Abstract
INTRODUCTION: Selective attrition due to interwave mortality can lead to bias in studies of cognitive decline. Long interwave intervals exacerbate challenges. METHODS: Using the two waves of the Longitudinal Aging Study in India-Diagnostic Assessment of Dementia (LASI-DAD) (N = 3546, from 2017-2019 to 2022-2024), we compared approaches to account for interwave mortality in estimating cognitive decline across age groups, including inverse probability weighting, joint modeling, and imputation of cognition before death using end-of-life interview data. RESULTS: All approaches produced steeper estimates of cognitive decline compared to estimates that did not account for interwave mortality. Imputations combined with joint modeling led to the steepest estimates of decline in the 60 to 69 and 70 to 79 age groups (22.0% and 35.9% steeper than the base model not considering mortality, respectively). DISCUSSION: Findings showcase the importance of accounting for interwave mortality and highlight the utility of imputation combined with joint models for studies with long interwave periods. HIGHLIGHTS: All approaches to accounting for interwave mortality led to steeper estimates of cognitive decline than approaches ignoring mortality. We developed adjustments to end-of-life cognition data to account for potential reporting bias. Imputations based on end-of-life data helped account for cognitive decline after the last study interview. Combining imputations from end-of-life data with a joint model led to the steepest estimates of cognitive decline in two of the three age groups. Accounting for interwave mortality and considering cognitive decline between waves is important in studies with long interwave intervals.