Objective: To systematically assess the quality of published reports of the prognosis of community-acquired pneumonia using a formal quality assessment instrument. Design: Retrospective review of studies published during 1966-1991. Articles: 108 articles related to the prognosis of community-acquired pneumonia retrieved by a computerized search. Intervention: All articles, blinded to author(s), journal title, year of publication, and study institution(s), were independently reviewed by two investigators using a ten-item quality assessment instrument designed to evaluate: 1) identification of the inception cohort (4 items), 2) description of referral patterns (1 item), 3) subject follow-up (2 items), and 4) statistical methods (3 items). Adherence to each of the ten individual quality items and an overall quality score were calculated for all articles and across three time periods. Main results: Among all 108 articles that underwent quality assessment, 30 were published from 1966 to 1979, 61 from 1980 through 1989, and 17 from 1990 through 1991. The mean total quality score of all articles was 0.55 (range 0.22-0.90). There was a significant trend toward improvement in total quality scores over the three time periods (0.50 to 0.56 to 0.65; p<0.001). However, several systematic errors in the study design or reporting of these studies were discovered throughout time: only 3.7% provided comparative information about nonenrolled patients, 28.7% determined whether the study institution was a referral center, 36.1% specified inclusion or exclusion criteria, and 45.5% used appropriate statistical analyses to adjust for more than one prognostic factor. Conclusions: Despite improvement in overall quality of published articles, systematic errors exist in the design and reporting of studies related to the prognosis of community-acquired pneumonia. The quality assessment tool employed in this study could be used to guide the development of high-quality outcomes research in the future.