During the last decade the pay-for-performance paradigm has increasingly attracted the interest of all parts involved in health care systems. There is no clear information about how will this affect in the short-term the quality of those systems.1 Despite improvement in quality indicators at patient-level from quasi experimental studies,2 it is difficult to demonstrate if this paradigm would deeply improve medical and non-medical processes in real life scene-ries.1,3
Moreover, real data from specific settings in which the pay-for-performance strategy is used, demonstrates an almost null correlation between the size of the economic benefit and the expected health gain. This could suggest that an expected health gain may increase the risk of skewing activity towards areas with high workload but relatively low benefits for health.4
In this issue Robert P. Myers, et al. demonstrate critically how difficult or biased could be to properly categorize a hospital. The cirrhotic population makes an interesting example, in which an improvement in survival has been observed during the last decades. In this manuscript it is shown that the analysis of mortality, as clinical outcome, is very important for the decision makers. However, other significant outcomes for the patients should be explored in the future, such as admission to emergency room, invasive procedures, quality of life, among others. This issue is very important because most of the hospitals considered in this article have average expected mortality rates, but when other liver related complications are taken into account, such as ascites management, only one third of the patients receive all recommended care.5 Additionally, some uncertainty is reported about which outcome is the most suitable to assess the quality of a hospital. Recent information suggests that statistical tools used to analyze this issue should be considered in detail.6 Commonly, death or liver transplantation are considered as a composite outcome in many clinical trials, and they are considered a metric value to assess prognosis in liver cirrhotic patients. However, this could be difficult to assess in this kind a quality assessment, as was previously reported regarding another paradigm (volume-based benchmarks), in which the combination of a valid benchmark and a powerful incentive (accreditation and reimbursement in this case) may be insufficient to ensure uniform hospital/provider compliance.7
The article from Myers, et al., brings to the fore the outliers in the health system, especially those hospitals with worse than expected mortality in comparison with hospitals with better than expected mortality, suggesting that it is necessary to find alternative ways to reduce mortality in the health care system.
This provocative article shows the difficulties for maintaining concordance between several models in order to adjust for mortality, raising questions about how the model selected to adjust for mortality (and assess quality) will affect medical care in patients with cirrhosis. Maybe in the future we will be able to make medical care fit economic and mathematical models so we can determine the quality of a health system. Nowadays evidence is lacking about how to increase the quality of hospitals that are below the median in quality indicators.8
Finally, assessment of quality is needed in order to take decisions when looking for the best possible care for our patients with a rational use of resources. However, this assessment is a very complex issue in constant analysis, and so other alternatives should be considered in the future without unfocu-sing our primary goal: the patients.
