Logistic Regression, Classification Tree, and Boosted Regression: A Comparison of Models in the Prediction of Pregnancy-Induced Hypertension

• Activity Number: 209
• Type: Contributed
• Date/Time: Monday, July 30, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Paul Kolm*+ and Deborah Ehrenthal and Claudine Jurkovitz and Matthew Hoffman and James Bowen
• Companies: Christiana Care Health System and Christiana Care Health System and Christiana Care Health System and Christiana Care Health System and Christiana Care Health System
• Keywords: Data mining ; Logistic regression ; Tree classification ; Boosted regression

Abstract

The development of data-mining methods, and software, have facilitated building regression models from large databases in terms of number of observations and/or number of predictor variables. In this study, we compare logistic regression, classification tree and boosted regression models of pregnancy induced hypertension (PIH) in 19,626 patients. Predictors included age, race, parity, pre-pregnancy BMI, chronic hypertension, pre-gestational and gestational diabetes and smoking history. Similarities and differences in models, identification of interactions and nonlinearities, prediction accuracy, and differences in presentation of results are evaluated.