To the Editor.
—We were surprised that Dr Brook and colleagues1 believe that "[q]uality-of-care reporting systems need to rely primarily on clinically valid process (vs outcomes) measures." Granted there remain technical and conceptual problems with outcomes measures: rare outcomes, like mortality, are insensitive measures of quality; many quality-of-life measures must rely on risk adjustment to form appropriate comparisons between various organizations and providers; and disease-specific outcomes measures are still in development. However, process measures are imperfect as well.Three facts argue for the use of outcomes measures at least as often as process measures for quality control. First, while outcomes are sometimes not sensitive as a measure of quality (the absence of a bad outcome does not necessarily imply good quality), the corollary is that process measures are not specific measures of quality (the mere presence of a process may not imply good quality). For example, a high level