[Introduction to risk adjustment methods in comparative evaluation of outcomes].

The increasing demand for comparative evaluation of outcomes requires the development and diffusion of epidemiologic research, the ability to correctly conduct analyses and to interpret results. When healthcare outcomes are used for comparing quality of care across providers, failure to use methods of risk adjustment to account for any variation in patient populations can lead to misinterpretation of the findings. The purpose of this paper is to provide a detailed but easy-reading review of different risk adjustment methodologies to compare health care outcomes. The paper is divided in two parts. Introduction describes the difference between experimental and observational studies, the role of confounding in observational studies and the ways confounding is identified and controlled (propensity adjustment and risk adjustment), Specific part on risk adjustment describes: (1) the methods for constructing the severity measures; (2) the methods that use the severity measures to obtain "adjusted" outcome measures for valid comparison between groups (stratified analysis, indirect and direct standardization); (3) identification and management of effect modification; (4) the methods to gain the precision of the estimates; (5) the risk adjustment methods used with multiple comparisons and (6) introduction to other models (multi-level models) used for risk adjustment. For policy makers and planners, epidemiologists and clinicians it is important to understand which factors can improve or worsen the effectiveness of treatments and services and to compare the performances of hospitals and healthcare providers. Decisions should be based on the validity and precision of study results, by using the best scientific knowledge available. The statistical methods described in this review cannot measure reality as it truly is, but can produce images of it, defining limits and uncertainties in terms of validity and precision. Since any risk-adjustment model used for comparative evaluation of outcomes must be time- and population-specific, only the studies that use credible risk adjustment strategies are more likely to yield reliable findings.