Best Pracitces

The toolbox can't solve all problems and depending on the use-case your requirements might be different. Here are some things to consider when working with our toolbox:

Warning

This page is incomplete and a work in progress.
Content might change, and some sections are missing.

Data gaps

Gaps in data are typically not reported. While we consider basal rates as active until the next rate, it should be clear that in cases of missing data, basal rate outdate at some point. At this moment we have not implemented any method to detect or deal with data gaps. Therefore, it is left to the user of this toolbox to detect data gaps and/or mark basal rates as outdated. As a general rule, in AID basal rates typically change very often unless micro bolusing is used. One value per day is the absolute minimum that we have seen (flat basal in some users).

Patient ids are strings

Sometimes confusion occurs when the patient ID is compared to an integer because we use strings to be compatible with data sources whose patient IDs are not numerical. It is important to ensure that patient IDs are always treated as strings to avoid any potential issues with data compatibility and integrity.

Patient ids are maintained

We don't change the original patient ID (even if they are non-numerical). Therefore, there is no guarantee that patient IDs are monotonically increasing from 1.

CGM out of range values.

Some datasets set glucose below/above ranges to 40/400, others use special indicators such as 39,401. We replace these with 40,400. However, some datasets like Loop contain sensors that provie readings above range (1 patient as far as we know). This information is also clipped to be consistent.