Objective To conduct a systematic review of studies that employ conjoint analysis methodology in outcomes research in medicine published between 1985 and 2006 in order to document: i. clinical areas of focus; ii. sample size; iii. method of design; iv. method of analysis and other quality parameters. Methods Papers published between 1985 and 2006 were identified on Medline, using the search terms conjoint analysis/analyses, stated preference(s), discrete choice analysis/analyses, and discrete choice modeling/experiments(s). All papers were then reviewed for content by three reviewers, with papers not actually related to conjoint analysis being deleted. Remaining papers were then classified as: i. a clinical application; ii. an application focusing on heath systems; or iii) papers focusing on methods. We classified all clinical applications by ICD-9 codes and identified key methodological characteristics such as sample size, design methodology and type of analysis when available. Results We began our review in 1985 due to the limited number of publications between 1970 and 1985 (n=4). Post 1985, 27% (n=48) discussed the methodology of conjoint analysis with no application, 25% (n=45) focused on health systems in medicine and 48% (n=86) were clinical and therapeutic applications of conjoint analysis. There is a steady increase in the use of conjoint in medicine between 1985 and 1999, most common clinical areas being HIV, cancer and STI. The average sample size is 267. Use of orthogonal arrays was the most common design, 74% (n=50) followed by adaptive conjoint analysis, 16% (n=11). Primary analysis techniques were probit and logistic regression, 39% (n=27) and 30% (n=21) respectively. Conclusions We find insufficient information on the methods used in a significant proportion of manuscripts reviewed. Given the importance of preference elicitation in medicine, we need to focus on developing standardized research practices for the application of conjoint analysis in outcomes research.