Evaluating the Heterogeneous Impacts of the Improving Access to Psychological Therapies (IAPT) Programme

心理健康问题在英国乃至全世界都是一个严重的问题。在英国，心理健康问题是导致残疾的最大单一原因，每年有四分之一的成年人至少经历过一种可诊断的心理健康问题。大流行和相关的封锁进一步导致心理健康大幅下降。我们预计，在从大流行病中复苏的过程中，心理健康将成为政策议程上的首要优先事项。在英格兰，国民保健制度实施了一项前所未有的大规模公共心理健康方案：改善获得心理治疗的机会方案，解决了人口中的心理健康问题，该方案以患有抑郁症和焦虑症的成年人为目标，是同类方案中最雄心勃勃的方案之一。自2008年在全国实施以来，该规划已被证明是有效的，平均有50%的患者在转诊后康复。然而，在不同的人口群体和地理区域之间，恢复率差别很大。减轻这些差异是NHS的首要任务。这可能变得更加重要，因为疫情已被证明大大加剧了不同患者群体和当地地区之间现有的心理健康不平等。我们的目标是系统地研究IAPT项目有效性中的这些不平等，以帮助政策制定者了解其潜在来源，并就如何减少这些不平等提供基于证据的建议。为此，我们首先将方案的因果效应（即治疗效果）与影响治疗结果的其他混杂因素（如自然恢复率）区分开来。然后，我们将确定与异质性治疗效果系统相关的因素。确定其中的个别治疗效果和不平等现象，将有助于深入了解哪种服务、治疗和患者特征的组合能最显著地改善心理健康。我们将把来自NHS Digital的详细服务和患者数据与当地的地理信息结合起来，这将使我们能够构建财富、教育、种族和其他特征的衡量标准，以研究由地方特征导致的不平等。然后，我们将在服务提供者水平上汇总患者水平的结果，以分析是什么导致不同特征的服务的恢复率不平等。这将有助于深入了解高绩效服务和低绩效服务之间的差异，以及如何改善低绩效服务的治疗结果。我们的项目有两个关键创新：第一个是评估IAPT项目的因果影响。对IAPT方案的现有分析是描述性的。我们更进一步，通过使用准实验技术来估计因果平均和异质性治疗效果。特别是，我们将利用治疗等待时间的差异来构建准实验对照组。第二个关键创新是，我们将用最先进的机器学习技术来补充标准的计量经济学方法来分析治疗效果的异质性，以揭示服务、患者和当地特征之间的复杂关系，以及IAPT项目有效性中的不平等。我们的目标是确定如何改善服务和治疗，以及患者和治疗如何更好地匹配，为英国及其他地区改善精神疾病治疗的循证政策提供信息。

Poor mental health is a serious concern both in the UK and worldwide. In the UK, mental health problems are the largest single cause of disability, with one in four adults experiencing at least one diagnosable mental health problem in any given year. The pandemic and associated lockdowns have further led to substantial reductions in mental wellbeing. We expect mental health to be a top priority on the policy agenda in the recovery from the pandemic.In England, mental health problems in the population have been addressed by an unprecedented, large-scale public mental health programme implemented by the NHS: the Improving Access to Psychological Therapies (IAPT) programme, which targets adults with depression and anxiety disorders and which is one of the most ambitious schemes of its kind. Since its nationwide implementation in 2008, the programme has been shown to be effective and, on average, 50% of patients show recovery after referral. However, recovery rates differ significantly between different population groups and geographical areas. Mitigating these differences is a top priority for the NHS. It is likely to become even more important as the pandemic has been shown to significantly worsen existing mental health inequalities between different patient groups and local areas.We aim at studying these inequalities in the effectiveness of the IAPT programme systematically, to help policy-makers understand its underlying sources and to inform evidence-based recommendations on how to reduce them. To do so, we will first distinguish the causal effect of the programme, i.e. the treatment effect, from other confounding factors that influence treatment outcomes, e.g. natural recovery rates. Then, we will identify the factors that are systematically associated with heterogeneous treatment effects.Identifying individual treatment effects and inequalities therein will provide insights into which combinations of service, treatment, and patient characteristics yield the most significant improvements in mental health. We will combine detailed service and patient data from NHS Digital with geographical information on local areas, which will enable us to construct measures of wealth, education, ethnic, and other characteristics, to study inequalities resulting from local area characteristics. We will then aggregate patient-level outcomes at the service-provider level to analyse what drives inequalities in the recovery rates of services with different characteristics. This will provide insights into the differences between high and low-performing services and how treatment outcomes of low-performing services can be improved.Our project has two key innovations: the first is to evaluate the causal impacts of the IAPT programme. Existing analyses of the IAPT programme are descriptive. We take this further, by using quasi-experimental techniques to estimate causal average and heterogeneous treatment effects. In particular, we will exploit differences in waiting times for treatment to construct quasi-experimental control groups. The second key innovation is that we will complement standard econometric approaches to analysing treatment effect heterogeneity with state-of-the-art machine learning techniques to uncover complex relations between service, patient, and local area characteristics and inequalities in the effectiveness of the IAPT programme.Our goal is to identify how services and treatments can be improved and patients and treatments better matched, informing evidence-based policy on how to improve treatment of mental ill-health in the UK and beyond.