Masters candidate in Biostatistics, Shruti Vempati, will present:

“A Comparison of Prediction Methods for Hypertension and Type 2 Diabetes in Living Kidney Donors”

Advisor: Erika Helgeson

Abstract: Living kidney donation has been successfully performed since 1954, but there are risks associated with living with one kidney. Previous literature has studied the risks of hypertension and type 2 diabetes development, among other outcomes, in kidney donors using Cox proportional hazard models. While useful in evaluating linear effects, or pre-specified non-linear associations, these models do not identify non-linear associations as well as other prediction models, such as random forests. Also, there has been limited work in translating individualized risks associated with kidney donation in an accessible format for prospective kidney donors. In this study, we evaluated the predictive performance of Cox proportional hazard models and random survival forests in prediction of type 2 diabetes and hypertension in the University of Minnesota living kidney donor cohort. Results show that random forest models have lower prediction error curves and higher prediction accuracy than Cox models. We also developed a Shiny app to provide prospective donors with an accessible way to visualize their estimated risks of these outcomes following kidney donation. Prospective donors can input their demographic and medical information, along with recipient characteristics, and obtain information about their predicted risk of hypertension, type 2 diabetes, eGFR, ESRD, and proteinuria following donation.

Refreshments will be served prior to the presentation.