SEARCH

SEARCH BY CITATION

Keywords:

  • length of stay;
  • public hospitals;
  • treatment centres;
  • private providers

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

We investigate differences in patients’ length of stay between National Health Service (NHS) public hospitals, specialised public treatment centres and private treatment centres that provide elective (non-emergency) hip replacement to publicly funded patients. We find that the specialised public treatment centres and private treatment centres have, on average, respectively 18% and 40% shorter length of stay compared with NHS public hospitals, even after controlling for differences in age, gender, number and type of diagnoses, deprivation and regional variation. Therefore, we interpret such differences as because of efficiency as opposed to selection of less complex patients. Quantile regression suggests that the proportional differences between different provider types are larger at the higher conditional quantiles of length of stay. Copyright © 2012 John Wiley & Sons, Ltd.

1 INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

The hospital sector has lagged behind other sectors of the economy in moving towards greater specialisation (Essletzbichler, 2003) but is beginning to catch up. The number of orthopaedic, cardiac or general surgery specialist hospitals in the USA increased from 29 in 1990 to 91 in 2005 (Shactman, 2005; Schneider et al., 2008) after which the government imposed a moratorium on further development, concerned primarily that hospitals were specialising merely on the most profitable procedures (Shactman, 2005). In contrast, the English government has been actively encouraging the creation of both public and private treatment centres, located in areas with insufficient existing capacity (House of Commons Health Committee, 2006), that specialise in a limited set of elective procedures.

In this study, we assess whether provision of care in specialised treatment centres is more efficient than in the more traditional hospital setting. Efficiency is examined by comparing differences in hip replacement length of stay (LOS) among public hospitals and public and private (specialised) treatment centres. LOS often is used as a good proxy of the costs of treating patients (Ellis and McGuire, 1996; Gilman, 2000; Norton et al., 2002). Efficiencies in treatment centres may derive from economies of scale, whereby the unit cost of treatment falls as volume increases, and from specialisation, where it is cheaper to concentrate on providing a limited set of activities, rather than a diverse range of services (Schneider et al., 2008). The relative efficiency of private and public provision also may be driven by the different profit motive, the expectation being that private providers have stronger incentive to contain costs and behave more efficiently.

We focus on hip replacement because it is a common elective procedure performed on more than 40 000 patients every year in England. Moreover, the number of hip replacements has been steadily increasing in the past and likely to increase in the future driven by ageing of the population (Stargardt, 2008). There also is policy interest in this procedure, given the long waiting times which characterise it (Siciliani and Hurst, 2004). Indeed, one of the reasons for introducing treatment centres was the potential reduction in waiting times for this and other elective procedures like cataract surgery. Note, however, that the move towards more specialised centres was never pursued aggressively, with the government anticipating only around 10% of elective procedures being undertaken by private treatment centres (House of Commons Health Committee, 2006). In 2006/2007, the actual figure was around 2.3% (Street et al., 2010). Hip replacement was one of the procedures most commonly provided in treatment centres.

Differences in LOS may be indicative not of efficiency but patient selection. Selection may be due to diverse causes. First, private providers may ‘cherry pick’ less severe cases within any reimbursement category to boost profits (Shactman, 2005). In contrast, even if public providers are able to retain surpluses, these must be re-invested, so the absence of external claimants to surpluses places them under less pressure to engage in selection of less costly patients. Second, treatment centres tend to be less well equipped than hospitals, making them less suited to provide complex care. For this reason, treatment centres usually apply exclusion criteria (Mason et al., 2008). Third, hospitals tend to be more prestigious and attract highly specialised doctors with the skills to treat more complex cases. As these factors may lead to differences in patient complexity across organisations, it is important to account for patient complexity in evaluating relative efficiency.

