Glooko releases Annual Diabetes Report based on data from over 1.5 million people with diabetes in 15 countries – February 23, 2018

Glooko released its Annual Diabetes Report (single-page infographic) based on an analysis of over 8.5 billion data points collected from more than 1.5 million people with diabetes around the world. Included in the analysis are many fascinating metrics split out among 15 countries. We summarize some of the key highlights below; the report is really worth a look. Belgium had the best glycemic metrics across the board.

We’re elated to see this growing trend of real-world, de-identified glucose and device data – something Abbott has done with FreeStyle Libre (now in over 250,000 users – ATTD 2018, DTM 2017), Tidepool is starting to do with its Big Data donation project, and just-shared ATTD 2018 analyses from Dexcom’s G5/Clarity (50,000 users) and Medtronic’s MiniMed 670G (~14,000 users).

There are so many variables and answerable questions, given the vast quantity of data now available: where are the bright spots in diabetes – what distinguishes individuals and clinics seeing high time-in-range and little hypoglycemia? What can we learn from those doing well that can be expanded? Where do disparities exist and what might be driving that? What clinics are doing really well and why? What are the unmet needs? Where is decision support most needed, especially if there are patterns we can count on?

Of course, as with any Big Data analysis, there are underlying population questions to answer – are the individuals in Glooko’s data set representative of the average person with diabetes? Or do they represent those with access to technology and motivation to upload their data?

The availability of Glooko by country was not provided – nor was device adoption or type of diabetes – but we note the large data set of over 1.5 million users likely helps to wash away some of the biases (but not all). Of course, a country’s low average blood glucose value could exist because Glooko is available at savvy clinics, meaning the numbers (if anything) could understate the true global reality of hypoglycemia and hyperglycemia.

As one calibration point, the average BG in the US in Glooko’s data set (183 mg/dl, eA1c of ~8%) comes quite close to the average A1c in the T1D Exchange (~8.4%; Miller et al., Diabetes Care 2015). Of course, we assume Glooko has a sizeable number of people with type 2 diabetes, so the T1D Exchange may not be the best comparator.

We cannot wait to see more analyses like these, and we salute all companies working to publish real-world data like this.

1. Average BG Varies by Country From 157 mg/dl (Belgium) to 194 mg/dl (Australia)

Results indicated a wide disparity for average blood glucose by country, ranging from 157 mg/dl (Belgium) to 194 mg/dl (Australia). Belgium has maintained the lowest average blood glucose since 2016, while Australia dropped from the second-to-highest to now-highest average BG. No population or clinic data were provided, so it’s hard to know what is driving Belgium’s low average – is it because Glooko is only offered at great Belgium clinics? Because CGM is widely used there? Interestingly, in Glooko’s 2016 edition, Canada had the highest average blood glucose (192 mg/dl) but is now middle of the pack at 175 mg/dl – quite a sizeable change. We wonder what drives these year-over-year shifts – for instance, has the Canadian data set meaningfully expanded or changed, or does it reflect a true underlying year-over-year change in care?

2. Running on Little Data Across the Board: 4.4 Fingersticks/Day in Netherlands is the Highest of Any Country

Despite improvements in diabetes technology, glucose data remains elusive: Patients in the Netherlands had the highest blood glucose reading frequency (4.4 readings/day), while South Africa, Belgium, and the US tie for the lowest fingerstick frequency (2.7 readings/day)! These rates are for fingersticks, though they are obviously affected by CGM use – for example, Belgium has one of the lowest fingerstick rates, yet it also has the lowest average blood glucose and reimbursement for CGM. Interestingly, Canada again demonstrated a major change: In 2016, Canadians averaged 5.5 readings/day, while in 2017, Canadians took readings 3.5 times/day – is the driver a year-over-year change in the data-set (e.g., it’s now reaching more type 2s or a broader number of clinics)? Or does this reflect policy changes, such as the test strip limit in Ontario (see Dr. Irl Hirsch’s ATTD 2018 Yearbook presentation). Across the gamut, average blood glucose reading frequency could be either over-represented or under-represented: we’re not sure of the type 1 vs. type 2 split, the CGM vs. non-CGM split, or how representative the clinics are. Nevertheless, these data are a reminder of little blood glucose data most people with diabetes have – this must change! See Adam’s presentation of dQ&A data from day #1 of ATTD, which aligns with these results.

