Algorithmic Detection of Adverse Bleeding Events among Percutaneous Liver and Renal Biopsies

To assess the performance of an adverse event (AE) detection algorithm designed to more accurately identify and classify bleeding AEs among all percutaneous liver and renal biopsies at a single institution over a 3-year period to a provider-reported quality tracking (QT) system.

All percutaneous liver and renal biopsies performed at a single institution over a 3-year period were evaluated. An algorithm developed to detect biopsy-related bleeding AEs using keywords and events in the electronic medical record was employed. Charts flagged by the AE detection algorithm, QT system, or biopsy database review were manually reviewed. AE incidence, sensitivity, and specificity of the algorithm based on retrospective chart review were compared with provider-reported QT data. Fisher test was used to calculate P values.

A total of 4,821 liver and 1,320 renal biopsies were performed during the study period. Retrospective review identified 57 (1.16%) and 104 (7.8%) bleeding-related AEs in the liver and renal biopsy groups, respectively. The detection algorithm reported 82 and 116 AEs, in the liver (sensitivity, 100%; specificity, 99.5%) and renal (sensitivity, 99%; specificity, 98.9%) biopsy cohorts, respectively. The QT system reported 37 and 31 bleeding-related AEs, in the liver (sensitivity, 63.2%; specificity, 99.9%) and renal (sensitivity, 29.8%; specificity, 100%) biopsy cohorts, respectively. The algorithm outperformed QT in AE detection in both liver and renal biopsy (P < .001) cohorts.

A procedure-specific detection algorithm improves identification of bleeding AEs among percutaneous liver and renal biopsies compared with a provider-reported system alone.