Skip to main content
Fig. 2 | BMC Medical Informatics and Decision Making

Fig. 2

From: Comparison of causal forest and regression-based approaches to evaluate treatment effect heterogeneity: an application for type 2 diabetes precision medicine

Fig. 2

Final treatment selection model performance for A Penalized regression and B Causal forest in CPRD validation data. Left panels show the distribution of predicted individualized treatment effects. Negative values reflect a predicted benefit on SGLT2-inhibitor treatment, positive values reflect a predicted HbA1c benefit on DPP4-inhibitor treatment. Right panels show calibration between observed and predicted treatment effects, across strata defined by decile of predicted treatment effect. Estimates are adjusted for clinical features in the treatment selection model (see Methods: Predictors section), and potential confounders (see Methods: Confounders section) to improve precision and control for potential differences in covariate balance within strata

Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us