LAUSR

BACKGROUND AND AIM: Multisystem inflammatory syndrome in children (MIS-C) associated with a recent SARS-CoV2 infection is a rare complication of the infection that often requires admission to the pediatric intensive care unit (PICU). Currently, there are no prediction models to assess severity in MIS-C patients as tools of risk assessment that may prevent decompensation out of the ICU setting. METHOD(S): We performed a retrospective chart review from 5/2020 to 6/2021 to collect presenting symptoms, lab and imaging results, and treatment regimens for all patients presenting to our institution with MIS-C. We imputed missing values using KNN Imputation and used a random forest algorithm to identify three lab test features for our model: ferritin, procalcitonin, and platelets. The echocardiogram ejection fraction (EF) was included for clinical relevance. We modeled these variables using multiple logistic regression and evaluated it through the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. RESULT(S): 78 patients were analyzed, of which 50 (64%) were admitted to the ICU. Platelet concentration was significant (OR 0.9930, CI 0.986-0.998, p-value = 0.021). However, procalcitonin, ferritin, and EF showed no significance. The AUC of our ROC curve was significant at 0.8043 (p-value < 0.0001). Fig1 shows the mean and standard deviation of platelets and EF to ICU admission and AUROC of the model. The negative predictive value (NPV) 75.0%, positive predictive value (PPV) 77.6%. CONCLUSION(S): Platelet concentration on admission was a significant predictor of ICU admission. Larger studies are required to develop predictive PICU admission models. (Figure Presented).