Occupational Stress Among Healthcare Workers and the Feasibility of a Machine Learning-Based Risk Screening Approach

Introduction: Occupational stress among healthcare workers is a major determinant of burnout, reduced quality of care, and adverse mental health outcomes. In Poland, evidence remains fragmented across professions, limiting system-wide understanding. This study aimed to identify sociodemographic and occupational predictors of stress among healthcare workers and to evaluate the feasibility of a machine learning (ML)-based proof-of-concept screening model for detecting high-risk individuals. Materials and methods: A cross-sectional survey of healthcare professionals in Poland (N = 178) was conducted between November 2025 and January 2026 using the Perceived Stress Questionnaire (PSQ-30). Participants were categorized into a high stress (PSQ ≥ 0.60) and a low–moderate stress group. Statistical analyses included reliability testing, non-parametric comparisons, MANOVA, and multivariable logistic regression. A Random Forest classifier with repeated cross-validation was developed to predict high levels of stress using demographic, occupational, and psychosocial variables. Results: The PSQ-30 demonstrated excellent reliability (Cronbach’s α ≈ 0.96). High stress prevalence was 14.6%. Night-shift work emerged as the strongest predictor of high stress (OR ≈ 3.45–12.4), while greater work experience was associated with lower levels of stress. The profession was not an independent predictor after adjustment, although nurses showed a higher proportion of stress. Significant differences across professions were observed only for the Joy subscale (p = 0.026), with paramedics reporting lower scores. Department differences were not significant after accounting for sample size variability. The ML model achieved an ROC AUC value of 0.728, with 70% sensitivity and a high negative predictive value (97%), supporting its utility as a screening tool. The key predictors included age, pediatric department, and workplace conflict. Conclusions: Occupational stress in Polish healthcare workers is driven primarily by work patterns, particularly night shifts, rather than the profession alone. ML-based screening models seem promising for early identification of high-risk individuals, enabling timely preventive interventions.