Key result
An extreme gradient boosted tree (XGBoost) model predicted 30-day readmission in patients with mental or substance use disorders with an AUROC of 0.737, outperforming a GLMNet model (AUROC 0.697).
Why the study?
Mental or substance use disorders carry a high risk for hospital readmissions, prompting the development of a machine learning-based readmission prediction model.
Does an XGBoost machine learning model improve the prediction of 30-day hospital readmission in patients with mental or substance use disorders compared to a generalized linear model?
Observational (n=65,426)
Does an XGBoost machine learning model improve the prediction of 30-day hospital readmission in patients with mental or substance use disorders compared to a generalized linear model?
Absolute Event Rate: 0.737% vs 0.697%
An XGBoost machine learning model outperformed a generalized linear model in predicting 30-day hospital readmissions for patients with mental or substance use disorders.
No takes yet. Share an insight, caveat, or question.
May aid readmission prediction in mental health; extends ML modeling but requires prospective validation before adoption.
Morel et al. (2020) conducted an observational in Mental or substance use disorders (n=65,426). Extreme gradient boosted tree (XGBoost) model vs. Generalized linear model with elastic net regularization (GLMNet) was evaluated on Readmission within 30-days from discharge (AUROC) (95% CI 0.732-0.742). An extreme gradient boosted tree (XGBoost) model predicted 30-day readmission in patients with mental or substance use disorders with an AUROC of 0.737, outperforming a GLMNet model (AUROC 0.697).
Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context: