Jon Moon1, PhD Senior Member, IEEE, Jared Sieling1 Member, IEEE, Katia Francelino-
Tomita1 Member, IEEE, Tyler Schenk1, Budi Wibowo2 and Susan J. Woolford2, MD
1MEI Research, Ltd., Edina, MN • 2University of Michigan, Ann Arbor, MI, USA
Conclusions
Adolescents develop food
preferences and eating behaviors
that can last well into adulthood.
Impacting these factors may
improve health outcomes both in
the short and long term. This
project is the first ever attempt to
reach adolescents at times when
they chose to eat with actionable
and appropriately tailored food
recommendations at eating venues
they selected. Further work should
test the intervention with more
participants and for longer periods
to determine whether persistent
changes in behavior occur and
whether these result in secondary
outcomes such as change in body
mass index (BMI).
Acknowledgement, Disclaimer and Contact
Jon Moon • jmoon@meinergy.com • +1 (952) 373-1636 • www.PiLRHealth.com
Delivering Individualized Food Messages by Location
Presented EMBC Milan, Italy August 2015
This work was supported by the US National Institute on Minority Health and Health Disparities [1R41MD008840-01].
Background Methods
Results
The prevalence of obesity (BMI > 95th percentile for age and sex) is
higher for African-American adolescents (AA) compared to White
adolescents (23.7% vs. 16%). Obesity is remarkably refractory to
treatment, with outcomes typically worse for AA. Adolescents carry
mobile phones during their daily activities and can receive tailored
interventions.
Our Michigan Pediatric Outpatient Weight Evaluation and Reduction
(MPOWER) program developed messages using a motivational
interviewing approach. Our experience with MPOWER suggests
that despite intensive treatment and individual tailoring, adolescents’
success making healthy choices is greatly impacted by their
environment. Adolescents were enthusiastic about the text
messages but said messages would be more effective if delivered
at times when they faced specific food choices.
We developed a personal location information system (pLIS) that
identified when AA students were in food-related venues and linked
this with tailored messages to encourage healthy food choices
specific to each venue. Geo-location technology applied on a
smartphone and controlled remotely identified when students arrived
in an eating venues with “intent to eat”. Messages were crafted from
the students known food preferences, cultural perceptions and pre-
evaluated knowledge of food available at each venue. These
messages that promoted healthy food choices were delivered to
students automatically. Students responded through the mobile app
with information about their eating experience by adding annotations
to a photo of their food selections.
We tested our system on five AA from a high school in Detroit over a study period of thirty
days. During the first week they registered venues. In the subsequent weeks we confirmed
venue detection and delivered foo recommendations.
On average, they were delivered 16.6 notifications and reported back on 8.4 of these
(50.6% compliance). Of the 42 reports received, only 3 were reported without a notification
being delivered upon arrival at the venue (7.1% proximity detection failure rate). Also of the
42 reports: 6 had just moved from a hot zone to at venue, 29 from warm, and 7 from cold.
Across all participants, reports were submitted on average 10.05 minutes after we detected
at a venue.
Students' responses to the effectiveness of food suggestions were:
21% - I did not receive a suggestion
0% - No, I didn't get the item
49% - Sort of, I got the item but did not follow the suggestion exactly
16% - Yes, I got that item and plan to follow the suggestion
3. If a student was detected
within the radius of a venue a
notification was made and
they were shown a food
suggestion (Figure 2, below).
After a delay, students were
asked to assess the meal
and submit a photo.
4. If students were in a hot
zone within times for breakfast,
lunch or dinner they could
access the app to be
presented a ‘location
verification’. Responding to the
location verification updated
the app (Figure 3, above). If
they indicated intent to eat,
they were given a food
suggestion. No further
notifications or questions would
be delivered if they had already
eaten or were not eating.
1. Each student built a list
of venues specific to
them, indicating places
they eat. When students
arrived at an eating
location and logged their
arrival it was added to
their pLIS list. The
latitude, longitude, and an
initial radius were set for
that location (Figure 1).
2. Location was tracked
by customizing a “fused
provider” service. When
the student was outside
the radius of an eating
venue, we categorized
their proximity to the
nearest venues and set
the next poll interval as
cold (> 15 km, next poll at
5 min), warm (>250 m poll
at 1 min), or hot (<100 m
poll 1 min). This
conserved battery, while
allowing fine location
polling near venues. Figure 1. Location management.