Automated Mapping of Real-world Oncology Laboratory Data to LOINC

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Clinical Data Management

Background:

• Manual mapping of laboratory data to standardized codes like LOINC is time-consuming and prone to errors.
• Electronic health records (EHRs) generate vast amounts of oncology-specific laboratory data requiring efficient coding.
• Standardization of laboratory data is crucial for data analysis, research, and clinical decision-making.

Purpose of the Study:

• To evaluate the effectiveness of automated mapping methods for laboratory data to LOINC codes.
• To assess the impact of machine learning classifiers on reducing the manual mapping burden.
• To determine the accuracy and efficiency of automated systems in an oncology EHR dataset.

Main Methods:

• Development of novel encoding methodologies to vectorize free-text laboratory data.
• Evaluation of logistic regression, random forest, and k-nearest neighbors (KNN) machine learning classifiers.
• Comparison of machine learning model performance against deterministic baseline algorithms.

Main Results:

• All evaluated machine learning models outperformed deterministic baseline algorithms.
• Random forest classifiers achieved 94.5% accuracy in predicting the correct LOINC code.
• Ensemble classifiers further enhanced accuracy, with the top model reaching 99% accuracy.

Conclusions:

• Automated laboratory mapping models are effective in reducing manual mapping time and improving mapping quality.
• Machine learning approaches offer a viable solution for standardizing laboratory data in real-world oncology datasets.
• Automated mapping enhances the efficiency and reliability of clinical data management in healthcare settings.