The performance of mental healthcare providers in England

This thesis investigates the performance of mental health providers in England on resource use (length of inpatient stay and costs) and quality (readmission rates and patient outcomes). Under a new payment system, it is intended that a national tariff (price) based on national average costs will be introduced and a part of future payments will be contingent on outcomes. Therefore, providers will have incentives to control costs and improve patient outcomes. We investigate the potential to achieve these aims using two nationally representative patient-level data sets: Hospital Episode Statistics (HES) and the Mental Health Minimum Data Set (MHMDS).

We utilise multilevel models, which allows us to isolate the residual variation in our response variable attributable to providers. Residual variation is quantified using Empirical Bayes (EB) methods and comparative standard errors are used to rank providers to make inferences about performance. We model length of stay (LOS) using a Poisson model; costs using a log-linear model and a generalized linear model (GLM) with a gamma distribution and log link; outcomes using ordered probit and linear models; and costs and outcomes simultaneously using a bivariate model. We employ a comprehensive range of patient and provider covariates.

Demographic, diagnostic, severity and treatment variables are key drivers of LOS and costs. Worse outcomes are associated with severity and better outcomes with older age and social support. Provider-level emergency readmission rates are associated with lower LOS and formal admissions with higher LOS. Provider-level variables have negligible effects on outcomes but a notable effect on costs. Ranking providers by residual variation suggests some providers can improve performance. Providers performing below average face financial instability under a national tariff and when a part of payment is linked to outcomes. The correlation in provider-level residual costs and outcomes is miniscule suggesting that cost-containment and outcome improving efforts by providers should not conflict.