Speaker: Pro. Marteen Speekenbrink, UCL

Title: Mental health states, not types: Hidden Markov models for identification and monitoring of mental health trajectories.

Host: Manos Konstantinidis

Abstract: Psychopathology is complex and heterogeneous. I will argue that clinical theory and practice is better served by viewing psychopathology as a journey through mental health states, rather than assigning patients to diagnostic types. I will present a general method, based on hidden Markov models, to identify and track patients' latent mental health states over time. After introducing the key ideas and principles, I will show results for a large sample (N=2944) of heterogeneous psychiatric inpatients, who were assessed over time on the Health of the Nations Outcome Scores (HoNOS). I will show that the method identifies clinically meaningful mental health states, with plausible connections to formal diagnoses. Moreover, a patient's mental health state at time of admission provides better prognostic predictions than using HoNOS ratings directly. Together, these results show that decoding routinely collected outcome data into mental health states offers a viable and useful alternative to traditional diagnosis.