(1498) Development of a Predictive Model for Persistent Acute Kidney Injury in Sepsis Acute Kidney Injury

Introduction: Several drugs are being developed to treat Acute Kidney Injury secondary to Sepsis (S-AKI), aiming to prevent progression to Chronic Kidney Disease (CKD). However, only 30 to 40% of patients with S-AKI will progress to persistent renal failure. Therefore, identifying the subgroup that will benefit from early treatment is crucial. This study aimed to develop a predictive model, used on the day of S-AKI diagnosis, to predict persistent renal failure.

Methods: Retrospective cohort study was conducted using data from the Hospital Italiano de Buenos Aires, Argentina. Adult patients with AKI (KDIGO criteria) due to sepsis (Sepsis-3) were included. The primary outcome was Major Adverse Kidney Events (MAKE), encompassing the need for new dialysis or a ≥25% Glomerular Filtration Rate (GFR) decrease within 60 days after S-AKI diagnosis. Predictors included demographics, comorbidities, laboratory, fluid balances, and critical support requirements. Due to death being a competing event with MAKE, the model was developed using the Fine and Gray competing risk regression. External validation was performed with data from the Hospital Agustin Roca, Argentina.

Results: The model was developed using 1436 patients with S-AKI (1005 generation; 431 internal validation). Final model predictors included GFR day of S-AKI diagnosis: GFR 60-90; sHR 2.11 (95%CI 0.93-4.78); GFR 30-60 sHR 3.02 (95%CI 1.39-6.59); GFR 15-30 sHR 3.90 (95%CI 1.71-8.90); GFR < 15 sHR 5.08 (95%CI 2.17-11.89); history of CKD sHR 1.84 (95%CI 1.21-2.78); no mechanical ventilation (MV) on the day of S-AKI sHR 3.6 (95%CI 2.55-5.07); and male sex sHR 1.96 (95% CI 1.32-2.92). The model's discrimination was 0,71 (IC95% 0.67-0.74) in the generation cohort and 0.71 ( 95%IC 0.67-0.78) in the external validation.
The model was transformed into a scoring system adding 1.2, 1.7, 2.2, and 2.8 points for each GFR grade; CKD history 1 point; no MV on the day of S-AKI 2 points; and male sex 1.1 points. The score was dichotomized into low or high risk (>2). With this cutoff, the Number Needed to Treat was 31.7 for the low-risk and 5.5 for the high-risk group in the external validation cohort.

Conclusions: The developed model accurately predicts MAKE in S-AKI patients, aiding in the identification of high-risk patients for targeted interventions.