Development and Validation of a Preoperative Risk Score for Difficult Laparoscopic Cholecystectomy: A Retrospective Study in Resource-Limited Settings

Study Design and Setting
This single-center retrospective observational cohort study analyzed data from Ibb University Hospitals. We reviewed surgical and radiological records of consecutive patients undergoing laparoscopic cholecystectomy from April 22, 2020, to November 23, 2024.

Ethical Approval
The Ibb University Institutional Research Ethics Committee approved the protocol on January 1, 2025 (Ref: IBBUNI.2025.1.001). Individual consent was waived given the anonymized, retrospective design, following Declaration of Helsinki principles.

Patient Population
We screened adults (≥18 years) with laparoscopic cholecystectomy for benign gallbladder disease (acute/chronic cholecystitis, symptomatic cholelithiasis). Exclusions: gallbladder malignancy, combined hepatobiliary/pancreatic procedures, or incomplete records. Final cohort: 301 patients (STARD Flow Diagram, Supplementary Figure 1).

Data Collection
A trained team extracted preoperative data via standardized electronic forms: demographics (age, sex), BMI, clinical history (prior acute cholecystitis, comorbidities), physical exam (RUQ rigidity), labs (WBC count), and ultrasound (gallbladder wall thickness, impacted stone at neck/cystic duct).

Primary Outcome
"Difficult LC" was a composite: (1) intraoperative conversion to open surgery, or (2) major complication (significant biliary/vascular/visceral injury). Of 384 procedures, 57 (14.8%) were difficult.

Statistical Analysis
Analyses used Python 3.10 (SciPy, pandas, scikit-learn, statsmodels). Continuous data: mean±SD or median (IQR) per Shapiro-Wilk test. Comparisons: t/Mann-Whitney U tests (continuous), χ²/Fisher's exact (categorical).

Model Development
Variables with p<0.10 univariately entered multivariable logistic regression (VIF<2.0, linearity checked). Final model by AIC, p<0.05, clinical relevance. β-coefficients derived integer risk score (divide by smallest β, round). Assessed: AUC (bootstrapped 95% CI, 1000 reps), calibration (Brier score, plot), DCA for utility. Bootstrap internal validation.