Risk adjustment for hospital report cards

Abstract

Numerous countries develop and publish hospital report cards with the aim of assisting patients, employers, and/or insurers in the selection of appropriate healthcare providers. The variations in quality presented in these reports may be an indication of actual differences in quality; however, it's plausible that the results are biased due to improper adjustments for case-mix or incorrect interpretations. This underlines the importance of using a statistical risk adjustment method when dealing with patient risk data. Use of such methods would improve the quality of comparisons across hospitals. Logistic regression and linear regression are commonly used for the adjustment of risk; however, hierarchical models have been appearing in more recent studies and published report cards. Demographic factors, prior healthcare utilization, severity of illness, physiological risk factors, and self-reported health status are widely used as risk adjustors. Taiwan has built local systems for the monitoring of clinical quality; however, the lack of a reliable mechanism for risk adjustment makes it difficult to compare the quality of care across hospitals. Thus, the development of local risk adjustment models is an urgent requirement for improving the quality of hospital report cards. In accordance with our review, we suggest that the government evaluate the reliability of quality indicators and be more selective with regard to public reporting, in order to prevent the misclassification of outliers, which would otherwise limit the usability of report cards.