Abstract: SA-OR10
Prediction of Postdischarge Kidney Disease Progression Among Patients with Hospitalized AKI: The ASSESS-AKI Study
Session Information
- AKI Research: Seeking New Paths to Progress
November 04, 2023 | Location: Room 118, Pennsylvania Convention Center
Abstract Time: 05:51 PM - 06:00 PM
Category: Acute Kidney Injury
- 102 AKI: Clinical, Outcomes, and Trials
Authors
- Menez, Steven, The Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
- Kerr, Kathleen F., University of Washington School of Medicine, Seattle, Washington, United States
- Cheng, Si, University of Washington School of Medicine, Seattle, Washington, United States
- Hu, David, The Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
- Thiessen Philbrook, Heather, The Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
- Moledina, Dennis G., Yale School of Medicine, New Haven, Connecticut, United States
- Mansour, Sherry, Yale School of Medicine, New Haven, Connecticut, United States
- Go, Alan S., University of California San Francisco, San Francisco, California, United States
- Ikizler, Talat Alp, Vanderbilt University School of Medicine, Nashville, Tennessee, United States
- Kaufman, James S., New York University Grossman School of Medicine, New York, New York, United States
- Kimmel, Paul L., National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, Maryland, United States
- Himmelfarb, Jonathan, University of Washington School of Medicine, Seattle, Washington, United States
- Coca, Steven G., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Parikh, Chirag R., The Johns Hopkins University School of Medicine, Baltimore, Maryland, United States
Background
Acute kidney injury (AKI) occurs frequently during hospitalization, but only a fraction of patients progress to chronic kidney disease (CKD) after discharge. Biomarkers of kidney injury, inflammation, and repair have been shown to be informative of long-term kidney disease risk.
Methods
We evaluated data from 723 hospitalized patients with AKI and post-discharge follow-up in the prospective Assessment, Serial Evaluation, and Subsequent Sequelae of AKI (ASSESS-AKI) Study. We investigated 75 candidate predictors, including 11 urinary and 28 plasma biomarkers measured at 3-month post-discharge follow-up. We employed both random forests and least absolute shrinkage and selection operator (LASSO) regression to predict major adverse kidney events (MAKE): CKD incidence, CKD progression, or development of end-stage kidney disease. The data was split into training (80%) and test (20%) datasets. We used multiple imputation to handle missing data and independent test data for unbiased estimates of model performance.
Results
A total of 235 patients developed MAKE over 3 years of follow-up. While a prediction model containing 9 key clinical variables yielded an area under the receiver-operating characteristic curve (AUC) of 0.75 (0.63-0.87), random forest and LASSO models using all 75 variables yielded AUC values of 0.81 (0.71-0.92) and 0.80 (0.70-0.91), respectively (Table 1). The top 5 predictive biomarkers based on random forest modeling were sTNFR1, sTNFR2, NT-proBNP, FGF-23, and Ang2, all measured in the plasma and yielded an AUC of 0.75 (0.63-0.88). A combination model leveraging both clinical variables and the top 5 biomarkers demonstrated an AUC of 0.77 (0.65-0.89).
Conclusion
A parsimonious prediction model using a combination of key clinical variables and top biomarkers offers moderately strong discrimination in identifying patients with hospitalized AKI at highest risk of progression of kidney disease long-term.
Funding
- NIDDK Support