Related Experiment Video
Updated: May 12, 2025
Author Spotlight: Advancing Early Detection and Treatment of Gastrointestinal Tumors
Published on: February 16, 2024
An App model that utilizes a logistic regression algorithm for predicting choledocholithiasis: A prospective clinical
F García-Villarreal1, L M Torres-Treviño2, C Herrera-Figueroa1
1Departamento de Medicina Interna, Servicio de Gastroenterología y Endoscopia Digestiva, Hospital Universitario "Dr. José E. González", Universidad Autónoma de Nuevo León, Monterrey, Nuevo León, Mexico.
Introduction And Aim:
The diagnostic yield of the current criteria for assigning the risk of choledocholithiasis (CL) is inaccurate. The aim of our work was to develop a logistic regression model for predicting CL diagnosis in patients catalogued as either intermediate or high risk for CL, according to the criteria of the American Society for Gastrointestinal Endoscopy (ASGE).
Material And Methods:
We conducted an analytic, observational, cross-sectional study for evaluating the diagnostic yield of a logistic regression model in adults with intermediate or high risk for CL. A receiver operating characteristic (ROC) curve analysis was done to determine the best cutoff point for predicting the diagnosis of CL. Endoscopic retrograde cholangiopancreatography (ERCP) was utilized as the gold standard for diagnosing CL.
Results:
A total of 148 patients suspected of presenting with CL were studied. In our cohort, 71 had immediate risk and 77 had high risk. CL diagnosis was confirmed in 102 patients (69%). Our model showed an area under the curve (AUC) of 0.68. In patients with an intermediate risk for CL, the AUC value was 0.72 and the positive predictive value (PPV) was 70%. In patients with a high risk for CL, the AUC value was 0.78 and the PPV was 89%.
Conclusion:
Our model appears to better predict the diagnosis of CL than the ASGE criteria for patients with an intermediate or high risk for the disease. Our model can guide clinical decisions in patients with suspected CL.
