05 August 2013

I'm a nerd

I confess: I'm a nerd! I love data! I love piecing stuff together!

Which totally makes me perfect for diabetes. I mean, this is such a data heavy condition.

I'm also an engineer (in training. Yes, that's the technical term for a person with a BS in engineering who passed the FE, an "engineer in training"). I'm also an electrical engineer (who emphasized in power and energy, but this part is irrelevant to this post).

Unsurprisingly, this meant I took a fair amount of math in college. I actually rejoiced when I finished my last math course, but not because I dislike math, but because I hated our math department. I also took a few statistics courses.

What may surprise a lot of people is how much statistics is involved in engineering. And it can be frightening how much hand waving-engineering magic actually goes in to very important things. If you ever are on a plane with my husband, I'm sorry for this exact reason. He will go on and on about how certain planes failed and engines failed and how terrifyingly thin the actual barrier is between you and the outside while you're in the air. Statistically, though, your greatest threat in the air is pilot error. Comforting.

Statistically, I'm know I should be more terrified of lightning striking me than being attacked by a shark. Statistically, I should be more stressed about getting in to a car accident when I'm close to home rather than when I'm driving somewhere new. Statistically, I should be more worried about the things that I'm not as worried about, just like I'm not as nervous driving as I am flying.

But I've already won the statistics lottery, many times. I have type one diabetes. All my siblings have type one diabetes. My body hates me in other weird ways. I've experienced drama and trauma that most people don't.

So, when statistics tells me I should be more worried about x, rather than y, I call it a liar.

And that's the thing about test strip accuracy. Even though it's supposedly guaranteed to be within 20% (and let's be honest, that range is large enough to drive a space shuttle through it) something like 98% of the time, I will worry about that 2%, not because it is common, but because it is disastrous when it happens.

Let's crunch some numbers, here. Let's pretend I check 10 times a day, to make the math easy, and that my 98% SWAG is correct***. The probability that all those numbers are within that 20% range for the day is .98^10, which is about 82%. That's not too bad, but it's still a B-, which is still passing. For a week? That's .98^70 which is about 24%. That is terrible. I'm supposed to make life altering decisions when there's about a 1 in 4 chance that all my tests results will be within a giant range, every week? (And, for those curious, for the year, that's .98^3650, which is pretty much not happening.) How am I supposed to not worry about something like that?

Let's look at that 20% number as well. I'm going to use mg/dL because that's what I'm familiar with. Let's pretend that my range is 80 to 120 and my target is 100 (because who doesn't love a wonderful 100?). So, I'll have my handy dandy correction formula which will look like this:

(BGreading - target)
ISF
 
 
Where ISF (insulin sensitivity factor) is the X when we general write it out like this: 1u:X>target (more diabetes notation! YAY!) For simplification, we'll call my ISF 50 (For example, if I was 150, I'd take 1 unit, using these numbers). Disclaimer: use the formulas and numbers you and your health professionals have decided on. I'm pulling these numbers out of engineering magic air, like the airplane.
 
So, let's look at how much a dose can change based on that 20%. Pull out your pre-algebra! We're looking at slopes (for my math people, derivatives) and math!
 
The slope is 1/ISF = 1/50, while our independent variable is our BG reading.
 
So, let's take a look at some readings:
 
If I'm 100, my meter could show 80 to 120, which would give me a correction range of -0.4 units to +0.4 units, which is technically an infinite error, since I should really be giving myself 0 units.
 
If I'm 150, my meter could show 120 to 180, which would give me a correction range of 0.4 units to 1.6 units, which is a 60% error, since I should be giving myself 1 unit.
 
If I'm 200, my meter could show 160 to 240, which would give me a correction range of 1.2 units to 2.8 units, which is an 40% error, since I should be giving myself 2 units.
 
If I'm 250, my meter could show 200 to 300, which would give me a correction range of 2 units to 4 units, which is a 33% error, since I should be giving myself 3 units.
 
And so on and so forth.
 
Percent error being calculated by (deviation from expected dosage)/expected dosage. So, for example, if I expected to give myself 1 unit, but the result said to give me 1.6 units, that's (1.6-1)/1 =.6 = 60%. For those of you interested, this simplifies down to*:
 
(1.2*RealBGreading - target)  -1
(RealBGreading-target)
 
The percent error does go down as the number increases, but that has to do more with the target value becoming smaller and smaller relative to the actual BG reading. What you will notice that is increasing, besides the BG number, is the range of dosages.
 
Part of this tells me I should really stay as close to my target as possible to minimize the range of dosing errors.
 
But what this tells me more than anything is that we are living with crap standards.
 
And then we add in all the other problems that come with test strips. They are less accurate less often than what the standards are supposed to be.
 
And we are expected to make decisions based on these numbers every hour, every day. Our doctors judge us based on these numbers every time we come in.
 
How are we supposed to feel safe or know how we're actually doing when there is more accuracy and precision put in the GPS on my smartphone**?
 
 
I want more accurate and more precise test strips (and meters), not only to indulge my perfectionist side, but because my life depends on it.
 
 
 
 
 
 
*This is the same as
1 - (0.8*RealBGreading - target)
     (RealBGreading-target)
 
 
** I don't know if there actually is higher precision standards for the GPS on my smart phone, and I'm too tired at this point to look it up, but is sure as hell feels like it. At least I know where I am and know where I'm going with that sucker.


***So, I looked it up. Apparently, it's 95% of the time, not 98%. This means, testing 10 times a day, there's a 59% chance that all the results are within the +/- 20% (or +/- 15 mg/dL for readings lower than 75 mg/dL) for a day, a 2% chance for the week, and and even sadder "this is not going to happen" for the year.


No comments:

Post a Comment