Derivation and validation of the ED-SAS score for early prediction of mortality in acute pancreatitis

Document Type

Conference Proceeding

Publication Date


Publication Title

Academic emergency medicine


Background and Objectives: Existing scoring systems to predict mortality in acute pancreatitis largely account for variables present within the first 24 or more hours of hospital admission. The objective of this study was to derive and validate a simple predictive score using variables readily available in the emergency department (ED) to predict mortality in acute pancreatitis.

Methods: We performed a retrospective observational study across 2 independent health systems, one used for the derivation cohort and one for the validation cohort. The health systems encompassed 10 community and academic EDs. Adult patients were eligible who presented to the ED, required hospital admission, and were confirmed to have a final diagnosis of acute pancreatitis. We excluded patients with chronic pancreatitis or hepatic failure and those with a recurrent episode of pancreatitis. We standardized and validated data collection instruments across sites. The analysis consisted of multivariable logistic regression, and candidate variables in the derivation cohort of a prediction score were selected a priori based on reliable availability in the ED and the existing literature. We applied the score to the validation cohort and report odds ratios associated with the primary outcome of 30-day mortality.

Results: The derivation cohort included 599 patients (mean age 53 years, 47% female), and the validation cohort 2,011 patients (mean age 56 years, 51% female). Thirty-day mortality in the derivation cohort was 4.2% and 3.9% in the validation cohort. From the derivation cohort, 3 variables were retained in a prediction score: ≥2 SIRS criteria, age >60 years, and SpO2 <96%. Summing the presence (1 point) or absence (0 points) of each variable yielded an ED-SAS score (SIRS, age, SpO2), ranging from 0 to 3. In the derivation cohort, mortality based on scores from 0 to 3 was 0%, 1.3%, 10.4%, and 15.4% respectively. Applied to the validation cohort, the OR for death increased substantially for each additional ED-SAS point: 4.4 (95% CI 1.8 - 10.8) for 1 point, 12.0 (95% CI 4.9 - 29.4) for 2 points, and 41.7 (95% CI 15.8 - 110.0) for 3 points. Model discrimination in the derivation and validation cohort for predicting mortality was good (C-statistic 0.80 and 0.77 respectively).

Conclusion: An ED-SAS score that incorporates age, SIRS, and ED SpO2 measurement provides a rapid method for predicting mortality in acute pancreatitis earlier than existing score systems.



First Page


This document is currently not available here.