Funding of mental health services: Do available data support episodic payment?

Background
The primary method of funding NHS mental health services in England has been block contracts between commissioners and providers, with negotiations based on historical expenditure. There has been an intention to change the funding method to make it similar to that used in acute hospitals (called the National Tariff Payment System or NTPS, formerly known as Payment by Results (PbR)) where fixed prices are paid for each completed treatment episode. Within the mental health context this funding approach is known as episodic payment. Patients are categorised into groups with similar levels of need, called clusters. The mental health clustering tool (MHCT) provides a guide for assignment of patients to clusters. Fixed prices could then be set for each cluster and providers would be paid for the services they deliver within each cluster based on these fixed prices, although the emphasis to date has been on local pricing. For this episodic payment system to work, the MHCT needs to assign patients to clusters, such that they are homogenous in terms of 1) patient need, and 2) resource use.

Objectives

We test whether the existing data collected on mental health activity amongst NHS providers would support this new payment system. Specifically we examine whether there is homogeneity within clusters in terms of 1) costs, and 2) activity/resource use, and 3) whether the MHCT effectively clusters people with similar levels of need.

Data

We use the Mental Health Services Dataset, a patient level dataset which records activity on the patient’s classification into clusters, for the financial years 2012/13 and 2013/14. We link cluster activity within each provider to Reference Costs (for 2013/14), which provide information on the cost of the activity performed by each provider and facilitates the calculation of average costs for the different clusters both at provider and at national level.

Methods

We calculate a cost index to observe the variation in costs across providers. We run multilevel regressions of activity within clusters to test whether the observed differences in length of clusterepisodes translate into differences in the activity performed within them. We perform latent class analysis to determine whether the clusters as currently defined correspond to mutually exclusive groups of patients based on their answers to the MHCT. We ran a workshop for mental health commissioners to ascertain the key challenges of implementing this payment system.

Results

There is substantial variation in costs across providers. Considering all activity together, the ratio between the provider with the highest costs and the one with lowest is around two, but in some clusters this ratio can be as high as ten. Longer cluster-episodes do not translate into proportionally more activity. The existing 21 clusters do not correspond to 21 different groups of patients as defined by their answers to the MHCT. There is also great variation across commissioners in terms of capacity to contract for and implement this payment system.

Conclusions/Discussion

The results indicate that if a new cluster-based episodic payment system were introduced it would have different financial impacts across providers as there is variation in the activity they perform ii CHE Research Paper 137 within clusters and the costs they report for them. However remaining under block contract incurs a risk of mental health services experiencing disinvestment. There is thus a need to invest in improving the quality of data collection, and refining the cluster classification system to facilitate sustainable payment system reform.