A56-15 Development and Validation of a Novel Risk Assessment Score - Cardiac Arrest Mortality Index (CAMI) to Predict In-hospital Mortality Following Cardiac Arrest Using the National Inpatient Sample (2016-2022)

Abstract Rationale Cardiac arrest remains a major contributor to in-hospital mortality in the United States. Existing comorbidity indices such as the Charlson Comorbidity Index (CCI) and Elixhauser Comorbidity Index (ECI) are not optimized to predict mortality in this high-acuity population. A risk score that integrates comorbidities, patient demographics, and in-hospital clinical events could better stratify patients at highest risk, support prognostic decision-making and to help intensivists guide goals of care. Objectives To develop and validate a comprehensive risk scoring system for predicting in-hospital mortality among patients hospitalized with cardiac arrest and compare its discriminatory performance with CCI and ECI using DeLong’s test for correlated ROC curves. Methods We performed a retrospective cohort study using the National Inpatient Sample (NIS) from 2016-2022. Adult hospitalizations containing ICD-10 codes for cardiac arrest were included. Candidate predictors were selected a priori and included demographics (age, gender), comorbidities, and key in-hospital events (e.g., shock, mechanical ventilation, acute organ dysfunction). A multivariable logistic regression model was developed to estimate the probability of in-hospital mortality, and model coefficients were converted into a point-based scoring system. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC). AUCs were compared with CCI and ECI using DeLong’s method. The data utilized weighed estimates for national-level representation of the sample. Results The study included 354,569 adult hospitalizations which are representative for approximately 1.77 million hospital admissions with cardiac arrest. About 62.4% of these patients died during hospital stay. Mean age was 65.47 (15.85) years. The novel risk score demonstrated strong discriminatory ability in the cohort (AUC 0.659) with a sensitivity of 62.65% and specificity of 59.03%, significantly outperforming both CCI (AUC: 0.517) and ECI (AUC: 0.510). DeLong’s test confirmed statistically significant differences between the new model and both comparator indices (p 0.001). The score showed excellent calibration across deciles of risk, with observed-to-predicted mortality demonstrating close agreement. Conclusions We developed and validated a robust, clinically interpretable risk scoring system that outperforms existing comorbidity indices in predicting in-hospital mortality after cardiac arrest. This tool may improve clinical risk stratification, guide resource allocation, and support in-hospital vigilance team structure. Limitations Using ICD-10 codes does not classify the severity of each comorbidity listed in the criteria used for the scoring model prediction. Further research is warranted for generation of better classifying scores to help the patients make healthcare decisions prior to cardiac arrest. This abstract is funded by: None