Algorithmic Prediction of Delayed Radiology Turn-Around-Time during Non-Business Hours

Repository accompanying manuscript

How to use this Repo

Installation Instructions

1. Clone this repository

git clone https://github.com/bdrad/report_delay.git

2. Install dependencies (install via pip or conda)

• numpy
• pandas
• sklearn
• matplotlib
• xgboost
• argparse

3. Install Rad Classify (utilized for preprocessing steps & fastText wrapper): https://github.com/bdrad/rad_classify

Data

Users will need to provide their own training data in 2 parts.

Main Data

The following column headers should be used for the input data and saved as a .xlsx file.

• "Minimum of Exam Ordered to Prelim/First Com": the interpretation time as seconds (float)
• "Minimum of Exam Completed to Prelim/First Com": the total time as seconds (float)
• "Report Text": raw clinical history (str)
• "Patient Status": inpatient or emergency (str)
• "Patient Status numerical": inpatient or emergency (int)
• "Time of Day Label": evening, morning, afternoon, late night (str)
• "Time of Day Label numerical": time of day (int)
• "Body Part Label numerical": body part examined (int)
• "Preliminary Report By": Trainee performing the report (str)
• "Preliminary Report By numerical": integer
• "Preliminary Report Date": datetime
• "Point of Care": the hospital campus/facility (str)
• "Exam Code": study description (str)

Secondary Data for PGY levels

With the following columns, create a key 'trainees.xlsx'. Place this file in the root directory.

• "Preliminary Report By": string
• "PGY": integer

Training and evaluation script

train_evaluation.py contain the training and test set evaluation scripts for their respective classifier.

To run the script:

python train_evaluation.py --data_path PATH_TO_DATA --time_delay CHOOSE_TIME_DELAY

• CHOOSE_TIME_DELAY must be interpretation_time or total_time
• PATH_TO_DATA must be a relative path to the the main data .xlsx file