Abstract: FR-OR05
Machine Learning-Guided Novel Subphenotypes of Sepsis-Associated Persistent AKI
Session Information
- AKI Prediction and Assessment: Traditional and Novel Tools
November 03, 2023 | Location: Room 118, Pennsylvania Convention Center
Abstract Time: 05:06 PM - 05:15 PM
Category: Acute Kidney Injury
- 101 AKI: Epidemiology, Risk Factors, and Prevention
Authors
- Oh, Wonsuk, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Kittrell, Hannah, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Kohli-Seth, Roopa D., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Nadkarni, Girish N., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Sakhuja, Ankit, West Virginia University, Morgantown, West Virginia, United States
Background
Sepsis associated acute kidney injury (SA-AKI) is common and is associated with high mortality. AKI lasting 48 hours or longer, known as persistent AKI (pAKI), has much worse outcomes. SA-AKI is a heterogenous disease, however, it is unknown whether such heterogeneity exists in SA-pAKI. We aimed to identify subphenotypes (SP) of SA-pAKI using routinely collected data in electronic medical records.
Methods
We conducted a retrospective study using MIMIC IV database. We defined AKI and pAKI using both creatinine and urine output based KDIGO criteria. We identified adult patients (≥18y) with sepsis who developed SA-AKI within 48h and SA-pAKI within 96h of ICU admission. We used available features for demographics, comorbidities, SOFA score, vital signs, labs, fluid balance & vasopressors to identify SPs. We used factor analysis of mixed data for dimensionality reduction followed by k-means clustering to identify SPs. Outcomes were 30-day in-hospital mortality and 30-day AKI recovery while adjusting for competing risk of mortality.
Results
Among 6,681 patients with SA-pAKI, we identified 4 distinct SPs. Each SP demonstrated distinct characteristics and outcomes (Fig 1a & b). SP1 (n=1,137) included patients with severe AKI, low systolic BP, high INR, and WBC counts. It had highest mortality(47%) and low AKI recovery(37%). SP 2 (n=1,231) had moderate to severe AKI but low vasopressor requirements. It had low mortality (19%) but also low rates of AKI recovery (32%). SP3 (n=1315) included patients with high comorbidity burden but low disease acuity. Their mortality was between that of first two SPs and had high rates of AKI recovery (54%). SP4 (n=1,678) included patients with mild to moderate AKI and low disease acuity. They had highest rates of AKI recovery (56%) and lowest mortality (13%).
Conclusion
We identified 4 distinct SPs of SA-pAKI with differing patient characteristics and outcomes. Early recognition of these SPs will allow for personalized management strategies.
Funding
- NIDDK Support