The Prognostic Value of Early Treatment Factors for Cognitive Behavioral Therapy for Insomnia (CBT-I) Outcomes.
Cognitive behavioral therapy for insomnia (CBT-I) is the first-line treatment for chronic insomnia, yet a significant proportion of patients fail to achieve favorable outcomes. Baseline-only prognostic models show limited predictive performance, and the added value of early treatment variables remains unclear. This retrospective study aimed to investigate this issue.
We retrospectively included 1059 patients who completed face-to-face group CBT-I. Post-treatment outcomes were assessed using the Insomnia Severity Index (ISI), and 12-month follow-up outcomes were evaluated using the Pittsburgh Sleep Quality Index (PSQI). The dataset was randomly split into training (70%) and test (30%) sets. Candidate predictors were selected using Least Absolute Shrinkage and Selection Operator, Random Forest, and eXtreme Gradient Boosting. Variables consistently identified across all three algorithms were entered into multivariable logistic regression models. Predictive performance of baseline-only and combined models (baseline plus early treatment variables) was evaluated using the area under the receiver operating characteristic curve (AUC) and Brier Score, with bootstrap resampling used for model comparison.
Unfavorable outcomes were observed in 32.4% of patients at post-treatment (neither ISI < 8 nor a reduction > 7) and in 44.5% at the 12-month follow-up (PSQI > 5). The combined model significantly outperformed the baseline-only model in predicting post-treatment outcomes in the test set (AUC: 0.741 vs 0.630, p < 0.001), with week 2 ISI change identified as the strongest predictor. For the 12-month follow-up, the combined model demonstrated a non-significantly higher AUC compared with the baseline-only model (0.742 vs 0.735, p > 0.05), with age emerging as the strongest predictor.
Incorporating early treatment variables, particularly Week 2 ISI score reduction, may provide incremental value in predicting short-term CBT-I outcomes. However, the added value of early treatment variables for long-term outcomes appeared limited.
We retrospectively included 1059 patients who completed face-to-face group CBT-I. Post-treatment outcomes were assessed using the Insomnia Severity Index (ISI), and 12-month follow-up outcomes were evaluated using the Pittsburgh Sleep Quality Index (PSQI). The dataset was randomly split into training (70%) and test (30%) sets. Candidate predictors were selected using Least Absolute Shrinkage and Selection Operator, Random Forest, and eXtreme Gradient Boosting. Variables consistently identified across all three algorithms were entered into multivariable logistic regression models. Predictive performance of baseline-only and combined models (baseline plus early treatment variables) was evaluated using the area under the receiver operating characteristic curve (AUC) and Brier Score, with bootstrap resampling used for model comparison.
Unfavorable outcomes were observed in 32.4% of patients at post-treatment (neither ISI < 8 nor a reduction > 7) and in 44.5% at the 12-month follow-up (PSQI > 5). The combined model significantly outperformed the baseline-only model in predicting post-treatment outcomes in the test set (AUC: 0.741 vs 0.630, p < 0.001), with week 2 ISI change identified as the strongest predictor. For the 12-month follow-up, the combined model demonstrated a non-significantly higher AUC compared with the baseline-only model (0.742 vs 0.735, p > 0.05), with age emerging as the strongest predictor.
Incorporating early treatment variables, particularly Week 2 ISI score reduction, may provide incremental value in predicting short-term CBT-I outcomes. However, the added value of early treatment variables for long-term outcomes appeared limited.
Authors
Hu Hu, Ni Ni, Wang Wang, Zheng Zheng, Weng Weng, Lai Lai, Huang Huang, Wei Wei
View on Pubmed