Including Deaths When Measuring Health Status Over Time

摘要: Measuring health status over time is problematic when some subjects die, because death does not have a defined value on most measures. This situation different from the usual missing data problem of dead is, in sense, known. We examined eight strategies for incorporating deaths into such analyses using three measures taken two sets, after which we used computer simulation to explore more fully effect deaths. The differed amount influence given deaths, varying none (deaths were discarded) complete (mortality itself was measure). that gave less tended show favorable changes time, and therefore, favor group had influenced by showed negative favored with fewer choice strategy should depend goals an intervention. For promotion studies, recommend recoding variables estimate probability person will be healthy 2 years (or other period can estimated data).