Machine learning for detecting wrong blood in tube errors

Abstract

Wrong blood in tube (WBIT) errors are a persisting patient safety issue in clinical laboratory practice. Post-analytically, these errors are identified by a combination of automated delta checks and manual review of results by laboratory staff. However, machine learning may offer a superior approach. Evidence is emerging that machine learning models outperform both delta checks and review of results by laboratory staff. This finding challenges the assumption of human superiority on the task and highlights the need to establish new laboratory workflows that optimise contributions from both machine learning and staff.