I wanna model.
No, not that kind of modeling. And I don’t want one of those other kinds of models, either. (Although it’s certainly nice to watch Heidi Klum every week on Project Runway.)
I want to develop a model that helps me figure out how to balance exercise, insulin, and food — a model that helps me have a pretty good idea what to do before and during exercise so that I can start exercising in a healthy range and end within it, too. It doesn’t matter to me whether it’s a set of more-or-less repeatable actions that are loose and fuzzy but get me to my goal or a table of values where I put in starting values and how much exercise I’ll be doing to end up with an action plan. Either one would work for me, and I suspect it’s going to require both. But consistency (and safety) are my goals.
I know this is possible. When I exercise in the morning — before giving myself any bolus insulin — I just have to lower my basal insulin rate to about 30% of normal and I can go for hours and hours. Of course, I usually eat a little something beforehand; and I need to eat about 30-40 grams of carbohydrate every hour from the second hour onward. But that’s easy enough to do.
But I know that it’s possible to do even better. Olympic nordic skier Kris Freeman seems to have developed something that works most of the time. (His hypo during the 15 kilometer pursuit notwithstanding.) I’m no Olympian, but I know that with the appropriate amount of personalization, I can have the same level of predictability, too.
And do I ever want the ability to predict better! And I don’t mean, “I predict that I’m going to have a low blood sugar event during tonight’s swim.” (That’s what happened tonight, when I started out at a very respectable 156 mg/dL a couple hours after dinner and ended at a very thin 47 mg/dL, complete with shiny spots in my vision provided by my glucose starved brain.) No I mean the ability to more accurately and precisely target all of my BG readings at all parts of the day.
Modeling isn’t always easy. It depends on the problem, how well it generalizes, how sensitive the phenomena are to small perturbations, how many variables there are, whether the relationships between variables are simple (e.g. linearly related) or complex, etc. But I work at a company that develops modeling software. The expertise I need is just down the hallway.
But first I need data. I need to identify the relevant independent variables and collect them. To that end, starting today I’m keeping track of much more data and being much more diligent at recording it. I hate experimenting on myself, but that’s diabetes. Soon, I’ll share more data and maybe ask for your help, too.