Diabetesnet.com pump algorhythm, is it any good?

Has anyone tried john walsh's new pump algorhythm on his website? It comes from some data from some boatload of pump users and has apparently a means to take your two week mean, current weight and current insulin use, to come up with ideal settings, based upon this large group of users segmenting out the ones who had good control...

we have tried it with mixed results, the basals seemed right but the carb ratios too low ( too much insulin)

he says that users adjust settings based on bogus information, not utilizing enough evidence based science, and rely on carb ratios out of habit, rather than systematic evidence.

take a look and let me know what you think?

he has some very nice presentations that support this, but obviously we'll want to talk to our endo's about it!

Our endo at vandy had not seen it yet, so he is taking a look and getting back to us!

thanks natalie

oh here is the site...

diabetesnet.com under " tools"

A few comments:

1) It's got one set of numbers for all day, which doesn't work for many of us who need
variable basal rates and corrections factor / carb ratios for different times of the day to avoid going too high or too low.

2) As a starting point for personal tuning, it is not that different from the tables in his books "Using Insulin" and "Pumping Insulin" ; they seem reasonable.

3) I consider his target avarage glucose level of 144 mg/dl (factored into his formulas if you read the study) grossly unacceptable for me. I would freak out if my average were running this high.

yep, my daughter kennedy uses around .55 u per hour at night and .85 from 0500 to 1700, so that basal in his algorhythm doesn't account for that, although the average of all these probably comes out the same for basals, I had been using that 50/50 table in his book prior and it does correlate pretty well, it seems, except the carb ratio seems to come out lower ( more insulin) in the algorhythm than it does in the table, for example kennedy uses 35 u a day, and the 50/50 ratio table has something like 12 to 14 as a probably carb ratio, but if you plug these same numbers into algorhythm, it recommends 6 for the carb ratio, I'm glad I used as a guide, being a little scared of that 6, it would have killed the kid with a hypo... she's 12 and things seem jut so erratic, what seem to be the perfect settings for 72 hours is just too much insulin for yesterday and today...
would you use an ap like that on your phone if it used that same open source algorhythm to help you adjust settings, or at least have a baseline estimate?

merry christmas!

The actual ratios vary according to, at least, the amount of excercise you get and possibly also according to ambient temperature - I need greater amounts of insulin in Winter, but that could just be because I'm more lethargic and excercise less.

If I get a lot of excercise then for the next few days I will be particularly sensitive to insulin - I need more food and/or less insulin.

The insulin is also required for digesting protein, and if you don't count protein as well as carbohydrate a change in the balance of your diet (ratio of protein to carbohydrate) will apparently change your carb factor.

So people with good control tend to lead very predictable lives.

I don't, so my Omnipod target blood sugar is set to 100mg/dl instead of 80, my wife always counts proteins in meals and I do delayed (typically 2 hour) boluses for the protein component and my blood sugars tend to vary a lot; my 90 day average is 149, but only 26% of readings are in the goal of 75-120.

My carb g/IU ratio is set to 12 and the program (Improved Practicies) suggests 15.5, my basal is set to 0.5IU.hour, except for a 2 hour slot in the morning where I boost it to 1.5IU/hour (avoiding the morning effect; it starts an hour before the time I typically get up.) The program suggest 0.6IU/hour, which actually matches my average.

My correction factor is 64mg/dl/IU, the tool suggests 65, but I think 64 is actually a little high if I'm getting reasonable excercise. Still I find over correcting is good because my blood sugar goes down faster - I can always eat after an hour or so.

The thing I find odd about the program output is that it is suggesting a smaller bolus amount than I use, and my carb count also includes an amount for proteins - so I'm deliberately over-estimating the carbs I eat. 12g/IU seems fine for me. The program also suggests my insulin sensitivity is 143%, so maybe that is where the error comes from. It may also reflect some lack of data entry on my part - I tend not to record low blood sugars or the carb correction that goes with them (I use a separate BG meter because my insurance covers only Bayer test strips.)

John Bowler

I seem to get the best results with a 60/40 basal/bolus ratio and use those tables in his book as a basis for starting my tuning.

I probably wouldn't use an app like that... I would just use the web page as a starting point and tune from there.

I would *love* to have an app that would take my CGM traces and pump logs for a couple of weeks or more and recommend fine tuning adjustments based on time of day and actual numbers for insulin, carb, and BG. I do this by eyeballing but would like to have something more data driven.

At 12 Kennedy may be having the hormone surges of early adolescence which may be complicating things...

I also find that the holidays really screw things up for me... due to the completely different schedules than typical work/school weeks and different food. Perhaps that factors into why it worked well for a while and then didn't ?

I agree with all the comments here. Basically every T1 has different insulin requirements, variable basal setting needs etc.
I would like to point at something else. You mentioned that you will run it by your endo. Frankly most endocrinologist know less then John Walsh and many CDEs about pumps, carbs etc. I am endocrinologist with T1, this is why I am personally interested, know all pumps, sensors, read T1 books etc. But there is not much in the training to prepare specialists to deal with tons of info, all the innovations etc. The best what you can do is search, learn, discuss with others and try/test. I have learned a lot from the forum, by asking questions or following discussions. Bare miimum from my formal education.
I am sure that all endocrinlogists have extensive knowledge about pathophysiology of T1, management of complications, testing needed, additional meds etc. John Walsh, his book and website is trully a great source of data.

