Risk prediction of biliary infection after endoscopic drainage for malignant perihilar biliary obstruction: A 10-year multicenter retrospective study

Figure 1 Flowchart of model patient enrollment. From January 2014 to October 2024, 337 patients with malignant perihilar biliary obstruction (MPHBO) successfully received 471 endoscopic retrograde cholangiopancreatography (ERCP) drainage procedures for MPHBO in three centers. Nineteen patients suffering from 32 ERCPs had pre-procedural biliary infection (n = 32); three individuals suffering from 5 ERCPs (n = 5) had severe organ dysfunction before ERCP; two of those enduring 3 ERCPs (n = 3) had infections in other systems present before ERCP; twenty-five individuals suffering from 28 ERCPs (n = 28) had incomplete data. After 49 patients enduring 68 ERCPs were excluded, 288 individuals suffering from 403 eligible ERCP clinical data were enrolled in the study. ANN: Artificial neural network; LR: Logistic regression.

Figure 2 Nomogram for postoperative biliary infection. The nomogram illustrates the risk of postoperative biliary infection (PBI) in individuals suffering from malignant perihilar biliary obstruction who underwent endoscopic retrograde cholangiopancreatography. To apply it, the clinician identifies the patient’s value on each variable axis and draws a line upward to assign points for that variable. After summing the points across all variables, the total is located on the overall points axis. From there, a line is drawn downward to the risk scale to estimate the probability of PBI. In this nomogram, a patient with hypokalemia is shown as 1 and the case without is denoted as 0; in the Bismuth-Corlett classification, “3” denotes type IIIa, “4” symbolizes type IIIb and “5” defines type IV. AST: Aspartate transaminase.

Figure 3 Receiver operating characteristic curves for the prediction models. A: Receiver operating characteristic (ROC) curve for the logistic regression (LR) modeling group; B: ROC curve for the LR validation group; C: ROC curve for the artificial neural network (ANN) modeling group; D: ROC curve for the ANN validation group. AUC: Area under the curve.