Basterfield C, Newman MG. Development of a machine learning-based multivariable prediction model for the naturalistic course of generalized anxiety disorder.
J Anxiety Disord 2025;
110:102978. [PMID:
39904097 PMCID:
PMC11875880 DOI:
10.1016/j.janxdis.2025.102978]
[Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 05/09/2024] [Revised: 01/16/2025] [Accepted: 01/19/2025] [Indexed: 02/06/2025]
Abstract
BACKGROUND
Generalized Anxiety Disorder (GAD) is a chronic condition. Enabling the prediction of individual trajectories would facilitate tailored management approaches for these individuals. This study used machine learning techniques to predict the recovery of GAD at a nine-year follow-up.
METHOD
The study involved 126 participants with GAD. Various baseline predictors from psychological, social, biological, sociodemographic and health variables were used. Two machine learning models, gradient boosted trees, and elastic nets were compared to predict the clinical course in participants with GAD.
RESULTS
At nine-year follow-up, 95 participants (75.40 %) recovered. Elastic nets achieved a cross-validated area-under-the-receiving-operator-characteristic-curve (AUC) of .81 and a balanced accuracy of 72 % (sensitivity of .70 and specificity of .76). The elastic net algorithm revealed that the following factors were highly predictive of nonrecovery at follow-up: higher depressed affect, experiencing daily discrimination, more mental health professional visits, and more medical professional visits. The following variables predicted recovery: having some college education or higher, older age, more friend support, higher waist-to-hip ratio, and higher positive affect.
CONCLUSIONS
There was acceptable performance in predicting recovery or nonrecovery at a nine-year follow-up. This study advances research on GAD outcomes by understanding predictors associated with recovery or nonrecovery. Findings can potentially inform more targeted preventive interventions, ultimately improving care for individuals with GAD. This work is a critical first step toward developing reliable and feasible machine learning-based predictions for applications to GAD.
Collapse