sugr: make it shiny

tl;dr:

• here’s a shiny app to explore CGM data
• here’s the code
• see this post for the data-cleanup procedure

Making sense of the CGM data

In a previous post I cleaned up the data provided by Medtronic, now let’s build a shiny app to visualise it. There are 3 areas of interest available from the data:

1. The glucose levels (GL) in 5 minute intervals. This is the bulk of the data.
2. The amount of carbohydrate intake. However, due to unfathomably bad user prompting, there are a lot of false positives.¹
3. The amount of insulin injected for GL correction and carb intake.

The Minimed & Guardian setup is more or less a closed loop system and injects bolus insulin according to the current GL and GL prediction. However, the bolus amount is not available from the data.

Structure

The app isn’t too big or complicated and therefore it’s suffice to put the whole of it into a single app.R script instead of splitting it into specific logic / UI / misc files. For the most part, it consists of only a few elements:

• a sidebar to control some of the elements to display, like the time range, weekdays or something specific to the visualization or data at hand
• a tab-selector for various aspects of visualization
• the actual plots / graphs / etc

Footnotes

1. There is no way to just manually tell the pump to inject insulin and the workaround is to do that via pseudo-carb-intake.