Back to manual mode --or, bye bye openAPS and looping (for now)

For reasons not under our control (PM me to ask what happened in more detail), we have stopped using openAPS for the past week and are back to manually pumping and giving corrections (essentially, using sugar surfing) instead of looping. I thought this would be a disaster but surprisingly, his blood sugar has actually been MUCH better and I’ve enjoyed a ton about it. The past 7 days his average BG was 116 with an SD of 40; the prior week with openAPS running it was 139 with an SD of 55 – so same number of lows basically but with many more highs.

Here are some of the unexpected benefits of manual-mode pumping:

a) Without seeing his IOB and without the potential to give him remote boluses, I am simply less tempted to make lots of little tweaks during the day, meaning I’m paying a little less attention to his BG (I still get low alerts, but that’s it). This has been so good for my mental state; just feeling like my brain was less cluttered.

b) Without openAPS (and in particular, autotune and autosens) running, I can see very very clearly which parameters are off. Before, with so many factors changing all the time it was really difficult to deduce which things were wrong or would make the most difference. We also had certain parameters that we knew were not optimal for manual mode ,but we kept them because they seemed to work best with looping.

  • For instance, we had his basal rate set to about 5 units per day, knowing this was too high most days but trusting that openAPS could ramp down from this higher rate when it was not needed and then ramp up on the days when his insulin resistance went way up. But this was more of a hack to get openAPS to work best for things like covering meals. For instance, openAPS will not issue microboluses for meals if your max basal rate in the day is lower than 0.3 units per hour, so we had periods with this rate.

But it’s become glaringly obvious that this basal rate is simply too high. My guess is that he’ll settle in around a basal rate of about 4 units per day; right now it’s at 4.3 units per day. Once we truly isolate a better basal rate, we can work on tweaking factors like sensitivity and carb factors. My hope is that we can somehow find a set of parameters that both work in manual mode AND for openAPS

c) THE BIGGIE: openAPS allowed us to be sloppy when bolusing for food because it did a good job of preventing spikes into the 300s. Instead of just bolusing for a whole meal upfront, we would bolus for half knowing that openAPS could pick up the slack by issuing super-microboluses. But often times this led to persistent highs or highs followed by lows. Now that we know we can’t go back and “revise” our meal boluses, we have become much more attentive. And not surprisingly the results have been much better!!! This also gives us confidence that our carb ratios aren’t totally off.

d) TEMP BASALS. openAPS issues temp basals to moderate insulin levels, and the algorithm updates every minute or two. So, you can’t manually set a temp basal for foods that are slow digesting, because openAPS would simply overwrite that temp basal a minute later.

But some foods really do need a slow, sustained, higher basal rate over a period of hours. The hacks that openAPS uses to mimic temp basals never worked well for us. For instance, you are told to set a temporary low target (such as 80 mg/DL) or log extra “phantom” carbs for foods like pizza 2, 4 or 5 hours after eating, and that should prompt the algorithm to pile on extra insulin. But it usually didn’t ramp up quite enough or at the right times; we needed to log something ridiculous like 300 g of carbs to get the desired effect! Now, I can just set a temp basal for 0.6 units per hour for 6 hours, and Samson’s BGs have been perfectly flat or with just a slight rise after eating 130 g of carbs!! The constant temp basals just seem to work better for pizza, cake and other fatty foods.

Which brings me to my last benefit

e) Flatter profiles with less swings. When openAPS predicts a low, it often “zero temps” a person for an hour or two after meals, effectively turning most boluses into super boluses. But we often had a problem with Samson being flat after a meal, only to rise an hour or two later. I now suspect hat is because openAPS was subtracting some of his meal insulin soon after he ate, predicting a low. Now we have much flatter landings to normal. Also, when he got low treatments, it was usually after he’d been zero temped for a while… meaning that after he ate the low treatment he usually spiked much higher… which meant openAPS predicted a high… which meant he got a pile of insulin after a low treatment. Unfortunately, fast-acting carbs are fast-in, fast-out for him and take less insulin than the equivalent number of carbs in a solid meal, so those quick spikes led to yo-yoing highs and lows with tons of low treatments. Without the zero temping before a low, he isn’t spiking as high. And without the automating high temping after the spike, he isn’t yo-yoing.

In general, it seems like Samson’s BGs have seen a lot less wandering – even without the highs and lows, when he’s flat he’s flatter than he was with openAPS. I think all the turning on and off of temp basals had some unpredictable effectsin him because he uses such low basal rates, and so it was creating a less flat insulin delivery profile than was optimal for him… At least, that’s my impression

The downsides:

NIGHTTIME: We currently have Samson’s night profile set to 0.075 units/hour, and last night I had to give him 3 carb treatments to keep him in range. But a nighttime rate of 0.05 units/hr has him rising steadily. Other nights he’s been high for several hours and there is nothing gently bringing him down. openAPS really did well at nights at keeping him in range. I suspect the optimal nighttime dose is something like 0.065 units/hr, which is not attainable on the pump but which openAPS can pretty naturally accomplish.

