Redefining diabetes for researchers

The Tidepool Big Data Donation Project helps students, academics, and industry partners innovate faster and expand the boundaries of our knowledge about diabetes. Any of our users can opt in (or out) of donating their data at any time. Our team anonymizes the data, stripping it of all identifying information including name, email address, custom labeled profiles, and device identifiers before including it in the pool of datasets that we can license to our data partners.

One of those data partners includes our friends over at Carleton College. We decided to check in with Sarah Meerts and Dave Musicant to learn more about their research efforts, the impact of your donated data, and how it has been put to good use.

What is your current research project?

Dave: We first started with Sarah’s big question about blood sugar management during periods. We wanted to see if we could detect fluctuations in glucose levels and insulin intake over an average 28-30 day cycle. Turns out they fluctuate a lot, and what we really need to be focusing on is how the Insulin Sensitivity Factor (ISF) varies. So that is what we are currently working on: figuring out what insulin sensitivity looks like. We are looking for cyclical patterns in blood glucose and insulin data clustered by demographics, based on what we are seeing. The datasets from Tidepool have allowed us to get our feet wet with prototype results.

Taking a step back, why diabetes? Why diabetes data?

Sarah: It is personal for me. I have a child who was diagnosed with type 1 diabetes when she was 8 years old. Diabetes is especially complicated with a kid, because you aren’t the one controlling all of the variables. Dealing with the complexities of caring for a child with diabetes, and then seeing a correlation between puberty and wild blood sugar fluctuations with tons of highs and lows, I noticed the patterns.

I am not a data scientist, but my background is as a behavioral neuroendocrinologist so I have an understanding of the biology involved, but I only have an n=1 sample size with my own kid’s data. I knew about Tidepool from using it to gain insight into my child’s glucose patterns and learned about the Tidepool Big Data Donation Project. I also had the opportunity to connect with Dave for an independent research opportunity with Carleton students. I asked him if we could ask this question about blood sugar fluctuations around hormonal cycles, and Dave said ‘yes, we can do this’. So here we are.

From a data science perspective, what are some of the biggest challenges with the research you are conducting today?

Dave: There is a lot of terminology that is used inconsistently by different fragments of the community: insulin on board (IOB), insulin absorption, duration of insulin, insulin action, etc. That has been challenging as a researcher and as a teacher because these terms are not used uniformly across patient, provider, and research communities and then it’s on us to determine what is assumed, what is approximated, and what is guessed.

The data dictionary that came with the Tidepool datasets was extremely helpful, but not having a dictionary for the standard meaning and use of terms used across the industry has been difficult. Not every research paper fits together, and it would be wonderful to standardize the usage of terminology.

More directly for us and the focus of Sarah’s research on female sexual health—ovarian hormones and how that relates to sexual function—there is a dearth of research on biological functions of women relative to men. It’s important that we look at women as a separate cohort because [men] have different biology and it is important to recognize and tailor medical treatments to appropriate populations.

How are you balancing research with your other responsibilities at Carleton College?

Sarah: We are teachers first. We signed up for a diverse array of responsibilities, only one of which is research. So our process involves setting up meetings; this term we had two meetings per week. We catch up with the students that we mentor on the current state of research and give them the tools to push the research forward.

We have a two-fold goal: 1) we would love to answer these research questions but 2) we are also training students to think critically. That is a much bigger endeavor than just answering the questions regarding blood glucose fluctuations around hormonal cycles. Due to our roles as professors, our process is slower than it might be otherwise.

Looking ahead, presuming success, what do future iterations of your research look like?

Dave: Our immediate goal is to extract the ISF from the data. We don’t have that extraction yet because we need an entire team of students to do the work on this. We are trying to produce synthetic data based on our own model. If we can model how we think sensitivity works, it can help us understand what data looks like. Our next goal is to extract some meaning of ISF from the data in correlation with hormonal cycles.

What would your research efforts/investigations look like if you had reliable menstrual cycle data included with the diabetes device data (like we are looking to provide through the Tidepool Period Project)?

Dave: Because we don't have actual data on the timing of hormonal cycles, we've had to look for cycles around the right length in hopes of finding something to analyze. Using Fourier analysis, we have been able to cherry pick results to indicate that it looks like there is a cycle here.

If Tidepool could give that to us, we would put aside the question of how to extract the cycle data. Having access to menstrual cycle data with diabetes data would enable us in a dramatic way and allow us to demonstrate that there is something measurably different at different stages of the cycle. Then we don't have to guess where in the cycle a person is. All of the Fourier extraction becomes irrelevant and unnecessary with that additional data. It would be a big leap forward.

As a nonprofit organization, we are able to focus on this larger impact on the diabetes community instead of on returns; but we need your help. We invite you to join us along the pathway we’re building to make an interoperable automated insulin dosing system a reality and drive change across all levels of the diabetes industry.

We can Redefine Diabetes together when you support Tidepool with a donation at tidepool.org/donate.