*Purpose: Despite the vital need, 3,553 (19%) deceased-donor kidneys were discarded in 2018 in the current US allocation system. The discard rate has remained at 18%-20% over the last decade.

*Methods: The cohort consisted of adult deceased donor kidneys donated between 12/3/2014-03/01/2019. Two logistic regression models (KROD-A and KROD-B) using Kidney Donor Profile Index (KDPI) with additional covariates were fitted.

*Results: Kidney discards are predictable (KROD-A AUC=0.857, KROD-B AUC=0.884, Figure 1a). KDPI-based models improve over the probability of discard models (PODD-r AUC=0.853, PODD AUC=0.875, P<0.01, Figure 1b and 1c). A KDPI-only discard model’s AUC=0.773. The advantage of using KROD models is more pronounced when considering KDPI ≥ 75% kidneys discarded as “No recipient found, list exhausted” (KROD-B accuracy=79.97% vs PODD accuracy=69.58%). In the KROD-A model, the most notable predictors of discard are hepatitis C (OR=7.43), KDPI >95% (OR=5.78), hepatitis B surface antigen (OR=4.03), tumor (OR=3.94), and KDPI (90%, 95%] (OR=3.04), and. The OR is >1.5 for donors with KDPI (80, 85%] and (85%, 90%], terminal creatinine >1.5 mg/dL, DCD, diabetes >10 years, and positive hypertension. Glomerulosclerosis >20% (OR=16.03), (15%, 20%] (OR=4.59) are the notable predictors in KROD-B model along with the variables in KROD-A model. Secondary analysis shows that KDPI >75% kidneys when biopsied and those on pump have a reduced risk of discard.

*Conclusions: KDPI-based risk scores can be incorporated in policies intended to reduce kidney discard rates. Further investigation of variables with effect additive to KDPI on discard is desirable.

« Back to 2020 American Transplant Congress