We contribute to the extensive literature which investigates differences in behaviour between types of providers. Efficiency studies tend to find either that public providers are more efficient or that differences are not related to ownership.1

2 ECONOMETRIC SPECIFICATION

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

The purpose of this study is to investigate differences in patients’ LOS between types of provider.2 First, we define two indicator variables:

  • Define d1j = 1 if provider j is specialised
  • Define d2 j = 1 if provider j is privately owned

Our linear regression model is

  • display math(1)

where LOSij is the LOS of patient i in provider j; sij is a vector of variables, which captures regional dummies, demographic variables (age and gender), and patients’ co-morbidities through recorded diagnoses.

All of the private providers in our sample are specialised (they are private treatment centres), so we only observe d2j = 1 if d1j = 1. The base case, d1j = d2j = 0, is public (non-specialised) hospitals, so we can interpret the coefficients:

  • Effect of specialisation (only) over public hospital = β1
  • Effect of private ownership over specialisation (only) = β2 – β1

The coefficients β1 and β2 capture the extent to which patients treated in public and private (specialised) treatment centres differ in their LOS from those treated in public hospitals. By comparing the estimates of β1 and β2 when the vector of covariates sij is included and when it is not, we can identify the extent to which differences in LOS are because of treatment centres treating patients of different complexity (selection) as opposed to differences in their efficiency. For example, suppose that when omitting the vector sij in the regression equation, we find that β2 is negative. Then, patients treated in private treatment centres have a shorter LOS. If after the inclusion of the vector sij, the coefficient β2 reduces, then part of the differences in LOS can be attributed to differences in the characteristics of patients being treated (i.e. to selection).

Because the distribution of LOS is skewed, we use the log transformation of LOS as the dependent variable. We estimate Equation (1) using OLS. To identify the differential impact of type of provider along the conditional distribution of LOS, we also apply quantile regression methods.

3 DATA

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

We use data from the Hospital Episode Statistics (HES) in financial year 2006/2007. HES comprises individual patient records about everyone whose care was funded by the English National Health Service (NHS). We focus on those patients who received a cemented or uncemented primary hip replacement (HRG H80 or H81). Each patient record contains a range of variables including demographic (e.g. age and gender) and clinical information (e.g. diagnosis, procedures performed).3 The estimation sample includes 42 948 patients, of which, 1841 were treated by public treatment centres and 938 by private treatment centres. The sample includes 173 public hospitals, six public treatment centres, and 14 private treatment centres.

We control for various patient characteristics including age, gender, and number and type of diagnosis. For age, we construct seven groups: 18–29 years old, 30–39, 40–49, …, 70–79, and above 80. We use the diagnosis fields in HES, which record up to 12 diagnoses using the International Classification of Diseases version 10 (ICD-10) codes. For the type of diagnosis, we include dummy variables for each individual diagnosis to allow a fully flexible (non-linear) specification. For primary diagnoses, because of the extremely large number of diagnoses recorded, we only include dummies for diagnoses with at least 40 observations; this gives 28 dummy variables. The three most common primary diagnoses are different types of coxarthrosis (arthritis of the hip). Similarly, for secondary diagnoses, to keep the number of variables to a manageable level, we only include dummy variables for the most common 37 individual secondary diagnoses, covering 80% of admissions. The four most common are hypertension (high blood pressure), presence of (existing) joint implants, type 2 diabetes and asthma. Because we do not have a dummy variable for every secondary diagnosis, and HES records up to 11 secondary diagnoses for each patient, we also control for the number of additional diagnoses (for which we do not have an individual diagnosis dummy) using a dummy variable specification. We control for the number of procedures for each patient in the same way. The average per-patient number of diagnoses and procedures is larger in public hospitals (2.8, 2.3) than in public treatment centres (2.5, 2.2) or private treatment centres (1.1, 1.5).

