Predicting Hospitalizations and Returns to the Emergency Department 30-Days Post Emergency Department Discharge Among Patients with Acute Bacterial Skin and Skin Structure Infections
MetadataShow full item record
BACKGROUND: Despite Acute Bacterial Skin and Skin Structure Infections (ABSSSI) being among the most common infections, and the requirement to track hospital quality measures associated with readmissions, there is no risk stratification tool to guide clinical decision-making processes in the Emergency Department (ED) for deciding when to hospitalize a patient with ABSSSI. The primary objective of this study was to develop a regression model to identify patient, treatment, and facility-level factors that would predict a composite outcome of either hospitalization or return to the ED for any reason within 30-days post initial ED visit discharge for a patient with ABSSSI. The secondary objective was to evaluate the role of an admission at the initial ED visit (Initial Episode of Care, IEC) discharge on the probability of experiencing the composite outcome. METHODS: A retrospective cohort database analysis was conducted using data collected from a retrospective manual medical chart review across 41 ED sites in the United States. The outcome of interest was a composite outcome of the occurrence of an unscheduled all-cause hospitalization or return to the ED within 30-days post IEC discharge for patients with ABSSSI. Predictors of interest included patient, treatment, and facility-level factors collected in the ED for patients with ABSSSI. Backward stepwise regression was used to select variables for inclusion in the multivariable prediction model and model discrimination and calibration were assessed. As a secondary aim, the role of the admission at IEC on the composite outcome was evaluated using recycled predictions. RESULTS: Of a total of 937 patients with ABSSSI included in the model, 205 (21.9%) experienced an unscheduled all-cause hospitalization or return to the ED 30-days post IEC discharge. The stepwise multivariable generalized linear model identified admission decision at IEC, Charlson Comorbidity Index (CCI) score, being immunocompromised, having a hospitalization within 90 days prior to IEC visit, and facility bed capacity as being best predictive of the composite outcome (Akaike Information Criteria [AIC] score = 973.39). Model discrimination was poor (AUC = 0.61) and calibration was good (χ2 Hosmer-Lemeshow = 12.26 [p=0.14]). There was no significant difference in the composite outcome between those that were admitted versus not at IEC, with those admitted experiencing a 7% lower predicted risk of experiencing the composite outcome, when compared to those not admitted at IEC (95% confidence interval [CI]: 0.68 to 1.29) in the study population. Using the method of recycled predictions, the difference in predictive probability of a hospitalization or ED return 30-days post IEC discharge for those not admitted at IEC vs those admitted at IEC was 5.6%. CONCLUSION: Our findings provide an estimate of the probability of admission or ED revisit within 30 days of IEC among patients with ABSSSI; and a preliminary prediction model for the same. Future work will include identification of a model with improved model discrimination and conducting a model validation exercise. Our work is a first step in the development of a risk stratification tools to ensure clinician efforts target ABSSSI patients with a need for admission at IEC, and one that identifies characteristics that predict hospitalizations or ED revisits within 30 days of IEC.