Causal Actionable Prediction and Target Trial Emulation in Psychosis: Leveraging Real-World Clinical Cohorts to Target Treatment Resistance and Cardiometabolic Disease

Background

People with psychosis and schizophrenia face profound health inequalities, experiencing a life expectancy up to 15 years shorter than the general population.(1) A major driver of this premature mortality is cardiometabolic disease – a critical presentation of multimorbidity that is heavily compounded by rapid, antipsychotic-induced weight gain.(2) While early intervention services improve initial trajectories, clinical management is frequently stalled by a bidirectional crisis: high rates of treatment-resistant schizophrenia (TRS) and an inability to safely balance psychiatric efficacy with cardiometabolic harm. Consequently, clozapine (the gold-standard treatment for TRS) is profoundly underutilised, delayed by clinical apprehension regarding severe metabolic side effects.(3) Current clinical prediction models are unable to implement a precision medicine approach to resolve this because they rely on static, associational population averages that cannot guide actionable, personalised treatment. Current prediction models cannot answer counterfactual “what if” questions or dynamically account for time-varying confounding. To reduce health inequality, clinicians require causal actionable prediction models capable of calculating the conditional average treatment effect (CATE) and simulating counterfactual outcomes for individual patients to meet the precision medicine ideal.(4)

Aims

This PhD project will leverage advanced causal inference methodologies and large-scale electronic healthcare records to pioneer precision intervention strategies that safely optimise early clozapine access while mitigating cardiometabolic risks. The specific aims are to:

1. Mathematically validate the hypothesised bidirectional causal structure linking metabolic impairment and treatment-resistant psychosis using global causal discovery algorithms applied to longitudinal healthcare data
2. Emulate parallel target trials within a strictly incident "new-user" first-episode psychosis cohort. Trial A will estimate the prognostic effect of baseline metabolic impairment on the time-to-onset of TRS, while Trial B will estimate the dynamic effect of incident TRS on time-to-incident cardiometabolic multimorbidity. This will utilise Inverse Probability of Treatment Weighting (IPTW) and doubly robust survival models to adjust for time-varying confounding. 
3. Develop and internally validate causal actionable prediction models using cutting-edge machine learning (e.g., Bayesian Additive Regression Trees, Causal Forests). These models will evaluate heterogeneity of treatment effect (HTE) and estimate CATE, thereby personalising both psychiatric response and physical health risk profiles for early interventions.

Training Outcomes

The student will receive world-class, interdisciplinary training at the interface of clinical psychiatry, epidemiology, and computer science. They will gain highly sought-after quantitative skills, mastering causal inference frameworks (including g-methods), Target Trial Emulation, and advanced machine learning techniques applicable to complex, longitudinal electronic healthcare records. The PhD will benefit from our existing collaboration with and grant funding from the UKRI Causality in Healthcare AI Hub and collaborating institutions, embedding the student in a vibrant, cross-institutional research environment. The student will be perfectly positioned to become a future leader in health equalities, driving the translation of trustworthy AI into clinical practice to break the cycle of multimorbidity, treatment resistance, and metabolic decline in marginalised psychiatric populations.

References

1. National Mental Health Intelligence Network. Premature mortality in adults with severe mental illness 2023 [Available from: https://www.gov.uk/government/publications/premature-mortality-in-adults-with-severe-mental-illness/premature-mortality-in-adults-with-severe-mental-illness-smi.
2. Kivimaki M, Strandberg T, Pentti J, Nyberg ST, Frank P, Jokela M, et al. Body-mass index and risk of obesity-related complex multimorbidity: an observational multicohort study. Lancet Diabetes Endocrinol. 2022;10(4):253-63.
3. Krishnadas R, Fernandez-Egea E. Clozapine Beyond Walls: Community Clozapine Initiation in Early Psychosis. Schizophrenia Bull. 2026;52(4).
4. Bica I, Alaa AM, Lambert C, van der Schaar M. From Real-World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges. Clin Pharmacol Ther. 2021;109(1):87-100.