Forecasting Population Need for Mental Health Care: A Bayesian Methodology Applied to the Epidemiology of Psychotic Disorders in England

Abstract

Background: Mental health service policymakers require evidence-based information to optimise effective care provision based on local need, but tools are unavailable. We developed and validated a population-level prediction model to forecast need for early intervention in psychosis [EIP] services in England up to 2025. Methods: We fitted six candidate Bayesian Poisson regression models, combining epidemiological data on psychosis risk, to predict new annual caseload of referrals, assessed, treated, and probable first episode psychosis [FEP] cases in EIP services, aged 16-64 years at small-area level. Models were validated against observed NHS Mental Health Services Data Set [MHSDS] data at Clinical Commissioning Group [CCG] and national levels for 2017. Projections were made up to 2025, based on small-area demographic forecasts. Outcome: In 2017, our best-fitting model predicted 8,112 (95% interval: 7,623 to 8,597) individuals with probable FEP in England, compared with 8,038 observed in the MHSDS (difference: n=+74; +0·92%), after accounting for psychosis risk by age, sex, ethnicity, small area-level deprivation, social fragmentation and regional cannabis use. In 2020, this model forecasted 9,066 new treated FEP cases (8,485 to 9,618), rising 1% annually up to 2025. For every ten treated cases, we forecasted that 23 and 21 people would be referred to and assessed by EIP services, respectively, for “suspected FEP”. Interpretation: Our methodology provides an accurate, validated toolkit to inform planners, commissioners and providers about future population need for psychosis care at different stages of the referral pathway, based on local determinants of need.

Gianluca Baio
Gianluca Baio
Professor of Statistics and Health Economics