Abstract: FR-PO1136
Decision Tree Analysis to Study the Short- and Long-Term Renal Outcomes in Patients Hospitalized During the First Year of the COVID-19 Pandemic
Session Information
- COVID-19 - II
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Coronavirus (COVID-19)
- 000 Coronavirus (COVID-19)
Authors
- Wang, Weihao, Stony Brook University, Stony Brook, New York, United States
- Hajagos, Janos G., Stony Brook University, Stony Brook, New York, United States
- Fochtmann, Laura J., Stony Brook University, Stony Brook, New York, United States
- Zhu, Wei, Stony Brook University, Stony Brook, New York, United States
- Koraishy, Farrukh M., Stony Brook University, Stony Brook, New York, United States
Background
COVID-19 has been associated with AKI in hospitalized patients and GFR decline after discharge. However, analysis of risk factors associated with renal outcomes has been limited to traditional machine learning models. Additionally, length of hospitalization (LOH) or days on the ventilator (DOV) have not been typically used for risk prediction.
Methods
In this retrospective single center study, we applied a Classification and Regression Trees algorithm (CART decision trees) to study renal outcomes in patients hospitalized in the United States during the COVID-19 pandemic in 2020 (pre-vaccination era). Outcomes: Moderate/severe AKI (Stage 2 &3) in the hospital and eGFR decline during a 24-month follow-up (until February 2023. To address the effect of confounding demographics in ICU vs. non-ICU and COVID-19 positive vs. negative patients, propensity score matching (PSM) was applied. Influential variables were selected based on feature importance in Random Forest plots.
Results
In our initial cohort of adult 6.933 patients (without baseline end-stage kidney disease), 10.4 % (719) were diagnosed with AKI-2/3. CART decision tree analysis on the primary cohort and the ICU and COVID-19 status PSM cohorts showed DOV, sepsis, age and vasopressor use as key factors for AKI-2/3 risk categorization. In the follow-up study hospital survivors (n = 1,747), those originally admitted to the ICU and those with COVID-19 had a greater mean GFR decline compared to control groups. CART analysis (Figure 1) showed LOH, age, AKI-2/3, body mass index, vasopressor use, and baseline CKD as the main features that risk categorized patients for rate of GFR decline.
Conclusion
To our knowledge, this is one of the first studies to report renal outcomes in patients with and without COVID-19 using CART decision tree analysis. Besides traditional risk factors, days in the hospital and on the ventilator were noted as key features that stratified patients’ risk for renal outcomes.