That is exactly how it works. Before O5 was released there was a Loop and Learn interview with some of the clinicians that conducted the clinical trials, who described how the algorithm works. Everyone gets the same formula. Hourly basal equals 1/48 the average total daily dose (aTDD) of the previous pod ( aTDD ÷ 24 hours * ½). It tries to adjust it from there based on 60 minute predictions.
So yeah, you could fake the result with a massive bolus, but that would waste so much insulin from a small pump.
I have big basal needs after nearly 4 decades, no detectable c-peptide, and midlife hormonal craziness, but don’t eat very much. That 50/50 split did not work for me at all. It was brutal. Couldn’t get the basal I needed to save my life, literally.
I was advised to factory reset the PDM (before the apps) between sessions. It’s basically an Android phone so you can access the boot menu during a restart by holding down the power button and one of the volume keys at the same time. (I don’t remember if it was the up or down button.) It’ll pop up a menu and factory reset was one of the options. When you set up the PDM, it asks you for your Total Daily Dose of insulin, because it has no data on you yet. If you program a TDD that is 48 times your average basal rate, you’ll get a better starting point. But what a pain in the butt… That’s why I went back to Tandem.
I have no experience with the O5 app, but it might be a lot easier to start fresh with it. Clear all the data/cache? Maybe uninstall/reinstall?
Yes, but whatever remains in the pod when it is removed is just going to be wasted anyway. (Unless someone is removing the insulin with a syringe and putting it into their next pod.)
So whatever remains in the pod when you remove it, rather than just throwing it out, you could at least do a fake bolus it and get some “use” out of it.
But yes, these are not ideal solutions. Would be much simpler to just tell the pod what your baseline basal rates are! Come on Insulet, we are not all 50%!
FWIW, in my approx six weeks on O5, it has never once come close to a 50/50 split. It may try to, but it doesn’t. Which is fine with me, because my split was never close to 50/50 before, either.
If you want to increase basal adjustments, I’d opt for adjusting settings rather than tricking the algorithm by dumping insulin. Unfortunately I made my algo too aggressive, and now I’m working on talking it down.
The Insulet algorithm, which seems amazingly close to the AAPS “Dynamic ISF” algorithm, is to deliver 50% of the total daily insulin requirement pre-emptively but gradually over the period of time.
When this doesn’t work BG either rises or falls and the closed loop kicks in and overrides the algorithm.
Insulet also uses the OpenAPS (or maybe AAPS) innovation (sadly not patented) of “micro boluses”; AAPS refers to these as “SMB”'s, “Super Micro Bolus” (they do love their TLAs!) These allow rapid, but incremental, adjustments to IOB to safely deliver insulin quickly; a fundamental condition of OpenAPS design.
In practice the test of the 50% solution is whether we go low; not whether we hit LGS (I do that quite often) but whether we, in fact, go low.
SMBs combined with pre-emptive delivery of insulin together solve many of the issues we face as T1s. Works for me but of course YDMMV.
I’ve never had it deliver, with consistency, more than 30% of my tdd as basal… which has been incredibly ineffective. I either have to run it in manual with a much stronger basal profile than it’ll automatically deliver, or I’ve recently discovered that if I deliver about 20% of my tdd separately with tresiba it picks up some of its slack. It simply makes nowhere near any sort of robust attempt to deliver half the daily dose as basal. I can take photos of my history on the device at some point to illustrate this…
Ah. So the reported algo is not, in fact, what happens; it’s a lot more complicated.
Can you run “autotune” on your data?
That’s the latest docs which is not helpful in this case but gives some background. Here’s “autotuneweb” which is the original implementation enclouded:
It does depend on NightScout, but that’s a part of TidePool (though I’ve yet to work out how to get them to communicate!)
Per the great Clinical Implementation guide Insulet wrote for the Omnipod 5 it uses a Model Predictive Control algorithm. I spent a couple hours googling and reading papers about it and now I know that I don’t know enough to try to explain it. Also I understand why 6 years after release there is no simple explainer. I took a look at the math in Model Predictive Control of Type 1 Diabetes: An in Silico Trial - PMC which was a part of the JDRF project to connect pumps to CGMs. No where in there is the basal/bolus ratio. I think Sam is right.
The FDA approved the Omnipod 5 for use by T2Ds in Aug 2024. Putting some long acting insulin in plus the pump is kinda sorta a little like being a T2D.
