acePrediction of Clinically Meaningful Improvement After Internet-Delivered Cognitive Behavioral Therapy for Depression and Anxiety Disorders: Machine Learning-Based Predictive Model Development and Temporal Validation Study
Abstract: Background Up to 50% of patients treated with internet-delivered cognitive behavioral therapy (ICBT) for depression and anxiety disorders do not experience clinically significant symptom reduction. Identifying these patients prior to the initiation of ICBT can support treatment planning. Objective The aim of this study was to enhance baseline prediction of clinically meaningful improvement in patients treated with ICBT for common psychiatric disorders in routine care, which could ultimately inform treatment plan…
PATIENT APPEALS IN HEALTH CARE by Committee on Ways and Means. Public domain
In a routine-care sample of 1790 patients with major depressive disorder, panic disorder, and social anxiety disorder, researchers built multimodal machine learning models from pretreatment clinical, sociodemographic, and genetic data to predict clinically meaningful improvement after internet-delivered cognitive behavioral therapy, testing them with nested cross-validation and temporal validation in 356 held-out patients as of the August 2026 publication.
The ability to stratify risk at intake could help clinicians identify patients unlikely to benefit from ICBT and adjust treatment planning, but the models have only been temporally validated in one cohort, genetic predictors showed no added value in this setup, and prospective validation of clinical utility is still needed.
- Study included 1790 patients treated with ICBT for major depressive disorder, panic disorder, and social anxiety disorder in routine care.
- Models integrated clinical, sociodemographic, and genetic data available pretreatment with nested cross-validation and temporal validation in 20% holdout test set (n=356).
- Primary performance measure was area under the receiver operating characteristic curve, with full phenotypic models achieving AUCtest 0.732-0.749.
- Polygenic scores added no independent predictive value in this cohort and modeling setup.
Up to half of patients receiving internet-delivered CBT for depression and anxiety disorders fail to achieve clinically significant symptom reduction, complicating treatment planning at intake.
The rundown
Researchers developed models using logistic regression, random forest, extreme gradient boosting, support vector machines, and ensemble methods, applying elastic net variable selection and multiple imputation on pretreatment data from 1790 routine-care patients.
In the 356-patient temporal holdout, all full phenotypic models performed comparably, but random forest and ensembles incorporating register data outperformed a benchmark based on self-reported screening data, while genetic data did not add value.
Sources
- Peer-reviewedJournal of Medical Internet Research2026-08-07
How should this claim be treated?
The debate