5. Competition and Quality in Health Care
1. Theory
a. Regulated prices – competition increases quality (if price >
marginal cost; e.g., regulated airline models).
i. Quality elasticity of firm’s demand increasing in the # of firms.
ii. Competition increases management effectiveness, thereby increasing
quality.
b. Market determined prices – anything can happen.
2. Evidence is mixed
a. Regulated prices
i. Medicare – competition improves quality (e.g., Kessler & McClellan,
2000)
b. Market determined prices
i. U.S. private markets – not so clear (Volpp et al., 2003)
ii. U.K. 90s reforms – competition reduced quality (Propper et al, 2008)
3. Little evidence from policies designed to introduce
competition
8. NHS Reforms, cont’d.
4. Key elements
a. ‘Choose and Book’ – patients must be offered choice of 5
hospitals.
b. Payment by Results (PbR) ‐ movement from negotiated
to fixed prices (HRGs ‐ similar to U.S. DRGs).
i. PbR accounts for almost 70% of activity.
c. Reward /Penalties for Performance.
i. Foundation Trust Status – Retain net income.
ii. Poor performance – management replacment, closure, merger.
9. Expected Effects of the Reform
1. Expected effects
a. ‘Choose and Book’ – increase elasticity of demand facing
hospitals.
b. PbR – change conduct .
i. Hospitals paid for activity.
ii. Focus on quality as prices fixed.
2. Do hospitals have incentives to respond?
a. Not for profit, annual budget constraint.
b. Poor financial & clinical performance heavily penalised.
c. PbR system is very highly geared.
i. Levels of prices key.
27. Econometric Issues
1. There may remain concerns over endogeneity of
concentration + patient heterogeneity.
2. Control for patient heterogeneity with observables.
a. Patient age, sex, severity (Charlson index) .
b. Local area health, income.
c. Include hospital fixed effects.
3. Replace actual HHI with a measure of market structure
based on factors unrelated to quality or unobserved
patient heterogeneity.
4. Also concerns about whether DiD assumptions are met.
a. Are there pre‐existing differences (observable and
unobservable) between hospitals with different market
structures?
31. DiD Estimates of Market Structure on
Outcomes and Waiting Times
(1) (2) (3) (4) (5) (6) (7)
28 day
AMI
mortality
rate
(in-
hospital,
ages 55+)
30 day
AMI
mortality
rate
(on or after
discharge,
ages 35-
74)
28 day all
causes
mortality
rate
(in-
hospital)
28 day
mortality
rate
(in-
hospital,
excluding
AMI)
MRSA
bacterae
mia
rate
Patients
waiting
3
months
or more
Attendance
s
spending
less than
4 hours in
A&E
DiD coefficient 0.246*** 0.313** 0.069** 0.066** -0.110 0.078 -0.005
(0.084) (0.116) (0.027) (0.028) (0.118) (0.167) (0.011)
Hospitals 133 133 162 162 161 162 150
Observations 250 250 323 323 318 323 299
39. Tests of DiD Assumptions
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Covariate
Total
admissions
AMI
admissions
(ages 55+)
Doctors
(share of
clinical
staff)
Qualified
clinical
staff
(share of
clinical
staff)
Area
standardized
mortality
rate
Case
mix
Index of Multiple
Deprivation
(average for
patients’ areas of
residence)
Charlson
index
(average for
admissions at
the hospital)
28 day AMI
mortality rate
(in-hospital,
ages 55+)
30 day AMI
mortality rate
(on or after
discharge,
ages 35-74)
28 day
all-cause
mortality
rate
(in-hospital)
Coefficient -0.624 -0.051 5.329 -10.233 -1.695 -0.001 81.732 -1.080 -1.808 47.847
(0.642) (0.086) (8.792) (7.947) (1.936) (0.003) (72.197) (4.751) (6.623) (42.840)
P-value for Wald
test 0.129
Observations 162 151 161 161 162 162 162 162 130 130 162
40. Robustness tests
(1) (2) (3)
28 day AMI
mortality rate
(in-hospital, ages 55+)
28 day all-cause
mortality rate
(in-hospital)
Mean
length-of-stay
(days)Robustness test
1. Baseline 0.246*** 0.069** 0.254***
(0.084) (0.027) (0.059)
Observations 250 323 323
2. Placebo DiD test for 2001-2003 -0.047 0.005 -0.036
(0.077) (0.027) (0.047)
Observations 250 309 309
3. Using time invariant pre-reform HHI level (2003) 0.216*** 0.066** 0.245***
as market structure measure (0.079) (0.028) (0.059)
Observations 250 323 323
4. Controlling for the Charlson index 0.246*** 0.067** 0.239***
(0.084) (0.027) (0.060)
Observations 250 323 323
