Photo Credit: prostock-studio
The following is a summary of “Predicting post-treatment symptom severity for adults receiving psychological therapy in routine care for generalised anxiety disorder: a machine learning approach,” published in the June 2024 issue of Psychiatry by Delamain et al.
Many patients with generalized anxiety disorder (GAD) don’t fully recover with standard treatments, and a lack of reliable tools to predict treatment outcomes.
Researchers conducted a retrospective study to develop and validate a precise and understandable prediction model.
They used data from 15,859 adults using 13 different machine learning (ML) algorithms undergoing treatment for GAD across 8 Improving Access to Psychological Therapies (IAPT) services. Patients were divided into training, validation, and holdout sets, 13 ML algorithms were tested using 10-fold cross-validation. The best model was then applied to the holdout dataset. Model-specific measures were used to determine which factors most influenced GAD symptom severity prediction.
The results showed that the Bayesian Additive Regression Trees model performed the best (MSE= 16.54 [95% CI: 15.58-17.51]; MAE=3.19, R2= 0.33), surpassing simpler models like linear regression (MSE= 20.70 [95% CI=19.58-21.82]; MAE=3.94, R2=0.14). Key predictors included PHQ-9 anhedonia, GAD-7 annoyance/irritability, restlessness, fear items, and referral-assessment waiting time. The model accurately predicted post-treatment GAD severity, outperforming clinical judgment approximations.
Investigators concluded the model reliably predicted how worse GAD symptoms will be after treatment using only pre-treatment data which may help doctors decide on treatments and give patients a clearer idea of what to expect during therapy for GAD.
Source: sciencedirect.com/science/article/pii/S0165178124001951
