The Challenge
Graduate medical education (GME) faculty and residents often face the challenge of deciding what aspects of their dataset are most suited for visual communication. Should they use the full dataset or a portion? How much data should be included and what metrics should be displayed? In addition to these challenges, authors must choose among the broad array of graph and chart options available to creators, from standard bar, line, and area charts to the more uncommon dot plots, slope charts, and Sankey diagrams.1
What Is Known
When working with raw data—be it large data sets with millions of records or small datasets with dozens of rows—data communicators must first determine the story they wish to convey. Visualizing all information contained in a dataset is unlikely to enhance the story. In presenting final analyses, it is critical to understand the main takeaway message(s) from the data prior to determining the optimal visual presentation. This is perhaps the most challenging task in data visualization, and it requires sifting and winnowing through the data to determine which fields will be most useful. To select the takeaway message(s), decide how visuals will inform the intended audience’s decisions or actions.
How You Can Start TODAY
Identify your audience. Start by identifying who you are trying to reach: residency applicants, learners, GME faculty, or those beyond GME such as health system leaders, researchers, publishers, or legislators. Once identified, estimate the audience’s familiarity with data and data visualization (eg, different chart types), the platform(s) they commonly use (eg, slideshow, social media), and the time they will have to interact with the findings. Will your intended audience have a few seconds on their mobile device or in a face-to-face presentation, or will they have several minutes to digest a written document? This will help you to identify the appropriate amount of detail for the data visualization and avoid the pitfall of saying, “I know this graphic is busy, but….”
Know your message. Determine the key points for your audience. For example, are you seeking support for a hypothesis, Clinical Competency Committee decision, or budget request? How are you going to convince your audience of your analyses and conclusions using visuals? In publications, how can you integrate the argument in the text along with the visual representations? For some audiences, the in-depth, sophisticated analysis is important and even necessary, but for others, a headline with the takeaway message may be the most important visual to provide.
Become familiar with a graph library. In the medical field, quantitative data visualizations are often shown as survival graphs, line charts, and bar charts. Qualitative data often uses word clouds and quotes. Resources like the Data Visualization Catalogue, Financial Times’ Visual Vocabulary, and the Graphic Continuum are libraries of different graphs, charts, and diagrams.
Test your visuals. Every visualization has strengths and liabilities. Test your data in several visualization formats to determine which is the most effective. If you are able, pilot the formats with members of the intended audience and have them “talk out loud” as they view each sample. Different visuals will often generate different questions and deliver different messages to an audience.
What You Can Do LONG TERM
Analyze engagement with your data visualization to better understand your audience. If you are sharing your data in a more traditional setting like a presentation, consider debriefing your audience to see what they found to be more or less effective elements of your data visualization. If you are sharing the graphic on social media, the platform will provide analytic data on variables such as interaction, likes, and shares. Compare the analytics between different data visualizations as a proxy for what your audience appears to find most engaging.
Create different data visualizations for different audiences or venues. Consider how integrating interactivity or animation into your visuals can enhance engagement and understanding. After you create a dense, rich graphic for a website or manuscript, consider a more straightforward, more digestible graphic for presentation slides or social media posts, with annotations (Figure).
Explore newer data and data science analytical methods. With advancements in AI, data visualization tools will get better. In the short term, consider prompting AI tools to generate additional ideas to analyze or visualize your data, while recognizing that these tools have their own built-in biases and errors. In the future, AI tools are likely to be integrated directly into data visualization tools that will enable users to directly query their data.
Continuously learn and improve your skills. There are many good books,1,2 and numerous online and offline courses to learn how to improve your data and data visualization skills. With respect to data visualization tools, Python and R have excellent graphic libraries, compared to traditional statistical packages like Prism and SAS, and offer direct integrations with online publication tools through R Markdown and Quarto.3 Other browser-based tools like Datawrapper and Flourish offer straightforward options to create data visualizations with little to no coding skills required.
Figure.
Annotating a Chart Can Help Readers Better Understand the Content More Quickly and Easily
Abbreviations: AOA, American Osteopathic Association; ACGME, Accreditation Council for Graduate Medical Education.
References and Resources
- 1.Schwabish JA. Better Data Visualizations: A Guide for Scholars, Researchers, and Wonks. Columbia University Press; 2021. [Google Scholar]
- 2.Boers M. Data Visualization for Biomedical Scientists: Creating Tables and Graphs That Work. VU University Press; 2022. [Google Scholar]
- 3.Wickham H, Çetinkaya-Rundel M, Grolemund G. R for Data Science: Import, Tidy, Transform, Visualize, and Model Data. 2nd. O’Reilly Media; 2023. [Google Scholar]