Brief depressive and anxiety symptom measures rank alongside comorbidity burden as predictors of health-related quality of life in 15 countries.

BACKGROUND

Population health surveillance increasingly relies on health-related quality of life to compare groups and countries, monitor inequalities and inform resource allocation. Survey space is finite and it is not known how much predictive information brief depressive and anxiety symptom measures add or whether that information is stable across countries.

OBJECTIVE

To quantify the predictive importance of depressive and anxiety symptoms relative to comorbidity burden, socioeconomic position and health behaviours for EQ-5D-5L utility in 15 countries and to test whether a pooled international model transports across settings.

METHODS

Cross-sectional analysis of EQ-DAPHNIE, a harmonised general population survey conducted in 15 countries between February and December 2024. EQ-5D-5L utility was scored uniformly with the UK value set. Random forests estimated the permutation importance of 18 demographic, socioeconomic, behavioural, clinical and mental health predictors, including the two-item Patient Health Questionnaire (PHQ-2) and two-item Generalised Anxiety Disorder (GAD-2) scale in pooled and country-specific samples. Transportability was assessed by leave-one-country-out validation.

FINDINGS: Of 67 178 respondents, 62 613 (93.2%) had complete data; mean age 45.9 years (SD 16.7) and 33 196 (53.0%) females. The pooled model explained 45.6% of the variance in utility (cross-validated R²=0.456 (SD 0.007)). Comorbidity count (importance 0.168), PHQ-2 (0.095) and GAD-2 (0.094) were the three leading predictors, and the same three comprised the top three positions in every country, although their order varied and, where the bootstrap CI for the difference included zero, the ordering between them was not determined. Both remained predictive when the outcome was rescored with the EQ-5D-5L anxiety/depression dimension held at level 1, although their importance attenuated (PHQ-2 by 47%, GAD-2 by 75%). The top-three predictor set was retained in 13 of 15 countries when national value-set or crosswalk scoring was used instead of the UK value set. Leave-one-country-out performance was moderate in 14 countries (R² 0.281–0.544) but poor in China (0.220 vs within-country 0.405).

CONCLUSIONS

Four questions on depressive and anxiety symptoms ranked alongside comorbidity burden among the leading predictors of population health-related quality of life, and the membership of this leading set was consistent across 15 culturally, economically and linguistically diverse countries.

CLINICAL IMPLICATIONS: PHQ-2 and GAD-2 are strong candidates for inclusion alongside morbidity measures in quality-of-life surveillance where questionnaire space is limited. The weaker transportability to China indicates that pooled international prediction models should be validated locally before use; the country-level differences observed here are hypothesis-generating.