Are mortality indicators good indicators of health care quality?

Importance of publishing mortality rates: Mortality rates for certain interventions or disease states have been used over the last decade as indicators of the quality of care provided by a given hospital, unit or, medical team. If published, these rates would be a useful tool for decision makers in the process of fund allocations, for public information, and for promoting improved care in hospitals or units with a low classification. Methodological limitations: It is difficult to adjust an indicator of mortality to disease-related risk factors and any modification of this adjustment can have major consequences on the validity of subsequent comparisons. The differences in mortality observed between hospitals and physicians can reflect not only differences in quality of care but also differences in approaches to disease-related risk factors, therapeutic choices, or coding practices. The lack of statistical power is a major limiting factor in interpreting differences in mortality rates. To evidence a statistically significant difference in mortality between two hospitals whose rates are respectively 0.5% and 1% (for example in total hip replacement patients), it would be necessary to include 4673 patients, a number which would correspond to 20 years data for a hospital performing 230 interventions per year. Consequently, the number of interventions performed in the most active hospitals would not be sufficient to make such comparisons. Limitations and counter effects: Some studies have demonstrated that the publication of mortality rates does not have a major influence on patients' decisions nor on physicians' choice of a referral hospital. It would have no effect on improving health care quality of the institutions cited. One the contrary, certain counter effects have been observed: modification in patient recruitment, higher-risk patients being referred to hospitals with unpublished mortality rates. For many authors, procedure indicators are more pertinent than outcome indicators for detecting differences in health care quality between different care structures.