Abstract: FR-PO069
Risk Prediction for AKI in Patients Hospitalized With COVID-19: Withstanding Variants Over Time
Session Information
- AKI: Epidemiology, Risk Factors, Prevention
November 04, 2022 | Location: Exhibit Hall, Orange County Convention Center‚ West Building
Abstract Time: 10:00 AM - 12:00 PM
Category: Acute Kidney Injury
- 101 AKI: Epidemiology‚ Risk Factors‚ and Prevention
Authors
- McAdams, Meredith C., The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Xu, Pin, The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Li, Michael M., The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Ostrosky-Frid, Mauricio, The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Willett, Duwayne L., The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Lehmann, Christoph Ulrich, The University of Texas Southwestern Medical Center, Dallas, Texas, United States
- Hedayati, Susan, The University of Texas Southwestern Medical Center, Dallas, Texas, United States
Background
Acute kidney injury (AKI) is common in patients hospitalized with COVID-19, predictive models for AKI are lacking. We aimed to develop the best predictive model for AKI and assess performance over time.
Methods
Patients with positive SARS CoV-2 PCR hospitalized between 3/1/2020 to 1/14/2022 at 19 Texas hospitals were included. Those with AKI present on admission were excluded. Comorbidities, demographics, baseline laboratory data, and inflammatory biomarkers were obtained from the EHR and used to build nested models for AKI in an inception cohort. Models were validated in four out-of-time cohorts. Model discrimination and calibration measures were compared to assess performance.
Results
Of 13,468 patients, 5,676 were in the Inception Cohort and 7,792 in subsequent validation cohorts grouped based on predominance of COVID variants, with cohorts 1 and 3 containing a mix of variants, cohort 2 corresponding to Delta predominance, and cohort 4 to Omicron. Prevalence of AKI was 13.7% in inception and 12.6%, 12.4%, 13.3%, and 14.4% in the validation cohorts. Proportion of AKI stages 2 or 3 vs. 1 was lower in the Omicron-dominant cohort 4 compared to the inception cohort (28/139 vs. 257/776, P=0.008), but was no different for cohorts 1-3. The final model containing demographics, comorbidities and baseline WBC, hemoglobin, hsCRP, ferritin, and D-dimer, had an AUC=0.781 (95% CI, 0.763, 0.799). Compared to the inception cohort, discrimination by AUC (validation 1: 0.785 [0.760, 0.810], P=0.14, validation 2: 0.754 [0.716, 0.795], P=0.14, validation 3: 0.778 [0.751, 0.806], P=0.14, and validation 4: 0.743 [0.695, 0.789], P=0.14) and calibration by ECI (validation 1: 0.116 [0.041, 0.281], P=1.0, validation 2: 0.081 [0.045, 0.295], P=0.64, validation 3: 0.055 [0.030, 0.162], P=1.0, and validation 4: 0.120 [0.043, 0.472], P=0.50) showed stable performance over time.
Conclusion
Using demographics, comorbidities, admission laboratory values, and inflammatory biomarkers, we developed and externally validated a model to accurately predict AKI in hospitalized patients with COVID-19. A lower proportion of patients hospitalized during the Omicron-dominant period of the pandemic experienced severe AKI, but our predictive model withstood changes in practice patterns and virus variants.
Funding
- NIDDK Support