Mission Impossible: Predicting ERCP Procedure Times to Facilitate Quality Improvement

Abstract

Purpose: ERCP procedure length remains quite variable, ranging from minutes to hours, given the heterogeneity of procedural indications and inherent technical difficulty. ERCPs are typically scheduled in block time without taking into consideration indication, patient anatomy, or other clinical factors. At times, the procedure may be longer or shorter than the allotted time. We aimed to retrospectively evaluate the utility of a priori prediction of ERCP procedural length with an overall goal to improve the scheduling efficiency and capacity of our therapeutic endoscopy unit. Methods: Our endoscopy database was utilized to identify patients who underwent ERCP procedures from January 1, 2012 to February 1, 2012 at a tertiary care medical center. Clinical and procedural data were reviewed. A brief summary was generated by a medical resident for each patient, which outlined recent clinical history and ERCP indication. Additional procedural information was provided regarding GI fellow assistance, sphincterotomy status, failure or success of previous ERCPs, and presence of surgically-altered anatomy or stents. A 2nd-year GI fellow and a staff therapeutic endoscopist with >20 years of ERCP experience attempted to retrospectively predict procedural length (0-30 min, 31-60 min, 61-90 min, >90 min) in each case blinded to other details about the case. Predictions were then compared to actual procedural length using the kappa statistic. Results: ONe consecutive ERCP procedures (51 men; mean age 62±15 years) were included. Indications for ERCP included a biliary etiology in 83 cases. The remaining cases were divided equally between pancreatic (N=10) and various other indications (N=7). Fellows participated in 70 procedures. Median ERCP time was 43 minutes (IQR 28-63). ERCP times were classified by blocks of time (min): 0-30 (n=28), 31-60 (n=44), 61-90 (n=16), >90 (n=12). Time blocks were accurately predicted in 40% cases by the 2nd year GI fellow and in (31%) cases by the staff endoscopist. When compared to actual ERCP times, both the fellow's and staff endoscopist predicted times had poor kappa agreement (κ=0.15 and κ=0.02, respectively). The kappa between fellow and staff was also poor (κ=0.08) with an overall agreement of 35%. Neither fellow nor consultant predicted correctly in 41% cases with both overestimating in 23/41 (56%), both underestimating in 13/41 (32%), and one overestimating/one underestimating in 5/41 (12%). Conclusion: Accurate prediction of ERCP procedure length appears to be very challenging. Inaccurate prediction is likely multifactorial with the inherent technical difficulty and unpredictability of ERCP being major limiting factors. Prospective studies with a priori prediction of ERCP length are needed to confirm our findings.