Concordance-Based Validation of Electronic Health Records and Modality Log Files to Improve MRI Exam Duration...

Area of Science:

• Healthcare Informatics
• Medical Imaging
• Data Science

Background:

• Integrating multi-source healthcare data for predictive modeling necessitates data quality validation.
• Assessing concordance between magnetic resonance imaging (MRI) modality log files (MLFs) and electronic health records (EHRs) is rarely done before model development.

Purpose of the Study:

• To evaluate inter-system concordance of MRI exam duration data between MLFs and EHRs.
• To develop a concordance-based data cleaning framework.
• To determine if machine learning models trained on validated multi-source data enhance MRI scheduling accuracy compared to template-based methods.

Main Methods:

• Extracted MRI exam duration data from MLFs and EHRs (February 2022-February 2024).
• Assessed concordance using Bland-Altman analysis and concordance correlation coefficients after fuzzy merging.
• Removed outliers based on inter-system agreement thresholds.
• Trained a Random Forest regression model on cleaned data to predict exam durations and compared it with template-based scheduling.

Main Results:

• Initial concordance correlation coefficient was 0.33, improving to 0.87 after concordance-based filtering.
• The Random Forest model outperformed template-based scheduling for 11 of 12 procedure codes, with mean absolute error reductions from 2.0% to 57.0%.
• For high-variability procedures, the proportion of exams completed within ±10 minutes of scheduled duration increased from 29% to over 79%.

Conclusions:

• Concordance-based validation is critical for integrating multi-source healthcare data.
• Machine learning models trained on validated multi-source data substantially improve MRI scheduling accuracy, especially for procedures with high variability.