The Impact of Duration on Disability Transitions: Evidence from the Cardiovascular Health Study

Liming Cai, National Center for Health Statistics (NCHS), CDC
James Lubitz, National Center for Health Statistics (NCHS), CDC
Nathaniel Schenker, National Center for Health Statistics (NCHS), CDC
Paula Diehr, University of Washington
Alice Arnold, University of Washington
Linda Fried, Johns Hopkins University

We used the Cardiovascular Health Study to investigate whether fitting a Semi-Markov Process (SMP) model using the stochastic EM algorithm (SMP-EM) to estimate total duration in left-censored data yields improved estimates of the effect of duration in a health state on the odds of recovery or decline. Health was categorized into two functional states (active and disabled) using measures of limitations in Instrumental Activities of Daily Living and Activities of Daily Living. The results indicated that the SMP-EM approach yielded better estimates of the relation of duration in a health state to the odds of recovery or decline, as compared to the multi-state life table model or SMP without the EM. The value of the EM is greater, the shorter the follow-up period.