We control for two characteristics of hospitals, which may influence LOS: Foundation Trusts and teaching status. The government has granted Foundation Trusts greater financial independence than other public hospitals, giving them a stronger incentive to contain costs. Teaching hospitals may have longer LOS because of sicker patients, higher quality of care and more time spent with patients for teaching purposes. We also control for geographical factors including dummy variables for the Strategic Health Authority (region) and the income deprivation of each patient's local area, the rationale being that timely discharge may be more difficult to arrange in more deprived areas.4

Table 1 provides the descriptive statistics. Public hospitals have the longest LOS (7.5 days) followed by public treatment centres (5.9 days) and private treatment centres (4.5 days). There is a higher proportion of uncemented hip prosthesis (HRG H81) in public treatment centres (48%) than in hospitals (29%) or private treatment centres (11%). Current NICE guidance favours cemented hip prosthesis as being lower cost and more viable in the long term (NICE, 2000). The descriptives suggest that patients differ in terms of age, gender and diagnostic characteristics across providers.

Table 1. Descriptive statistics
VariableNHS public hospitalPublic treatment centrePrivate treatment centreAll
 MeanSDMeanSDMeanSDMeanSDMinimumMaximum
Length of stay7.4554.7805.8662.5724.4811.4947.3224.6891155
HRG H81 (uncemented)0.2910.4540.4950.5000.1130.3170.2960.45601
Foundation Trust0.3130.4640.1740.3800.0000.0000.3010.45901
Teaching Trust0.1180.3230.0000.0000.0000.0000.1110.31401
IMD—income deprivation0.1190.0980.1060.0880.0950.0680.1170.09700.83
Age68.67510.91669.34410.68469.9988.69968.73310.86518102
Female0.6190.4860.6430.4790.5860.4930.6190.48601
Individual primary diagnoses:          
M169 ‘Coxarthrosis, unspecified’0.4990.5000.1930.3950.4450.4970.4850.50001
M161 ‘Other primary coxarthrosis’0.2940.4560.5520.4970.5140.5000.3100.46301
M160 ‘Primary coxarthrosis, bilateral’0.0620.2400.0240.1530.0090.0920.0590.23501
Individual secondary diagnoses:          
I10X ‘Hypertension’0.3190.4660.3530.4780.0580.2330.3140.46401
Z966 ‘Presence of orthopaedic joint implants’0.1010.3010.1540.3610.0050.0730.1010.30101
E119 ‘Non-insulin dependent diabetes’0.0540.2260.0610.2400.0120.1080.0530.22501
J459 ‘Asthma’0.0510.2200.0370.1900.0070.0860.0490.21601
Number of diagnoses2.8281.8572.4941.5311.1430.5572.7771.843112
Number of procedures2.3210.8082.1520.5121.4650.5082.2950.802110
Transfers-in0.0010.0380.0000.0000.0000.0000.0010.03701
Transfers-out0.0230.1490.0100.0980.0090.0920.0220.14601
Observations40169184193842948

4 RESULTS

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Table 2 provides the OLS estimates of the model described in Equation (1).

Table 2. Log-linear regressions
 [1][2][3]
 CoefficientSE CoefficientSE CoefficientSE 
  1. Notes: OLS regressions of ln(length of stay) on three different sets of regressors. Reference category patient is in an NHS public hospital, HRG H80, Age70–79, male, primary diagnosis = M169 (coxarthrosis, unspecified), no secondary diagnosis, 1 procedure, 1 diagnosis, North-East STHA. Models 2 and 3 contain 9 additional dummy variables for Strategic Health Authorities. Model 3 includes 26 additional dummy variables for primary diagnoses, 33 additional dummy variables for individual secondary diagnoses, 6 additional variables for 5–10 procedures and 8 additional dummy variables for 4–11 additional secondary diagnoses for which the coefficient estimates are not shown. Standard errors are adjusted for clustering at the hospital/treatment centre level.