SICK DAYS: So far, Samson has not been sick so his daily use of insulin is on the lower end of his habitual use - maybe 7 to 9 units. But I have no idea how manual mode will fare when he all of a sudden starts needing more insulin. When he’s sick he can use 14 units easily. I suspect I’ll have to go to daycare a lot more when he comes down with a cold.

MEAL SPIKES: Right now Samson’s basal rate is clearly too high, which means that there’s always something pulling his BG down if we, say, bolus for his food after a meal. But I wonder if we’ll have to start doing things like pre-bolusing or risk seeing much higher spikes once his basal rate is dialed in properly. It also means that when we miscalculate in a restaurant, we have to step in much sooner to prevent that 300 mg/DL spike that lasts for hours.

I’m not sure if the problems I noted with openAPS are truly intractable, or are due to the fact that Samson uses so little insulin each day and so the openAPS safety settings, etc. just don’t work well (in which case, it might make sense to dilute insulin). Or it may be that openAPS simply exacerbated poor tendencies on our part (like sloppy bolusing).

in any case, we will probably set up a new rig for him in a few days…but I’m feeling a bit ambivalent. I’d love to get my nighttime sleep back but other than that, life has been a lot simpler without all the extra fiddling with knobs on openAPS. If we do get another rig up and running, perhaps I should create a thread about reevaluating all his settings and troubleshooting it anew…


Can you turn this system on just at night? It seems like this type of system would be incredibly helpful at night, but seems to have drawbacks during the day (per your post- I have no experience with this system at all).

1 Like

WE definitely could do that.

The teachers at daycare, though, feel much more comfortable with it running – it means they don’t have to give insulin corrections during the day because the program is running. And no one there knows how to count carbs or even use the bolus wizard on the pump…and giving more than 2 doses of insulin a day (for lunch and snack) seems to be pushing them past their comfort level.

I am hoping we can have a solid three weeks without it running during the day so I can get all of his parameters set to a level I truly think is correct before we update them.


Ah, that makes sense. As a parent, I bet it’s comforting to run the system while he’s at daycare too.


@TiaG, your post is a fascinating one, and opens up a lot of questions. I will mull it over and think through it carefully first.

As an aside, like @Katers87, I see huge benefits to Looping at night: that’s primarily why we are gearing up for it. And for college, too, of course. But I look at you as a pioneer, so your post is full of reflection for me. Thanks so much for sharing it, and for discussing all your thoughts. Tremendously valuable.


Agreed. I should say, however, that I do not think Loop or any other presently available closed-loop system would be able to very successfully deal with extreme variability associated with growth hormones or such. Out-of-nowhere rapid spikes or substantial day-to-day basal swings are simply way beyond the models currently assumed by closed-loop systems. But, by college time, those extreme variability times should be behind, I hope.


@TiaG, thank you for the very thorough write up on your experience with OpenAPS. I haven’t looked much into the program since we are getting acceptable (for us) results without it. Your write-up confirmed that while there are parts of the equation that are easier with the technology, you are still spending time fiddling, and tricking the system to get the results that you want.

It seems that the best of both worlds would be a system that would go automatic (easy, like one button push) between meals or at night, when management is something of a hassle, but that lets you take control during meals, or when you have the time and want to surf during the day.

Clearly, none of the systems are here yet, but it doesn’t seem like it will be long. I could see us easily getting caught up in trying to understand the inputs and results without getting the benefit of the system.


In Loop, going open-loop or closed-loop is as easy as sliding a switch in the app. I just keep it closed-loop all the time, but one can manually surf or do however one wants manually if so desired.


Simply incredible. It’s this kind of insight that someone would be hard pressed to find on other platforms. @TiaG, the way you lay out all the facts with pros and cons is truly appreciated. I absolutely love reading everything you post. I have internally debated if I should abandon Omnipod and try OpenAPS. I suppose no system works 100% and doesn’t work the same for everyone.


I actually think it’s pretty easy to switch between auto and manual with openAPS – if you are controlling your own rig. All our problems stem from the fact that we are away from Samson (either in a separate room, or a separate location) for much of the day, yet trying to alter his insulin profile remotely. But if Samson was old enough to turn his program on and off himself, I think a lot of the hassle and fiddling with openAPS could be circumvented.


Not directly attainable via the basal settings, but have you heard of the trick of having the basal switch between 0.075 and 0.05 every hour or so to average to the value you want?


We set that up last night :slight_smile:. Well, not exactly every hour; we are doing two to three hour stretches oscillating between 0.05 mg/DL and 0.075 mg/DL, based on when he is likeliest to crash. But the same idea as you describe. He was pretty flat last night but it’s hard to say if the new settings are right: His sensor is also on the fritz. For instance he was 73 last night and I tested and he was 171 and 148 mg/DL on two different finger sticks. So it’s still a little too early to see if that works.


