on ownership of practices
• Encourage new entry (Darzi health centres)
• NHS Choices website with data on practices
• Patients to be given right to register with any practice
8

9. 9
ONS Neighbourhood Statistics
Socio economic data at LSOA
level
General Medical Services Statistics
Practice and GP characteristics
General Practice Patient Survey
Patient satisfaction by practice
Quality and Outcomes Framework
Clinical quality indicators by
practice
Hospital Episode Statistics
ACSC emergency admission rates
by practice
Attribution Data Set
Patient registered with each
general practice
LSOA by age by gender by
practice cells
Data sets

10. Sample
• Unit of analysis: LSOA
• Population: mean 1500, min 1000
• Attribution Data Set
• Number of patients in each of 8300 practices resident
in each of 32,482 LSOAs in 36 age/gender bands
• At 1 April 2010
• East Midlands Strategic Health Authority
• Mix of urban, rural areas; high proportion of non-UK
qualified GPs; relatively high proportion of pop of Asian
origin
• Not adjacent to Wales or Scotland
• 2875 LSOAs, 3.372M individuals aged ≥25
• 994 practices with 1235 surgeries 10

11. 11

12. Data: practice QOF quality
• Quality and Outcomes Framework
– National P4P scheme introduced 2004
– Points awarded (max 1000) for achievements of indicators
• Clinical domain
• Organisational domain
• Patient experience domain
• Additional services
• Holistic care
– £125 per point
– Data extracted from practice electronic patient records
• Main quality measure: (lagged) total QOF points in 2006/7
• Also use
– total QOF points 2009/10
– average QOF points 2006/7-2009/10
– domain points
– other measures derived from clinical indicators
12

13. Data: non-QOF quality measures
• Emergency hospital admissions for Ambulatory
Care Sensitive Conditions
• Patient satisfaction
• General Practice Patient Survey 2009: 5% sample of
all patients on lists
• “In general, how satisfied are you with the care you get
at your GP surgery or health centre?”
• “How satisfied are you with the hours that your GP
surgery or health centre is open ?”
• “Would you recommend your GP practice or health
centre to someone who has just moved to your local
area?”
13

14. Data: LSOA - practice distances
• Practice locations
• Main surgery – GMS statistics
• Branch surgeries – NHS Choices
• LSOA population weighted centroids
• Linear distance from each LSOA centroid to nearest
branch of each practice within 50km
• Dummy variable = 1/0 as practice located in
different/same PCT as LSOA
14

15. Data: LSOA choice sets
• Practices with branch within 10km of LSOA
centroid
– 99.3% patients choose practice within 10km
• If more than 30 such practices restrict choice
set to 30 practices with largest number patients
from LSOA
15

16. 16
Characteristics of patients in LSOAs
Mean SD Min Max
Proportion female 0.507 0.022 0.276 0.618
Proportion in fair or good self-rated health 0.907 0.032 0.760 0.983
Proportion adults without qualification 0.231 0.071 0.035 0.430
Proportion non white 0.065 0.130 0.000 0.948
Income deprivation score 0.143 0.110 0.013 0.830
Urban 0.731 0.444
Proportion of LSOA registered at nearest
practice
0.399 0.263 0.001 0.998

17. Practice characteristics
17
Mean SD Min Max
Average GP age 47.9 6.7 31.5 72.5
Proportion female GPs 0.362 0.248 0 1
Proportion GPs trained outside Europe 0.267 0.354 0 1
Opted out of out of hours care 0.613 0.487
PMS contract 0.479 0.500
Dispensing practice 0.204 0.403
Patients aged ≥ 25 registered with
practice
4886 3063 653 24988

