Abstract: TH-PO039
Developing a Risk Index Predicting Kidney Transplant Dropout for African American and Hispanic Patients Using Artificial Intelligence/Machine Learning
Session Information
- AI, Digital Health, Data Science - I
November 02, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Augmented Intelligence, Digital Health, and Data Science
- 300 Augmented Intelligence, Digital Health, and Data Science
Authors
- Waterman, Amy D., Houston Methodist, Houston, Texas, United States
- Hsu, Enshuo, Houston Methodist, Houston, Texas, United States
- Axelrod, David, University of Iowa Hospitals and Clinics, Iowa City, Iowa, United States
- Jones, Stephen L., Houston Methodist, Houston, Texas, United States
- Patel, Shefali A., Houston Methodist, Houston, Texas, United States
- Al Awadhi, Solaf, Houston Methodist, Houston, Texas, United States
- Meinders, Andrea M., Houston Methodist, Houston, Texas, United States
- Pulicken, Catherine, Houston Methodist, Houston, Texas, United States
- Parsons, Faith, Houston Methodist, Houston, Texas, United States
- Cassell, Victoria, Houston Methodist, Houston, Texas, United States
- Kakadiaris, Ioannis A., University of Houston, Houston, Texas, United States
Background
On average, kidney patients who receive a transplant live longer and have a better quality of life than those who remain on dialysis. Since African American (AA) and Hispanic patients are more likely to drop out before receiving a transplant than White patients, identifying patients at higher risk of dropout can improve care.
Methods
We created a dataset from Houston Methodist sources, including EPIC Clarity, Phoenix, and United Network for Organ Sharing (UNOS), and external databases including Health Resources Services Administration (HRSA), Centers for Disease Control and Prevention (CDC), and the US Census Bureau for community-level characteristics. The final transplant analytical registry included 4,245 kidney patients pursuing transplants from 6/2016 to 5/2022 and contained clinical, demographic, social, and transplant variables. The data was preprocessed to apply AI/ML methods to predict the risk index of patients who began transplant evaluation but were never listed for transplant within 12 months. F1 and AUROC were the evaluation metrics used.
Results
Of the 4,245 individuals who presented for transplant evaluation, 1,999 (47.1%) were waitlisted after 12 months, and 550 (13.0%) were transplanted. AA (7.23%) and Hispanic (12.20%) patients received fewer transplants compared to White patients (19.70%).
Our AI/ML-based risk index predicted the probability of not being listed for transplant within 12 months (AUROC=0.732). Patients predicted to be low dropout risk (0-30%) had an 81.3% listing rate, middle risk (30-60%) had a 54.3% listing rate, and high risk (>60%) had a 27% listing rate within 12 months (Figure 1). Compared to low-risk patients, high-risk patients were more likely to be African American (44.2% vs. 19%), Hispanic (31.5% vs. 25.3%), unemployed (91.2% vs. 25%), spent more time on dialysis (488 vs. 131 days), without intended living donors (3% vs. 64.5%), and live in areas with more residents living below the poverty line (28% vs. 17%).
Conclusion
Applying the risk index, we will identify patients at higher risk of dropout and explore health delivery improvements within the transplant care process to slow or eliminate transplant dropout for underserved communities.
Funding
- Other NIH Support