As the consumers and regulators of health care have become more concerned with quality of delivered services, interest has focused on hospital- and physician-specific mortality rates as an index of quality. Mortality rates have several characteristics that promote their use as a performance indicator. The numerator, death, is generally (but perhaps incorrectly) accepted as an adverse outcome of health care. Death is thought to be easily measured, and is recorded in several locations, including the medical record abstract and death certificates, where the information is accessible without provider consent. The denominator, persons or patients, is also available from several public sources. The desirability of mortality data is further enhanced by the wide variety of statistical methods to manipulate and compare rates and proportions. The conceptual validity of mortality rates as reflecting quality is supported by a long tradition of using mortality rates at the “macro” level to compare the quality of national health care delivery systems (eg, infant mortality rates) and at the “micro” level to compare the outcome of different therapies (eg, thrombolytics for acute myocardial infarction). However, despite face validity, ease of measurement, and widespread acceptance in other areas, hospital-specific mortality rates, as calculated from current data sources, have a variety of potential problems.This article will explore the clinical, administrative, and information-based difficulties in using mortality rates as an indicator of the quality of medical care delivered by specific hospitals or physicians.