HPB SO46 - Risk Stratification for Acute Pancreatitis: A Retrospective Comparative Analysis between Glasgow-Imrie and BISAP Scoring Systems within a District General Hospital

Abstract

Abstract Background Risk stratification for acute pancreatitis (AP) is an important tool in clinical decision making. The BISAP score has been validated to be a comparable risk prognostication tool to current models[1, 2]. Current trust practice is the Glasgow-Imrie score, which involves more invasive and non-routine investigations, resulting in an increased cost and labour burden. Additionally, a previous local audit demonstrated poor compliance with only 55% of admissions having a documented score, limiting utility and prognostication. By demonstrating non-inferiority, we aim to implement BISAP as the standard of practice with earlier severity assessment, improved patient experience and minimising unnecessary investigations. Method A retrospective analysis of 153 admissions with acute pancreatitis admitted to a single-site district general hospital over a 12-month period was performed, reviewing electronic patient records for documentation, observations and investigations. Both Glasgow-Imrie and BISAP scores were calculated within 48 and 24 hours of admission retrospectively and respectively. Descriptive statistics was used to summarise data and compare trends between both scores and the severity of acute pancreatitis, ICU admissions and mortality. Logistic regression was utilised to model the relationship between both systems and binary outcomes with area under the curve (AUC) and non-inferiority tests derived to determine efficacy. Results The cohort had a mean age of 57 ± 18.53 (52:48 male:female) with an average LOS of 7.25 ± 10.12 days. 6.54% had severe AP, 3.92% required ICU admisison and 3.26% mortality rate. The AUC value for both models was 0.868 in predicting severity (p<0.001) with Glasgow-Imrie performing better in predicting ICU admission (AUC 0.954 > 0.855, p<0.01) and BISAP outperforming Glasgow-Imrie in predicting mortality (AUC 0.903 > 0.867, p<0.01). Bootstrapping resampling demonstrated no significant difference between models in predicting severity (δAUC 0.0007, 95%CI -0.01181, 0.1167), ICU admission (δAUC 0.1010, 95%CI -0.0113, 0.2140) and mortality (δAUC -0.0366, 95%CI -0.0788, 0.0045). Conclusion Both the Glasgow-Imrie and BISAP scoring systems are accurate in predicting the severity of pancreatitis, ICU admission and mortality. The BISAP score is similar in prognostic accuracy to the Glasgow-Imrie score and is advantageous in being simpler to obtain in clinical practice with a higher rate of completeness. We recommend the BISAP score as a non-inferior, pragmatic risk stratification tool in acute pancreatitis.