Give me carbohydrates and then give me death
About three months ago I started publicly charting my blood-sugar levels. Since then I updated two charts daily in order to track a single measure: how well I’m doing at managing my Type-1 diabetes. This got me thinking about how to quantify blood-sugar control in one number instead of two or more.
If you’re diabetes chic then you may be thinking I’m on the fool’s errand of trying to reinvent the HbA1c test.
And if diabetes isn’t a part of your life then you frantically clicked your browser’s Back button and are now scanning Google listings of other pages that mention pop-music temptresses.
It’s true that the HbA1c test is the gold standard for measuring diabetes control. But it does have its limitations. The test only predicts long-term complications from high blood-glucose levels while ignoring short-term risks associated with low blood-glucose. Equally important is the evidence that HbA1c test results fail to reflect variability in blood-glucose levels.
Diabetes hipster and fellow software geek Scott Hanselman likes to use the analogy of air travel to explain what it’s like to live with Type-1 diabetes. Think of the level of glucose in a person’s bloodstream as an airplane. Every day is like a transcontinental flight. A person with a healthy pancreas has a sophisticated autopilot that continuously monitors the plane’s altitude, magically keeping it level and at a safe distance from the ground at all times. When sudden upward pressure or a downdraft pushes on the wings, the autopilot compensates and keeps the passengers from dumping their rum tonics in their laps. Diabetes Type 1 takes a ball-peen hammer to both the autopilot and the altimeter, leaving the cockpit to rely on intuition and on periodic updates from an air-traffic controller (that is, a blood-glucose meter which I operate by drawing blood from a finger). Although they are still cheaper than continuous glucose monitoring, test strips are expensive, so diabetics can only afford to test a few times per day. This means that moderate highs and lows are unavoidable.
But how best to make use of only a handful of blood-sugar test results? What’s needed is a risk index—a single number that tells me how bad a diabetes day it is compared to another day. When my blood-sugar runs high there’s an increased long-term risk of health problems. When my blood-sugar dives I could pass out without warning: a short-term risk. I’m having a good day if I test often—about five times daily—and keep the numbers in a tight range around 100 mg/dL.
The formula for such an indicator should take into account:
- How close my blood-glucose readings are to my target
- How often I’m testing compared to my target testing rate
Because a bad diabetes day can be bad on the high side or bad on the low side, it makes sense for a risk index to start at zero—representing the perfect day—and to rise with increased risk of either high or low blood-glucose. (Think of the airplane again. Dip too low and you die in a fiery or watery crash; soar too high and the oxygen masks pop out, or something like that.) A blood-glucose reading of 200 mg/dL is bad on the high side. 
A blood-glucose reading of 50 is comparably bad, but on the low side. Likewise, 500 is very bad on the high side even as 20 is on the low side. A ratio of actual to target—a ratio that inverts when the actual is less than the target—allows us to average blood-glucose highs and lows without having them cancel each other out; a high and a low are just as undesirable as two highs, and the average of a proper risk index should reflect this.
In addition to an average of individual ratios, we need to measure the variance of the readings from the target. No need for standard deviations here; a simple range of the minimum and maximum readings will do.
The rate of blood-sugar testing should factor heavily into the risk index. If on a given day I take the desired five tests then the risk index should be determined largely by the values of the readings I got. If on another day I test only two times then the risk index should jump, even if the two readings I got were spot-on at 100.
You get the idea. I’ve put together a spreadsheet that calculates a risk index from a set of blood-glucose test results. You can tweak the two parameters of target blood-glucose (the default is 100 mg/dL) and target number of tests per day (the default is 5).
I’ve also made it work like a report card so you know if you’re acing or flunking the diabetes test. Enjoy!
Update: On the advice of a physician friend who’s a diabetes specialist I’ve changed the risk index formula slightly. I no longer take off a full letter grade (1 point) for every missed test during a day, but only two thirds of a point. For example, on Sat 4/01 I tested only twice instead of my target of five times. Previously I would have received a D– (a risk index of 3.8), but now it is a C– (2.8). By the way, just as you get dinged for not testing as often as you plan to every day, there is also a slight grade advantage to testing more times per day. Examine the Excel formula for details.
Update: I showed this to Greg while he and I were in an M. D. Anderson waiting room before hearing the less-than-encouraging news. He suggested that some patients might not want the number of blood-glucose tests to factor into the risk index. So I made another change in the spreadsheet. If you leave the Target Count cell blank, the risk index formula ignores the number of tests and works with the blood-glucose levels only. 
Dig this post? Then 
Digg it!
