Trajectories of informal care intensity among the oldest-old Chinese.

Abstract

Countries around the world face increasing demand for long-term care in the older population. Yet, the longitudinal patterns of long-term care use and the underlying predictors have not been well understood, which impedes efficient care planning and timely service delivery. This study investigates the trajectories of informal care intensity in the oldest-old Chinese population and identifies the most important predictors of care trajectories. The data come from four waves of the Chinese Longitudinal Health Longevity Survey (CLHLS 2005-2014, N=10,292). We conducted the latent trajectory analysis (LTA) to cluster people's diverse trajectories into a finite number of groups. We built machine learning (ML) models to predict people's care trajectories and ranked the relative importance of the predictors. The LTA identified three distinct trajectories of informal care intensity: the low, increased and high intensity trajectories. Care intensity increases in all three trajectories. Older people with more severe limitations, females, urban residents, people with a higher income, and people with more daughters in the first wave are more likely to follow the increased or high intensity trajectory rather than the low intensity trajectory in the following decade. The random forest classifier has the best overall prediction performance among the four machine learning models. Its prediction accuracy can be further improved via model optimisation. Oldest-old people in China follow divergent trajectories of care utilisation, and inequality of informal care intensity is discernible across time, demonstrating the need for timely and targeted delivery of government support to those who need it most. Accurate prediction of care trajectories will be of great value to policy makers and practitioners in relation to the planning of personalised care and the equitable allocation of care resources.