Methods to evaluate risks for composite end points and their individual components.

摘要: Abstract Objective Both randomized and observational studies commonly examine composite end points, but the literature on model development criticism in this setting is limited. Study Design Setting We examined approaches for evaluating heterogeneity effects of risk factors different components point, determining impact ability to predict point. A specific example considered cardiovascular disease point Physicians' Health that occurred 1,542 (myocardial infarction, n  = 716; stroke,  = 557; death,  = 269) 16,688 participants with complete information baseline covariates. The strategy compared alternative polytomous logistic regression models assuming a comparable common effects. Results Likelihood ratio tests identified age, alcohol consumption, diabetes across outcome, comparability other factors. However, uniform explained over 90% log-likelihood change best model, two also performed similarly based comparison ROC curves. Conclusion overall may be helpful validity analysis identifying