Community mental health workforce capacity and official early psychiatric readmission indicators in Italy: a longitudinal ecological panel study
Pasquale Scognamiglio, Donato Morena, Vittorio Fineschi, Valeria Iniziato, Vassilis MartiadisAbstract
Aims
Workforce counts and early-readmission indicators often appear side by side in health-system dashboards, encouraging the assumption that more staff should accompany fewer returns to hospital. We examined whether annual variation in previous-year medical-nursing headcount density was associated with Italy’s official provider-side indicators of unplanned readmission within 30 days after eligible ordinary psychiatric discharges from public psychiatric wards and, secondarily, within 7 days.
Methods
Using national mental-health, hospital-discharge and population data, we assembled a retrospective 2015–2024 panel of 21 Italian territorial units (210 unit-years). Exposure was medical and nursing headcount at the end of the preceding year per 100,000 adults. We modelled 30-day readmission counts using Poisson pseudo-maximum likelihood (PPML) with an eligible-discharge offset and territorial-unit and year fixed effects. Primary inference used a restricted wild cluster score bootstrap with 9,999 Rademacher draws and 21 clusters. Secondary and sensitivity analyses were prespecified, source blanks were not imputed, and analytical materials and scripts are openly archived.
Results
Outcome counts and eligible-discharge denominators were complete for all 210 observations; exposure was available for 205. The incidence rate ratio (IRR) per 10 additional medical-nursing staff per 100,000 adults was 1.006 (95% CI 0.934–1.034; p = 0.8095), compatible with relative changes from a 6.6% reduction to a 3.4% increase. Across linear PPML sensitivity analyses, IRRs ranged from 0.995 to 1.042, with all confidence intervals including 1. Across 31 leave-one-unit or leave-one-year-out re-estimates, IRRs ranged from 0.974 to 1.044, and every confidence interval included 1. The nonlinearity test gave p = 0.8572. The 7-day IRR was 1.032 (95% CI 0.929–1.075; p = 0.6743), while the weighted linear model estimated +0.215 percentage points (95% CI −1.159 to +0.704; p = 0.7203).
Conclusions
The data did not provide evidence of a detectable longitudinal ecological association between year-to-year variation in medical-nursing headcount density and the official 30-day readmission indicator. This does not imply that workforce is unimportant: headcount does not capture full-time-equivalent availability, deployment or transition processes, and the provider-side design cannot support patient-level or causal inference. Territorial policy should combine adequate resources with a whole-system view of their deployment, post-discharge organisation, and the social, economic and health context in which care is delivered.