From early risk to preventive action: Empowerment based digital triage and intervention for burnout, anxiety, and depression in Scandinavian healthcare workers

Burnout, anxiety, and depression among healthcare workers are associated with long-term sickness absence, reduced quality of care, and impaired patient safety. In Sweden and other Scandinavian countries, common mental disorders account for a substantial share of sickness absence, with costs borne by employers and national insurance systems (the “payee side”). Early identification of workers at elevated risk, followed by proportionate preventive support, is therefore central to sustainable hospital management and occupational health. The Empowerment for Participation (EFP) assessment is a 110-item web-based system that maps everyday participation, demands, and self-expectations and derives validated risk indices for burnout, anxiety, and depression, together with constructs such as motivation, stress, and defence routines. The EFP battery has been used as a triage tool and an outcome framework in web-based psychotherapy trials targeting burnout risk and as a component in AI-driven digital triage concepts within the My-E-Health ecosystem. Prior Swedish evidence also links burnout and stress symptoms in healthcare workers to subsequent long-term sickness absence, supporting the relevance of risk indices for occupational prevention. Empowerment-related constructs have further been shown to be responsive to interventions such as mindfulness-based treatment.

To integrate empirical data from a cohort of healthcare personnel assessed with the EFP battery into a broader preventive model linking early risk identification, digital and therapist-delivered support, and the Scandinavian sick-pay context, while clarifying the methodological limits of non-randomised, real-world data.

I analysed EFP data from 325 healthcare workers with baseline (T1) and follow-up (T2) assessments who, in routine practice, either (1) received no structured treatment or AI support (no structured support/control), (2) engaged in human-delivered psychotherapy (human therapy), or (3) used an AI-supported, web-based intervention grounded in the EFP architecture (AI-only support). Outcomes included the EFP Empowerment index and EFP-derived risk indices for burnout, anxiety, and depression, plus stress burden, motivation, and defence routines. Change over time was examined within groups (paired t-tests; Cohen's d_z) and explored between groups using change-score ANOVAs (η²). Because groups were not randomised and differed at baseline, between-group comparisons were interpreted cautiously as descriptive.

At baseline, participants who went on to use human therapy or AI-only support showed higher EFP-derived risk for burnout, anxiety, and depression and lower empowerment than those who did not engage in structured support, consistent with risk-driven help-seeking. Under naturalistic conditions, human therapy and AI-only support were associated with large within-group improvements in empowerment and substantial reductions in EFP-derived risk indices, whereas the no-structured-support group showed only small changes. Defence routines decreased far more in the human therapy group than in the AI-only group, suggesting potentially distinct mechanisms. Improvements in empowerment were strongly associated with reductions in risk indices; however, the observational design did not support causal inference.

The EFP battery functions as a practical framework for early risk identification and change-sensitive outcome monitoring that can support scalable digital triage and stepped-care pathways. However, because allocation to support was non-randomised and sickness absence was not measured in this cohort, claims about prevention of sick leave should be treated as hypotheses to be tested in prospective studies that include baseline-balanced comparisons and registry-based sick-leave endpoints.