The research, published Feb. 19 in JAMA Psychiatry, analyzed data from 24,449 patients ages 15-60 who had at least two psychiatric outpatient visits in Denmark’s Central Denmark Region between 2013 and 2016.
The learning model, which was trained on nearly 400,000 outpatient contacts, used predictors such as medication history, clinical notes and previous diagnoses to assess the likelihood of a patient developing schizophrenia or bipolar disorder.
The results showed that the model performed better at predicting schizophrenia than bipolar disorder. The findings also suggest machine learning can help reduce diagnostic delays and improve early intervention, the researchers said.