Early Intervention Services (EIS) for People With Serious Mental Illness: An Econometric Analysis
Posted: 16 Jun 2007
Date Written: November 22, 2006
Background: Services for people aged between 14 and 35 years who have serious mental illness are generally of poor quality in the UK. The government has established Early Intervention Services (EIS) which are teams that prioritise care of people with a first episode of serious mental illness in this age group. Evidence on aspects of organisation and delivery of EIS is scarce. In order to fill this information gap, a project evaluating the implementation and impact of EIS over a 2 year period was commissioned by the UK National Health Service (NHS). This paper reports the findings from one part of that project: the econometric modelling.
Objectives: The objectives of the research were to assess the relationship between different service configurations, fidelity to national guidance and pre-determined outcomes such as duration of untreated psychosis (DUP) and sectioning patterns. We also explored the variability of the outcomes and sought to identify the factors associated with such variation.
Methods: Data on resource use and outcomes were collected from a sample of 479 episodes of mental illness in five EIS case study evaluation sites. An econometric framework has been used to explore variability. The first model explores the relationship between DUP and a vector of explanatory variables including baseline characteristics, process measures and outcome descriptors. A Generalised Linear Model (GLM) was used for this analysis. The second model explores factors associated with sectioning and used a Logistic regression approach. The study uses baseline data to predict DUP and probability of sectioning.
The analysis highlights recognised problems with such data: skew to the right, heavy tails and heteroscedastic error terms. Results using OLS with heteroscedastic retransformation and the GLM are compared. Issues pertaining to the measurement of the DUP are also examined. In addition, there exist missing data - multiple imputation techniques were employed.
Results: The mean DUP for all five EIS sites was 15.6 days (range 1 to 144). About 125 (57%) of all individuals assessed under the Mental Health Act (MHA) were sectioned. No baseline characteristics or outcome descriptors were significant in explaining DUP. Some process measures were however significant: service users referred to an EIS by their General Practitioner (GP) or primary care team had the shortest DUP (about 70 % shorter) when compared to those referred by a Psychiatrist, home treatment team or other sources. In addition, DUP for those assessed under the MHA was also approximately 39% lower than that of those who were not assessed under the Act. In the second model, economically active students were more likely to be sectioned than the general group of employed service users. The results also indicate that being sectioned was associated with fewer parasuicide events.
Conclusions: The results of these analyses contribute to the policy debates around EIS as they highlight patient characteristics and service measures that are associated with higher levels of DUP and sectioning. In addition, the paper highlights important methodological issues, notably around the use of the GLM in health econometric analyses and how missing data problems can best be addressed.
Keywords: GLM, multiple imputation, DUP
JEL Classification: C19
Suggested Citation: Suggested Citation