18. Practice quality measures
18
Mean SD Min Max
QOF 2006/7 total points 956.1 63.6 426.5 1000
QOF 2006/7 clinical points 632.8 36.4 330.5 655
QOF 2006/7 organisational points 166.5 21.0 13.2 181
QOF 2006/7 patient experience points 103.3 16.1 0 108
QOF 2006/7 additional services points 35.3 2.8 6 36
QOF 2006/7 holistic care points 18.3 3.1 0 20
QOF 2009/10 total points 940.5 46.9 545.5 1000
Average QOF total points 2006/7-2009/10 954.6 44.8 545.5 1000
ACSCs 2006/7 per 10,000 259 76 28 679
Overall patient satisfaction 2009 0.89 0.06 0.57 0.99
Satisfaction with opening hours 2009 0.80 0.06 0.45 0.97
Prop patients would recommend practice
2009
0.82 0.10 0.38 0.99

19. Distances
19
Mean SD Min Max
Distance to practices in choice set (km) 4.83 1.65 0.35 9.89
Distance to chosen practice (km) 1.89 1.34 0.13 9.87
Distance to nearest practice (km) 1.20 1.16 0.02 9.81
Proportion practices in different PCT 0.27 0.45 0 1
Prop practices chosen in different PCT 0.19 0.39 0 1
Prop nearest practices in different PCT 0.05 0.22 0 1
Prop LSOA pop registered at nearest
practice
0.40 0.26 0.00 1

20. Distance to practices
in choice set
Distance to practice
chosen
20

21. Model: conditional logit
21
All patients in an LSOA have same choice set
All patients derive same utility from observed practice characteristics
unobserved utility iid type 1 extreme value distribution
P
a
iaj aj iaj
aj
iaj
a C
u 

 x β
x
 
1
1
1
robability patient in LSOA choose practice
exp( ) exp( )
Log likelihood ( patients in LSOA choose practice )
ln ln exp( ) exp( )
a
A
a a
iaj aj ajj C
aj
n
aj aj aja j C j C
i a j
P
n a j
L n



   
 
 
 
 

  
x β x β
x β x β

22. Model: sensitivity checks
• Alternative quality measures
• Distance specification
• Patient heterogeneity
• Restrictions on choice sets
• Endogeneity
22

23. Model: interpretation of results
• Report average of marginal effects
• AMEs small
• Potential incentive for practice to increase quality
depends on change in demand for practice
– Depends on change in probability patients will want to
join practice and number of patients in whose choice
set practice falls
• 75,000 individuals  25 yrs within 5 km of av practice
• 25,000 individuals  25 yrs within 2km of av practice
– Small changes in individual probabilities can
translate into large demand increases
23
ˆˆ ˆ ˆ/ (1 )aj kaj k aj ajP x P P   

24. Results: preferred model
Average Marginal
Effect
z
QOF 2006/7 Total points 0.00013 6.87
Distance (cubic) -0.06778 -14.18
Practice in different PCT -0.04751 -10.12
GP age -0.00144 -13.31
Prop female GPs 0.01508 6.12
Prop GPs non Europe trained -0.03029 -10.36
Opted Out 0.00543 2.49
PMS contract 0.00564 2.95
BIC 11714907
McFadden R2 0.3955
N LSOA 2,870
N GP practices 987
N patients 3,364,263
Increase 10 QOF pts increase pr(choice) by 0.0013
Mean QOF points = 957, sd = 64. 24

25. Results: other quality measures
• 2006/7 total QOF points performs at least as well as
– domains of QOF
– 2009/10 QOF total points
– av 2006/7-2009/10 total QOF points
• QOF clinical measures adjusted for exceptions and thresholds have no
explanatory power given covariates
• Patient satisfaction
– Overall patient satisfaction insignificant when practice characteristics
and QOF 2006/7 total points included
– Patient satisfaction summarises effect of practice characteristics
(Robertson et al, 2008)
• ACSC admission rates
– Negatively correlated with QOF quality but no additional explanatory
power
25

26. Results: distance
• Cubic preferred to other polynomial
specifications and to log distance
• Distance effects similar (except linear)
26

27. Plot of the average marginal effects of distance for linear (km),
quadratic (km2), cubic (km3), quartic (km4), and quintic (km5)
specifications for baseline model specification.
Average marginal effects of distance
27
Linear specification

28. Model estimation: patient heterogeneity
• Baseline models assume patients homogenous
• Test for observed and unobserved patient preference
heterogeneity
– estimating models for each age and gender group
– stratify LSOAs (separately) by characteristics of the
proportion of pop who are non-white, deprived etc.
– estimate random coefficient model (mixed logit model)
28

29. Results: patient heterogeneity
Separate estimates for 10 age and gender groups
• Young men (25-35) less sensitive to quality
Separate estimates for LSOAs stratified by patient characteristics
(rurality, income deprivation, education deprivation, self
assessed health, ethnicity; top vs other quintiles)
• Rural areas: less sensitive to distance and quality, ratio similar
• More deprived LSOAs more affected by distance and less by
quality (marginal rate of substitution halves)
Unobserved heterogeneity (mixed logit model)
• Similar means, only significant variance in distance and quality
29

30. Mixed logit Conditional logit
Mean of
coefficients
z Coefficient z
QOF 200607 Total points 0.0029 11.55 0.00224 14.58
Practice in different PCT -0.891 -10.55 -0.826 -19.00
Distance -1.556 -38.51 -1.563 -40.06
Distance squared 0.109 9.06 0.121 10.55
Distance cubed -0.00417 -4.58 -0.00432 -4.88
GP age -0.0254 -15.52 -0.025 -15.68
Proportion female GPs 0.262 7.77 0.262 7.85
Proportion GPs non Europe trained -0.522 -18.95 -0.527 -19.33
Opted out 0.0998 2.73 0.0943 2.61
PMS 0.104 3.28 0.098 3.13
Standard deviation of coefficients
QOF 200607 Total points 0.00317 7.02
LSOAs from different PCTs -0.478 -1.76
Distance 0.214 8.28
Distance squared km -0.00439 -1.56
Distance cubic km 0.000341 1.98
GP age 0.00633 1.23
Female GPs 0.0071 0.20
GPs trained outside Europe -0.048 -0.85
Opted out 0.308 2.09
PMS 0.0759 0.34
30

31. Catchment areas and closed lists
• Do data reflect patient or practice choices?
• GPs have obligation to make home visits if medically
required. Higher costs to practices if patient lives
further away.
• Practices can refuse to enrol patient only if
– patient lives outside catchment area agreed with PCT
• If our LSOA choice set radius greater than radii of catchment areas
then zero patients registered with a practice may reflect
catchment area not patient demand
– practice has notified PCT that its list is closed: it will not
accept any more patients
• List turnover averages 8% pa so no practice can have permanently
closed list
• Our data is patients on list at 1 April 2010: stock not flow 31

32. Results: choice set radius
32
Choice set
radius
QOF 2006/7
total points
Distance
(cubic)
10km
AME 0.00013 -0.06778
z 6.87 -14.18
8km
AME 0.00014 -0.07751
z 6.94 -14.2
6km
AME 0.00019 -0.09964
z 7.07 -14.72
4km
AME 0.00029 -0.15248
z 7.76 -16.76
2km
AME 0.00044 -0.26363
z 7.98 -23.16
Quality effects increase as choice set shrinks

33. Results: choice sets only with practices
with LSOA patients
33
Minimum patients from
LSOA in practice
QOF 2006/7
total points
Distance
(cubic)
0 patients
AME 0.00013 -0.06778
z 6.87 -14.18
1 patient
AME 0.00019 -0.0969
z 7.18 -15.6
5 patients
AME 0.00026 -0.1268
z 7.47 -18.01
10 patients
AME 0.00029 -0.14037
z 7.82 20.94
20 patients
AME 0.00032 -0.14739
z 7.99 -24.43
50 patients
AME 0.00031 -0.13713
z 7.07 -25.88

