mHealth tools for smoking cessationJacobs Center - CCWD 2018 Workshop

Séverine Toussaert (U. of Oxford)

November 5, 2018

Acknowledgments

Study Team:I Justin S. White, PhD, University of California, San FranciscoI Johannes Thrul, PhD, Johns Hopkins UniversityI Lee Westmaas, PhD, American Cancer SocietyI Jeuneviette Bontemps-Jones, MPH, American Cancer SocietyI Lorien Abroms, ScD, George Washington University

(Some) Study Partners:I Erik Augustson, National Cancer InstituteI Sydney Sattell, Wellpass (previously Sense Health), Inc

Funding: American Cancer Society

Motivation

I Over last two decades, widespread efforts to expand access tosmoking cessation aids.

I Yet, 2/3 of quit attempts still occur without treatment.

I Barriers: restricted assess, high costs, inconvenience.

I mHealth technologies can reach large population at low cost.

I Smokers are increasingly using their smartphone for cessationsupport:

I 1.2 million Android app downloads per month (Abroms 2013)

MotivationI Typical app (Smoke Free):

I Rarely evaluated, poor concordance with medical guidelines.I Support often less targeted and less personal.

What we propose

Our long-term goal is to design a custom smartphone app with 2innovative features:

1. Peer mentoring

I Users supported through messages by peer mentor.I Peer mentor = former smoker who quit successfully.

Status: Results from first RCT presented today.

2. Game-based environment

I Users enroll in 7-day quit challenges.I Mentor and users accumulate points for engaging with the app.I Points translate into non-monetary rewards.

Status: Ongoing (design stage)

What we propose

Our long-term goal is to design a custom smartphone app with 2innovative features:

1. Peer mentoring

I Users supported through messages by peer mentor.I Peer mentor = former smoker who quit successfully.

Status: Results from first RCT presented today.

2. Game-based environment

I Users enroll in 7-day quit challenges.I Mentor and users accumulate points for engaging with the app.I Points translate into non-monetary rewards.

Status: Ongoing (design stage)

Why interesting? (1)

I Widespread adoption of peer mentoring for health promotion.I AA, Weight Watchers

I Lack of rigorous evaluation and existing evidence is mixed(Webel et al. 2010)

I Westmaas (2010) points to ceiling effects, different definitionsof social support, mechanisms not examined

I Yet, reasons to believe in efficacy (Gale 2018):1. Informational channel2. Emotional pathway3. Social pressure and accountability

I More so if strong connection between mentor and mentee:

) study how similarity affects engagement and likelihood of quitting.

Why interesting? (2)

I Incentive-based behavioral interventions can promote cessation.I Volpp et al. (2009), White et al. (2013), Halpern et al. (2015)I Leverage behavioral science concepts such as loss aversion or

present bias.

I Gamification for health behavior change still fairly uncommon.I SuperBetter (.5M users), KwitI But lack of rigorous evaluation.

I Two distinguishing features of our incentive scheme:1. Incentives with surprise elements: keeping novelty up2. Focus on identity and image concerns: keeping esteem high

Outline

1. Results from RCT focused on peer mentoring

2. Discussion of next steps involving gamification

Peer mentoring via text messages

I We conducted an RCT to test the effectiveness of peermentoring for smoking cessation delivered via text messages.

I We study (i) preferences for peer mentors; (ii) whether successdepends on the similarity between mentor and mentee.

1. Preference for similarity:I Do smokers have a preference for a mentor more similar to

themselves? If so, what dimensions of similarity matter?

2. Importance of similarity for success:I How does similarity impact engagement and likelihood of

quitting?

Sample and design

I Recruit 200 U.S. adult smokers whoI want to quit within 30 daysI do not use NRT or e-cigsI agree to take a salivary cotinine test

IControl group: receives automated text messages used inSmokefreeTXT (developed by National Cancer Institute)

I 1-5 messages per day for up to 8 weeksI Messages include encouragement, advice, and tips for quitting

ITreatment group: receives similar automated messages +personalized messages from a peer mentor.

I Follow-up at 3 months (self-report + saliva test)

Text messaging intervention

Our peer mentors

I Peer mentor: former smoker who quit successfully at least 6months ago.

I Good diversity in terms of socio-demographic characteristics.

I 36 mentors assigned 3 smokers each. Assignment is random.

I Had to complete a 2-hour online training and create a profilepage.

I Incentives: $50 for training, $150 for mentoring + entry in$1,000 lottery drawing (additional entries if mentees quit)

Preferences for peer mentors

I Before assignment to treatment, smokers complete a baselinesurvey gathering data on their preferences for a peer mentor.

I 7 characteristics considered:

1. Demographic variables: gender, age, race/ethnicity2. Smoking-related variables: heavy/light smoker, reason for

smoking, reason for quitting, method used to quit

I Smokers asked to rank 7 dimensions in order of importance.

