Comparison of methods for calculating healthy life expectancy in a Japanese population: An analysis of Nippon Data90

Abstract

Introduction There are several methodological approaches to calculating healthy life expectancy (HLE). Sullivan's method (prevalence-based life tables), which uses age-specific disability prevalence, is conventionally employed to estimate HLE. The multistate life table method presents an alternative approach that permits transitions from one health state to another using Markov models. Although these two approaches are frequently applied to calculate regional and national HLE, little is known about the differences in their estimates. To examine these differences, we applied both approaches for estimating HLE in a nationwide cohort study of the Japanese population and compared the resulting estimates. Methods The National Integrated Project for Prospective Observation of Non-communicable Disease And its Trends in the Aged (NIPPON DATA90) is a nationwide cohort study involving nine thousand Japanese participants. Data from this project were used in our analysis. We identified participants who were aged 60 years or older at the baseline survey and had provided activities of daily living (ADL) information based on the Katz index. Two-wave interview surveys of ADL were conducted in 1995 and 2000. Participants who answered “not independent” to any of five stipulated items in the Katz ADL index were designated “disabled” in the analysis. Both Sullivan's method and the multistate life table method were used to estimate HLE in participants aged 60 to 99 years. In both estimations, smoking status was categorized into three groups (never-smokers, ex-smokers, and current smokers). Blood pressure was fixed at an optimal level (systolic blood pressure Results Our analysis was conducted on 6676 individuals (2840 men and 3836 women). The HLE estimates at age 60 from Sullivan's method in never-smokers, ex-smokers, and current smokers were 23.24, 22.55, and 19.33 in men and 25.04, 20.84, and 20.74 in women, respectively. The corresponding HLE estimates from the multistate life table method were quite similar to those of Sullivan's method both in men (23.32, 21.88, and 19.71, respectively) and in women (25.39, 23.37, and 22.22, respectively). The disparity in HLE estimates at age 60 between these two methods was virtually negligible but widened gradually as the index age increased. The HLE estimates at age 70 from Sullivan's method in never-smokers, ex-smokers, and current smokers were 14.95, 14.36, and 11.58 in men and 16.26, 12.68, and 12.41 in women, respectively. The corresponding HLE estimates from the multistate life table method were 14.98, 13.68, and 12.06 in men and 16.66, 14.82, and 14.02 in women, respectively. The HLE estimates at age 90 from Sullivan's method in never-smokers, ex-smokers, and current smokers were 3.45, 3.21, and 1.94 in men and 3.44, 2.16, and 1.55 in women, respectively. The corresponding HLE estimates from the multistate life table method were 4.06, 3.46, and 3.13 in men and 4.57, 3.70, and 3.67 in women, respectively. Conclusions Two different approaches to calculating HLE were compared using a nationwide Japanese cohort study. The HLE estimates at ages 60 and 70 were generally similar between Sullivan's method and the multistate life table method, but the latter produced higher estimates when the index age was 90 years. Analysts should be aware that the choice of method may influence HLE estimates.