Causal mediation analysis of contact with COVID-19 patients and burnout syndrome in physiotherapists in Japan: secondary analysis of a multicentre observational study.

Background

Physiotherapists played central roles in high-risk 'red-zone' care during COVID-19 and reported substantial distress, yet mechanisms linking exposure to burnout remain unclear.

Objective

To examine whether the association between direct patient contact and burnout operates through a stress-related behavioural mediator (feeling like quitting/taking a break) and whether organisational resources moderate this pathway.

Methods

We conducted a secondary analysis of a nationwide, multicentre, web-based survey of Japanese physiotherapists engaged in COVID-19 care (March 2021). Burnout was defined using the Japanese MBI-GS case algorithm. Counterfactual mediation analysis with bias-corrected and accelerated (BCa) bootstraps (5,000 resamples) was used to estimate the average causal mediation effect (ACME), average direct effect (ADE), and total effect on the risk-difference scale. Models were adjusted for age, sex, years in practice, and cohabitation.

Results

Burnout prevalence was 19.4%. A small indirect effect was observed (ACME = 0.00932; 95% CI 0.00188-0.0200), whereas the direct and total effects were not statistically significant. Appreciation/respect was associated with attenuation of the indirect effect (ACME = 0.00007), whereas endorsement of enhanced reimbursement for two-therapist care of critically ill patients was associated with a larger indirect effect (ACME = 0.00637; 95% CI 0.00122-0.0100). Because the primary mediation analysis used a binary mediator and binary outcome with logistic link functions, a formal ρ-based sensitivity analysis for unmeasured mediator-outcome confounding was not available under this specification.

Conclusions

Among Japanese physiotherapists, the association between direct COVID-19 patient contact and burnout was consistent with an indirect pathway operating through a stress-related behavioural mediator. Recognition appeared to buffer this pathway, whereas reimbursement-linked resource strain appeared to amplify it. These findings suggest potentially relevant organisational targets for future intervention, but should be interpreted cautiously given the cross-sectional design and the causal assumptions required for mediation analysis.