Externally validated prediction models for incident mental health disorders in young adults: A systematic review of methods, performance and clinical utility
Abstract Mental disorders constitute a substantial public health concern globally. Prediction models may improve early identification of high-risk individuals. We conducted a systematic review to identify externally validated prediction models of depression, anxiety, bipolar and psychotic disorders. We searched MEDLINE, APA PsycINFO and Embase up to April 2026 along with backward and forward citation tracking. Out of 692 records yielded, 14 unique prediction models were identified for anxiety (n = 1), depression (n = 3), bipolar disorder (n = 1) or psychosis (n = 9). Most models used Cox proportional hazards (n= 6) and 13 models used solely psychosocial predictors. Model discrimination ranged from poor (AUC = 0.51) to considerable performance (AUC = 0.82). Only 1 model tested clinical utility. To date, there is a lack of accurate and clinically applicable models for predicting incident mental disorders.
Authors
- Jennifer L. Dykxhoorn (ORCID: https://orcid.org/0000-0001-6391-8626)
- Naomi Launders (ORCID: https://orcid.org/0000-0003-4241-938X)
- Anna Tsaligopoulou (ORCID: https://orcid.org/0000-0002-0570-4299)
- Joseph Hayes (ORCID: https://orcid.org/0000-0003-2286-3862)
- Layla Sadek (ORCID: https://orcid.org/0009-0004-0513-1728)
Institutions
- University College London (GB)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-12
- DOI
- https://doi.org/10.21203/rs.3.rs-10819677/v1
- Primary Topic
- Mental Health Treatment and Access
- Type
- preprint