Auto calculate QA from BG - let's get it right

Hi Novorapid,

What are the 4 correction factors you use? Currently you specify a normal correction factor (QA:BG ratio) and then add overrides which specify a BG range or time range to use a different QA:CP ratio.

Cheers

I currently use:

below 10mmol/l = 1 unit drops 2.5
10-13mmol/l = 1 unit drops 2
13-17 = 1 unit drops 1.5
above 17 = 1 unit drops 1

you can see the pattern........this could probably be tweaked much more accurately, but I feel this would be going to far.........

There is a danger of a discontinuity here...

BG is 16.9, corr is (16.9-5.5)/1.5 = +7.6
BG is 17.1, corr is (17.1-5.5)/1.0 = +11.6

Logically the different corrective ratios should be applied separately to each band and accumulated, but nobody will already be doing that in their head. A computer can do it, though. And that will avoid the discontinuity problem.

Eg:
BG is 17.1, need (17.1-17)/1.0 = 0.1 to get to 17,
plus (17-13)/1.5 = 2.7 to get to 13,
plus (13-10)/2.0 = 1.5 to get to 10,
plus (10-5.5)/2.5 = 1.8 to get to 5.5.
So corr = 0.1+2.7+1.5+1.8 = 6.1.

This is lower than novorapid would want, I suspect, but that's due to the corrective ratios being applied to each band. By using slightly different ratios the desired correction can be made in all cases without any discontinuity problem.

I recommend that we should use this calculation method, since it will give reliable results.

I think I remember you or someone else asking me if I used all the bands of blood sugars when correcting, as obviously, my blood sugars would be dropping and therefore would need less.......

But for me, the method I use, which is just to use one band, has worked perfectly, and its simple, however this could just be specific to me.........its all about insulin resistance, hence the varying effectiveness of 1 unit....

My theory is that if I am above 17mmol/l then the cells of my body have a higher resistance to insulin than I would at lower levels, but its likely my blood sugar levels would drop quicker than that of my resistance and therefore trying to calculate a correction by using all the bands wouldn't really apply.....

I conclude this from my own personal experience and blood sugar results, quite simply, I am on target by the next meal time......

Its interesting though and I would be interested to see someone testing this method...

What do you mean when you say discontinuity? If I was 16.9, I would treat that as if it was 17 or above, not an exact science by any means...........

I'm the same as Novorapidboi26. Since beginning DAFNE I've identified 2 bands so far, when I'm in the higher band my correction was being calculated at 1QA for each mmol so with a 15 reading I'd be taking 15QA. No consideration enters into it from the reduced correction of the lower band.

Yes, it was me who asked, and this discontinuity issue is exactly what I was investigating when I asked. Let me explain. We will be using an equation to compute the QA correction we need for any given BG level. In mathematics, a discontinuity is when the result of applying an equation to values that are a small distance either side of some particular value results in a large jump in the result of the equation.

As I showed above, there is a discontinuity at BG 17 (also at 13 and 10). BG 16.9 gives +7.6 while BG 17.1 gives +11.6. So if you have a BG of close to 17 (say 17.5, or 16.5, perhaps), your QA correction will be either +7 or +11, approx. Now, clearly, both of these cannot be the right corrective dose to use. If +7 is right for a BG of 16.5, then +11 is clearly way too much for a BG of 17.5.

What my previous post says is that we can avoid this discontinuity problem by applying the corrective factors only within the bands they apply to. This will assist in obtaning good and consistent corrective adjustment values, and will help the app users to achieve better BG control.

No doubt some personal tuning by trial and error will be needed to arrive at the QA:BG ratios for each band. But I believe it will be a simpler task to arrive at these ratios if we use this banded application, in spite of the method being harder to understand. It will be safer, too: the discontinuity issue would have the potential to cause unexpected and unanticipated overcorrections.

I'll go with the consensus on this one - if at the moment people use novorapidboi's method by applying one qa:BG ratio to the whole correction dependent on the current BG then I'll go with that, whereas if people use the banded method that Podacris describes then ill go with that.

The aim of this is to aid people in their own calculations, rather than forcing a new (and potentially not well understood method) on them, which could be dangerous.

I was actually under the impression that it was only me who used such a method of correction...........

Do others use it?

To come to these corrective ratios I, made sure my BI and QA were right, then recorded how much my correction dropped me when at different levels......so using the standard between 2-3mmol/l for every 1 unit.........

So for me being at 17.2 would require the most insulin, if I go by the observations I made.........

So it all depends on what each individual observes when testing I suppose.....that's how I have arrived at these values.....

There would be know evidence that i possess that would suggest that my resistance would change at the same rate as my blood sugar returned to normal........its a fine line being 16.9 and taking 7.6 units and being 17.1 and taking 11.6 I agree, but for me so far it has worked....

So my final word would be that people need to test and record to see how the behave......

Implementing my method would be easier from an equation point of view I would imagine, so if your can be done, surely both can be put in place.....?

Sorry Simon for suggesting you should do both......

Yes, I agree with Novorapid that you should do both, Simon, and it is not too much extra work. Both calculation methods require the storing of the same data: band limits and a QA:BG ratio for each band, and that data can be used differently depending on the calculation method the user chooses to apply.

All you need is to have one additional radio button question, to allow the user to choose which calculation method he wants: "Apply your banded QA:BG corrective ratios to (1) the whole difference between BG and target; or (2) each band separately to accumulate a total." An if then else block can apply the user's chosen calculation method.

No doubt this question will need some help message to explain what is meant. Something like this, perhaps: "If you choose to apply the QA:BG corrective ratio to the whole difference, the corrective QA adjustment will be the difference between your actual BG and your target BG mutiplied by the QA:BG ratio of the band in which your actual BG falls. If you choose to apply the QA:BG corrective ratio to each band separately and accumulate a total, the corrective QA adjustment will be calculated in the same way if it is in the lowest band but differently if it is not in the lowest band. If it is in a higher band then the corrective QA adjustment will be the difference between your actual BG and the lower limit of the band in which it falls times the QA:BG ratio of that band, plus the sum of the BG adjustment needed in all lower bands times the QA:BG of each band, and this accumulated value will be your QA adjustment."

You should probably choose (1) as the default, because that is how anyone using the bands currently will already be doing it, as Novorapid demonstrates. But I would change my app settings to choose to use method (2), because I think that method (1) is unsound due to the discontinuity problem I have described.

Hi everyone, thanks for all your input so far. iOS users can now upgrade to version 1.6 from the app store, which contains this auto-calculate functionality (amongst other new bits) with both option 1 (called Simple) and 2 (banded) as described by podacris above.

As always, let me know what you think, and more importantly if there's any issues that you can see with the calculation (there's only so much testing a man can do!)