Association of heatwave exposure and multimorbidity with depression trajectories among older adults: evidence from the China Health and Retirement Longitudinal Study

Comprehensive Summary

This unicenter, retrospective study investigates whether machine learning algorithms could predict older adults’ long-term depression trajectories using baseline demographic, health, and environmental factors. Data was sourced from the China Health and Retirement Longitudinal Study (CHARLS) across 2011–2018, restricting the sample to adults aged 60 or older in 2011 (final N = 3819). It is important to note that no temporal or external validation was performed, as the study was not replicated across different time windows, platforms, languages, or cultures. Depressive symptoms were assessed using the 10-item Center for Epidemiologic Studies- Depression scale. Researchers linked these outcomes with baseline predictors - including temperature data derived from ERA5 climate data, multimorbidity, demographics, health status, social environment, and childhood adversity - and used latent growth curve modeling (LGCM) to identify five distinct depression trajectories (consistently low, low‑increasing, high‑decreasing, consistently high, and high‑increasing). The dataset was then partitioned into an 80% training set and a 20% testing set, and four learning algorithms (Decision Tree, Random Forest, XGBoost, and Support Vector Machine) were trained to classify individuals into the depression trajectories. Support Vector Machine achieved the highest test set accuracy (0.629). More importantly, higher average temperatures increased the relative risk ratio (RRR) of belonging to the more favorable depression trajectories (consistently low: RRR = 1.281; low-increasing: 1.127) whereas the presence of heatwaves reduced this likelihood (consistently low: RRR = 0.820; low-increasing: 0.886). Because RRRs above 1 indicates increased likelihood of belonging to a certain trajectory, these results suggest that moderate temperatures are protective while extreme heat is detrimental to mental health. Additionally, greater multimorbidity reduced the likelihood of stable mental health, highlighting the combined role of health vulnerability and heat exposure.

Outcomes and Implications

This study emphasizes the overlooked effects of climate change and rising heat extremes on worsening mental health trajectories, emphasizing the compounding impact of heat extremes and multimorbidity in older adults. This underscores the need for intensifying public health strategies that account for climate-related risks. Furthermore, this study confirms the importance of machine learning in detecting complex, nonlinear relationships between climate exposure and health, allowing researchers to use similar approaches to identify who is most vulnerable as environmental conditions continue to change.