Derivation and Validation of Clinical Subphenotypes of AKI with Prognostic Implications in Critically Ill Children

Background: AKI is a heterogeneous syndrome. Identification of distinct clinical subphenotypes (SPs) may allow more precise therapy and improve care. We aim to derive and validate AKI SPs and determine whether the SPs were relevant with respect to outcomes in critically ill children. Methods: This is a secondary analysis of our prospective multicenter cohort study, including AKI children met KDIGO criteria or predicted by our prediction model. Medical characteristics easily obtained within the first 24h after PICU admission were employed to select features using the SHapley Additive exPlanation. Latent profile analysis (LPA) was used to identify distinct SPs in derivation cohort and validated in external cohort. Results: Three SPs were derived among 332 children and validated in 242 using LPA applied to 23 features. Children in SP1 (74.9%) exhibited the least severe illness and organ injury and responded well to milrinone; those in SP2 (19.5%) suffered severe inflammatory response and homeostatic imbalance; those in SP3 (5.6%) suffered more severe hepatic and cardiovascular dysfunction and the highest risk of mortality. Conclusion: Three SPs were identified in critically ill children with AKI that correlated with outcomes. Further research is needed to determine the utility of the SPs in clinical care and trial design.Comparison of variables contributing to SPs. Chord diagrams showing abnormal variables by SPsComparison of the incidence rate. Kaplan-Meier curves