Imperfections of chart review as a data collection method are detailed, with particular emphasis on the issue of reliability. The authors identify 125 journal articles published between 1991 and 1992 in which the authors' conclusions were based principally on chart review findings and were either clinical or health care epidemiology studies, or quality assessment studies. Eight percent of epidemiology studies and 56% of quality assessment studies presented data on interrater reliability. Forty-four percent of epidemiology studies and 20% of quality assessment studies did not describe the number of chart reviewers involved. Forty-three percent of epidemiology studies and 28% of quality assessment studies used supplementary data sources. The authors conclude that the validity and utility of studies based on chart review would be enhanced by attention to interrater reliability and the use of supplementary data sources.