Patients undergoing a percutaneous kidney biopsy with real-time ultrasound guidance.
Overall assessment: A potentially useful calculator that has not been validated externally.

Prediction accuracy: Moderate. Bias-corrected C-statistic 0.70 for minor bleeding and 0.83 for major bleeding, both in the development sample.

Risk to reproducibility: Moderate. Based on 617 patients and 79 events (67 minor; 12 major). The authors used strategies to reduce risk of overfitting, including shrinkage estimators and selection of predictors on clinical grounds.

Risk of data reporting bias: Low. Few patients had missing data and they were imputed.

Risk to transportability: High. The model was developed using retrospectively collected data from a single academic centre in London, Ontario, Canada, where all biopsies were performed by Nephrologists. It may not perform well in a different setting.