Psychological characteristics and stress differentiate between high from low health trajectories in later life: a machine learning analysis

Abstract

Objective: This study set out to empirically identify joint health trajectories in individuals of advanced age. Predictors of subgroup allocation were investigated to identify the impact of psychological characteristics, stress, and socio-demographic variables on more favorable aging trajectories.

Method: The sample consisted of N = 334 older adults (M_AGE=68.31 years; SD = 9.71). Clustered health trajectories were identified using a longitudinal variant of k-means and were based on health and satisfaction with life. Random forests with conditional interference were computed to examine predictive capabilities. Key predictors included psychological resilience resources, exposure to childhood adversities, and chronic stress. Data was collected via a survey, at two different time points one year apart.

Results: Two different clustered health trajectories were identified: A ‘constant high health’ (low number of health-related symptoms, 65.6%) and a ‘maintaining low health’ profile (high number of symptoms, 34.4%). Over the one-year study period, both symptom profiles remained stable. Random forest analyses showed chronic stress to be the most important predictor in the interaction with other risk and also buffering factors.

Conclusion: This study provides empirical evidence for two stable health trajectories in later life over one year. These results highlight the importance of chronic stress, but also psychological resilience resources in predicting aging trajectories.

Keywords:

• Successful aging
• random forest
• conditional interference
• longitudinal k-means
• trajectories

Acknowledgments

We express our gratitude to all participants. During the work on their dissertations, Jan Höltge and Shauna L. Mc Gee were pre-doctoral fellows of LIFE (International Max Planck Research School on the Life Course; participating institutions: MPI for Human Development, Humboldt-Universität zu Berlin, Freie Universität Berlin, University of Michigan, University of Virginia, University of Zurich). This study was supported by the University Research Priority Program (URPP) “Dynamics of Healthy Aging” at the University of Zurich.

Disclosure statement

All authors report no conflict of interest.

Additional information

Funding

This work was supported by the Swiss Government Excellence Scholarship (ESKAS-Nr. 2016.0109) which funded Shauna L. Mc Gee’s position.