Public treatment centre−0.1710.018*** −0.1990.031*** −0.1770.031*** 
Private treatment centre−0.4630.049*** −0.4910.062*** −0.4010.065*** 
HRG H81−0.1290.017*** −0.1180.015*** −0.0330.016** 
Foundation Trust   0.0110.028 0.0150.026 
Teaching Trust   −0.0130.036 −0.0200.033 
IMD—Income deprivation   0.2170.050*** 0.2030.044*** 
Age 18–29      −0.2120.028*** 
Age 30–39      −0.2000.023*** 
Age 40–49      −0.1810.015*** 
Age 50–59      −0.1580.008*** 
Age 60–69      −0.1170.006*** 
Age 80+      0.2150.008*** 
Female      0.0900.006*** 
Individual primary diagnosis (26 dummy variables)         
M161 ‘Other primary coxarthrosis’      −0.0040.020 
M160 ‘Primary coxarthrosis, bilateral’      0.0010.024 
Individual secondary diagnoses (33 dummy variables)         
I10X ‘Hypertension’      0.0030.006 
Z966 ‘Presence of orthopaedic joint implants’      −0.0480.009*** 
E119 ‘Non-insulin dependent diabetes’      0.0690.010*** 
J459 ‘Asthma’      0.0370.010*** 
Number of additional secondary diagnoses         
1 Additional diagnosis      0.0650.007*** 
2 Additional diagnoses      0.1000.010*** 
3 Additional diagnoses      0.1720.016*** 
Number of procedures         
2 Procedures      −0.0500.058 
3 Procedures      −0.0190.059 
4 Procedures      0.0530.062 
Transfer-in      −0.0840.149 
Transfer-out      0.0320.047 
Constant1.9330.013*** 1.8970.014*** 1.8560.067*** 
Observations429484294842948
R20.0450.0650.267
F-statistic (p-value)87.9 (0.000)15.63 (0.000)98.46 (0.000)

Looking first at column [1], with no controls (except for a dummy for HRG H81), patients treated in public treatment centres have 17% shorter LOS, and those treated in private treatment centres have 46% shorter LOS than those treated in hospitals. As shown in column 2, these differences are little changed after controlling for hospital characteristics, regional dummies and local income deprivation: public and private treatment centres have 20% and 49% shorter LOS, respectively, compared with hospitals. Foundation and teaching status of hospital appear to have no effect on LOS, but income deprivation does have a substantial impact, increasing LOS.

Column 3 shows results when controlling for patient's age, gender, type and number of diagnoses. These patient characteristics have substantial explanatory power, increasing the regression R2 by 20 percentage points. However, conditioning on these variables does not substantially change differences in LOS across providers: public treatment centres still have 18% shorter LOS than hospitals. This suggests that the overall contribution of patient characteristics does not vary sufficiently between these provider types to explain differences in LOS. The difference between hospitals and private treatment centres is reduced to 40% (6 percentage points down). This suggests that patient selection explains a small part of the shorter LOS achieved by private treatment centres

Table 3 and Figure 1 provide quantile regression results. The specification is similar to model [3] in Table 2 (in terms of variable choice). To allow quantile regression models to converge, we simplify the specification of dummy variables for diagnoses to include only the most common ten diagnoses for primary and secondary diagnoses. The results suggest that the proportionate difference between (non-specialised) public hospitals, public (specialised) treatment centres and private (specialised) treatment centres are larger at the higher conditional quantiles of LOS and smaller at the lower quantiles. Public treatment centres have 26% shorter LOS compared with public hospitals at the 90% quantile, falling to 9% at the 10% quantile. Similarly, private treatment centres have 52% shorter LOS compared with public hospitals at the 90% quantile, reducing to 35% at the 10% quantile. Figure 1 plots the effect of provider type on LOS over the five conditional quantiles. Given that there may be a ‘minimum’ LOS for hip replacement patients, at low quantiles, doctors and nurses have little scope to substantially reduce LOS even further. At high quantiles, the LOS is much longer, and there is more scope for treatment centres to be able to discharge patients sooner. However, we must interpret the very large effects for private treatment centres at the highest quantiles with some caution, given the small number of observations with high LOS in private centres.

