The long-term goal is to develop effective ways to improve glycemic control in youth with
T1D. To that end, we propose to specifically evaluate the effectiveness, feasibility, and
usability of two new behavioral economic interventions in diabetes care: an automated
meal detection and patient engagement device to promote meal-time insulin dosing behavior
(AIM2DOSE) and an economic incentive (COIN2DOSE) to promote patient engagement with
mealtime insulin dosing behavior. If initially efficacious, our results will serve as the
basis for an R01 submission(s) to conduct a fully powered efficacy trial of LOAN2DOSE and
COIN2DOSE. This study is significant as omission of mealtime boluses is one of the common
reasons for suboptimal diabetes control in youth. This project is highly innovative
because it will: 1) target youth who are predicted by artificial intelligence to
experience a worsening in glycemic control, 2) use a just-in-time intervention to improve
mealtime insulin dosing behavior, and 3) use an economic incentive intervention to
improve mealtime insulin dosing behavior.
Study Design The study is an unblinded, 3-arm, randomized, controlled trial. After
successful screening and consent, individuals will wear a blinded Dexcom G6 Pro CGM
inserted 1 week after consent. . At the baseline visit (visit 1; day 1), participants
will be randomized to the LOAN2DOSE, COIN2DOSE, or control treatment arms and will be
trained on the procedures appropriate to their treatment arm. Participants will also
complete baseline measures by questionnaire. Participants will complete procedures
specific to their treatment arm weekly. Visit 2 will occur after 12 weeks (90 days).
Participants will stop any treatment interventions after this visit. A1C and
questionnaire measures will be completed, and participants will again wear the Dexcom G6
Pro CGM for 10 days (the device will be returned by mail post-visit). Participants will
receive no treatment intervention from week 13 to week 24. Participants will return for
visit 3 at week 24. Again, A1C and questionnaire measures will be completed and
participants will wear the Dexcom G6 Pro CGM for 10 days.
All study participants will also have the option to participate in a focus group
discussion (FGD) on general AI ethical issues (control group) or intervention experiences
(intervention group).
Parents of children who meet inclusion criteria may also elect to participate in a focus
group.
Study Visits Study visits may occur in-person or remotely (i.e., from home). Procedures
will be adapted to accommodate both scenarios.
Individuals who have been predicted via an artificial intelligence-intelligence based
model to experience a rise in A1C in the near future (90 days) will be approached for
recruitment. 36 participants will be recruited from the Children's Mercy Diabetes Center
(any clinic or hospital location). Based on preliminary data review of our clinic
population, there were at least 81 individuals who met inclusion criteria who were seen
in clinic in the month of May 2019 alone. Patients will be randomized to LOAN2DOSE,
COIN2DOSE, or control group using a 1:1:1 randomization scheme.
Intervention:
To identify an at-risk population for more intensive intervention opportunities, the
Children's Mercy Diabetes Center now routinely uses a validated prediction model based on
advanced machine learning (random forest method) and natural language processing to
identify individuals who are predicted experience a rise in A1C in the next 90 days. The
model analyzes all patients who presented for a diabetes visit in the prior week. To
accomplish this task, the complete health record for the CMH Diabetes Center registry is
analyzed. We will select patients from this cohort for recruitment into the present
study.
COIN2DOSE intervention: From one-week post-randomization to the 12-week study visit,
youth randomized to this treatment arm will receive personalized feedback via monetary
incentives for dosing insulin at mealtimes. We will define mealtimes based on hour of the
day and the presence of a carbohydrate entry associated with the insulin bolus. Breakfast
will be 0600-1000, lunch will be 1100-1500, and dinner will be 1600-2000. Thus, we will
reimburse youth per mealtime with at least one meal-associated (carbohydrate-associated)
insulin bolus completed. We will offer the opportunity for youth to earn a bonus
reimbursement for weeks during which they achieve at least 5 days of 3 mealtime insulin
boluses. Finally, we will pay youth per week for sharing their insulin use data at least
two times per week with the study team during the three-month treatment phase.
LOAN2DOSE intervention: From one-week post-randomization to the 12-week study visit,
youth randomized to this treatment arm will receive personalized feedback via monetary
deductions from a virtual bank of $210 for missed doses of insulin at mealtimes. We will
define mealtimes based on hour of the day and the presence of a carbohydrate entry
associated with the insulin bolus. According to the methodology for calculating BOLUS(1)
breakfast will be 0600-1000, lunch will be 1100-1500, and dinner will be 1600-2000. Thus,
we will deduct $0.50 per mealtime with at least one meal-associated
(carbohydrate-associated) insulin bolus missed (maximum -$1.50/day). Youth can also lose
an additional amount of up to $5.00/week for weeks during which they don't achieve at
least 5 days of 3 mealtime insulin boluses. Finally, we will deduct the virtual account
up to $2.00 per week for failing to share their insulin use data at least two times per
week with the study team during the three-month treatment phase (maximum deduction of
$24.00). Therefore, maximum total deductions is $210.