Development and Internal Validation of a Prediction Model for Ureteral Access Sheath Insertion Depth During Flexible Ureteroscopy Based on Patient Anthropometric and Anatomical Parameters

Backgound:Optimal placement of UAS during fURS) relies largely on surgeon experience. Insertion that is too deep may cause ureteral injury, whereas insertion that is too shallow necessitates repeated fluoroscopic adjustment. This study aimed to develop and internally validate a machine learning model that predicts UAS insertion depth using the L1 vertebra–pubic symphysis distance and readily available anthropometric parameters. Methods This single-center cohort study enrolled 124 consecutive patients who underwent fURS with UAS placement between April 2025 and July 2026. The L1–pubic symphysis distance was measured on preoperative plain radiographs or CT scout images. Candidate predictors included height, age, sex, side, BMI and the L1–pubic symphysis distance. The outcome was the actual UAS insertion depth (cm), measured intraoperatively from the perineal skin to the ureteropelvic junction. The dataset was divided into training (70%, n = 86) and test (30%, n = 38) sets using stratified random sampling by sex. Six regression algorithms were compared, including Ridge regression, least absolute shrinkage and selection operator (Lasso), Bayesian ridge, random forest, support vector regression, and multilayer perceptron. Hyperparameters were tuned by five-fold cross-validation. Model performance was assessed using R², mean absolute error (MAE), root mean square error (RMSE), and the proportion of predictions within ± 2 cm of the actual depth. The final model was refitted on the entire cohort, and bootstrap 95% confidence intervals (CIs) were obtained for all coefficients. The study followed the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guideline. Results The mean UAS insertion depth was 33.7 ± 2.8 cm (range, 26.5–40.0 cm). Height (r = 0.749), male sex (r = 0.661), L1–pubic symphysis distance (r = 0.563), BMI (r = 0.251), and younger age (r = − 0.498) were correlated with insertion depth. Ridge regression with regularization parameter α = 10 achieved the best balance of accuracy and interpretability, with a test-set R² of 0.661, MAE of 1.24 cm, RMSE of 1.56 cm, and 78.9% of predictions falling within ± 2 cm. Five-fold cross-validation yielded an R² of 0.642 ± 0.067. Leave-one-out cross-validation across all 124 patients produced an R² of 0.649, MAE of 1.39 cm, and 77.4% accuracy within ± 2 cm, substantially outperforming an L1-distance-only model (R² = 0.291; MAE = 1.86 cm). Bootstrap internal validation (1,000 resamples) confirmed minimal overfitting (optimism = 0.026), with an optimism-corrected R² of 0.659, MAE of 1.39 cm, and a calibration slope of 1.033. Standardized coefficients identified male sex (β = 0.927; 95% CI, 0.651–1.205), height (β = 0.811; 95% CI, 0.444–1.148), and L1–pubic symphysis distance (β = 0.740; 95% CI, 0.396–1.086) as the strongest predictors, with BMI also contributing independently (β = 0.410; 95% CI, 0.164–0.674). A freely accessible online calculator was developed for clinical use. Conclusion A multivariable Ridge regression model incorporating the L1–pubic symphysis distance and anthropometric parameters can predict UAS insertion depth with clinically acceptable accuracy. The novel L1–pubic symphysis radiographic measurement, obtainable from routine preoperative imaging without additional examination, contributes independently to depth prediction. External validation in larger and more diverse cohorts is warranted before broad clinical implementation.