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

Fig. 2

From: Dirichlet process mixture models to impute missing predictor data in counterfactual prediction models: an application to predict optimal type 2 diabetes therapy

Fig. 2

A Posterior predictive distributions for HbA1c outcome at 6-month post-treatment for three synthetic but representative patients. For patient A, SGLT2i has a 98% probability of performing better than DPP4i. B Predicted treatment response difference or conditional average treatment effect (CATE) at 6-months. A negative value corresponds to a benefit on SGLT2i, and a positive value corresponds to a benefit on DPP4i. For patient B, SGLT2i has a 58% probability of performing better than DPP4i. For patient C, DPP4i has a >99% probability of performing better than SGLT2i. Patient [A,B,C]: Number of Past Drugs [4,3,4], Number of Current Drugs [2,2,2], HbA1c [67,75,65], eGFR [84.2,66.6,67.9], ALT(log) [3.4,2.8,2.6], BMI [26.1,33.4,28.5], Age [68,79,81]

Back to article page

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

Contact us