Toward personalized care for insomnia in the US Army: a machine learning model to predict response to cognitive behavioral therapy for insomnia

Frances H Gabbay,Gary H Wynn,Matthew W Georg,Sarah M Gildea,Chris J Kennedy,Andrew J King,Nancy A Sampson,Robert J Ursano,Murray B Stein,James R Wagner,Ronald C Kessler,Vincent F Capaldi,Frances H. Gabbay,Gary H. Wynn,Matthew W. Georg,Sarah M. Gildea,Chris J. Kennedy,Andrew J. King,Nancy A. Sampson,Robert J. Ursano,Murray B. Stein,James R. Wagner,Ronald C. Kessler,Vincent F. Capaldi

DOI: https://doi.org/10.5664/jcsm.11026

2024-02-01

Journal of Clinical Sleep Medicine

Abstract:STUDY OBJECTIVES: The standard of care for military personnel with insomnia is cognitive behavioral therapy for insomnia (CBT-I). However, only a minority seeking insomnia treatment receive CBT-I, and little reliable guidance exists to identify those most likely to respond. As a step toward personalized care, we present results of a machine learning (ML) model to predict CBT-I response.METHODS: Administrative data were examined for n=1,449 nondeployed US Army soldiers treated for insomnia with CBT-I who had moderate-severe baseline Insomnia Severity Index (ISI) scores and completed one or more follow-up ISIs 6-12 weeks after baseline. An ensemble ML model was developed in a 70% training sample to predict clinically significant ISI improvement (reduction of at least two standard deviations on the baseline ISI distribution). Predictors included a wide range of military administrative and baseline clinical variables. Model accuracy was evaluated in the remaining 30% test sample.RESULTS: 19.8% of patients had clinically significant ISI improvement. Model AU-ROC (SE) was 0.60 (0.03). The 20% of test sample patients with highest probabilities of improvement were twice as likely to have clinically significant improvement as the remaining 80% (36.5% versus 15.7%; χ²₁=9.2, p=.002). Nearly 85% of prediction accuracy was due to ten variables, the most important of which were baseline insomnia severity and baseline suicidal ideation.CONCLUSIONS: Pending replication, the model could be used as part of a patient-centered decision-making process for insomnia treatment. Parallel models will be needed for alternative treatments before such a system is of optimal value.

clinical neurology