Kidney Week

Abstract: FR-PO0410

Clinical Phenotyping of Drug-Induced Kidney Injury Using Large Electronic Health Data

Session Information

• AKI: Biomarkers, Diagnostics, and Risk Prediction
  October 23, 2026 | Location: Exhibit Hall A, Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Acute Kidney Injury

• 102 AKI: Clinical, Outcomes, and Trials

Authors

• Awdishu, Linda, University of California San Diego, La Jolla, California, United States

Linda Awdishu, PharmD, MA, FASN

• Zhuang, Yonghua L., University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States

Yonghua L. Zhuang, MD, PhD

• Brasher, Maizy S., University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States

Maizy S. Brasher, MA

• Yousif, Zaid, University of California San Diego, La Jolla, California, United States

Zaid Yousif, PharmD

• Cole, Joanne B., University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States

Joanne B. Cole, PhD

• Joy, Melanie S., University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States

Melanie S. Joy, PharmD, PhD, FASN

Background

Medications cause acute kidney injury (AKI) in ~26% of adults.¹ Using a published DI-AKI clinical risk model from DIRECT (PMID 29142931, 38025217) we modified and applied a phenotyping tool and adjudication dashboard within University of Colorado Health Data Compass to evaluate patients exposed to high risk drugs.

Methods

Electronic health record study of adults (1/1/2013–7/1/2023). DIRECT definitions were modified to include Stage 1 AKI: reference SCr within 90 days pre-drug start; qualifying SCr (≥1.5x reference) within 14 days post-start for vancomycin, tacrolimus, ibuprofen, and ketorolac. Cases required SCr at admission, drug start, AKI day, and discharge with a valid temporal sequence. Variables included demographics, SCr, KDIGO staging, AKI risk factors, drug dosing/concentrations, dialysis, and recovery.

Results

Interim results from 2225 cases: median age 62 years (IQR 47–71), 51.4% female and median BMI 27.62. SCr rose 1.75 x reference; median reference SCr 0.78 mg/dL (IQR 0.60–1.00) and AKI Day SCr 1.48 mg/dL (IQR 1.10–2.09). KDIGO staging: 68.4% Stage 1, 21.5% Stage 2, and 10.1% Stage 3. Drugs: vancomycin 59.3%, NSAIDs 49.7%, tacrolimus 6.9% (not mutually exclusive). Comorbidities were hypertension 60.5%, diabetes 31.9%, heart failure 22.3%, CKD 20.4%. Admission AKI risk factors: anemia 77.6%, hypoalbuminemia 83.4%; in-hospital: contrast exposure 37.4%, hypotension 22.2%, sepsis 9.5% and major surgery 3.2%. Median drug-to-AKI onset 2 days (IQR 1-3), 92.5% within 7 days. Discharge outcomes: complete renal recovery 37.7%, partial recovery 26.8%, no recovery 35.6%.

Conclusion

A DI-AKI phenotyping model was successfully modified and externally validated to identify temporally consistent cases. Future work targets drug-exposed controls and AI-based risk scoring.