Many of the previous studies of rehospitalization are surveyed and those findings are abstracted that are relevant to three basic methodological issues: the selection of predictor content, the source of predictor information, and the length of the follow-up interval. An empirical examination is made of several questions surrounding each issue, including the relative predictive power of a) patients' background characteristics vs. their behavior; b) symptomatic behavior vs. instrumental role behavior; c) patients' own reports vs. others' reports; d) hospital behavior vs. community behavior; and e) each domain of predictor content at a 6-month vs. a 1-year follow-up interval. Finally, individual predictors are selected from each domain to produce a composite picture of the person who is at high risk for rehospitalization.