1091. Diagnostic Performance of an Algorithm Based on Clinical and Sonographic Parameters for the Diagnosis of Acute Cholecystitis
Authors * Denotes Presenting Author
  1. Andrew Sill *; Mayo Clinic Arizona
  2. Maitray Patel; Mayo Clinic Arizona
  3. Nirvikar Dahiya; Mayo Clinic Arizona
  4. Frederick Chen; Mayo Clinic Arizona
  5. William Eversman; Mayo Clinic Arizona
  6. J. Scott Kriegshauser; Mayo Clinic Arizona
  7. Scott Young; Mayo Clinic Arizona
Identify a diagnostic algorithm using both clinical and sonographic parameters resulting in the highest diagnostic performance for acute cholecystitis (AC), and compare that performance with radiologist interpretation prior to implementation of the algorithm.

Materials and Methods:
Retrospective review of consecutive emergency department patients with possible AC from 4/1/2019 to 12/31/2019, with assessment of ultrasound (US) and non-US parameters. Non-US parameters evaluated were: (1) clinical pain location; (2) white blood cell (WBC) count; and (3) results of any CT or MR performed within 1 day before US. US parameters evaluated were: (1) GB diameter; (2) GB wall thickness; (3) GB contents; (4) pericholecystic fluid; (5) sonographic pain assessment; and (6) presence of choledocholithiasis. Outcomes were categorized as either: (1) AC; by pathology; or (2) negative AC; established by either pathology showing only chronic cholecystitis (CC) or no cholecystitis, or by clinical follow-up. Patients treated with cholecystostomy tube or extended antibiotics were excluded, since AC could not be pathologically assessed. The radiologist prediction (based on US report) and algorithm prediction was categorized as either (1) positive for AC; (2) uncertain AC, recommend hepatobiliary scintigraphy (HIDA); or (3) negative AC. Diagnostic performance was measured by: (1) sensitivity for AC without HIDA recommendations; (2) sensitivity for AC with HIDA recommendations (presuming 100% sensitivity of HIDA); (3) specificity for negative AC; (4) diagnostic rate (all certain predictions divided by all predictions); and (5) adverse outcome rate (positive AC predictions that did not have AC or CC, combined with negative AC predictions that had AC outcomes, divided by all predictions). Algorithm diagnostic performance was statistically compared to radiologist performance using chi-square tests with significance at p < 0.05.

366 studies on 357 patients met the inclusion criteria. 10.9% (40/366) of US studies had AC outcome, 12.6% (46/366) had pathologically identified CC without AC, and 76.5% (280/366) were negative acute cholecystitis (3 by pathology, 277 by clinical outcome). Diagnostic algorithm compared to US report performance was as follows: (1) sensitivity without HIDA: 87.5% vs 55.0%, p < 0.01 ; (2) sensitivity with HIDA: 100% vs. 89.5%, p = 0.03; (3) specificity: 93.6% vs 94.8%, p = 0.50 ; (4) diagnostic rate: 96.4% vs 93.2%, p = 0.05; (5) adverse outcome rate: 0.0% vs 1.6%, p = 0.01.

Use of a diagnostic algorithm incorporating clinical and US parameters when assessing suspected acute cholecystitis in ED patients results in high diagnostic performance. In our study, the algorithm was statistically better than radiologist assessment without the algorithm in terms of sensitivity for predicting acute cholecystitis, both with and without using HIDA for uncertain cases, and in terms of the rate adverse outcomes. Fewer studies were assessed as uncertain with use of the algorithm, though diagnostic rate barely missed statistical significance.