Deriving clinical prediction rules from stroke outcome research.

BACKGROUND AND PURPOSE Our purpose was to determine whether clinical prediction rules could be derived from current stroke outcome research. SUMMARY OF REPORT We reviewed 92 articles on stroke outcome research to determine their suitability for implementation as a clinical prediction rule. Methodological problems in many of these studies made implementation of their results as a clinical prediction rule difficult. CONCLUSIONS Implementation of stroke outcome research as clinical prediction rules would be facilitated by description of patient population demographics; precise definitions of predictor and outcome measures; stratification of patients by stroke mechanism; use of adequate patient sample sizes; and description of the mathematical methods used, including coding schemes, cutpoints, beta coefficients, constant terms, and a priori probabilities.