Researchers: Hekselman Idan¹, Vardi Noa ¹, Markovitz Nechama¹, Steiner Amit², Menkes-Caspi Noa¹, Krivoy Amir1,2^1,3

Background: Metabolic syndrome (MetS) is a global epidemic that may lead to debilitating cardiovascular complications and excess mortality. Patients diagnosed with psychiatric disorders have a higher incidence of MetS, increasing their cardiovascular risk. Unfortunately, the complexity of metabolic exacerbation in psychiatric patients hinders identification of patients at higher risk.

Objectives: In this retrospective cohort study, we used machine learning to predict the risk of MetS in patients with major psychiatric disorder.

Method: We collected data on over half million Clalit members diagnosed with major psychiatric disorders between 2000 to 2024. Using these data, we generated 278 clinical features that were used to train predictive models, including demographic, diagnostic, anthropometric and laboratory features, as well as patterns of medication prescription and purchase. We developed survival prediction models for the diagnosis of MetS in this population.

Findings: These models showed high discriminative abilities. Characteristics that impact MetS risk were metabolic characteristics, as well as older age at the diagnosis of the major psychiatric disorder. Intriguingly, schizophrenia-spectrum disorders were a notable risk factor for the development of MetS in patients with no previous MetS risk.

Recommendations: Such models may serve as a tool for metabolic risk assessment in mental health patients, allowing early detection and intervention. Considering the increased metabolic risk associated with schizophrenia-spectrum disorders, we suggest proactive metabolic monitoring in these patients.

Research end date: 07/2026