The Early Prediction of Mortality in Acute Pancreatitis: A Large Population-based Study

Bechien U Wu ^1*, Richard S. Johannes ¹, Xiaowu Sun ², Ying Tabak ², Darwin L. Conwell ¹ and Peter A. Banks ¹

¹ Brigham and Women's Hospital, United States
² Cardinal Health, United States

^* To whom correspondence should be addressed. E-mail: buwu{at}partners.org.

Accepted 13 May 2008

Abstract

Background: Identification of patients at risk for mortality early in the course of acute pancreatitis (AP) is an important step to improving outcome.

Methods: Using Classification and Regression Tree (CART) analysis, we developed a clinical scoring system for prediction of in-hospital mortality in AP. We derived the scoring system on data collected from 17,992 cases of AP from 212 hospitals in 2000-2001. We validated the new scoring system on data collected from 18,256 AP cases from 177 hospitals in 2004-2005. Accuracy of the scoring system for prediction of mortality was measured by the area-under the receiver-operating curve (AUC). We further validated the performance of the new scoring system by comparing its predictive accuracy to APACHE II.

Results: CART analysis identified five variables for prediction of in-hospital mortality. One point is assigned for the presence of each of the following during the first 24 hours: BUN>25 mg/dL, Impaired mental status, SIRS, age>60 years or the presence of a Pleural effusion (BISAP). Mortality ranged from >20% in the highest risk group to <1% in the lowest risk group. In the validation cohort, the BISAP AUC was 0.82 [95% CI 0.79, 0.84] vs. APACHE II AUC of 0.83 [95% CI 0.80, 0.85].

Conclusions: We have derived and validated a new mortality-based prognostic scoring system for use in acute pancreatitis. The BISAP is a simple and accurate method for the early identification of patients at increased risk for in-hospital mortality.