Omnipod 5 Live Brain Stream

I have been jotting down notes about Omnipod 5 every time a new thought hits me. This feels a bit like blindly splattering paint all over my brain and I've been unable to shift into reverse to turn my scrambled thoughts into written prose that is easily understandable.

So today, you get a live stream from my brain with all of my pent up musings on Omnipod 5.

Expectations

Omnipod 5 may not be the right fit for everyone. I've found it works best when you accept variations in average BG between days instead of constantly beating yourself up to get to 5.x A1C. But we have seen the worst days on Omnipod 5 are much improved over the worst days on Loop.

Overall Comparisons

I didn't know what to expect with Omnipod 5 versus Loop but after 80 days, you wouldn't be able to distinguish between the graphs. A1C, TIR, Std Dev are all nearly identical. But I'm making a new trackable statistic called MFPB, short for Mental FPhysical Burden (or another acronym of your choosing). Our combined MFPB is down by more than 90% and that's the biggest win of Omnipod 5. While Loop was fabulous for 3.5 years, it takes much more micro-managing to keep on track. On the flip side, I've seen Omnipod 5 do much worse when trying to micro-manage it.

Myth Busting

I was fortunate to be able to ask Dr. Trang Ly, Insulet CMO, some questions to confirm a few misunderstandings that I've been seeing posted a lot.

Myth 1: "It takes months to learn your body". 
O5 is not a learning algorithm in the sense of AI or anything like that. The entire algorithm is based on Total Daily Insulin with no variations for day and night, highs or lows at the same times of day, etc. Since most people I've heard mention this talk about O5 leaving them too high, I'll base this tip on that.  The only reason it would need to take a while to "learn" you is that TDI is nowhere near where it needs to be. You have either started pod 1 with too low of basal or you need to make your carb ratio and/or correction factor more aggressive. If your settings aren't aggressive enough, you are increasing TDI very slowly between pods so it takes a long time to increase the average TDI.

Myth 2: "Change setting [fill in the blank]"
I have seen people say that the algorithm worked better by changing DIA, max basal, or a number of other settings. To be clear, there is exactly 1 setting that affect the automation: Target Glucose.

• Automated mode uses its own decay curve and DIA to calculate IOB and determine dosing.
• The system calculates its own adaptive basal rate that is TDI/48.
• After the first pod, the maximum automated dose it will give you is 4x the adaptive basal rate.

Tips

And now I proceed to the point where I still have half of my random notes and have lost the ability to cultivate them properly. So in no apparent order, here are my parting thoughts.

• If you are worried about variations (sickness, cycle, etc), O5 can handle a huge range of TDI. My daughter's lowest TDI was 25 and the highest was 77.
• Glooko has been a great replacement for Nightscout except it is not real-time like Loop. Embrace this lack of information and let the system do its job because this will improve your MFPB.
• Glooko also syncs all of your O5 settings. Just check under Devices and then pick the Omnipod 5 item.
• If you want to pull the O5 carbs and boluses into Nightscout, it can be done with about a 2 hour delay. You can check out my Github repo for the Glooko Bridge that I'm using.
• I do wish it allowed profiles for carb ratio and correction factor to easily switch weekdays, weekends, etc. Since it does not, we actually use a single ratio and correction factor to keep things simple for when they change.
• If you want to know what O5 thinks your basal is, divide TDI by 48, voila!
• Hey Loopers! Carbs are irrelevant! We never enter low carbs, there is no adjustment of past carbs, it just doesn't care! In fact, there are times where she just doses an amount without using the calculator. Remember back to MDI and dosing for an egg? You just knew it would take half a unit so you dosed and didn't worry about dividing the protein by the square root of the fat times the phase of the moon.
• Carb ratio most likely needs to be drastically more aggressive with Omnipod 5. Hers changed from 1:6.5-7.5 down to 1:5.
• Remember, you can use decimals in carb ratios. I have changed by as little as 0.1. 
• I have no idea what her actual correction factor is but 60-70 seems to work well at meals when she uses the calculator. For actual corrections, she usually just doses what she knows she needs and it's close enough for the algorithm to handle it.
• The last bullet brings up a very important point. You are allowed to enter whatever you need in the insulin field for a bolus. The calculator doesn't know if you just ran a marathon or had a 5 hour Stranger Things binge with a bag of potato chips. It also doesn't know if your correction factor or carb ratio settings are incorrect. So if you are high and it is not recommending a dose, you may need to pull out your slide rule and figure it out on your own.
• IOB is complicated. Manual doses use the user specified DIA, automated doses use the built-in decay curve DIA and amount over the calculated adaptive basal rate. So the IOB shown to you is a mashup of these 2 that is inherently difficult to untangle.
• Speaking of DIA. There is a chart below on why 4 hours is what you should be using. The green straight line decay at 4 hours most closely aligns to the actual Novolog/Humalog non-linear decay curve over the bulk of insulin action. For Loopers accustomed to 6 hours, the blue line shows why this doesn't work well on Omnipod 5.
• We found it works better to give smaller corrections more frequently rather than waiting for a high and giving a larger one. Remember, it isn't stacking if you actually need the insulin.
• It has been helpful to me to determine a simple math correction factor based on trial and error. Her correction factor is somewhere 60-70, I think, maybe, but possibly not. That was also based on a testing with a basal rate that we setup which is no longer valid. So rather than rely on the calculator that may not be accurate or trying to figure out how much it has already dosed over what it thinks she needed for basal, I realized that 100 for manual corrections worked well. This made the math quick and easy: she needs to come down about 25, so that's 0.25 units. And I realized the higher value allowed some wiggle room for what the algorithm might have already dosed on its own.
• For Loopers that wonder about low overrides, use Activity Mode to replace these for extended low periods.

Last But Not Least: Michael Faraday

A day doesn't go by without someone posting that the new pod won't connect to Dexcom. This happens because the last pod did not disconnect from the transmitter correctly. Remember, Dexcom can only have 1 receiver connected at a time. No, you don't need to restart your controller 100 times to fix it. No, you don't need to delete the transmitter and re-add it. When you take off the old pod and before activating the new one, wrap it in foil and put it in the microwave. This blocks the signal from the old pod to the transmitter and allows the new pod to connect reliably, every time.

r