Table 3. Quantile regressions
  1. Quantile regressions of the 10th, 30th, 50th, 70th and 90th conditional percentiles of ln(length of stay). Reference category patient is in an NHS public hospital, HRG H80, Age70–79, male, primary diagnosis = M169 (coxarthrosis, unspecified), no secondary diagnosis, North-East STHA. Additional dummy variables for six age groups, seven individual primary diagnoses, six individual secondary diagnosis, the number of additional secondary diagnoses (one to six), five procedures and nine Strategic Health Authorities also are included, but coefficient estimates are omitted to save space.

 10%30%50%70%90%
 CoefficientSE CoefficientSE CoefficientSE CoefficientSE CoefficientSE 
Public treatment centre−0.0900.015*** −0.1600.010*** −0.1920.011*** −0.2180.012*** −0.2640.017*** 
Private treatment centre−0.3500.030*** −0.4230.019*** −0.4910.020*** −0.5020.017*** −0.5200.018*** 
HRG H81−0.0480.007*** −0.0390.005*** −0.0340.005*** −0.0230.005*** −0.0050.008 
Foundation Trust0.0230.006*** 0.0080.005 0.0160.005*** 0.0140.006** 0.0160.009* 
Teaching Trust0.0010.009 −0.0180.008** −0.0210.009** −0.0130.008 −0.0130.013 
IMD–Income deprivation0.0680.034** 0.1620.023*** 0.1950.023*** 0.2640.023*** 0.3580.040*** 
Female0.1220.006*** 0.0990.005*** 0.0900.005*** 0.0750.005*** 0.0600.008*** 
Individual primary diagnosis (10 dummy variables)               
M161 ‘Other primary coxarthrosis’−0.0100.006 0.0010.006 0.0120.005** 0.0130.005** 0.0010.009 
M160 ‘Primary coxarthrosis, bilateral’0.0110.014 0.0040.009 0.0120.010 0.0110.011 0.0160.014 
Individual secondary diagnoses (10 dummy variables)               
I10X ‘Hypertension’0.0070.007 0.0010.004 −0.0020.005 −0.0010.005 −0.0060.008 
Z966 ‘Presence of orthopaedic joint implants’−0.0580.009*** −0.0410.008*** −0.0530.007*** −0.0480.008*** −0.0580.012*** 
E119 ‘Non-insulin dependent diabetes’0.0570.016*** 0.0580.010*** 0.0650.01*** 0.0720.013*** 0.0850.020*** 
J459 ‘Asthma’0.0420.011*** 0.0380.010*** 0.0360.01*** 0.0320.009*** 0.0380.020* 
Transfer-in−0.2920.061 −0.2010.155 −0.0430.085 0.0540.048 −0.0540.128 
Transfer-out−0.1130.024 −0.0030.017 0.0300.015 0.0930.023 0.2070.043 
Constant1.5060.030 1.7630.023 1.9360.020 2.1230.023 2.5650.049 
Observations4294842948429484294842948
Pseudo R20.0740.1020.1170.1290.172
image

Figure 1. Marginal effects across quantiles of ln(length of stay)

Download figure to PowerPoint

5 DISCUSSION AND CONCLUDING REMARKS

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

Government policy in England has been to encourage delivery of care to NHS patients in specialised treatment centres rather than in traditional hospital settings. This policy has been subject to criticism, particularly pertaining to the generous contracts awarded to private treatment centres whereby they were paid around 11% more per patient than their public counterparts (House of Commons Health Committee, 2006; Pollock and Godden, 2008; Mason et al., 2010; Street et al., 2010). Despite these criticisms, the evidence presented here demonstrates that the LOS for people having a hip replacement is lower in treatment centres than in hospitals. This result holds even though private treatment centres were more generously funded, which would have dampened their incentive to reduce LOS. Nor is the result because of the different characteristics of patients, which we have taken into account using routine data. This is despite poor depth of coding by some private treatment centres (Mason et al., 2010), which would imply that the complexity of their casemix is under-estimated and hence, that the LOS differential could be even greater than that observed.

