A data-driven approach for quality assessment of radiologic interpretations

Area of Science:

• Medical Informatics
• Radiology Quality Assurance
• Health Services Research

Background:

• Growing demand for high-quality, cost-efficient healthcare necessitates improved methods for evaluating diagnostic accuracy.
• Current quality assessment in radiology often lacks objective, data-driven metrics for individual performance and departmental oversight.
• Integrating diverse clinical data sources presents a significant challenge in healthcare quality improvement initiatives.

Purpose of the Study:

• To develop and present a data-driven methodology for automated quality assessment of radiologic interpretations.
• To utilize clinical information, such as pathology results, as a reference standard to evaluate radiology performance.
• To enable quality assessment at the level of individual radiologists, subspecialty sections, imaging modalities, and entire departments.

Main Methods:

• A data-driven approach was implemented to automatically extract and compare patient medical data from electronic health records.
• Radiologic diagnostic conclusions were compared against downstream diagnostic conclusions (e.g., pathology) serving as the 'ground truth'.
• Information extraction tools were applied to characterize concordance between disparate clinical data sources.

Main Results:

• The system was initially applied to breast imaging, analyzing 18,101 radiologic interpretations against 301 pathology diagnoses.
• Achieved a precision of 84% and a recall of 92% in matching radiologic interpretations with pathology results.
• Demonstrated the feasibility of automated quality assessment by comparing radiology diagnoses with established clinical outcomes.

Conclusions:

• The presented data-driven method offers an automated solution for assessing radiologic interpretation accuracy and utility.
• Highlights the potential of integrating multiple data sources and leveraging information extraction for healthcare quality improvement.
• This approach can provide objective feedback for radiologists and departments, contributing to enhanced patient care and cost-efficiency.