34. Endogeneity?
• Better educated patients care more about quality?
• Quality indicators easier to achieve with better educated
patients?
• Implies quality AME biased upward
• Sicker patients care more about quality?
• Quality indicators harder to achieve with sicker patients?
• Implies quality AME biased downward
• Easier to achieve quality in larger practices (econ of scale)?
• Implies quality AME biased upward
34

35. Endogeneity: two stage residual inclusion
• First stage
– Linear practice quality model
– IV: average quality of two nearest practices
– F statistic for IV: 18.5
• Second stage choice model
– Add residual from stage 1 quality model
– SEs : SD of second stage AMEs from 100 bootstrap
samples LSOAs
35

36. Results: 2SRI model
36
Baseline model 2SRI model
AME z AME
z
bootstrap
QOF 2006/7 points 0.00013 6.87 0.00074 4.35
QOF 2006/7 residuals -0.00044 -3.02
Different PCT -0.06778 -14.18 -0.10942 -9.85
Distance (cubic) -0.04751 -10.12 -0.11108 -18.82
GP age -0.00144 -13.31 -0.00281 -11.50
Female GPs 0.01508 6.12 0.03084 6.52
GPs non Europe trained -0.03029 -10.36 -0.06328 -10.26
Opted out 0.00543 2.49 0.00803 1.78
PMS 0.00564 2.95 0.01021 2.36
Quality AME larger when instrumented

37. Effects on demand
37
AME Extra
metres
Patients
gained
Elasticity
2006/7 QOF points
(1/10th SD increase: 6.4 points )
0.00082 12.4 103.6 1.44
SE 0.00012 0.9 9.4 0.06
Av age GPs
(1/10th SD increase: 0.7 yrs)
-0.00096 -14.6 -120.6 0.003
SE 0.00007 0.9 5.3 0.00
Prop female GPs
(1/10th SD increase: 0.025)
0.00374 56.7 468.1 0.07
SE 0.00061 7.2 47.3 0.01
Prop non-European trained GPs
(1/10th SD increase: 0.035)
-0.01072 -162.7 -1342.4 -0.08
SE 0.00103 -8.7 93.6 0.00

38. Conclusions
• Issue of whether choice will increase quality
current and important
• A pre-requisite for increased competition to
increase quality is that demanders are
responsive to quality
• Find that this does appear to be the case for
choice of GP
38

39. Thank you
39

40. Other QOF based quality measures
40
Table 3
QOF 2006/7 total
points
(1)
QOF 2006/7 total
points
(2)
QOF 2006/7
clinical points
(3)
AME z AME z AME z
QOF 2006/7 total points 0.00013 6.87
QOF 2006/7 clinical points 0.00015 2.71 0.00019 5.69
QOF 2006/7 organisational points 0.00018 4.90
QOF 2006/7 patient experience -0.00003 -0.78
QOF 2006/7 additional services 0.00139 4.12
QOF 2006/7 holistic care points -0.00020 -0.47
Distance (cubic) -0.06778 -14.18 -0.07626 -9.83 -0.06695 -12.00
BIC 11714907 11710963 11724753
Av 2006/7-09/10
total points
(4)
QOF 2009/10 total
points
(5)
AME z AME z
Av QOF total points 2006/7-2009/10 0.00026 6.947
QOF 2009/10 total points 0.000017 2.569
Distance (cubic) -0.09046 -17.89 -0.03005 -7.334
BIC 11783137 11806537

41. Patient satisfaction
41
Table 4
Quality measure: overall satisfaction Quality measure: access satisfaction
(1) (2) (3) (4)
AME z AME z AME z AME z
QOF 2006/7 total points 0.00012 6.34 0.00012 6.32
Overall Satisfaction 0.24984 41.6 -0.00725 -0.89
Access Satisfaction -0.18348 -5.50 -0.01072 -1.57
Distance (cubic) -0.06504 -10.93 -0.06419 -11.37
BIC 19439563 11714835 19466172 11714636
Quality measure: would recommend
(5) (6)
AME z AME z
QOF 2006/7 total points 0.00013 6.70
Would recommend 0.27598 225.71 0.02226 3.12
Distance (cubic) -0.07284 -15.02
BIC 19374984 11713209

