Population mental health monitoring in and out of crisis: Lessons from the MAP-studies
A Baumer, V Yasenok, S Ruegger, M Kaufmann, A Frei, T Ballouz, V Von Wyl, A Kostenko, M Puhan,Abstract
Background
Real-time population mental health data are largely absent during crises, limiting timely policy responses. We developed and implemented a longitudinal monitoring framework to assess mental health during the war in Ukraine and identified key methodological and operational challenges.
Methods
We conducted a two-year population-based longitudinal study with assessments every three months starting in Spring 2024. The primary cohort was recruited in Ukraine, complemented by cohorts of Ukrainian refugees and the general population in Zurich, Switzerland, enabling cross-context comparison. Depression, anxiety, and post-traumatic stress disorder were assessed using validated self-report instruments. To address overestimation of prevalence inherent to screening tools, we applied Bayesian corrections based on published diagnostic performance estimates.
Results
7459 participants were recruited across the three cohorts. Recruiting a representative sample and managing attrition was challenging and required adaptive recruitment strategies (e.g., instant messaging, phone calls). One-year retention rates ranged from 44.3% to 62.1% across cohorts. Bayesian correction highlighted substantial uncertainty in prevalence, e.g. mean width of 95%CI interval for depression prevalence estimates was 20.3% (SD = 7.4%). In Ukraine, incomplete population distribution data prevented aggregation of strata-level estimates, complicating interpretation and communication of findings.
Conclusions
Population mental health monitoring in crisis settings is feasible but requires adaptations beyond traditional surveillance models. Key challenges include achieving initial sample representativeness, participant retention, and addressing diagnostic uncertainty. Hybrid approaches combining scalable self-report questionnaires with interviews to validate diagnostic performance and statistical correction offer a practical pathway for near real-time mental health surveillance in future crises.