Our blog

The latest thoughts and announcements from Tidepool's team.

Back

How Your Donated Data is Helping in a Hospital Setting

July 6th, 2018

I recently spoke with Brad, a clinical pharmacist and father of a daughter with type 1 diabetes about how our recent analysis of insulin pump data donated to the Tidepool Big Data Donation Project can be used in a teaching setting. Brad and I walked through our original blog post and he shared insights that meant the most to him, his students, and peers and how this data can help illustrate the difference between type 1 and type 2 diabetes, and the value of patient-centered care, particularly in a hospital.

-Christopher Snider, Community Manager


In 2016, my daughter was diagnosed with type 1 diabetes at the ripe age of 7. I read probably 200 hours of diabetes-specific literature within a few weeks of her being diagnosed – I set my sights on reading every study there was and learning every technological advancement that was happening in the world at the time. I’m a clinical pharmacist in a hospital, so having that medical background and a tenacious personality helped me better understand what was happening to my daughter.

Since my daughter’s diagnosis, people have come to me with their questions. Our diabetes educator, who’s a nurse practitioner on staff, will oftentimes come to me for collaboration on different cases. Thought-leaders in our organization invite me to glycemic control meetings because I have such a strong interest along with that special perspective of being somebody who’s a full-time caregiver and an unofficial “insulin-dosing guru”, as you inevitably become when you’re the parent of a child with diabetes.

Here’s my analysis of the data Tidepool shared and how the information on each of these charts was useful to me from a clinical pharmacist’s point of view.

Total Insulin Per Day

Tidepool Big Data Donation Project - Median Total Daily Insulin Tidepool Big Data Donation Project - Total Daily Insulin Distribution

This median total insulin usage graph was great because in the inpatient environment, we give most patients a weight-based dose of basal insulin. That means they get a total daily dose calculation based on their weight and expected insulin sensitivity. But that approach can run into problems. Look at how many patients a weight-based basal dose calculation may not properly serve based on the distribution in the 25-29 age group. A weight-based dose covers the median, but the variance for the 25-29 age group suggests that calculated dose would not work well for most people.

One thing I’ve been advocating for in my health system, is to start looking at insulin pumps and home regimens to take into account a patient’s current insulin usage, rather than just giving a weight-based, “one size fits all” dose to every patient. If you have the information, use it!

There’s another important point to consider from this chart: impact of development on insulin needs. You can see from the graph that the insulin needs tend to peak at a certain life stage and then kind of taper off and hold steady. If a health system is used to treating patients that are type 2 or that fit this kind of “holding steady” pattern shown at 30-years-old and above, and then you try to apply that information to younger age groups – or to pediatrics – you’re going to have a bad result because you’re not taking into account different independent factors that are known to exist in that demographic like how insulin to carb ratios or carbs per day can vary by age.

So when I do a consult – especially for a type 1 patient on a pump – I like to review their pump settings with them and talk to them about their management. I try to understand the level of engagement in their diabetes self-management, lifestyle habits, and eating habits. Then I see if I can individualize their insulin dosing based on what I can garner from their home dosing. And that makes a big difference.

Insulin to Carb Ratio

Tidepool Big Data Donation Project - Median Insulin to Carb Ratio  

Tidepool Big Data Donation Project - Insulin to Carb Ratio Distribution

It’s important to recognize that every patient is going to have a different need for carb coverage, and also to emphasize the importance of accurate carb counting from our dietary team. Sometimes the hospital will send out a lunch and it’ll say “60 carbs,” but you’ll look at the meal and you can see the rice on its own is 60 carbs and then everything else is another 60 carbs – the carb count is way off. In other instances, a patient may be given half their meal coverage because they only ate half of their meal. However, you look at the tray and you can see all the high carb foods have been eaten.

This is a pervasive issue in health systems. So when I’m teaching nursing staff about estimating carbs, I talk to them about the importance of accurate carb counting and how it impacts insulin dosing. Then I can use this graph with my colleagues and students to illustrate the vast amount of variability. These patients shouldn’t just get a “set it and forget it” five units with each meal and then just never reassess their blood glucose. If they’re running high after every meal, we should be increasing those meal doses and looking at the variability here.

Insulin Sensitivity, a.k.a. Correction Factor

Tidepool Big Data Donation Project - Median Insulin Sensitivity (mg/dL)  

Tidepool Big Data Donation Project - Insulin Sensitivity Distribution (mg/dL)

Correction factor is pretty new in most hospital systems. Everybody used to use a sliding scale – if they’re this high, give this much insulin, if they’re that high, give that much, and so on – and the degree of dosing difference used to be one or two units between each scale. What we’ve moved to in our system is to use a correction factor (aka ISF) based dosing calculation for corrections. It’s set to a default of 50 and you can see from the graph that this fits for most of the age groups.

There’s a high degree of variability in patients that tend to be hospitalized patients – people in their 30s and beyond. If you have a demonstrable piece of information that tells you that correction factor is wrong for a specific patient, you shouldn’t just leave it at 50.

I try to encourage colleagues and students to dial this data in, because this is something that is not likely to change between the hospital and home. We have the ability to calculate this stuff and by taking these steps, we are improving transitions of care and we’re improving blood glucose within the hospital and hopefully beyond. It’s a huge point I harp on about, so this information is great to help illustrate the importance of correction factors.

