DATA ENVELOPMENT ANALYSIS APPROACH TO PREDICT RISK FACTORS FOR POST-ENDOSCOPIC RETROGRADE CHOLANGIOPANCREATOGRAPHY PANCREATITIS

Aims Determining the characteristics of patients who develop post-endoscopic retrograde cholangiopancreatography pancreatitis (PEP) is an important direction to improve ERCP performance. However, the rarity of specific adverse events such as PEP mostly violates the sample size requirement of multivariate analysis, leading to inaccurate risk prediction. To overcome this issue, we propose an approach based on data envelopment analysis (DEA), a data-driven method, to identify reliable predictive factors associated with the incidence of PEP.

Methods The DEA-based approach is implemented to a set of 10 inputs including the indications and cannulation-related variables, with PEP as an output within a decision-making system that retrospectively analyzes the evolution of ERCP patients. Using the clustering technique of DEA, we generated an overall inefficiency index to classify patients based on the relative efficiency performance across different variables and identify specific variables that potentially contribute to the inefficient performance.

Results PEP was developed in 32 /615 patients who underwent ERCP with native papilla (5.2%). The attached figure provides a concise description of the patients’ performance across the input variables, with higher values representing relatively worse performance. We noted that patients who had PEP showed highly suboptimal performances for cannulation duration and unintended PD cannulation. However, using precut to achieve biliary cannulation was not directly associated with PEP.

Fig. 1

Conclusions To our knowledge, this is the first attempt at developing a PEP prediction model using mathematical approaches, DEA. Herein, we found that long cannulation time and unintended PD cannulation>1 contribute as factors significantly related to PEP.

Publication History

Article published online:
14 April 2022