Free to choose?
Free to choose?
Greater choice and competition in healthcare is a popular reform model. This column discusses recent research suggesting that once restrictions on choice in the UK’s NHS were lifted, patients receiving cardiac surgery became more responsive to the quality of their care. This saved lives and gave hospitals a greater incentive to improve quality.
A central plank of the NHS reforms implemented by the UK Labour government of the 2000s was the introduction of patient choice. For the first time in the history of the NHS it was mandated that patients should have a say in the choice of hospital when being referred for an elective treatment. Rather than relying entirely on their general practitioner (GP), patients were now offered a set of five hospitals to choose from. At the same time, GPs were no longer tied to a particular hospital through selective contracting agreements and could refer patients more easily to any available hospital in the country. The intention behind the reform was to make referrals more responsive to hospital quality. The argument was that this in turn would increase hospitals’ incentives to improve quality. Although most economists subscribe to the idea that more choice always constitutes an improvement, things are slightly more difficult in the case of healthcare. Contrary to most consumer goods markets, evaluating hospital quality is not a trivial task, and patients might indeed find it hard to pick the best hospital for a particular treatment. This is presumably even more of an issue in the UK, where patients are not accustomed to makingchoices in the context of healthcare.
Do referrals respond to quality?
We set out to analyse whether referral patterns did indeed become more responsive to quality after the introduction of the reform (Gaynor, Propper and Seiler 2012). As a first test for whether this is the case, we analyse whether relatively better hospitals attracted a larger number of patients for one particular procedure: elective coronary artery bypass graft (i.e., heart bypass surgery). We look at the relationship between hospital quality, as measured by patient survival rates, and market shares separately for the time periods before and after the reform. Interestingly, we find that market shares are not correlated with hospital quality pre-reform, however they show a significant correlation with patient survival in the post-reform period. This gives us a first piece of evidence suggesting that patients were indeed allocated to relatively higher-quality hospitals after the reform (but not before). The magnitude of the effect of quality on market shares is economically significant. Post-reform, a one percentage point lower mortality rate led to the hospital attracting 20 more patients every year. This corresponds to about a 5% increase in market share.
In a second step we model patient behavior at a more micro level. Instead of analysing only aggregate shares of patients at each hospital, we model hospital choice individually for each patient. This allows us to incorporate the effect of the patient’s location relative to the hospital as well as to analyse how reactions vary across different patient groups.
We find, also at this level of analysis, that patients became more sensitive to the quality of service as measured by patient survival. However the effect differs substantially across patient groups. Our results show that more severely ill patients react more strongly to the reform, i.e. they are even more likely than the average patient to end up at a high-quality hospital post-reform. In other words, we see the reform having the strongest effect for the group of patients that is presumably most in need of high-quality treatment. Similarly, we find a stronger effect of the reform on patients who reported in a survey that they were informed about the choice reform at the point of referral.
Finally, we also analyse whether poorer patients reacted differently to the reform. A major concern of skeptics of the reform was that only affluent, well-educated patients would be able to process the necessary information and make an educated choice. According to this logic the reform would therefore effectively increase inequality in quality of healthcare. Our analysis however, shows that fortunately these concerns were unfounded. We find that poorer patients reacted no differently from other income groups to the introduction of choice. They did indeed react slightly more strongly, but the difference is not statistically significant, i.e. we cannot confidently tell it apart from noise in our data.
We then employ our statistical estimates to quantitatively evaluate the impacts of the reform. We perform the following thought experiment. We compare the actual hospital choices patients made post-reform with the hypothetical choices the same set of patients would have made prior to the reforms, i.e., had referral patterns not changed in response to the freeing of choice. Because we know the sensitivity of referrals to quality pre- and post-reform, we can calculate the probability of visiting each hospital that is available to the patient under either level of responsiveness to quality. Artificially depriving patients of the benefits of the reform allows us to see to what extent patients would have ended up in lower quality hospitals in the absence of patient choice. Using the hospitals’ patient survival rates (adjusted for differences in case-mix, i.e. the severity of the cases treated at each hospital) we find that nine fewer patients (relative to slightly over 300 deaths a year, i.e. around 3%) would have survived every year had the reform not been implemented.
