WEDNESDAY, Feb. 26, 2025 (HealthDay News) — It may be possible to predict diagnostic transition to schizophrenia and bipolar disorder from machine learning models using routine clinical data extracted from electronic health records, according to a study published online Feb. 19 in JAMA Psychiatry.
Lasse Hansen, Ph.D., from Aarhus University Hospital in Denmark, and colleagues investigated whether machine learning models using routine clinical data from electronic health records can predict diagnostic progression to schizophrenia or bipolar disorder among patients undergoing treatment in psychiatric services for other mental illness. The analysis included 24,449 patients aged 15 to 60 years with at least two contacts at least three months apart for psychiatric services.
The researchers found that transition to the first occurrence of either schizophrenia or bipolar disorder was predicted by the XGBoost model, with an area under the receiver operating characteristic curve (AUROC) of 0.70 on the training set and 0.64 on the test set. Using a predicted positive rate of 4 percent, the XGBoost model had a sensitivity of 9.3 percent, a specificity of 96.3 percent, and a positive predictive value (PPV) of 13.0 percent. Better performance was seen for predicting schizophrenia separately (AUROC, 0.80; sensitivity, 19.4 percent; specificity, 96.3 percent; PPV, 10.8 percent) than for bipolar disorder (AUROC, 0.62; sensitivity, 9.9 percent; specificity, 96.2 percent; PPV, 8.4 percent). Prediction was aided by clinical notes.
“These findings suggest that detecting progression to schizophrenia through machine learning based on routine clinical data is feasible, which may reduce diagnostic delay and duration of untreated illness,” the authors write.
Several authors disclosed ties to relevant organizations.
Copyright © 2025 HealthDay. All rights reserved.
