ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO036

Predictive Modeling of Graft Failure Risk in Deceased Donor Kidney Transplants: Leveraging Machine Learning for Improved Outcomes and Data-Driven Insights

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Salybekov, Amankeldi, Shonan Kamakura Sogo Byoin, Kamakura, Kanagawa, Japan
  • Yerkos, Ainur, Al-Farabi Kazakh National University, Almaty, Almaty, Kazakhstan
  • Kunikeyev, Aidyn Dauletovich, Qazaq Institute of Innovative Medicne, Astana, Astana, Kazakhstan
  • Buribayev, Zholdas, Al-Farabi Kazakh National University, Almaty, Almaty, Kazakhstan
  • Mochida, Yasuhiro, Shonan Kamakura Sogo Byoin, Kamakura, Kanagawa, Japan
  • Ishioka, Kunihiro, Shonan Kamakura Sogo Byoin, Kamakura, Kanagawa, Japan
  • Hidaka, Sumi, Shonan Kamakura Sogo Byoin, Kamakura, Kanagawa, Japan
  • Kobayashi, Shuzo, Shonan Kamakura Sogo Byoin, Kamakura, Kanagawa, Japan
Background

Machine learning (ML) has shown its potential to improve patient care over the last decade. The 1-year graft failure rate remains a major concern among ethnic disparities in the deceased donor kidney transplantation (KT). The aim of the study was to evaluate graft failure within a year and determine risk factors associated with these events in ethnic groups.

Methods

The KT data between 2000 to 2020 was obtained from Organ Procurement and Transplantation Network. After data preprocessing n=75000 White, n=35000 Black, and n= 5000 Asian KT were qualified for further data analysis. The random forest (RF) and light gradient-boosting machine (LGBM) models were used to classify the most important variables for further 1-year GF-associated risk factors.

Results

Thus, the number of initial variables was reduced from 490 to 40 in every three groups after significance-test validation. LGBM demonstrated better discrimination of GF-related risk factors among three groups ((AUC=0.78; White), (AUC=0.77; Black), and (AUC=0.73; Asian)) in fivefold cross-validation. Interestingly, the top features for prediction of graft failure were such as donor hepatitis C-virus (HCV) antibody, donor and recipient tattoo, type of perfusion solution, age, arterial blood gas PH level, diabetes, hypertension, etc. Then, LGBM predicted variables were further analyzed in Cox proportional hazard ratio (HR), and the leading risk factor for graft failure was donor HCV antibody level ((HR 1.9; White, HR 2.2; Black, and HR 1.75; Asian, all p values were 0.005). In contrast, donor and recipient tattoos were independent risk factors (HR 1.2; P <0.02) for Asian while arterial blood gas PH (HR 1.2; P <0.01), posttransplant malignancy (HR 1.25-2.2; P <0.005), and organ flushing solution (HR 1.2; P <0.005) associated with 1 -year graft failure for all three groups. Interestingly, Cox proportional HR analysis demonstrated that ML-predicted risk factors such as diabetes, age, BMI, and hypertension were not associated with 1-year graft failure in all three groups.

Conclusion

Taking together, ML algorithms predicted risk factors accuracy almost the same as traditional statistical methods on prediction of 1-year graft failure among ethnic disparities.