The purpose of this study is to help patients decide whether to accept a kidney from a donor with a potential disease. We consider infectious risk donors (IRD) for HIV, HCV antibody positive and HBV core antibody positive donors. We will refer to these donors as high-risk donors.

We first build a kidney transplant survival model for the general population using machine learning and an ensemble of statistical methods. Our model obtains greater performance based on Harrell's concordance index, than the EPTS model used by the kidney allocation system in the U.S. Using a similar methodology we build transplant survival models for recipients with high-risk donors. We then use our models to build an interactive program that shows the survival curves for a patient either accepting a particular high-risk organ or remaining on the waiting list for a non high-risk organ. Using our program we simulate thousands of different patient scenarios and compare the survival for both decisions.

We find that patients have greater survival accepting the high-risk organs in the majority of simulated scenarios. In conclusion, we find that all three types of high-risk organs can offer lifesaving transplants and increase a patient's survival over staying on the wait list in many scenarios. Our interactive program can be used to help physicians and patients make this decision.

CITATION INFORMATION: Mark E., Goldsman D., Keskinocak P., Gurbaxani B. Using Machine Learning and Simulation to Compare High Risk Kidney Transplant Survival to Waiting for a Non High Risk Organ Am J Transplant. 2017;17 (suppl 3).

« Back to 2018 American Transplant Congress