Abstract: PUB173
Joint Modeling of Longitudinal Ferritin Trajectories and COVID-19 Infections in Native American, Hispanic, and White Patients on Hemodialysis
Session Information
Category: Dialysis
- 801 Dialysis: Hemodialysis and Frequent Dialysis
Authors
- Hu, Mingzhao, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Quandelacy, Talia, University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States
- Wang, Yuedong, University of California Santa Barbara, Santa Barbara, California, United States
- Kotanko, Peter, Renal Research Institute, New York, New York, United States
- Conway, Baqiyyah, University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States
Background
American Indians/Alaska Natives (AIAN) and Hispanic individuals in the U.S. are overrepresented in the hemodialysis (HD) population and were hardest hit by COVID-19. Early in the pandemic, hyperferritinemia was linked to COVID-19 infections. Joint modeling learns from both longitudinal and time-to-event data for predictive accuracy and dynamic estimation. We investigated the predictive utility of ferritin w/ COVID-19 infection in a cohort of AIAN, Hispanic and White HD patients.
Methods
We followed 2,242 HD patients from 1/2020 to 3/2022. Baseline ferritin was the average between days 365 and 14 till COVID-19 diagnosis (day 0). In non-infected control, a random date matching the distribution of diagnosis dates was chosen as the “test date”. Longitudinal ferritin trajectories were modeled with a linear mixed effects model and COVID-19 infection with a Cox model. A joint model linked these to measure the association between COVID-19 infection and ferritin.
Results
Cumulative COVID-19 incidence was 36%. In exploring ferritin’s link with potential confounders, the longitudinal submodel coefficients highlighted age had a strong positive link while albumin, creatinine, hemoglobin and WBCs were not linked. The joint model revealed a marginal association between ferritin and infection risk (HR = 1.001, p < 0.001). Males had lower COVID-19 risks while diabetes was not linked. AIAN patients had almost twice the risk of White patients (HR = 1.948, p < 0.001) while Hispanics also had higher risk than White (HR = 1.328, p < 0.001). Vaccination greatly reduced the risk (HR < 0.001, p < 0.001).
Conclusion
Joint modeling indicated a weak link of ferritin with COVID-19 infection, however, dynamic predictions via joint models did not strongly support early detection via ferritin alone. AIAN and Hispanic patients had higher infection risk than White patients, while vaccination nearly eliminated the risk.
Dynamic prediction for a typical patient. Left: Longitudinal. Right: Survival.
Funding
- Other NIH Support