Modeling Disability-Free Life Expectancy With Duration Dependence: A Research Note on the Bias in the Markov...

Area of Science:

• Demography
• Epidemiology
• Biostatistics

Background:

• Traditional demographic studies on healthy life expectancy (HLE) often use the Markov assumption, which overlooks the impact of exposure duration on health transitions.
• Duration-dependent multistate life table (DDMSLT) models account for exposure duration but are challenging to apply to left-censored survey data due to unknown initial state durations.

Purpose of the Study:

• To present a flexible approach for applying DDMSLT to left-censored survey data for estimating multistate life expectancies.
• To compute disability-free/healthy life expectancy (HLE) for older U.S. adults using this novel approach and compare it with traditional Markov-based models.

Main Methods:

• Developed a method to adapt DDMSLT for left-censored survey data by partially excluding observations and truncating duration.
• Applied the approach to the U.S. Health and Retirement Study (HRS) to estimate HLE.
• Compared findings from duration-dependent models with standard multistate models employing the Markov assumption.

Main Results:

• Transition probabilities exhibit duration dependence.
• Despite duration dependence, its overall impact on healthy life expectancy (HLE) is minimal, averaging out across the population.
• The Markov assumption yields a plausible and parsimonious estimate of HLE in this context.

Conclusions:

• The proposed method allows for the use of left-censored survey data in DDMSLT models.
• For HLE estimation in older U.S. adults, the bias introduced by the Markov assumption is minimal.
• The Markov assumption offers a practical and sufficiently accurate approach for estimating HLE in similar demographic studies.