The results suggest that all three proposed service changes would provide health benefits, accounted for by increasing the use and completion of psychotherapy services. As well as the immediate benefit of reducing the length of a depressive episode, the service changes have the long-term benefit of reducing the probability of future episodes of depression. These additional benefits however come at an increased cost for the NHS, due to increased utilisation of services and psychological therapies. The sensitivity analyses (probabilistic, and scenario analysis) undertaken suggest that these results are robust to the model assumptions. The factorial design provides evidence for policy makers when considering multiple potential changes, but also highlights that unexpected interactions may occur and should be carefully considered.
Due to the local nature of the research and decision-maker environment, there is not a defined cost-effectiveness threshold based on local spending and outcomes. This means that interpretation of the cost-effectiveness results cannot be undertaken by this research team, but will need consideration by NHS commissioners and providers. Whilst some national guidance from NICE is mandatory, NICE has only considered a sub-set of possible interventions and treatments and so local policy-making requires evidence to inform it. The use of UK specific data and taking a UK perspective does limit the results of this analysis to a UK context; however we anticipate that the methodologies and concepts will be transferable to other countries.
The results of the analysis highlight the flexibility and appropriateness of the Whole Disease Modelling framework when looking to evaluate service changes which will have a wide and long-term impact. Each of the three service changes was evaluated in terms of their impact across the NHS, and for a patient’s lifetime.
The quantified model also provides useful service performance metrics, such as flows of people through different parts of the services, and changes can be evaluated in terms of their impact on specific resources. This is important because the local NHS is currently going through a period of change, and in general budgets are being restricted.
One important result is that the model suggests only low numbers of people with longer-term depression ‘step up’ to secondary (CMHT) and tertiary (SPS) services. This is because only a small proportion will present, and also because by the time they reach these higher steps, the depressive episode may have passed without a healthcare intervention. In particular, the self-referral service change highlights that if people can quickly receive appropriate therapy then there are likely to be significant health benefits both immediately, and in the longer term.
As with any health economics modelling, there are limitations of the analysis which should be noted. First, the analysis of the local NHS dataset has only been to identify broad flows of patients in the service, and covariates such as ethnicity and severity of depression have not been incorporated into the analysis. Secondly, health economic modelling allows evidence from different sources to be synthesised so that an estimate of the long term costs and benefits of new interventions can be found. As such, the model is limited by the quality of evidence that is available, and by the assumptions used when evidence is missing. The basecase model results suggest that patients experience on average nine depressive episodes in their lifetime. This is derived from longitudinal studies which estimate a time to relapse after a particular number of episodes, however evidence regarding relapse beyond five episodes was not identified, and so the model simulates a potentially inaccurate number of episodes. Experts suggest that this number may be too high, however it should be noted that there is still a low probability of patients presenting, and so these episodes may go undetected.
There is a significant evidence gap regarding the effectiveness of therapies at the higher steps (CMHTs and SPS). For this attempt at modelling the service, the effectiveness of SPS psychological therapy has been assumed no better than CMHT input. Because the IQuESTS project includes a specialist depression research clinic to pilot these service improvements, many of these assumptions can be tested and this will benefit the model in terms of collecting long term data regarding patients’ experiences with depression, and the impact of specific service improvements. This data will be used in an updated version of the model once the pilot study has been completed.
Mortality and serious adverse events have not been incorporated into the model. People who suffer a severe depressive episode have a significantly increased suicide risk, as well as increased risk of self-harm resulting in the need for hospitalisation . Also many people with depression will never access health services. Better access to services for these people will enable effective care and services to be provided, leading to improved health outcomes.
Because the service changes that have been evaluated are novel and un-evidenced, assumptions have been required to incorporate these changes into the model. Relatively simplistic assumptions have been required, such as the cost implications of the changes, and how they may impact on the flows of people through the system. The next step of the IQuESTS project is to explore fully how these service changes will be implemented in a pilot study. These results will inform the further development of the service changes, but may also allow an opportunity for refining the assumptions in the model and improving the accuracy of the results.
The Whole Disease Model framework provides a useful set of stages for a research project of this type. Because a change to a system can impact on other services, it is important to have a wide boundary and ensure that all possible costs and benefits are incorporated into the analysis. This aids decision-makers and commissioners, who are often challenged when provided with economic analyses which have a very specific boundary (i.e. a pairwise evaluation of treatments which does not capture the full care pathway). Also, the Whole Disease Modelling framework allows multiple evaluations of service changes to be undertaken. At this stage, three different service changes have been evaluated, however more can be added relatively easily, which means the Whole Disease Modelling framework allows efficient use of one (potentially more complex) model, rather than requiring several models.