Number of Boluses Per Day

Tidepool Big Data Donation Project - Median Boluses Per Day  

Tidepool Big Data Donation Project - Boluses Per Day Distribution

I like to use this graph because it helps people to understand disease burden and some significant differences between type 1 and 2 diabetes. When you have no insulin production (or variable production as seen in early diagnosis), it requires a lot more of a hands-on approach. This can also be used to highlight the difference between multiple daily injections (MDI) and what you’re willing to do with MDI versus what you might be willing to do with insulin pump therapy. A lot of patients aren’t willing to give themselves nine injections a day. And while we have some patients who are averaging nine a day, I think most parents aren’t real keen on giving their kid that many injections.

I also think it’s useful to know we’re never giving patients seven or more injections a day in the hospital. We look at our glycemic control numbers as a population and we run these reports and we’re scratching our heads. Well, it’s not that difficult to understand more intensive management and more intensive approaches at home yield better results.

The same thing can be applied in inpatient care. When we check a blood glucose, it would be ideal to respond to that immediately. If a patient’s high, we’d prefer to give them a correction now. We know it is not ideal to wait forty-five minutes for their meal tray to be delivered and toss a little bit of extra insulin with their meal coverage. That’s not going to benefit the patient the same way. They may have climbed higher by then, they may have dropped lower – we may not  even know because we’re not re-checking – and this is common practice in inpatient care. Unfortunately, at this time it is very difficult to operationally support the more intensive approach outside of endocrinology units (usually in pediatric centers) and ICU. Perhaps this will be an area of improvement that can be undertaken in the future.

Total Carbs Per Day

Tidepool Big Data Donation Project - Median Carbs Per Day  

Tidepool Big Data Donation Project - Carbs Per Day Distribution

This was really eye-opening for me. Even though there’s a wide degree of total carbs per day in the spread here, it was interesting to see how few carbs a lot of these patients are eating. In the inpatient care environment, we have a lot of patients receiving a set number of carbs per meal. What’s common is around 40 to 60. Those patients will be limited to 120 to 180 carbs a day. If you look at the amount of carbs people are taking in adult age groups, it does kind of match with that – 130 to 150 roughly. The pubescent age groups tend to take in more and the younger kids less. But this may be specific to type 1, so it would be interesting to see this graph for people with type 2 diabetes as well.

Carbs Per Meal

Tidepool Big Data Donation Project - Median Carbs Per Meal Bolus  

Tidepool Big Data Donation Project - Carbs Per Meal Bolus Distribution

It’s good to show how much people are taking in because I don’t think that all CDE’s or dietitians really have an firm grasp of that. I’ve asked our CDE and our dietitian how many carbs is normal for a pediatric type 1 patient. Unfortunately, I don’t think this is something that is widely known. I would use this graph to demonstrate for patients how many carbs is typical for a someone living with type 1 diabetes on a pump in their age group. This is especially useful for a parent, because you can feel kind of lost in the fray. You don’t know whether to go low carb or whether you should restrict carbs or what is normal. This graph just visualizes that information instead of leaving them guessing.

Basal Rate

Tidepool Big Data Donation Project - Median Basal Rate  

Tidepool Big Data Donation Project - Basal Rate Distribution

These charts help us consider the high degree of variability between two patients in the same age group who take different amounts of basal insulin, for us in the hospital this equates to how much long-acting insulin we give a patient. It’s all connected – variability, individualizing insulin dosing, taking the patient’s home doses into account, and their individual factors can make a difference in the quality of care they receive.

Basal Insulin Ratio

Tidepool Big Data Donation Project - Median Basal Insulin Ratio  

Tidepool Big Data Donation Project - Basal Insulin Ratio Distribution

I talked to a student the other day, about how in the beginning of type 1 diabetes disease progression, you talk to the patient about the background of how the disease presents, how it progresses through the partial remission stage, and the background insulin production that occurs and sometimes persists until the older ages depending on the age of diagnosis and the individual. That’s a nuance that’s not seen in other diabetes cases and it’s important to point out. Especially if they’re accustomed to the “one size fits all” approach – as I’ve explained, this doesn’t work for many patients. Healthcare providers are taught that patients will require about 50% of their total daily insulin as basal insulin, so they’ll be calculating it that way.

I’m always advocating for improving our diabetes management and trying to find tangible means to illustrate points about how we manage patients, especially in the hospital. For certain events like time delays between treatments, timing of meals, and individualization of insulin dosing, there are plenty of areas we can improve. A little bit of teaching could really go a long way.

With Tidepool’s blog post about insulin pump data, I have been able to illustrate these teaching points with some of my students and colleagues and emphasize both the differences between type 1 and type 2 diabetes and why we should do some things differently.
-Brad


Learn more about the Tidepool Big Data Donation Project and how your data donation can support diabetes research, innovation, and other diabetes nonprofit organizations.

One Comment

  1. Larry Martin

    I have finally got to see an Endo after 3 months of using a Dexcom and discovering post meal extreme highs that I had no idea were happening since my A1C is 5.7. I would really like to see of you can explore something as I am playing around with Carb ratios, basal rates, and sensitivity. I wonder if you can show a correlation between the need for carb ratios to be lower (2.5 for example) and the need then to lower basal rates past 2 hours or so after a meal. I am a firm believer, altho not convinced my Endo yet, that Novolog and Humalog are lasting for 4-5 hours when in our minds they should start within 15 mins and not last more than 2-3.

Leave a Reply

Your email address will not be published. Required fields are marked *