I’ve know Jessica for many many years … we go way back and we were about the first T1’s to start on Dexcom STS, first generation. Jessica went on to work for Dexcom and then Abbott (when they were selling the Navigator CGM). She’s now a technology consultant based out of Berkeley. Her comments are excellent and I must tell you, she manages her diabetes with a great deal of technical skill and attention.
Jessica’s Take Away after 3 or so months of looping with the RileyLink
The Good:
- Average has dropped about 5-10 points (not much, but it’s a bit)
- Time spent above 160 reduced by more than half (now around 13% vs used to be 29%). That is awesome
- Less swings; std deviation was 52 but is now 36. I think this will improve, even before we take the next step
The Bad:
- I spend a lot more time managing diabetes, mainly because RileyLink goes from online to offline throughout the day. So I have to constantly be checking on it. If the red circle was there for a while sometimes I manually compensate (I couldn’t agree with her more. It does require a lot of attention and rebooting and retuning. But I am reminded that this is still a do-it-yourself project!)
- Pump battery only lasts 5-6 days. I’m using cheap Duracells from Costco. (Very true … I’m trying to make my AAA battery in my Medtronic 723 last 7 days but just barely.)
- Exercise is problematic as my basal manipulation pattern for exercise doesn’t work any more. My software partner told me he thought it was because my body was accessing and using glucose differently. My goal is to tell Loop up to 4 (?) hrs in advance when I’m going to exercise, then have it apply my old pattern, while at the same time overlaying the Loop algorithm. Then I can exercise with starting BG 120, BG 120 during exercise, and ending BG 120. No post-exercise spike. And NO food needed; my body should get its fuel from fat. I did this for 17 years but Loop again causes my body to fuel itself differently, so my BG is no longer perfect before/during/after exercise as it was before. (See, I told you she was WAY techie! I don’t track my exercise like that but thank goodness for thinkers/doers like Jessica!)
The Verdict:
- Worth it to learn more! And learning I am. (I totally agree!)
- Worth it as we are now starting to build AI (artificial intelligence)
into the prediction algorithm. It’s pretty basic now; the algorithm just follows a set of instructions and doesn’t learn from its mistakes. Using RNN (recurrent neural networks) and DL (deep learning), we are collecting data to be able to write a program that will teach Loop to learn and adapt to my body, my activities, and my food. (Bring it on, Jessica and Mike!!!)