UPDATE: Well so far this week has worked less well. He was running low for the first half, and then on Wednesday we did an unintentional basal test in the morning because he refused to eat anything for breakfast –
and it was clear that his basal rate at that point, 0.225 units per hour, was too high.

So this morning I lowered his rate to 0.2 units/hr, which means it is now quite a bit lower than what it was with openAPS last week: 0.35 units per hour. But wouldn’t you know it, it seems we’re entering period of higher insulin need; this morning he started high and spiked to 300 from a 15g carb breakfast; I needed to give him a 0.8 unit correction to bring him in range in time for lunch. So either this basal rate is too low, or it could be that with his lower basal rate he now needs a much, much higher carb ratio.

For the nights, we’ve split it into about half 0.075 units/hr and half 0.05 units per hour. But this hasn’t been working so well either: Samson was was low, low low the first half of the night (basal 0.075 units/hr), then his sensor died, he switched into the 0.05 units/hr mode and when we calibrated in the morning he was 299 mg/DL.

So this was a much less stellar performance compared to the norm for openAPS. Clearly when the settings aren’t quite right it is much more obvious in manual mode than with openAPS and it requires a lot more intervention.


We alternate between .10 and .05 all night. We have

  • .10 from 12AM - 3AM,
  • .05 from 3 - 4AM,
  • .10 again from 4 - 7AM,
  • .05 from 7 - 8AM
  • .10 from 8 - 9AM
  • .05 from 9 - 10AM
  • And our 10AM is the biggie because he usually wakes up late and eats breakfast late (between 1030 - 1100). We have .20 from 10AM - 1PM.

Having said that, though, Liams nights are sporadic…this profile works perfectly one night and can keep a perfect flat line of around 110 - 120, but the next night he may be a roller coaster (up over 200 and down under 70). I find myself turning on the custom temp basal ‘off’ for 2 or 3 hours at a time sometimes. It’s so hard to iron down specific basal rates for Liam. It’s easier for us to just make the “on the fly” changes as they occur rather than trying to change his basal rates. Inevitably, with us…there is no “good” setting that is permanent. What works tonight, and possibly for a week, is going to have problems tomorrow, or next week. We just roll with the punches and change as necessary, on the fly.


While our basal settings are simpler because we aren’t trying to deliver such a small dose, I can tell you our nights are just like you describe. No setting is truly right, and its a guess if he is going to have a growth hormone dump and go high.


I can attest to the same with our son. I am afraid that you might need to reach your 20s before things stabilize more :frowning:

1 Like

SECOND UPDATAE: So, am really ready to go back to openAPS… but we’re hoping we can get his settings fine-tuned before we do so.
He’s been low, low low for much of the day – despite us dropping his basal rate down to 3.4 units per day up from his usual of 5.6. So we have halved the baseline basal rate, and for every meal we are delivering less than half of the usual insulin dose – and yet he’s crashing constantly. Last night my husband had to give him 10 low treatments!!! Our nights are horrible, as well.
I’m pretty sure the basal rate is a little too high still – but I am also wondering about our carb ratios and sensitivity factors. Now that I am looking at the data, I’m realizing that openAPS zero temped Samson for hours of the day…so our settings were essentially designed to have every meal be a superbolus. So we may have to increase our carb ratios and weaken our ISFs.

Then again, he’s also using much less insulin per day than was typical before; he’s running between 5 and 7 units this past week, so perhaps he’s about to get a tummy bug as well. We shall see.

His current basal rate is now 0.175 units per hour all day, and then mostly 0.05 units per hour at night.


Do you expect that you would be able to get better results from OpenAPS once you get him reasonably stable without it?

This would be a happy prospect.

That is really interesting. I would never have guessed. When you look at the resulting control, do you feel that the algortihim is actually doing the right thing there?

That is really interesting. I would never have guessed. When you look at the resulting control, do you feel that the algortihim is actually doing the right thing there?

I’m not sure. I suspect the algorithm was doing the best it could with not-so-optimal parameters. But on the other hand it’s possible that once we get to his optimal dose for the average day (say 7 units daily) the algorithm will be completely unable to ramp up to the levels necessary for sick days, high carb counts and other higher needs.

I’ve definitely noticed that other people seem to have flatter profiles using openAPS, and this is my hope for the eventuality when we fine tune the parameters

UPDATE: Our latest A1C is 6.7 – Up 0.2 from his past A1C, but not too bad considering that we just spent a week in Tahoe with terrible control and a bunch of super highs, which is likely overrepresented in his numbers. Then again, his A1C likely includes a lot longer periods of low BG at night.

Our doctor also mentioned a few things she has noticed that confirm our suspicions – she now has a few patients on openAPS, and she has found that their basal rates need to be significantly dialed up in order for the algorithm to work properly. Which makes me think that you can’t actually translate between the two types of basal.
If you can’t directly translate between your manual and automatic parameters, that makes me wonder whether an algorithm that’s completely untethered from these traditional parameters (ISF, IC, Basal Rate) might actually be better for people, on the whole.