3. % of Days with hyperglycemia (>200): Ranges from 36% (Belgium) to 72% (Australia); Days with Hypo Range from 6% (Belgium) to 23% (Netherlands)

Hyperglycemia frequency (percentage of days with readings >200 mg/dl) ranged from 36% in Belgium to 72% in Australia, while hypoglycemia frequency (percentage of days with readings <70 mg/dl) varied from 6% in Belgium to 23% in the Netherlands. Interestingly, in 2016 the Netherlands had the highest rate of hyperglycemia at a whopping 60% of days. Not only has its highest frequency decreased to 36%, but the Netherlands has switched to reporting the highest hypoglycemia rate. What is driving that change? Notably, Belgium is in the best shape with respect to average glucose, hyperglycemia, and hypoglycemia around the world. What can clinics around the world learn from those in Belgium?

4. Higher Mean BG Reported on Valentine’s Day (168 mg/dl) than Halloween (158 mg/dl) and Christmas (140 mg/dl)

A bit surprisingly, out of Valentine’s Day, Halloween, Christmas, and New Year’s Eve, the highest average blood glucose was reported on Valentine’s Day (168 mg/dl) – we might have expected it to be Halloween, though given that most people with diabetes are adults, the trend makes sense. Interestingly, average blood glucose for all four holidays fell below the vast majority of countries’ overall average glucose values reported above – yes, that means holidays in most countries seem to drive better diabetes outcomes than most other days. In fact, only two countries (Belgium and Czech Republic) had average blood glucose values less than 168 mg/dl. Average blood glucose on both Christmas Day and New Year’s Eve was far lower than any averages reported by country – does this indicate more vigilance on those days? Less stress? More free time for exercise and self care? Of course, these results could also be explained by some people taking a break from checking when celebrating holidays, meaning only the most engaged are checking often on these days. September 28 was the day with the highest percentage of readings “in range” (3.9-8 mmol/l or 70-144 mg/dl ) and January 1 was the day with the lowest percentage in range.

5. Glooko User Base Rises from Nearly 1M to Over 1.5M in One Year

Glooko’s user base has almost doubled, increasing from nearly 1 million at the end of 2016 to over 1.5 million at the end of 2017. Its database size has also nearly doubled since 2016, rising from 4.3 billion data points to over 8.5 billion in just one year. As further indication of Glooko’s expanding reach, Glooko was used in 4,000 “clinics” in 2016 and is now offered by 7,000 “providers.” (We assume those metrics are comparable, but aren’t positive.) The Diasend merger and Insulet partnership are presumably helping to drive penetration – we anticipate that 2018 will be another big expansion year, as CEO Mr. Rick Altinger has hinted at a deal with a top three US payer.

Questions To Consider In Future Big Data Analyses

• By country, what is the distribution of devices and how is it trending over time? How many patients are on CGM, closed loop systems, and pumps, and what type of device? How (if at all) does device adoption contribute to average blood glucose and time-in-range?
• How do different patient cohorts compare on glycemic metrics – type of diabetes, age, fitness/activity levels, app engagement?
• What’s in the tool kit of those countries with the lowest average blood glucose and lowest frequency of hyperglycemia/hypoglycemia ? How can we learn from countries like Belgium (best in all three categories this year) and Canada, which saw a major reduction in average blood glucose?
• What’s in (or not in) the tool kit of those countries with the highest average blood glucose and high frequency hyperglycemia/hypoglycemia? Are countries like Australia reporting high blood glucose because a wider variety of clinics/providers offer Glooko as opposed to a few elite institutions? Is there something fundamentally different about the diabetes management processes or reimbursement statuses of these countries?
• What are clinics/providers with good outcomes doing to manage their patients on a population level? How do these tactics differ from clinics/providers with poorer outcomes?

-- by Maeve Serino, Brian Levine, Adam Brown, and Kelly Close