Key result
Machine learning models combining multiple domains achieved an AUC-ROC of 0.669 for predicting 9-year cardiometabolic disease onset, not significantly outperforming demographic variables alone.
Why the study?
Depression and anxiety disorders increase the risk of cardio-metabolic diseases, but whether adding detailed psychiatric or biological variables improves machine learning prediction of disease onset was unknown.
Does adding detailed psychiatric or biological variables to machine learning models improve the prediction of incident cardio-metabolic disease in a cohort with and without depressive and anxiety disorders?
Cohort (n=2,071)
Yes
Does adding detailed psychiatric or biological variables to machine learning models improve the prediction of incident cardio-metabolic disease in a cohort with and without depressive and anxiety disorders?
Answers are grounded in this paper and cite their sources.
Machine learning models incorporating detailed psychiatric and biological variables did not substantially improve the prediction of 9-year cardio-metabolic disease onset compared to basic demographics.
PURPOSE: There is a global increase of cardiovascular disease and diabetes (Cardio-Metabolic diseases: CMD). Suffering from depression or anxiety disorders increases the probability of developing CMD. In this study we tested a wide array of predictors for the onset of CMD with Machine Learning (ML), evaluating whether adding detailed psychiatric or biological variables increases predictive performance. METHODS: We analysed data from the Netherlands Study of Depression and Anxiety, a longitudinal cohort study (N = 2071), using 368 predictors covering 4 domains (demographic, lifestyle & somatic, psychiatric, and biological markers). CMD onset (24% incidence) over a 9-year follow-up was defined using self-reported stroke, heart disease, diabetes with high fasting glucose levels and (antithrombotic, cardiovascular, or diabetes) medication use (ATC codes C01DA, C01-C05A-B, C07-C09A-B, C01DB, B01, A10A-X). Using different ML methods (Logistic regression, Support vector machine, Random forest, and XGBoost) we tested the predictive performance of single domains and domain combinations. RESULTS: The classifiers performed similarly, therefore the simplest classifier (Logistic regression) was selected. The Area Under the Receiver Operator Characteristic Curve (AUC-ROC) achieved by singe domains ranged from 0.569 to 0.649. The combination of demographics, lifestyle & somatic indicators and psychiatric variables performed best (AUC-ROC = 0.669), but did not significantly outperform demographics. Age and hypertension contributed most to prediction; detailed psychiatric variables added relatively little. CONCLUSION: In this longitudinal study, ML classifiers were not able to accurately predict 9-year CMD onset in a sample enriched of subjects with psychopathology. Detailed psychiatric/biological information did not substantially increase predictive performance.
No takes yet. Share an insight, caveat, or question.
Rydin et al. (2025) conducted a cohort in Cardiometabolic disease (n=2,071). Machine learning prediction models (multiple domains) vs. Demographic variables alone was evaluated on Prediction of 9-year cardiometabolic disease onset (AUC-ROC). Machine learning models combining multiple domains achieved an AUC-ROC of 0.669 for predicting 9-year cardiometabolic disease onset, not significantly outperforming demographic variables alone.
Synapse has enriched 10 related papers. Consider them for comparative context: