Background
Depression and anxiety are the most common mental health conditions worldwide. NHS Talking Therapies treat over one million patients each year, yet only around half achieve recovery. Current treatment allocation—largely based on symptom severity—fails to reflect the diverse needs, comorbidities, and contexts of individual patients. To improve outcomes, personalised approaches are needed that match each patient to the treatment most likely to benefit them.
Novelty and Importance
This PhD will be the first to apply prediction under interventions—a causal inference framework estimating counterfactual outcomes—together with machine learning to stratify psychotherapy at scale within NHS Talking Therapies. It will combine evidence from randomised controlled trials with large-scale electronic health records to build causal prediction models capable of identifying which treatment is most effective for each patient. A tiered strategy will evaluate both routine clinical decisions (e.g. therapy intensity, LTC pathway, therapy ± medication) and broader innovations (e.g. in-person vs. digital delivery, clinical support tools). Models will be developed using modern causal machine learning approaches, validated in linked datasets, and assessed for clinical acceptability through a vignette-based evaluation with practitioners, ensuring translational impact.
Aims & Objectives
The project will:
1) Define clinically meaningful intervention contrasts;
2) Develop causal prediction models using NHS and linked cohort data;
3) Integrate RCT treatment effects to strengthen causal inference;
4) Evaluate calibration, discrimination, and decision utility in external datasets;
5) Assess clinician perspectives on model recommendations.
References
Hernán MA, Robins JM. Causal Inference: What If. Chapman & Hall, 2020.
Lipkovich I, Svensson D, Ratitch B, Dmitrienko A. Modern approaches for evaluating treatment effect heterogeneity from clinical trials and observational data. Stat Med. 2024;43(21):3989–4009. https://doi.org/10.1002/sim.10167
Buckman JEJ, Cohen ZD, O’Driscoll C, Fried EI, Saunders R, Ambler G, et al. Predicting prognosis for adults with depression using individual patient data from multiple randomized controlled trials: development and validation of multivariable prognostic models. JAMA Psychiatry. 2022;79(5):449–59.
Cristea IA, Karyotaki E, Hollon SD, Cuijpers P. What is personalised psychotherapy? Moving from evidence-based practice to practice-based evidence. Behav Res Ther. 2025;174:104657.