Unexpected drop in mortality rates
The drop in mortality post-reform is an important effect of the introduction of patient choice that was not emphasised by policymakers. Even if the increased responsiveness of referrals to quality did not change hospitals’ performance in any way, the reallocation of patients still leads to better health outcomes. This is due to the fact that patients now visit on average a higher quality hospital from the existing distribution of quality across hospitals.
The hope of the reform was that the quality distribution itself would change due to hospitals’ increased incentives to improve quality in order to attract patients. We next analyse whether there is reason to believe that such a change in hospital quality might have happened. To get a sense of the effect of the change in referrals on hospitals, we compute the change in patient admissions to each hospital if patients had been choosing their hospitals pre-reform according to post-reform referral patterns. This thought experiment – what would have happened had the reform been adopted earlier – gives us a direct sense of how much competitive pressure hospitals found themselves facing when the choice reform kicked in. When calculating the change in market shares we find a very substantial impact for some hospitals, with one hospital losing almost 10% of its market share due to the effects of the reform. There is substantial heterogeneity in the impacts across hospitals however, with most hospitals experiencing more modest changes in market share of around 2% or 3%. There is therefore good reason to believe that at least a significant subset of hospitals had substantial incentives to improve quality in order to retain (or enhance) their market shares of patient admissions.
Do hospitals pay attention to the new system?
In a final step we directly analyse whether there is any evidence of hospitals reacting to the change in referrals by increasing quality, as measured by a fall in mortality rates. To this end we look at whether hospitals that faced the strongest pressure post-reform (as measured by the potential loss in admissions) saw a bigger decline in their mortality rates than other hospitals. We do indeed find that this is the case. Hospitals which had the biggest increases in the responsiveness of patient admissions to their mortality rates had the largest declines in mortality rates, and vice versa. This result closely mirrors related work by Cooper et al. (2011) and Gaynor et al. (2012) who show that areas with more competition experienced a larger increase in patient survival rates after the introduction of patient choice.
In summary, we assess that the reform reduced cardiac bypass surgery mortality by 3% by re-allocating patients to better hospitals. This is clearly a lower bound on the beneficial effect one might expect from allowing choice, as we look only at the effect for one particular procedure. Secondly, we find evidence suggesting that hospitals responded to increased choice by improving their quality. If this is mirrored as a hospital-wide effect, there may be substantial additional positive benefits for patients. Finally, our findings add support to earlier evidence that indicate that the choice reforms led to falls in mortality in other treatments and shorter lengths of stay without increasing hospital total costs (Gaynor et al., 2012; Cooper et al., 2011) and also to work by Bloom et al. (2010) that indicates that more competitive environments have better management practices that are in turn associated with better hospital performance.
Bloom, N, C Propper, S Seiler, J van Reenen (2010), “The Impact of Competition on Quality: Evidence from Public Hospitals”, NBER Working Paper, 1630.
Cooper, Z, S Gibbons, S Jones, and A McGuire (2011), “Does Hospital Competition Save Lives? Evidence from the English Patient Choice Reforms”, Economic Journal, 121 (554), F228-F260.
Gaynor, M, R Morena-Serra, C Propper (2012), “Death by Market Power: Reform, Competition and Patient Outcomes in the British National Health Service”, forthcoming in American Economic Journal: Economic Policy, See also University of Bristol Centre for Market and Public Organisation, Working Paper, 10/242.
Gaynor, Martin, Carol Propper and Stephan Seiler (2012), “Free to Choose? Reform and Demand Response in the English National Health Service”, CEP Discussion Paper, 1179, November.