Our estimates might be contaminated by endogeneity if there is selection on unobserved patient characteristics. Public and private treatment centres could find ways to select ‘healthier’ patients, such as those with better pre-operative physical functioning, this not being recorded in our data. A standard approach to dealing with this form of endogeneity is to construct instrumental variables based on the patient's travel distance to the hospital or treatment centre (Kessler and McClellan, 2000). Unfortunately, we are not able to apply this approach as we do not have access to patients’ residential location in our data.

The location of treatment centres in areas in need of extra capacity also could be a source of endogeneity if these areas are correlated with lower LOS within each region (the region being accounted for with regional dummies). There also is the possibility that high socio-economic status patients, with low average LOS, select themselves into treatment centres to reduce their waiting time. We cannot rule out that these endogeneity concerns could affect our estimates. However, having controlled for a rich set of diagnosis and geographical variables in our regressions and still found very large effects, we believe it is highly unlikely that endogeneity explains the large differentials in LOS that we have observed.

Treatment centres may be able to deliver care more efficiently than hospitals because of their ability to benefit from specialisation, avoiding the disruption that hospital staff face by having to re-schedule elective work to accommodate patients requiring emergency care, and economies of scale in the production of a limited set of procedures (The Royal College of Surgeons of England, 2007). We have been unable to examine scale economies, particularly because no data are available about privately insured patients (or patients who pay out of pocket) treated in private treatment centres, thereby making it impossible to determine the full caseload in these providers. An examination of economies of scale also might require the ‘production line’ to be defined more broadly, recognising that specialised treatment centres usually also provide knee replacements and arthroscopies as well as other elective surgery alongside hip replacements (Mason et al., 2008).

We also find that LOS is lower in private treatment centres than in their public counterparts. This may be because they have a greater incentive than public providers to restrain costs through better use of resources and by ensuring timely discharge. It will be important to ensure that these lower lengths of stay are not due to shifting of responsibilities and costs to earlier or later stages of the care pathway and that they do not come at the expense of reduced outcomes. However, early evidence for these patients is reassuring (Browne et al., 2008), and payments have recently been refined to incentivise best practice across the care pathway (Department of Health, 2011).

ACKNOWLEDGEMENTS

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES

This paper arose out of a project originally undertaken for the English Department of Health. The views expressed are those of the authors and are not necessarily those of the Department of Health.

  • 1

    Herr (2008) finds that in Germany public hospitals are more efficient than private and non-profit hospitals. Farsi and Filippini (2008) in Switzerland find no significant differences between public, for-profit and no-profit hospitals. Marini et al. (2008) investigate the change in England of hospital status from 'public hospital' to 'Foundation Trust', a status which confers more financial independence and less monitoring and find that the new status had limited impact on behaviour. Barbetta, Turati and Zago (2007) find that the mean efficiency of public and non-profit hospitals converged after the introduction of the DRG system in Italy. In his review of 317 published papers on frontier efficiency measurement, Hollingsworth concluded that public hospitals tend to be more efficient than their private counterparts (Hollingsworth, 2008).

  • 2

    Length of stay is measured as the difference between the dates of the patient's admission to and discharge from the hospital.

  • 3

    All public providers routinely provide HES data for every inpatient and day case patient they treat. Private treatment centres are contractually obliged to submit HES data for the NHS funded patients they treat.

