Teaching Medical Decision Making with RStudio Cloud

Chirag Patel / Associate Professor of Biomedical Informatics at Harvard Medical School

“Have you ever Googled a health-related question and been dumbfounded by the hits? Wondered what an ‘odds ratio’ for a genetic variant you inherited is? Explanations of why we are who we are, and what diseases we might get, and why some of us are at risk, are often unsatisfactory.“

Harvard’s Data Science for Medical Decision Making course uses RStudio Cloud to conduct investigations and discover new associations with disease and health. Students are introduced to statistical decision theory and how modern data science and machine learning approaches can help improve medical decision making.

The course takes students through critical readings in the field and gives them the experience of analyzing the data from those papers to understand how the data is processed and might impact medical decision making.

Taking heart disease risk scores for example:

  • What are the variables that are input into the model and why?
  • What statistical techniques are used to estimate the risk scores?
  • What happens if you adjust, or “tweak” the input variables and how might it influence a decision?

The Challenge

When Chirag Patel began leading this course it quickly became difficult to replicate his own daily work on a student’s desktop. Chirag codes, writes papers, and deploys data tools like Shiny apps in R, so consequently R and RStudio have become an important part of his teaching. While R, the RStudio development environment, and Shiny are all open source and freely available, configuring these tools, and keeping them in sync across an entire class, presented significant challenges for Chirag.
Aside from teaching the course, Chirag also previously had to worry about:

  • The version of R are students running
  • The packages and versions they are using
  • Maintaining their own cloud instance for the course
  • Ultimately, maintaining a synonymous experience for all students

This was a lot of overhead and took a significant amount of time to work through. These challenges were soon accelerated with the pandemic as handling remote logins was even more difficult.

Our Solution

In just a month after learning about RStudio Cloud, Chirag was able to quickly move the entire course to it, by copying his GitHub repos directly into the workspace in RStudio Cloud.

Chirag wrote his own tutorial, inspired by RStudio education resources, and included medical examples from his own group, so students had content to work with from the start. The ability to send workspaces via a link, the classroom feel of the assignments view, and built-in tutorials soon became very helpful for students.

"Students basically had a cluster at their disposal without having to worry about package installation and compute availability. From start to finish it was a great experience.”

- Chirag Patel

The time and overhead needed to manage an environment quickly diminished. The team no longer had to worry about running and maintaining their own cloud instance - giving out logins, ensuring everything was up, and handling occasional crashes.

“From the instructor’s point of view, the things that stood out for us were the very efficient onboarding of students and deployment of assignments and projects. That’s spectacular.”

- Chirag Patel

Why RStudio?

In explaining his reasoning for choosing RStudio, Chirag shared, “because it makes everything transparent and code first. All of the analytic choices you make are clear to the instructor and the other students.” Reproducibility and a code-first methodology is essential. This is a central theme for our course. Biomedicine is facing a reproducibility crisis; therefore, code-first is vitally important. Using RStudio Cloud, students are able to present a paper in class and show their R Markdown code executing in RStudio Cloud. Co-Instructors are also able to work on course assignments together and deploy them easily to students. A unified view and quick deployment with RStudio Cloud is a huge win for instructors and students.

About The Department of Biomedical Informatics

The Department of Biomedical Informatics (DBMI) is one of 11 basic and social science departments of Harvard Medical School. The Master of Biomedical Informatics (MBI) program at Harvard Medical School is designed to advance the use of biomedical data, information, and knowledge for scientific inquiry, problem-solving, and decision making. This program is created for students who are looking to develop skills in data science in the context of medicine and biological sciences to improve human health. Our students graduate with the computational, methodological, and data science skills to contribute to an ongoing revolution in biomedical discovery.

RStudio provides open source and enterprise-ready professional software for data science teams using R and Python.