External Validation and Recalibration of a Risk Calculator for Major Bleeding After Diagnostic Kidney Biopsy

Bleeding is a serious complication of diagnostic kidney biopsy. A risk calculator for major bleeding associated with kidney biopsy exists, but its performance has not been tested externally.

Retrospective cohort study.

We collected data from all adults who underwent diagnostic kidney biopsy from 2012 to 2019 at a quaternary care center in Halifax, NS to generate an external validation cohort for the "Risk of bleeding complications after kidney biopsy" prediction model (https://perioperativerisk.com/kbrc), which includes the predictors of age, weight, height, platelet count, hemoglobin, kidney size, and native versus transplant graft kidney.

Major bleeding was defined as the need for blood transfusion, surgical intervention/embolization, or death.

Predicted probabilities from patients in the validation cohort were used to assess the model, which was then re-examined using a larger combined cohort of patients from both model derivation and validation datasets.

External validation was performed using 715 patients in an external validation cohort from Halifax, NS. Major bleeding occurred in 48 of 1,732 patients in the combined cohort, with a total major bleeding risk of 2.7%. The model demonstrated good overall discrimination performance in the validation cohort (C-statistic, 0.78; 95% CI, 0.69-0.85) and was well calibrated (calibration slope, 0.72; 95% CI, 0.41-1.02). Recalibration improved performance in the combined cohort; model performance improved with the combined dataset (optimism-corrected C-statistic, 0.84 [95% CI, 0.79-0.88]; bias-corrected calibration slope, 0.82 [95% CI, 0.73-0.88]).

Although model extension was performed with a relatively large patient cohort, the precision of performance remained limited by the small number of major bleeding events.

In this study, a risk calculator for predicting major bleeding complications from kidney biopsy demonstrated good model performance in an external validation that improved with recalibration in a combined cohort. This model can be used to individualize patient risk assessment for kidney biopsy-related major bleeding events, facilitating preprocedural optimization and more efficient allocation of postprocedural monitoring.