Yep, it worked perhaps during school with high Gi foods but not so much at home with my low Gi cooking! Man, considering protein I had read about that before, but I hadn’t even considered that!!!

Thanks! Wow you are a t1 endO! I understand what you are saying, I seem to b more uP On the artificial pancreas stuf than they are! And it’s Vanderbilt for god sake! I am a pediatrician, and having my own child with t1 has really thrown me for a looP, I love this concept of an ap to create data driven decisions based upon pump info, we would need API (programming interface development ) from dex and Omnipod, (which they will never do because of their business models) and then algorhythmic predictions of improved settings… I did enjoy his poster research also!!

Artificial pancreas is very exciting. In Portland they are quite far in the project. Apparently FDA have approved this year some project letting the project advance further. I know that they are doing some inpatient testing of system. For me it is a bit difficult to imagine. I will have a hard time giving up the control and trusting insulin delivery to the device. I am fully aware that this could be the argument against pump use in the past. :-)
It is amazing how many advances in diabetes management we are enjoying, starting with Banting and, ability to check bs with strips, and now pumps, sensors, .... I am very thankful for all the time and energy people continue to put in treatment of T1.

Wow, the calculations I got were a complete change. I wouldn't use it.

My original thought was that in diabetes, science, politics etc etc there's always another "expert" with an opinion. I'd rather use the averaged opinion of many rather than just one person's opinion. Much safer that way it seems to me. I took my doctor's original settings and then called Insulet to get their suggestions. They were almost identical. And, have worked since day 1. My BG's would be in the 500's using Mr Walsh's idea of settings for me.

I’m so thankful too! I talked to a lead investigator at UVA about their ap studies they are doing 12 to 18, pod and dex, and the algorithm runs through an android phone… I think they do two days in a hotel or something like that!! I’ll dig up the pic for you… They are not sure about commercializing and the university has a hand in the equity, sO they may do tech transfer…

If you access the tool via opensourcediabtes.org:


the disclaimer actually requires you to state that:


(Scare quotes from me but ridiculous capital letters from them.) The best two physicians I've dealt with (one a T1 endocrinologist in the UK, one a simply wise family doctor in Oregon) know well enough never to tell a T1 what to do.

It's really not that we know best, it's simply that it is our life to do with as we choose.

John Bowler

Dexcom didn't work for me, I sent it back and disputed the charge (on my credit card, my insurer didn't give a damn). Not a single one of the Dexcom devices lasted the stated time, Dexcom agreed to replace every one I had, and the readings compared to blood tests showed no agreement (correlation maybe, but no agreement.)

The reports are that other CGMs are better, but they have been discontinued in the US because of patent issues.

John Bowler

I had bad experience with old DEXCOM and minimed sensor. I have been using the new Generation 4 DEXCOM since November, a huge difference. The accuracy is superb! It is such a nice surprise.

Great to hear! We just got dex four for kennedy, she has very little fat, so I’m just hoping its comfortable to wear, wish us luck!!

Have you tried ManageBGL.com? It can import CGM traces and pump logs and makes recommendations 'as they happen'.

I am looking at ManageBGL. I have seen it before but my impression was that I have to log everything in. I would love to import data form my Omnipod and DEXCOM and see how this works. Unfortunately I can't figure out how to import omnipod data in .tab format. I have been searching in Co-pilot with no results. I called the IT support, they will get back to me but I have no idea when.

DO you know how to do it? Do you use ManageBGL? If so, what is your experience? It looks that it could be used to transfer data to MD, this would be interesting.

I just looked at the web site - I see you are the creator.

I'd love to read some information on how it makes it's calculations... whether it is a straight linear interpolation based on insulin action times or whether you do some more sophisticated modeling.

In CoPilot, after selecting the patient and choosing File Menu\Export, you can use the file type drop-down to select a tab input type (.tab). Then use ManageBGL to import the .tab file. Since I designed it and started using it, my HbA1C has fallen from 7.2% to 6.5%, and my daughter's from 9.5% to 8.2%. I expect these numbers to get even better!

Unfortunately you do have to log everything in, but we have made this as simple as possible by allowing BGLs, Carbs and Insulin to be entered as they happen - as entering times and dates is always tedious.

When the pump and meter people get their act into gear we are ready - we already have an open access API so that data can be pushed into our system with just a few lines of code.
You can even retrieve your Active Insulin into your own application very easily!

ManageBGL uses the same types of calculation as a pump, using your current carbohydrate ratio, current insulin sensitivity, active insulin time and any active factors such as stress, sickness, 3 levels of exercise and more.

But instead of using a pre-calculated table of insulin remaining, it uses a far more accurate and sophisticated absorption/elimination curve which varies with insulin type, and this same curve type when applied to food also varies with different GI factors (if you choose to enter them).

We also use a correct-to-target rather than a correct-to-range scheme, the recent literature indicates that this is far preferable.

Through this, we have also been able to predict BGLs 2-3 hours ahead - including hypos.