Development and Validation of a Simple Tool for Predicting Pandemic-Related Psychological Distress Among Health Care Workers

Abstract

We aimed to develop a simple predictive model that enables health care workers (HCWs) to self-assess pandemic-related psychological distress in order to assist them to find psychological support to avert adverse distress-related outcomes. In a pilot study, we recruited and followed longitudinally 220 HCWs at the Hospital of the Ludwig Maximilian University Munich (H-LMU) during the first wave of the COVID-19 pandemic (March–July 2020). In this sample, we evaluated whether a machine-learning model with sociodemographic, epidemiological, and psychological data could predict levels of pandemic-related psychological distress. To maximise clinical utility, we derived a brief, 10-variable model to monitor distress risk and inform about the use of individualised preventive interventions. The validity of the model was assessed in a subsequent cross-sectional study cohort (May–August 2020) consisting of 7554 HCWs at the H-LMU who were assessed for depressiveness after the first wave of the pandemic.The model predicted psychological distress at 12 weeks with a balanced accuracy (BAC) of 75.0% (sensitivity, 73.2%; specificity, 76.8%) and an increase in prognostic certainty of 41%. In the derivation cohort, the brief model maintained a BAC of 75.6% and predicted depressiveness (P < .001), resilience (p.001), and coping (p < .001). Furthermore, it accurately stratified HCWs’ psychological trajectories of global and affective burden as well as behavioural adaptation over the 12-week follow-up period. Our clinically scalable, 10-variable model predicts individual COVID-19 pandemic-related psychological distress outcomes. HCWs may use our associated predictive tool to monitor personal and team-based risk and learn about risk preventive interventions based on an intuitive risk stratification.