I wrote recently about a tailored messaging intervention to help people
living with HIV manage fatigue, but it was a pretty labor-intensive approach.
One of the research projects that I’m working on right now will automate the
tailored messages, with a slightly different but related focus on exercise
behavior. With modern antiretroviral medication treatments (ART), people living
with HIV can expect to live a relatively normal life, but they are at higher than
average risk for problems like heart disease and diabetes as they get older.
The latest research on HIV shows that people with the virus not only need
regular exercise to prevent heart and kidney problems, they need 30 minutes of higher-intensity
exercise on most days of the week. That’s more exercise than most Americans
without HIV manage to achieve.
In a new article, just published in the Journal of the Association of
Nurses in AIDS Care, my colleagues and I describe the intervention that
we’re testing to help people with HIV be more active. The first part of
the intervention is an exercise group that meets for 16 weeks at the campus
health center, providing personal-training support as well as group
interactions to help people develop their exercise habits. The
tailored-messaging component comes in after the intensive group phase is
finished, and is designed to help people maintain their exercise habit over
time. In addition to the tailored messages, people in our study will receive
individual calls from a coach who will use motivational interviewing strategies to support them.
As I have written elsewhere, the things that help us develop a new healthy habit are often
different from the things that help us to maintain that habit for the long
term. We start doing something new because it’s important to us, but we are
most likely to keep doing it when it is easy. The tailored messages, therefore,
focus on various barriers that might get in the way of exercising. When we
brainstormed possible barriers with our experienced research team, and also
asked focus groups that included people living with HIV, we came up with a list
of barriers that probably sounds pretty familiar: things like being tired, bad
weather, conflicting priorities, or not feeling well that day. In fact, none of
the barriers that our team identified had anything specifically to do with HIV!
The resulting tailored messages, therefore, might end up being useful to anyone
seeking to maintain an exercise habit.
After developing a list
of barriers, we brainstormed ideas to help people overcome each of them. In
keeping with Two Minds Theory, we used six basic principles of behavior change, targeting both the Intuitive and the Narrative
minds:
1.
Training the Intuitive mind through practice,
2.
“Tricking” the Intuitive mind into a different response
by modifying environmental conditions and stimuli,
3.
Engaging the Intuitive mind in creative problem-solving to generate ideas that the Narrative mind can then sift
through,
4.
Practicing mindfulness strategies
that delay the Intuitive response and give the Narrative mind more time to
engage,
5.
Using the Narrative
mind’s capacity for social imagination to create new incentives and enlist others’ aid, or
6.
Offering reframing ideas
that give the Narrative mind different ways of thinking about the problem.
We developed multiple
messages using each of these principles, relying on the team’s collective
wisdom and experiences. Team members contributed advice, web links, videos,
self-help tools, and inspirational quotes. We found that all of these could be
linked to one of the six behavior-change principles, although of course I have
never claimed those six to be an exhaustive list. (Prochaska originally
listed 10 processes of change, for example, and Susan
Michie has 16, so there
is almost certainly more work to be done in this area). Sometimes a message for
one barrier could be adapted to also be useful for people who are experiencing
a different barrier (e.g., exercising with a friend might help to alleviate
boredom, and also serve as a reminder for people who forget). Other times a
message was only applicable to one specific barrier to exercise (e.g., weather,
where we had to further differentiate between advice for hot weather,
advice for cold weather, advice for when air
quality was poor, etc.).
Once we had an initial
draft – over 170 messages, each linked to one of 12 barriers – we sent the
messages themselves to two different advisory groups to review. The groups
again included experts on physical activity research as well as people living
with HIV. We asked whether any of the messages struck them wrong, whether any
should be edited, and whether they had other suggestions. A few messages were
dropped at this stage, a few were added, and a number were edited based on
feedback from the advisory groups.
Finally it was time to
program the messages into a computer algorithm. For this stage we used a
web-based database design program called Caspio, which I had experience using in other
research. In principle, any web-enabled relational database would work. We
created a survey that asked people about their barriers to exercise, using
validated scales to measure things like pain, mood, and fatigue. Then I created
some formulas that compared scores on the various scales and identified the
person’s top barrier for that day. There were also categories for “no barriers
today” and for a control group that we intend to send only supportive, non-tailored
messages. Finally, we programmed each message possibility into the system. A
random-number generator was used to select from the pool of messages available
to address each barrier, so that the participant essentially selects the topic
of the message based on their survey responses, but the computer selects the
specific message. This was considered important because boredom and repetition
were cited as the most important obstacles among participants in a previous
study of tailored
messaging that I conducted.
Before releasing the
tailored messaging tool to actual patients in our study, seven team members
tried it out ourselves. This led to some additional changes – a few more
messages dropped or edited, but more importantly a shortening of the survey
tool. We hid the validated survey items until a person said they actually had a
particular concern, which shortened the overall length of the daily survey to
just a few questions. We added an extra screen where people could see feedback
every day on a graph of their active minutes (at this point self-reported;
ideally in a future iteration the data would come from sensor devices). We
added a behavior-change coach’s picture along with their contact information.
And we removed some explanatory text in order to get to the point – the actual
tailored message – more quickly. In this more technical design phase of the
project, we focused on usefulness of the system in terms of
doing what it was designed to do, and also on ease of use for
the end user, because that type of consideration is likely to play an important
role in whether or not the tool actually gets used.
You can generate a test
message from our system here, or enroll to receive daily messages here
(don’t worry: they’re 100% free and you’ll have a way to cancel at any time).
Something that the
system doesn’t currently do is to adjust its message content based
on users’ demographic characteristics or psychological states, other than the
main barrier identified for that day. In my last
tailored-messaging study I
did that kind of intensive “sculpting” of messages, with word choice and
grammatical nuances that were carefully curated in response to a participant’s
daily survey responses. In general, people didn’t notice the effort! They got
the gist, the take-home message, but they didn’t pay much attention to how it
was delivered. So this time around we are letting a random-number generator
make those choices. We will still collect data on specific
message characteristics, however – for instance, which of the six
behavior-change strategies was used. And we will ask people to rate which messages
had the most utility. This could lead to future iterations of the algorithm in
which more carefully tailored messages can be created by the computer, matching
content to individuals based on additional characteristics that we identify
from the data as we go along.
Psychological tailoring of health-promotion messages is still pretty wide-open territory for research, and I’m excited to see what we find in this new study. Watch this space for updates once we have some data to report!
Comments
Post a Comment