Development and validation of the discharge severity index for post-emergency department hospital readmissions

Hospital readmissions often result from a combination of factors, including inadequate follow-up care, poor discharge planning, patient non-adherence, and social determinants of health (SDOH) that impact access to healthcare and follow-up resources, many of which are beyond provider control. Enhanced post-discharge strategies, including risk stratification, are essential. This study aims to develop and validate the Discharge Severity Index (DSI) to predict readmission risk and optimize resource allocation for effective follow-up care.

This single-center retrospective study analyzed ED visits from the Medical Information Mart for Intensive Care IV, dividing the data into derivation (75 %) and validation (25 %) cohorts. Univariate analyses were conducted on factors commonly available for most discharges, including patient age, the latest vital signs recorded, medical complexity, and ED length of stay (LOS). Multiple logistic regression (MLR) was employed to identify independent risk factors of patients revisiting the ED within a week and being subsequently admitted to the hospital. Adjusted parameter estimates from the MLR were used to develop a predictive model.

Among 229,920 patients discharged from the ED, 1.92 % were readmitted. The analysis identified seven variables correlated with this outcome, with six significant risk factors pinpointed through MLR: age above 65, heart rate over 100, and oxygen saturation below 96 % (assigned 1 point each), along with having more than five active medications administered during the hospital stay or a LOS exceeding 3 h (assigned 2 points each). Using these scores, we categorized patients into five DSI groups, reflecting escalating readmission risk from DSI 5 (lowest risk) to DSI 1 (highest risk): DSI 5 (0; OR: 1.0), DSI 4 (1-2; OR: 3.49), DSI 3 (3-4; OR: 8.44), DSI 2 (5-6; OR: 11.65), and DSI 1 (>6; OR: 14.63). The seven-day readmission rates were comparable between the development and validation cohorts. For instance, for DSI 1, the rates were 5.16 % in the development cohort and 4.67 % in the validation cohort. For DSI 2, the rates were 4.16 % and 4.04 %, respectively.

This study seeks to develop and validate the DSI, proposing its effectiveness as a tool for healthcare providers to categorize patients by their risk of post-discharge admission from the ED. The utilization of this tool has the potential to lead to a more informed allocation of resources after discharge.