Approved Research

Replicating published diagnostic and prognostic models for severe mental disorders

Lay summary

Only 5% of prediction models in psychiatry have been tested in more than one place. There are some that hold promise but cannot be used in clinical care until they have been shown to predict well using new data in new settings. For example, a recent genetic diagnostic test for bipolar disorder performs well but has been shown to work in only a small number of people. By showing that this model, and other similar models, performs well in a new dataset, we can support the use of them in real-world clinical care and help people get the best care for them. As such, we aim to use UK BioBank data to test several models for classifying or predicting the onset of severe mental disorders and potentially support their use in clinical practice.

Project duration: 3 years

Public health impact: This will provide evidence for the use of models that can support clinical decision making in psychiatry, helping many people better access the support they need.