Background: While the majority of acute pancreatitis (AP) patients fulfill diagnostic criteria with characteristic abdominal pain and serum lipase ≥3x normal at presentation, early imaging is often utilized for confirmation. A prior diagnostic model and corresponding point-based score was developed to identify AP risk in emergency department (ED) patients using non-imaging predictors.

Aim: To prospectively validate the AP diagnostic score at two large academic centers.

Methods: Consecutive adult patients presenting to the ED at each center between January 2020 – March 2021 with serum lipase ≥3x normal were prospectively identified. Those transferred from outside institutions, or with established intra-abdominal malignancy, acute trauma, or altered mentation were excluded. Patients were scored using 8 non-imaging parameters (Table 1) by a research coordinator blinded to the final diagnosis. A diagnosis of AP was established by expert review of full hospitalization records, including relevant imaging. Discriminatory accuracy of various score cut-offs was assessed.

Results: Prospective scores in 349 patients demonstrated excellent AP prediction, with an area under the curve (AUC) of 0.91. A score ≥6 points achieved highest AP diagnostic accuracy (F-score 82.0), which corresponded to sensitivity 81.5%, specificity 85.9%, positive predictive value 82.5%, and negative predictive value 85.1%. Only one serious non-AP diagnosis scored ≥6 (8 points): a 56 year-old patient who was hospitalized for 2 days for recurrent small bowel obstruction and did not undergo imaging due to repetitive symptoms. Early CT or MR imaging (within 24 hours of ED presentation) was performed more often in those predicted to have AP (75.5% when scoring ≥6 points vs 57.2% when <6 points, p=0.0004). When utilized, early imaging revealed an alternative diagnosis in only 6.8%, 1.0%, and 1.4% of patients who scored ≥6, ≥7, or ≥8 points, respectively.

Conclusions: the AP diagnostic score demonstrated excellent discriminatory accuracy when applied prospectively to a multicenter cohort. Its application may be used to avoid unnecessary confirmatory imaging.

Peter Banks

Appearances