Abstract: TH-PO1015
Integrated Day-30 Electronic Health Record Trajectories Identify Clinically Distinct Recovery Patterns After Kidney Transplantation
Session Information
- Transplantation: Clinical - Outcomes, Malignancy, and Pathology
October 22, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Transplantation
- 2002 Transplantation: Clinical
Authors
- Dhamdhere, Rohan Narendra, Emory University, Atlanta, Georgia, United States
- Tokuyama, Naoto, Emory University, Atlanta, Georgia, United States
- Karadkhele, Geeta, Emory University, Atlanta, Georgia, United States
- Abe, Yohei, Kagawa Daigaku, Takamatsu, Kagawa Prefecture, Japan
- Krishnamoorthy, Anusha, Emory University, Atlanta, Georgia, United States
- Larsen, Christian, Emory University School of Medicine, Atlanta, Georgia, United States
- Madabhushi, Anant, Emory University, Atlanta, Georgia, United States
Group or Team Name
- Madabhushi Lab
Background
Early recovery after kidney transplantation is often assessed using isolated estimated glomerular filtration rate (eGFR) or creatinine values, which may miss recovery pace, direction, and concurrent changes in proteinuria or tacrolimus exposure. The ability of integrated day 0–30 electronic health record (EHR) trajectories to define clinically meaningful recovery phenotypes associated with subsequent inflammatory allograft injury remains unclear.
Methods
We performed a day-30 landmark analysis using day 0–30 EHR data. Unsupervised k-means clustering identified recovery phenotypes from 20 features spanning kidney function, proteinuria, tacrolimus, and transplant factors; serial signals were summarized as slopes, landmark-nearest values, deltas, extrema, and target-range percentages, excluding biopsy/rejection variables. The outcome was biopsy-detected rejection/inflammation (borderline or grade≥1 rejection) from day 30 to 1 year. Cox models were adjusted for age, donor type, cold ischemic time, human leukocyte antigen (HLA) mismatch, and delayed graft function (DGF). Day-0 sensitivity analysis tested whether P3 was P1- or P2-like.
Results
Among 663 recipients, three phenotypes emerged: P1 stable recovery (n=299), P2 hypofunctioning graft (n=183), and P3 tacrolimus-overexposed recovery (n=181), with distinct clustering and eGFR/tacrolimus trajectories (Figure A–C). Over median 276-day follow-up post-landmark (P1:306, P2:223, P3:278 days), P2 had higher risk than P1 (adjusted hazard ratio [aHR] 1.93, 95%CI=1.14–3.29; p=0.015), persisting after DGF/transplant-factor adjustment (C-index=0.65; multivariate p=0.002; Figure D–E). P3 showed no increased risk versus P1; 120/181 P3 recipients were P1-like at day 0, supporting P1-adjacent pattern.
Conclusion
Day-30 EHR phenotyping identified a hypofunctioning graft phenotype linked to higher rejection/inflammation, independent of DGF and transplant factors, supporting trajectory-based first-year surveillance beyond single-timepoint kidney-function assessment.
Acknowledgment
N.T. is supported by the Winship Cancer Institute Postdoctoral Award.