Determining Follow-Up Imaging Study Using Radiology Reports.

Area of Science:

• Radiology
• Medical Informatics
• Artificial Intelligence

Background:

• Radiology reports frequently include follow-up imaging recommendations.
• Non-adherence to these recommendations can result in delayed treatment, adverse patient outcomes, and increased healthcare costs.

Purpose of the Study:

• To develop a scalable, automated approach for identifying the completion of follow-up imaging studies recommended in prior radiology reports.
• To improve adherence to crucial follow-up imaging recommendations.

Main Methods:

• Utilized a dataset of 559 follow-up imaging recommendations and subsequent reports from an academic practice.
• Developed a ground-truth dataset by having three radiologists identify appropriate follow-up examinations.
• Trained an Extremely Randomized Trees classifier using recommendation attributes, study metadata, and text similarity.

Main Results:

• The classifier achieved an F-score of 0.807 in identifying follow-up examinations.
• Inter-annotator agreement among radiologists ranged from 0.853 to 0.868.
• The methodology demonstrated a high degree of accuracy in linking recommendations to completed studies.

Conclusions:

• Automated methods can effectively identify follow-up imaging completions, supporting integration into management applications.
• This technology can enhance adherence to follow-up imaging recommendations, leading to better patient outcomes.
• Radiology administrators can leverage this system to monitor compliance and implement proactive reminder strategies.