Insights from Big Data (1): Viewing of Real-Time Continuous

Glucose Monitoring Data and its Impact on Time in Range

Andrew Scott Parker, John B. Welsh, Laura J. Dunn, Annika Jimenez, Andrew K. Balo
Dexcom, Inc., San Diego, CA, USA

Background Methods
• Real-time awareness of glucose levels • The Dexcom Mobile App allows users to view current and up to
can inform short-term diabetes 24 hours of historical data from the Dexcom G5 Mobile CGM
management decisions that prevent or System.
reduce hypo- and hyperglycemia. • After a user’s consent, it also uploads glucose concentration
• Users of real-time CGM systems can and app usage data to a database, where they are available for
check glucose levels and trends retrospective analysis.
throughout the day and may benefit • Data are a convenience sample of 50,000 anonymized users
from real-time alerts. who employed an iPhone as their display device and
• Intermittently-scanned CGM devices spontaneously uploaded data in August 2017.
do not allow for real-time alerts and • Time in range (TIR) is the percentage of sensor glucose values
require a high frequency of scanning to from 70-180 mg/dL (3.9-10.0 mmol/L). Time in hypoglycemia
reduce hypoglycemia. (<54 mg/dL, <3.0 mmol/L) and hyperglycemia (>250 mg/dL,
• Screen view (SV) frequency may be a >13.9 mmol/L) were calculated similarly.
surrogate marker for patient • App usage was expressed as SV/day. Parameters for each of 20
engagement with their CGM data. equally-sized rank-ordered groups were calculated.
• We examined the relationship between • Multiple SVs within any 10-minute interval were counted as a
SVs and glycemia. single SV.

Figure 1.
The data platform makes sensor glucose
values available to patients and their
clinicians. At the heart of the data platform is
the data pipeline that moves data between a
server and distributed apps via an API. The
anonymized aggregated convenience sample
is drawn from the data pipeline for big data
insights.

Results
• Mean screen view frequency was 9.1 per day. The median (IQR) was 5.7 (2.7-11.3) per day.

22 22
15.5
64
4.94.9 Figure 3.
>250 mg/dL (>13.9 mmol/L), hours / day

20 20

Time in range 15
Hyperglycemia People with
>250 mg/dL (>13.9 mmol/L), hours / day
>250 mg/dL (>13.9 mmol/L), %

62
4.44.4
>250 mg/dL (>13.9 mmol/L), %

14.5
18 18
frequent screen
Time in Range, hours / day

60

views had
Time in Range, %

3.93.9
16 16
14
58

3.43.4
fewer sensor
Each point 14 14

56 13.5
glucose values
represents
54 2500 users. 13
12 12 2.92.9
>250 mg/dL
10 10 2.4
2.4
(>13.9 mmol/L).
52 12.5 0
0
5
5
10
10
15
15 20
20
25
25
30
30
Screen Views / Day
Screen Views / Day

50 12
0 5 10 15 20 25 30 1.4 20
Screen Views / Day 1.4 20

18 Figure 4.
1.2 Hypoglycemia 18

Even people
<54 mg/dL (<3.0 mmol/L), minutes / day

1.2
16
16
<54 mg/dL (<3.0 mmol/L), minutes / day
<54 mg/dL (<3.0 mmol/L), %

Figure 2.
1
1 14
14
with very few
<54 mg/dL (<3.0 mmol/L), %

0.8
0.8
12
12
SVs averaged
Higher screen view frequencies were 10
10
<15 min/day
0.6

strongly associated with greater TIR.

0.6 8
8

0.4 0.4 6 6
with sensor
The favorable trend was apparent between 4 4 glucose values
0.2
<54 mg/dL
0.2

all groups, even those that viewed their 2 2

screens at relatively high frequencies.

0
0
0
0 5 5 10 10
Screen
15 15
Views / Day
2020 25
25
0 0
3030 (<3.0 mmol/L).
Screen Views / Day

Conclusion
• Compared to users with fewer SVs, those with more SVs had higher proportions of sensor glucose values
in the target range and lower proportions in hyperglycemia.
• Even patients with infrequent SVs had very little clinically meaningful hypoglycemia, potentially related to
low glucose alerts.
• Screen view frequency may be a useful surrogate marker of patient engagement with their own CGM
data and diabetes management decisions.

THE 11TH INTERNATIONAL CONFERENCE ON

ABSTRACT ATTD8-0060
ATTD
ADVANCED TECHNOLOGIES & TREATMENTS
FOR DIABETES

POSTER BOARD 100 14-17 FEBRUARY, 2018 | VIENNA, AUSTRIA