Heterogeneity in the timing of instrumental activities of daily living disability onset and associated factors among community-dwelling older adults in Korea: evidence from the Korean Longitudinal Study of Ageing, Waves 1–9 (2006–2022)

Longitudinal studies have identified heterogeneous patterns in the timing of instrumental activities of daily living (IADL) disability onset, yet the factors associated with membership in specific classes remain poorly characterized. This study identified latent classes of the timing of IADL disability onset among community-dwelling Korean older adults and examined baseline factors associated with class membership. Using data from Waves 1–9 (2006–2022) of the Korean Longitudinal Study of Ageing, we analyzed 2,459 adults aged ≥ 60 years who were independent in IADL at baseline. Parametric (log-normal) and nonparametric (piecewise exponential) mixture survival models were compared for one to four classes. The log-normal specification fitted best among the parametric distributions and a three-class solution was retained; three classes were also specified for the piecewise exponential model for comparison, although its BIC favoured two. Multinomial logistic regression examined associated baseline factors, and the analysis was repeated with midpoint-imputed event times and with participants weighted by their posterior class probabilities. Three latent classes were identified; under the log-normal model these comprised a long-term maintenance class (62.8%), a gradual decline class (27.7%), and an early-event class (9.5%). Older age (≥ 70 years), fear of falling, and impaired vision and hearing were associated with higher odds of membership in both high-risk classes. Higher education, being married, economic activity, good self-rated health, and normal grip strength were associated with lower odds of membership in both high-risk classes. Under the piecewise exponential model, associations with age, fear of falling, education, self-rated health, and grip strength were retained; several others were not. Under the log-normal model only, chronic disease and depressive symptoms were associated with higher odds of membership in the early-event class. Class sizes differed substantially between models. The timing of IADL disability onset is heterogeneous and can be summarized by at least three distinct latent patterns, with different factors associated with membership in each. Under the log-normal model, chronic disease and depressive symptoms were associated specifically with membership in the early-event class. Class assignment and class sizes depended substantially on model specification and should be cross-validated before use in individual-level risk stratification.