Nonparametric Estimation of Disability-Free Life Expectancy Using Period Life Table and Cross-Sectional Disability Survey

Samir Soneji, Princeton University
Kosuke Imai, Princeton University

Increased prevalence of chronic impairment characterizes aging populations. Are additional years of life spent in poor health? Robust estimation of disability-free life expectancy (DFLE) is essential for addressing this question. Thirty years after publication, Sullivan's method remains the most widely used method for estimating DFLE. Sullivan's method simply partitions the total number of person-years lived in an interval by the proportion disabled in that interval. We prove Sullivan's method imposes previously unnoticed assumptions. To improve upon Sullivan's method, we relax this assumption and derive nonparametric bounds of DFLE. A bootstrap method is used to compute balanced confidence intervals for the bounds. Finally, we apply the proposed methodology to estimate DFLE for the 1999 U.S. population using the data from the period life table, National Health Interview Survey, and National Nursing Home Survey. We find important race, sex, and education differentials in DFLE and proportion of remaining life spent without disability.