New meal algorithm

Ahhh - yes.

I get early rise, late fall.
I use the 3 hour taco exclusively. That used to work great in July with the retroactive correction on. I don’t eat protein. High carb exclusively.

I have increased bolus and basal to the maximum, now. So, I’m just crashing all the time. This could easily be improved, but I’m exploring the meal algorithm. I notice that I am taking significantly more insulin on Loop than on traditional Omnipod (which I was using recently).

I don’t pre-bolus. Never have. Could try.

Last install was literally hours before they posted Loop 2.0, so I don’t know if I have retroactive correction in my implementation. But, I’ll have it in Loop 2.0, so I’ll probably update again. I :hearts: retroactive correction.

I’m gonna have to take the system down to do some physical stuff, but I’m open to starting her up again in a few days. I’m getting pretty physically run down. I’m not eating like I used to and need to sleep a lot.

Loop is more exacting than omnipod pdm. High carb requires prebolus. That’s just how the mechanics work if you want to avoid a significant rise (150+).

Taco will short your bolus. Play with absorption times, using the emojis as a guide. Most food is 2 hr. That will give about 100% bolus unless you are trending down below the suspend threshold.

Prebolusing is a game changer. That will help eliminate the rise as well.

Not sure what retroactive correction is. Retrospective correction has been a part of Loop for a long time. Integral Retrospective Correction (IRC) is experimental and was part of the Jojo branch. It would have helped compensate a little for the fast early rise. I’m hopeful micro/auto boluses will be added in the future and the need for IRC will be gone.

From what you have shared, seems like carb ratio can come down, use 2 hr more often for complete boluses. Best results will include prebolus.

If you still find the boluses are not complete, I have been reducing the absorption time overrun setting that makes the prediction more aggressive with what it expects carbs to do. Then it doesn’t short the boluses because it expects faster (and closer to real life) hit from carbs.

Hope that helps.

It used to work in July. Maybe I am just moving at too great of a rate of change.
I think I’m getting too physically run down. I have put into place pretty serious lifestyle limitations to accommodate. But, not eating as much makes me tired and I sleep a lot.
I’m having trouble reading and thinking. Having trouble getting up the energy to walk the dog.

Let me do a little Omnipod vacation and then I’ll talk to you again before I install 2.0.
I might just need a little time to get my strength back. I’ll get back to where I can do physical things.

I appreciate the feedback. You make good suggestions. I’ll start from scratch in a couple weeks.

What happened to @BradP? You still out there somewhere, Brad?

I don’t understand how you can eat no protein, carbs only. What about vegetables? And do you have no milk in your coffee? Maybe this is why you are so very tired lately. Just a thought.

1 Like

@mohe0001 What are you trying to do? Obviously something is not working right as you are so tired. Even cutting way back on eating you shouldn’t be so tired. And to keep crashing isn’t good either. So what’s up?

Hehehe @Pastel, don’t fret. I’m vegetarian. So, I’m just trying to be succinct and say that I’m a ‘big’ eater, lots of 120g meals.

@Marie20, Its a very complicated bolus calculator. I had a way to successfully operate it in July, but haven’t found one since October. The algorithm changes. This is a lot of trial and error. I think that not eating enough makes most people tired. At some point, eating becomes not worth it, if the system falls out of whack. I used to be able to run the system open loop (without any system automation) and still be able to see increased system stability because the meal bolus worked so well.

Here’s today’s sentiment about idea of whats going on (this changes) - I’ll run this past you and see if it jives. So, if you set one part of the system settings to make something work smoothly, then it impacts other parts and those parts fail. So, its taking a lot of effort to keep things running within a survivable range. There’s not as much flexibility because all parts of the system are sorta dependent on each other. Its like this:

OK, I think I know how to explain one of my main difficulties. Here goes…Its a simple question, now that I thought through it a bit.

In Omnipod,

I might want to deliver 12 units, with 85% delivered upfront, and the remainder delivered over the next hour and a half.

In Loop, what is that?

Maybe its something like,

Candy button/apple button/lemon button/watermelon button/cookie button/etc (one of the 2 hour absorption buttons), (12/2)*15 gram input = 90 gram input. But, how that last 15% gets delivered is what I really fail to understand and that makes a big difference.

What I need is a Omnipod --> Loop translator.

image

1 Like

I think I got it.

This is the math that I need to be able to do on the fly.
I need a calculator. I should never have been trying to do this in my head.
That was a mistake.

I’m not a looper and can’t help with your problem, but I have to applaud your meme use here.

After reading this:

That is exactly how I felt!

Turns out that those buttons are all, essentially, the same.

1 Like

I see it. I see what Loop is dealing with and why its struggling. (At least, I’m 80% sure.)
I can see whats going on when I run Omnipod.

I think that what I need to know is where Loop pulls its correction insulin from.
Is that max bolus? Is that where it pulls to correct on meals?
I think that setting needs to be increased quite a bit to deal with the ‘off days.’

I need to run some tests on sensor accuracy, just to be prudent, but I think that if I increase that upper limit (well beyond where I ever dreamed it might need to be), this will allow the meal algorithm an opportunity to work as intended. Meal algo needs to be able to deliver at least another 10 units.

Basal increase isn’t gonna cut it. I’m not taking a unit or two. I might have to increase that setting in small increments, maybe 10 units total…maybe more.

I’ve been mainlining insulin via Omnipod all day and that barley kept the system in check. I think this is the result of seasonal change.

Loop doses based on your settings. Around meals, it is mostly carb ratio. Absorption time determines how much it will recommend up front.

No matter what, you can enter whatever bolus amount you want. Loop will deliver the rest via temp basal changes unless the prediction shows you don’t need it (trending down too fast, for example).

Loop also has some correction logic when your BGs don’t match prediction values, so it will give some minor bumps if you are going high or cut basal if going lower than expected. Its not much, but its something.

Max basal rate just limits how much loop is allowed to give. In theory, if your settings are accurate, you shouldn’t need a max basal rate limit as loop will give what you need when you need it. Its for safety, not part of the algorithm.

That said, if you have a high max basal rate, give 80% of your bolus upfront for a 2hr entry, loop will ramp up the temp basal rate as high as it takes to deliver that last 20% over 30 minutes, unless capped by the max basal rate. So, if the last 20% is 2 units, it will pick a temp basal rate of [scheduled basal] + 1 u/hr = enacted temp basal rate.

If you are trending lower or higher, loop will modify the temp basal rate up or down from there.

Take a read of the algorithm pages. Start here and click the NEXT button to go to the next pages. Probably will need to skim it and then read it through a few times.
https://loopkit.github.io/loopdocs/operation/algorithm/overview/

For carb entries, you can use 1 entry for all your fast food. If you want to use pictures, you can add more than 1 to an entry. If you select a picture, it will change the absorption time to match the one you entered. For best results, enter fast/medium/slow entries separately and set the time for when you are actually eating.

1 Like

I see a lot of meal dosing variability. Always have, I think that’s been my problem with the algorithm. I think I overcome that by allowing Loop to supplement with max basal.

My max basal is 10 u/hr, but I think that needs to go up to 20 units/hr so Loop can do what it do. This isn’t unreasonable, I do 15 unit basal supplements in Omnipod sometimes. I am being too cautious. Its time to step things up a bit.

I gotta give Loop the tools to work.

I’m scared to increase max bolus any further (14 units) because I see so much sensor inaccuracy…but I’m working on that.