Predictive performance of self-perceived health for depressive symptom development in community-dwelling older adults: a European population-based study

Abstract

Background

Depressive symptoms in later life are common and linked to adverse outcomes, yet early identification remains challenging. Self-perceived health (SPH), a simple subjective measure, may offer prognostic value for depressive symptoms.

Objective

To assess the predictive performance of SPH for depressive symptoms in older adults compared with objective health indicators.

Methods

We analyzed data from 25,985 community-dwelling participants aged 50 + from the Survey of Health, Ageing and Retirement in Europe. Individuals living in nursing homes were excluded. Mean age of participants was 64.4 years (SD = 9.9), 54% were women, and average education was 9.8 years (SD = 4.5). Depressive symptoms were assessed using the 12-item EURO-D scale and defined as a score ≥ 4, in accordance with established cut-offs. Predictive performance of SPH was assessed cross-sectionally and longitudinally over two years by calculating the classification error rate (CER) and the area under the receiver operating characteristic curve (AUC), and compared with number of chronic diseases, cognitive function, grip strength, and walking speed. Analyses were stratified by age and sex.

Results

SPH demonstrated good predictive performance both cross-sectionally and longitudinally. At baseline, SPH had the highest AUC (0.72[0.70,0.73]) and lowest CER (22.4% vs. 24.0% for morbidity burden). At follow-up, performance remained stable (AUC = 0.69[0.67,0.71], CER = 21.8%), comparable to morbidity burden (AUC = 0.64[0.62,0.65], CER = 22.6%), and grip strength (AUC = 0.66[0.64,0.68], CER = 22.3%). Combining SPH with objective indicators modestly improved discrimination.

Conclusions

SPH provides meaningful prognostic information for depressive symptoms in older adults, with performance comparable to objective health measures. Its simplicity supports a potential use in screening and research contexts.