JDRF's Tom Robinson on Working With Partners to Gather Data for an Accurate T1D Index

Tom Robinson, vice president of global access at JDRF, talks about the data collection process in creating the Type 1 Diabetes (T1D) Index.

Transcript

How did you ensure the data being used to create the T1D Index were accurate and representative?

We were fortunate in that we had partners coming in who've done a lot of foundational work in the past. For our incidence data, for example, the IDF Atlas has been collecting and vetting good quality incident studies for 20 years. So we went back and took all 20 years worth of incidence studies, it took us all the way back to the 1930s, but we knew that they had all passed community review and academic muster. For mortality studies, again, there have been a few studies that looked at mortality rates and tried to predict what they might be using regression models. And we said, okay, well we could take those studies, let's generalize them to an adult group—they had looked in pediatric populations—let's take them into an adult group as well, and then let's, instead of using kind of a traditional regression model, let's use a random forest regression model using all the latest machine learning technology and cross-validation—all these things that are possible.

So the new model was far more powerfully predictive than the previous one, but we couldn't have done it without the previous one as a foundation. So constantly what we've been doing is taking really great work that epidemiologists have been doing, updating it a little bit and expanding it with the latest techniques or the latest data or the latest studies and then stitching it all together into 1 super model. That's essentially how we made sure that we were using good information, standing on the shoulders of giants, that kind of thing.