  • 4

    The IMD income deprivation score provides the proportion of the local population in the area where the patient lives living in households reliant on one or more means-tested benefits (Noble et al., 2004). Patients are from more deprived areas in public hospitals (12%) than in public or private treament centres (both 10%). We have nine dummy variables to account for the ten regions in England.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. 1 INTRODUCTION
  4. 2 ECONOMETRIC SPECIFICATION
  5. 3 DATA
  6. 4 RESULTS
  7. 5 DISCUSSION AND CONCLUDING REMARKS
  8. ACKNOWLEDGEMENTS
  9. REFERENCES
  • Barbetta GP, Turati G, Zago AM. 2007. Behavioral differences between public and private not-for-profit hospitals in the Italian national health service. Health Economics 16(1): 7596.
  • Browne J, Jamieson L, Lewsey J, van der Meulen J, Copley L, Black N. 2008. Case-mix & patients' reports of outcome in Independent Sector Treatment Centres: Comparison with NHS providers. BMC Health Services Research 8: Doi:10.1186/1472-6963-8-78.
  • Department of Health. 2011. Payment by Results Guidance for 2011–12. Department of Health: Leeds.
  • Ellis RP, McGuire T. 1996. Hospital response to prospective payment: moral hazard, selection, and practice-style effects. Journal of Health Economics 15(3): 257277.
  • Essletzbichler J. 2003. From mass production to flexible specialization: the sectoral and geographical extent of contract work in US manufacturing. Regional Studies 37: 753771.
  • Farsi M, Filippini M. 2008. Effects of ownership, subsidization and teaching activities on hospital costs in Switzerland. Health Economics 17(3): 335350.
  • Gilman BH. 2000. Hospital response to DRG refinements: the impact of multiple reimbursement incentives on inpatient length of stay. Health Economics 9(4): 277294.
  • Herr A. 2008. Cost and technical efficiency of German hospitals: does ownership matter? Health Economics 17(9): 10571071.
  • Hollingsworth B. 2008. The measurement of efficiency and productivity of health care delivery. Health Economics 17(10): 11071128.
  • House of Commons Health Committee. 2006. Independent sector treatment centres: fourth report of session 2005–06. Volume I: Report, together with formal minutes. The Stationery Office: London.
  • Kessler DP, McClellan MB. 2000. Is Hospital Competition Socially Wasteful? Quarterly Journal of Economics 115: 577615.
  • Marini G, Miraldo M, Jacobs R, Goddard M. 2008. Giving greater financial independence to hospitals. Does it make a difference? The case of English NHS trusts. Health Economics 17(6): 751775.
  • Mason, A, M Miraldo, L Siciliani, P Sivey, A Street. 2008. Establishing a fair playing field for payment by results. CHE Research Paper 39. York: University of York.
  • Mason A, Street A, Verzulli R. 2010. Private sector treatment centres are treating less complex patients than the NHS. Journal of the Royal Society of Medicine 103: 322331.
  • National Institute for Clinical Excellence. 2000. Guidance on the Selection of Prostheses for Primary Total Hip Replacement, Technology Appraisal Guidance – No. 2. National Institute for Clinical Excellence: London. http://guidance.nice.org.uk/TA2
  • Noble M, Wright G, Dibben C, Smith GAN, McLennan D, Anttila C, Barnes H, Mokhtar C, Noble S, Avenell D, Gardner J, Covizzi I, Lloyd M. 2004. Indices of Deprivation 2004 (Revised). Report to the Office of the Deputy Prime Minister. Neighbourhood Renewal Unit: London.
  • Norton EC, Van Houtven CH, Lindrooth RC, Normand ST, Dickey B. 2002. Does prospective payment reduce inpatient length of stay? Health Economics 11(5): 377387.
  • Pollock AM, Godden S. 2008. Independent sector treatment centres: evidence so far. British Medical Journal 336(7641): 421424.
  • Schneider JE, Miller TR, Ohsfeldt RL, Morrisey MA, Zelner BA, Li P. 2008. The economics of specialty hospitals. Medical Care Research and Review 65: 531553.
  • Shactman D. 2005. Specialty hospitals, ambulatory surgery centers, and general hospitals: charting a wise public policy course Health Affairs 24: 868873.
  • Siciliani L, Hurst J. 2004. Explaining waiting times variations for elective surgery across OECD countries. OECD Economic Studies 38(1): 96122.
  • Stargardt T. 2008. Health service costs in Europe: cost and reimbursement of primary hip replacement in nine countries. Health Economics 17(S1): S9S20.
  • Street A, Sivey P, Mason A, Miraldo M, Siciliani L. 2010. Are English Treatment Centres treating less complex patients? Health Policy 94: 150157.
  • The Royal College of Surgeons of England. 2007. Separating emergency and elective surgical care: Recommendations for practice. The Royal College of Surgeons of England: London.