42. Distance specification
42
Linear Quadratic Cubic
AME z AME z AME z
QOF 2006/7 Total points 0.00019 7.89 0.00014 7.10 0.00013 6.87
Distance -0.07747 -16.36 -0.06962 -14.60 -0.06778 -14.18
BIC 11886627 11720567 11714907
Quartic Quintic Log distance
AME z AME z AME z
QOF 2006/7 Total points 0.00013 6.96 0.00013 6.94 0.00026 6.45
Distance -0.06914 -14.34 -0.06894 -14.24 -0.21689 -13.44
BIC 11713827 11713833 12363341

43. Table 7. Rurality, income deprivation
Socio-economic status stratified models
Rurality Income deprivation
Urban LSOAs Rural LSOAs Lowest 4 quintiles Top quintile
AME z AME z AME z AME z
QOF 2006/7 points 0.00017 7.98 0.00003 1.80 0.00014 6.10 0.00010 3.40
Distance (cubic) -0.08518 -16.35 -0.01548 -3.47 -0.06056 -12.19 -0.09852 -8.22
BIC 9447553 2247370 9039531 2648384
N LSOA 2100 770 2295 575

44. Table 7: Self assessed health, Asian ethnicity
.
44
Proportion with poor self assessed health Proportion of Asian ethnicity
Lowest 4 quintiles Top quintile Lowest 4 quintiles Top quintile
AME z AME z AME z AME z
QOF
2006/7
Total
points 0.00016 6.51 0.00007 2.77 0.00011 5.18 0.00021 5.93
Distance
(cubic) -0.07075 -13.60 -0.06246 -6.06 -0.05234 -10.40 -0.12061 -13.22
BIC 9322153 2383773 8469297 3221563
N LSOA 2296 574 2294 576

45. Choice set restricted radii
45
Choice set radius 10km 8km 6km
AME z AME z AME z
QOF 2006/7 total points 0.00013 6.87 0.00014 6.94 0.00019 7.07
Distance (cubic) -0.06778 -14.18 -0.07751 -14.2 -0.09964 -14.72
BIC 11714907 11262610 10556052
McFadden R2 0.3955 0.3621 0.3094
N LSOA 2870 2806 2670
N GP practices 987 931 855
Choice set radius 4km 2km
AME z AME z
QOF 2006/7 total points 0.00029 7.76 0.00044 7.98
Distance (cubic) -0.15248 -16.76 -0.26363 -23.16
BIC 9096439 5279364
McFadden R2 0.2267 0.1056
N LSOA 2428 1925
N GP practices 729 605

46. Choice set with min nos patients
46
Minimum patients from
LSOA in practice
10 patients 20 patients 50 patients
AME z AME z AME z
QOF 2006/7 total points 0.00029 7.82 0.00032 7.99 0.00031 7.07
Distance (cubic) -0.14037 20.94 -0.14739 -24.43 -0.13713 -25.88
BIC 10056703 9195685 7367098
McFadden R2 0.1711 0.1284 0.0717
N LSOA 2802 2765 2670
N GP practices 774 751 719
N patients 3220697 3118525 2850124
Minimum patients from
LSOA in practice
0 patients 1 patient 5 patients
AME z AME AME z
QOF 2006/7 total points 0.00013 6.87 0.00019 7.18 0.00026 7.47
Distance (cubic) -0.06778 -14.18 -0.09690 -15.60 -0.12680 -18.01
BIC 11714907 11351519 10649668
McFadden R2 0.3955 0.2983 0.2126
N LSOA 2870 2865 2842
N GP practices 987 889 814
N patients 3364263 3353087 3298413