A population-level prediction tool for the incidence of first episode psychosis

 
00:00
Presentation First Author: 
James Kirkbride
Abstract: 

Background: Health commissioners require precise information on local population needs that vary enormously according to social and demographic factors. These are often unrecognised or ignored, as was the case when early intervention in psychosis services [EIS] for young people with first episode psychosis [FEP] were commissioned in England and Wales. We sought to develop a realistically-complex, population-based prediction tool to forecast incidence rates, using FEP as proof-of-principle. Methods: Data from 1028 FEP participants from two epidemiological studies were fitted in negative binomial regression to estimate risk coefficients for various sociodemographic and socioenvironmental factors. We applied these to the population at-risk of a third, socioeconomically different region to predict expected caseload over a 2.5 year period, where observed rates had been concurrently ascertained. We tested several models, comparing observed with predicted counts (with 95% prediction intervals) at regional, EIS and local authority district [LAD] levels. Results: A model with age, sex, ethnicity and population density had greatest predictive validity, forecasting 508 (95% PI: 459, 559) compared with 522 observed participants over the same period in our third region. It predicted correctly in 5 of 6 EIS (83.3%) and 19 of 21 LAD (90.5%). All models performed better than the current gold standard for EIS commissioning in England and Wales (716 cases; 95% PI: 664-769). Conclusions: We have demonstrated it is possible to accurately forecast the incidence of first episode psychosis according to local need, when the underlying epidemiology is wellcharacterised. This approach could improve resource allocation and commissioning of health services, and may be extended to many other disorders and many settings. For FEP we have developed a free tool (www.psymaptic.org) to predict expected annual incidence for all major sociodemographic groups in England and Wales at LAD and national levels.

Conference Name: 
Presentation Date: 
November, 2013
See other presentations in this session: