BACKGROUND: Efficient detection of individuals at risk of developing psychotic disorders or bipolar disorders is a crucial step to improving mental health outcomes in young people. A novel, transdiagnostic approach to jointly detect individuals at risk for either psychotic disorders or bipolar disorders would maximise the effect of prevention. The aim of this study is to develop and validate an individualised prediction model to detect the risk of developing psychotic disorders or bipolar disorders in the UK. METHODS: This RECORD Statement and TRIPOD+AI compliant study describes the development and validation of a clinical prediction model to estimate the risk of developing psychotic or bipolar disorders using data from patients of all ages with an index diagnosis of a non-organic, non-psychotic and non-bipolar mental disorder recorded in electronic health records from South London and Maudsley (SLaM in the UK) secondary mental health care between Jan 1, 2008, and Aug 10, 2021. Exclusion criteria included receiving long-acting injectable antipsychotics or clozapine before a diagnosis of bipolar or psychotic disorders, no recorded contact with SLaM services after the index date, and an index date falling within the washout period Jan 1, 2008, to June 30, 2008. A least absolute shrinkage and selection operator-regularised (LASSO) Cox proportional hazards model was developed to estimate the 6-year risk of developing psychotic disorders or bipolar disorders, incorporating sociodemographic and clinical predictors at index date (five predictors), and medication (four predictors), hospitalisation (two predictors) and natural language processing-derived signs and symptoms and substance use (66 predictors), derived using a 6-month look-back period. Model performance was assessed using internal-external validation, sequentially leaving out one borough from the SLaM area for testing and averaging performance across all five boroughs. The final model was fit with data across all the boroughs. Performance was assessed via discrimination (C-index), calibration (calibration slope and calibration-in-the-large), and potential clinical utility (decision curve analysis) during internal-external cross-validation. Individuals with lived experience of bipolar disorders or psychotic disorders were not involved in the research or writing process. FINDINGS: In total, data from 127 868 patients were included. 64 980 (50·8%) of the dataset were male, 62 711 (49·0%) were female, and 89 (0·1%) were other gender. For self-assigned ethnicity, the dataset was 71 390 (55·8%) White, 18 025 (14·1%) Black, 7257 (5·7%) other, 6270 (4·9%) Asian, and 5022 (3·9%) mixed (19 904 [15·6%] were missing ethnicity data). The mean age was 33·4 years (SD 18·8 [IQR 17·9-44·9]). The cumulative risk incidence of psychotic disorders or bipolar disorders was 0·0827 (95% CI 0·0784-0·0870) within 6 years (mean follow-up 622 days [SD 687]). The model showed the following performance in internal-external validation: C-index 0·80 (95% CI 0·78-0·81); calibration slope 1·02 (SD 0·14); calibration-in-the-large 0·06 (SD 0·02). Decision curve analysis showed that use of the model would detect three additional cases of psychotic disorders or bipolar disorders early per 100 patients screened compared with default assessment strategies. INTERPRETATION: This study shows that the transdiagnostic clinical prediction model can identify patients at risk of developing psychotic disorders or bipolar disorders and displayed excellent performance. Such a novel approach would enable systematic early detection of young people at risk of psychotic disorders or bipolar disorders, advancing preventive care in real-world clinical practice. FUNDING: UK Medical Research Council (MR/N013700/1), National Institute for Health Research (NIHR) Biomedical Research Centres at South London and Maudsley NHS Foundation Trust, and Oxford Health NHS Foundation Trust.