From Clinic to Code: Developing Interpretable ML Models for Treatment Selection in Major Depressive Disorder
DOI:
https://doi.org/10.64229/zb2r9r26Keywords:
Major Depressive Disorder, Interpretable Machine Learning, Treatment Selection, Explainable AI, Digital Phenotyping, Precision Psychiatry, Cognitive Behavioral Therapy, AntidepressantsAbstract
Background: Major Depressive Disorder (MDD) is a heterogeneous condition with a wide range of treatment options, yet achieving remission remains a challenge due to the trial-and-error nature of treatment selection. While machine learning (ML) promises to personalize this process, "black-box" models often lack clinical trust and actionable insights, limiting their adoption.
Objective: This study aims to develop and validate an interpretable ML pipeline for predicting optimal first-line treatment selection between Selective Serotonin Reuptake Inhibitors (SSRIs) and Cognitive Behavioral Therapy (CBT) for patients with MDD.
Methods: We utilized a dataset of 1,250 patients from the [Anonymized] Neuropsychiatric Dataset, featuring comprehensive clinical, demographic, and digital phenotyping data. We trained and compared several ML models, including a black-box Gradient Boosting Machine (GBM) and an interpretable Explainable Boosting Machine (EBM). Model performance was assessed using accuracy, F1-score, and Area Under the Receiver Operating Characteristic Curve (AUC). Interpretability was achieved through global feature importance and local explanations for individual predictions.
Results: The GBM model achieved the highest performance (AUC = 0.87), with the EBM model performing comparably (AUC = 0.85). Crucially, the EBM provided transparent insights, identifying key predictors of treatment success, such as baseline anxiety severity, sleep disturbance patterns, cognitive performance scores, and actigraphy-derived physical activity levels. A novel, clinically-actionable visualization, the "Treatment Suitability Scorecard," is presented for individual patient guidance.
Conclusion: Interpretable ML models can achieve performance comparable to black-box models while providing crucial transparency for treatment selection in MDD. The proposed pipeline and visualization tools facilitate the transition of ML from a research tool to a clinically-deployable decision-support system, fostering trust and enabling personalized, evidence-based care in digital neuropsychiatry.
References
[1]Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Nierenberg, A. A., Stewart, J. W., Warden, D., ... & Fava, M. (2006). Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: A STAR*D report. American Journal of Psychiatry, 163(11), 1905-1917. https://doi.org/10.1176/ajp.2006.163.11.1905
[2]Chekroud, A. M., Zotti, R. J., Shehzad, Z., Gueorguieva, R., Johnson, M. K., Trivedi, M. H., ... & Corlett, P. R. (2016). Cross-trial prediction of treatment outcome in depression: A machine learning approach. The Lancet Psychiatry, 3(3), 243-250. https://doi.org/10.1016/S2215-0366(15)00471-X
[3]Rudin, C. (2019). Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nature Machine Intelligence, 1(5), 206-215. https://doi.org/10.1038/s42256-019-0048-x
[4]Lou, Y., Caruana, R., & Gehrke, J. (2012). Intelligible models for classification and regression. In Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 150-158). ACM. https://doi.org/10.1145/2339530.2339556
[5]Insel, T. R., Cuthbert, B. N., Garvey, M. A., Heinssen, R. K., Pine, D. S., Quinn, K. J., ... & Wang, P. S. (2010). Research domain criteria (RDoC): Toward a new classification framework for research on mental disorders. American Journal of Psychiatry, 167(7), 748-751. https://doi.org/10.1176/appi.ajp.2010.09091379
[6]Dwyer, D. B., Falkai, P., & Koutsouleris, N. (2018). Machine learning approaches for clinical psychology and psychiatry. Annual Review of Clinical Psychology, 14, 91-118. https://doi.org/10.1146/annurev-clinpsy-032816-045037
[7]Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems 30 (pp. 4765-4774). https://doi.org/10.48550/arXiv.1705.07874
[8]Torous, J., Kiang, M. V., Lorme, J., & Onnela, J. P. (2016). New tools for new research in psychiatry: A scalable and customizable platform to empower data driven smartphone research. JMIR Mental Health, 3(2), e16. https://doi.org/10.2196/mental.5165
[9]Cabitza, F., Rasoini, R., & Gensini, G. F. (2017). Unintended consequences of machine learning in medicine. JAMA, 318(6), 517-518. https://doi.org/10.1001/jama.2017.7797
[10]Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). "Why should I trust you?": Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 1135-1144). ACM. https://doi.org/10.1145/2939672.2939778
[11]Friedrich, M. J. (2017). Depression is the leading cause of disability around the world. JAMA, 317(15), 1517. https://doi.org/10.1001/jama.2017.3826
[12]Kessler, R. C., van Loo, H. M., Wardenaar, K. J., Bossarte, R. M., Brenner, L. A., Cai, T., ... & Zaslavsky, A. M. (2017). Using patient self-reports to study heterogeneity of treatment effects in major depressive disorder. Epidemiology and Psychiatric Sciences, 26(1), 22-36. https://doi.org/10.1017/S2045796016000020
[13]Trivedi, M. H., McGrath, P. J., Fava, M., Parsey, R. V., Kurian, B. T., Phillips, M. L., ... & Weissman, M. M. (2016). Establishing moderators and biosignatures of antidepressant response in clinical care (EMBARC): Rationale and design. Journal of Psychiatric Research, 78, 11-23. https://doi.org/10.1016/j.jpsychires.2016.03.001
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Mercado Protacio (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
