TP: Applying the “Tidy Patient” approach to medicare claims

TP: Applying the “Tidy Patient” approach to medicare claims

Example using Medicare

Let’s do an example using Medicare data

Data Description

Research Question

Using medicare data, create a cohort satisfying the following I/E criteria:

  • Race
  • Diabetes status
  • Inpatient claims
  • prescription drugs
  • Survival

Non-Tidy approach

  • Classic example using like 150 lines of code

Tidy Patient approach

  • Show it in 10 lines of code

Compare approaches

  • Results are the same.
  • Tidy benefits
    • 90% less user code
    • easier to update
      • If you wanted to update the analysis to do X, almost anyone could do it with the tidy approach, only experts could in the non-tidy way.
    • Fewer opportunities for errors
    • Ready for automatic feature updates if the underlying modular functions improve.
    • Ready for use in external tools, such as a dashboard

How did we do it?

  • Modular, functional, documented, tested code

What’s next?

  • Wrap tidy code within a dashboard or API
  • Create methods for different datasets

Call to Action

  • To get support in building scaleable RWD solutions at your organization, reach out to Plinth.

Internal Notes

  • Show examples in R and Python? What about pseudo-code?
  • Link to a github repo with actual functions?
    • Pro: Could increase engagement
    • Con: Takes work