I For each dimension, also asked to express a preference.I choice from a list of categoriesI option “no preference” available

Baseline characteristics

Intervention effects on abstinence

10.6 points *

13.1

23.8

0

10

20

30

Perc

ent a

bstin

ent

ControlGroup

MentorGroup

A. Self-report(unadjusted)

12.7 points **

12.8

25.5

ControlGroup

MentorGroup

B. Self-report(adjusted)

4.9 points

3.0

7.9

ControlGroup

MentorGroup

C. Verified(unadjusted)

6.5 points **

2.5

9.0

ControlGroup

MentorGroup

D. Verified(adjusted)

Abstinence over time

0.00.10.20.30.40.5

Prop

ortio

n re

spon

ded

0 20 40 60 80

Panel A. Any Response to Abstinence Questions

0.000.050.100.150.20

Prop

ortio

n ab

stai

ned

0 20 40 60 80

Panel B. Abstinence (Intent-to-Treat)

0.00.20.40.6

Prop

ortio

n ab

stai

ned

0 20 40 60 80Day

Panel C. Abstinence (Responses Only)

Control Mentor

Mood over time

0.00.10.20.30.40.5

Prop

ortio

n re

spon

ded

0 10 20 30 40

Panel A. Any Response to Mood Questions

0.000.050.100.150.20

Prop

ortio

n in

goo

d m

ood

0 10 20 30 40

Panel B. Good Mood (Intent-to-Treat)

0.00.20.40.6

Prop

ortio

n in

goo

d m

ood

0 10 20 30 40Day

Panel C. Good Mood (Responses Only)

Control Mentor

Cravings over time

0.00.10.20.30.40.5

Prop

ortio

n re

spon

ded

0 10 20 30 40

Panel A. Any Response to Craving Questions

0.000.050.100.150.20

Prop

ortio

n lo

w cr

avin

g

0 10 20 30 40

Panel B. Low Craving (Intent-to-Treat)

0.00.20.40.6

Prop

ortio

n lo

w cr

avin

g

0 10 20 30 40Day

Panel C. Low Craving (Responses Only)

Control Mentor

Engagement

Preference for similarity with peer mentor

54 ***

35

21 **

49 ***

50 ***

52 ***

36 ***

17

28

12

32

23

25

17

Age

Gender

Race/ethnicity

Smoking intensity

Reason for smoking

Reason for quitting

Quit method

0 10 20 30 40 50 60 70 80 90 100Percent preferring similarity

Observed Random shuffle

Heterogeneity in preferences: age

Heterogeneity in preferences: gender

Summary (1)

1. It is feasible to train former smokers to act as peer mentors

2. Peer mentoring improves smoking abstinence at 3 monthsrelative to automated texting

I Verified abstinence: ↑ 5-7 points over 3% baseI Self-reported abstinence: ↑ 10-13 points over 13% base.

3. Smokers display a strong preference for mentors who aresimilar to themselves.

I Though we find heterogeneity in preferences, e.g., by gender.I “Smart” matching algorithms may help attract and retain users

of dyadic programs.

Summary (2)

I Similarity of own vs. mentor attributes did not strongly predictabstinence or engagement.

I However, the study findings need to be replicated in a larger,more diverse sample.

I The social interactions may not have been intense enough foreffects to appear.

I Mentees did not remember their mentor’s attributes very wellwhen quizzed in endline survey.

I A wider range of attributes also merit attention.

I Currently analyzing content of conversations to gain moreinsights.

Next steps

I Develop a prototype for a gamified app

I Game based on “quit challenges”I Commitment for short duration of 7 daysI Focus on early stage, small steps: “foot in the door”

I Engaging with the app unlocks rewards:I Points for completing various “missions”I Missions emphasize both processes and outcomes

I Integration with a carbon monoxide sensor connected to phoneallowing for personal monitoring.

CO sensor

Virtual rewards

Points from missions completed allow to:

1. Unlock new territories and play new “mystery” gamesI Curiosity, preference for non-instrumental information

2. Make digital pet grow:I Identity: symbol of transformationI Increases sense of accountability (good for teenagers?)

Virtual rewards

Points from missions completed allow to:

1. Unlock new territories and play new “mystery” gamesI Curiosity, preference for non-instrumental information

2. Make digital pet grow:I Identity: symbol of transformationI Increases sense of accountability (good for teenagers?)

Surprise and curiosity

Combining peer mentoring and gamificationI Importance of building self-esteem capital both for users and

mentors.

I For mentors:I Of course, social preferences - stronger for mentees who feel

“closer”I Commitment value of being a role model: “keep being

awesome”I Status concerns: leaderboard with list of top mentors

Discussion

I Game play as substitute for smoking?I “Temptation bundling”I Substituting one addiction for another?

I Building confidence through game playI Transfer from virtual to real world?

I MonitoringI Necessary? Detrimental?I Importance of feedback, but danger of reminders