5. Controlling for the Index of Multiple Deprivation 0.278*** 0.067** 0.263***
(0.085) (0.029) (0.061)
Observations 250 323 323
6. Controlling for surpluses/deficits 0.242** 0.076** 0.229***
(0.093) (0.030) (0.066)
Observations 236 302 302
7. All hospitals (weighted by number of admissions) 0.138** 0.069*** 0.261***
(0.069) (0.024) (0.061)
Observations 299 323 323
8. Using levels of the dependent variable and HHI 0.170** 0.069*** 0.197***
(implied elasticity)
Observations 250 323 323
9. Controlling for income (male wage in area) 0.247*** 0.061** 0.258***
(0.086) (0.029) (0.062)
Observations 248 319 319
10. Controlling for the share of urgent ambulance calls 0.238**
responded within eight minutes (0.100)
Observations 233
41. Magnitude of Effects
41
28 day
mortality rate
(all causes)
Panel A - Observed magnitudes
Average number of admissions (2003/04) 63,094
Average number of deaths (2003/04) 1,135.1
Average mortality rate (2003/04) (%) 1.799%
Average number of admissions (2007/08) 72,558
Average number of deaths (2007/08) 1,053.5
Average mortality rate (2007/08) (%) 1.452%
Average change in deaths (2003-07) (positive = deaths averted) 81.5
Average decrease in predicted HHI (2003/04-2007/08) -118
Panel B - Continuous HHI: magnitudes implied by estimated coefficient (summary stats refer to HHI 03)
Baseline coefficient (elasticity) (%) 0.069
Scenario 1: Average decrease in HHI (Policy impact)
Implied counterfactual percentage increase in the mortality rate 2007/08 per hospital (for elasticity calculated at mean HHI) 0.2%
Total lives saved for the whole sample of hospitals (N = 162) 327
Total number of years of life saved for the whole sample of hospitals 3,354
Total savings in £million (value of year of life = £60,000/p.a.) £201
Scenario 2: One standard deviation increase in HHI 03 (= increase of 1928 units, from 4,353 to 6,281 in the whole sample)
Implied counterfactual percentage increase in the mortality rate 2007/08 per hospital (for elasticity calculated at mean HHI) 3.1%
Total lives saved for the whole sample of hospitals (N = 162) 5,336
Total number of years of life saved for the whole sample of hospitals 54,771
Total savings in £million (value of year of life = £60,000/p.a.) £3,286
42. No change in patient type except for
IMD for good hospitals
42
Change 2007-2005
(1) (2) (3) (4) (5) (6)
Elective
admissions
Number of
MSOAs
(electives)
Mean
distance
travelled
(electives)
Mean
IMD
ranking
Charlson
index
Number of
diagnoses
Mean waiting time 56.434* 0.694** 0.004* 13.836*** 0.00002 0.001
(elective admissions) (33.680) (0.295) (0.002) (3.061) (0.00013) (0.001)
Overall quality of services 1165.859 21.762** 0.078 207.347** -0.00030 0.024
(score) (897.260) (9.294) (0.076) (95.395) (0.00519) (0.031)
In-hospital mortality rate -942.177 14.816* -0.175* 14.927 -0.01229* -0.005
(all causes) (1255.291) (8.718) (0.103) (91.709) (0.00638) (0.050)
In-hospital mortality rate -91.980 -1.035 0.001 26.533 0.00086 0.002
(AMI) (227.052) (1.029) (0.011) (21.365) (0.00131) (0.005)
Teaching hospital status 1234.181 20.211 -0.098 335.378** -0.00041 -0.034
(2088.132) (17.763) (0.133) (143.230) (0.01348) (0.042)
Note: Index of Multiple Deprivation (over all patients), where patients in the most deprived locality in the year are attributed
the ranking of 1 and higher values are attributed to patients living in less deprived areas.
43. Competitive hospitals in 2005 not attracting observably
sicker/different patients but are attracting more patients
43
Change 2007-2005
(1) (2) (3) (4) (5) (6)
Elective
admissions
Number of
MSOAs
(electives)
Mean
distance
travelled
(electives)
Mean
IMD
ranking
Charlson
index
Number of
diagnoses
Level of HHI -483.360 -2.620 0.032 41.200 0.002 -0.003
(505.402) (2.788) (0.025) (43.781) (0.002) (0.015)
Indicator for bottom 3156.728** 18.499* -0.084 -139.657 0.001 0.004
quartile of HHI (1513.533) (10.949) (0.096) (162.770) (0.008) (0.056)
Number of hospitals 162 162 162 162 162 162
Note: Index of Multiple Deprivation (over all patients), where patients in the most deprived locality in the year are attributed
the ranking of 1 and higher values are attributed to patients living in less deprived areas.