Using Machine Learning-Based Multianalyte Delta Checks to Detect Wrong Blood in Tube Errors.

Area of Science:

• Laboratory Medicine
• Clinical Pathology
• Machine Learning in Healthcare

Background:

• Wrong Blood in Tube (WBIT) errors occur when patient specimens are mislabeled.
• These errors can lead to significant patient harm.
• Current detection methods are insufficient.

Purpose of the Study:

• To develop a machine learning-based algorithm to detect WBIT errors.
• To mitigate patient harm caused by WBIT incidents.
• To improve the accuracy of specimen identification in laboratory medicine.

Main Methods:

• Simulated WBIT errors in routine inpatient chemistry test results.
• Developed and trained five machine learning-based WBIT detection algorithms.
• Evaluated algorithm performance using metrics like area under the curve.

Main Results:

• A support vector machine-based algorithm incorporating 11 analytes showed the best performance.
• Achieved an area under the curve of 0.97.
• Outperformed traditional single-analyte delta checks significantly.

Conclusions:

• Machine learning-based multianalyte delta checks offer a practical solution for WBIT error detection.
• This approach can identify WBIT errors before test reporting.
• Implementation can enhance patient safety in laboratory medicine.