Abstracts

RETURN TO ABSTRACT LISTING


1578. Building a Quantitative Index for Gender Disparities in Radiology Using a Twitter Big Data Approach: Development of the RadX Index
Authors * Denotes Presenting Author
  1. Kevin Seals *; Olive View-UCLA Medical Center
  2. Edward Zaragoza; University of California - Los Angeles Medical Center
  3. Justin McWilliams; University of California - Los Angeles Medical Center
  4. Tom Le; Olive View-UCLA Medical Center
  5. Monica Deshmukh; Olive View-UCLA Medical Center
  6. Denise Andrews-Tang; Olive View-UCLA Medical Center
  7. Margaret Lee; Olive View-UCLA Medical Center
Objective:
A wide body of literature has documented the persistence of gender disparities in radiology, with disproportionately low female representation on journal editorial boards [1], women representing only 22% of the radiology workforce [2], and fewer than 15% of radiology residency and fellowship positions held by women [3, 4]. Although various strategies have attempted to reduce these disparities [5], the problem persists and there is a need for novel solutions. Following the adage "what gets measured gets managed," we use a medical informatics approach to build the RadX Index (RXI), a quantitative measure of female engagement in radiology derived from the Twitter data stream.

Materials and Methods:
RXI data was collected over a period of 12 weeks from February to May of 2020. Twitter data was organized and stored to a cloud database in an automated fashion using an online natural language processing platform (FlowXO, Padiham, Lancashire, UK). Tweets focused on gender diversity and female empowerment in radiology were identified using #RadXX, a popular hashtag used to denote radiology gender issues. The broader general radiology discussion was identified using the hashtag #Radiology. Tweets using these hashtags were compiled in a cloud database, and the RXI was calculated by taking the ratio of the number of #RadXX tweets to #Radiology tweets within a given 24 hour period. Data processing was performed using a variety of techniques for smoothing and normalization, identifying an effective algorithm for generating a stable, meaningful RXI.

Results:
We compiled 35534 tweets in total, including 1341 #RadXX tweets and 34193 #Radiology tweets. Over the period of analysis there were on average 15.8 #RadXX tweets and 402.0 #Radiology tweets per day. The RXI ranged from a low of 0 to a high of 0.65, with an average value of 0.05 and a standard deviation of 0.08. The highest RXI value was seen on International Women's Day, on which there were 127 #RadXX tweets and 196 #Radiology tweets. A second outlier RXI value of 0.34 corresponded to the highest engagement #RadXX tweet of all time.

Conclusion:
Gender disparities in radiology are an important problem, with the lack of gender diversity both stifling innovation and lowering the quality of patient care. New technology platforms allow us to approach this problem in innovative ways and find new solutions. The current work uses technology to organize the massive, chaotic data stream of Twitter to derive meaningful insights in an automated way. The RXI allows us to track our progress around radiology gender diversity in an objective, transparent manner that can both measure our current progress and motivate positive change. Although the current RXI demonstrates significant value, it can be improved in future iterations with the inclusion of sentiment analysis, tweet engagement, a broader range of hashtags, and other algorithmic refinements. In addition, it is critical that future work builds similar indices for other key issues, including racial and socioeconomic disparities in radiology and other fields.