The individual and combined impact of two social cognitive

The Impact of Two Social
Cognitive Smartphone Modules
on App Usage
JA S O N FA N N I N G , P H D, S A R A H R O B E R T S , M S , C H A R L ES H I L L M A N , P H D, S E A N M U L L E N ,
P H D, L E E R I T T E R B A N D, P H D, E D WA R D M C A U L E Y, P H D
Background
PHYSICAL ACTIVITY
EHEALTH & MHEALTH
A complex, dynamic behavior
Web and mobile technologies are attractive
vehicles for intervention delivery
Subject to many barriers and changing motivators
Healthy behavior theories are helpful
Many Challenges
Background
FACTORIAL TRIALS COLLINS ET AL 2014
Two key advantages in the context of this
study
1. Test the individual and combined impact
of individual “ingredients”
2. Reduce the sample size needed for these
comparisons
Group A
Group B
Group C
Group D
Goals
Points
+
+
+
-
-
+
-
MAPS: Experimental Design
Four-arm, 12 week randomized factorial trial
◦ Evidence-based mHealth components Schoeppe 2016
◦ Social Cognitive Theory
◦ Ritterband’s Model for Internet Interventions
Outcome
Expectations
Objective
◦ Examine effectiveness of two theory-based tools
◦ Guided Goal-Setting
◦ Points-Based Feedback
Self-Efficacy
Goals
Sociostructural
Factors
Behavior
MAPS: Base App
MAPS: Goals
MAPS: Points-Based Feedback
◦ Incremental, instantaneous feedback
◦ Points are weighted by difficulty or importance
of activity
◦ Points earn levels
◦ Levels earn badges
◦ Badges earn titles
MAPS: Allocation
A
(Goals + Points + Basic App)
B
(Goals + Basic App)
C
(Points + Basic App)
D
(Basic App)
Participants
Low active adults
◦ N = 116
◦ Aged 30-54
◦ Own an iPhone or Android smartphone
◦ Access to mobile web
◦ Access to text messaging
◦
◦
◦
◦
English speaking
Free from cognitive impairment
Physically able to engage in ambulatory exercise
Willing to be randomized into any intervention
condition
Variable
Mean (SD)/Frequency (%)
Age
41.38 (7.57)
Gender
Female
93 (80)
Male
23 (20)
Married
89 (77)
White
101 (87)
College Graduate
98 (84)
Earning ≥$75,000/year
60 (52)
Measures
Physical Activity
◦ Actigraph
◦ Time (p < .01)
◦ Points (p = .04)
Efficacy
◦ Exercise Self-Efficacy McAuley, 1993
◦ Time x Points (p = .01)
◦ Lifestyle Self-Efficacy McAuley, 2009
Goal Setting
◦ Exercise Goal Setting Questionnaire Rovniak, 2002
◦ Time (p < .01)
◦ Points (p = .01)
Outcome Expectations
◦ Multidimensional Outcome Expectations for
Exercise Scale Wojcicki, 2009
◦ Physical outcome expectations: Time x Points (p = .04)
◦ Time x Goals x Points (p = .01)
◦ Barriers Self-Efficacy McAuley, 1992
◦ Time x Points (p = .03)
Barriers
◦ Perceived Barriers Scale Rogers, 2005
◦ Time (p = .01)
Mean Weekly App Usage Usage
Post-Program Feedback
Results: App Usage
Points
Raw
Transformed
Goals
Present
Not Present
Effect
Intercept
Time
Points
Goals
𝑩
2.72
-.04
.38
.36
P
<.01
<.01
.01
.02
Results: Key Feedback
WITH GOAL SETTING
WITHOUT GOAL SETTING
77% found goal setting fairly to very easy
◦ +Progressive structure
◦ +Easy of entry/editing
◦ +General motivation
47% found goal setting fairly to very easy
◦ Handbook was practical
◦ Most did not use
◦ Found it easy to misplace
Results: Feedback
POINTS
FAVORITE FEATURE
+Motivation for increased activity
Tailored Bi-Weekly Feedback
+Feeling of progress related to badges/levels
+Informative feedback on progress
-Desire specific information on point values
-More frequent badge and level delivery
◦ Sense of accountability
◦ Desired more frequent contacts
Discussion
Points-based systems can be effective for promoting engagement and highlighting mastery
experiences
In-app goal setting fosters structured goal-setting practice
Additional research implementing a more conservative “base app” over a longer follow-up
period is needed
Thank You
Jason Fanning, PhD
[email protected]
Supplement
Social Cognitive Theory
Used to study and
influence a wide number
of health behaviors
• Sexual health
Mastery Experiences
Outcome
Expectations
Bandura, 1994
Vicarious Experiences
• Nutrition
Hebert et al., 2001
• Physical activity
White, Wojcicki, McAuley, 2012;
Gothe et al., 2015; Fanning, 2016
Social/Verbal
Persuasion
Perceptions of Physical
and Psychological
Responses
Self-Efficacy
Goals
Behavior
Sociostructural
Factors
Bandura, 1986, 2007, 2004
Full Model
Design
Web app
◦ Built using Perl, PHP, HTML, CSS, and JavaScript
◦ Housed on a commercial server
◦ Text-messages delivered via Twilio
SMS Server
Web Server
Participant Device
Results: Physical Activity
Points
Goals
Results: Barriers Self-Efficacy
Points
Goals
Results: Exercise Self-Efficacy
64
62
60
58
56
54
52
50
Baseline
Follow-Up
Points (+)
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
5.269
.916
.480
.230
.384
.019
4.25
Points (-)
Goals*Points
𝜼𝟐
.05
.01
.0
.00
.00
.00
.04
P
.02
.34
.49
.63
.54
.89
.04
*adjusted for gender
Results: Lifestyle Self-Efficacy
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
11.86
.229
.110
2.38
.513
.001
.007
𝜼𝟐
.10
.00
.00
.02
.01
.00
.00
P
<.01
.63
.74
.13
.48
.97
.93
Results: Barriers
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
36.99
.246
.115
.511
.259
.680
.106
𝜼𝟐
.25
.00
.00
.01
.00
.01
.00
P
<.01
.62
.74
.48
.61
.41
.75
*adjusted for gender
Results: Exercise Goal Setting
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
41.285
7.332
.020
.534
3.313
2.878
.001
𝜼𝟐
.27
.06
.00
.01
03
.03
.00
P
<.01
.01
.89
.47
.07
.09
.98
*adjusted for race
Results: Physical Outcome Expectations
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
.052
.752
1.014
3.881
.020
1.036
.686
𝜼𝟐
.00
.01
.01
.03
.00
.01
.01
P
.82
.39
.32
.05
.89
.31
.41
Results: Self-Evaluative Outcome Expectations
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
.317
1.098
.355
3.166
.119
.012
1.283
𝜼𝟐
.00
.01
.00
.03
.00
.00
.01
P
.57
.30
.55
.08
.73
.91
.26
Results: Social Outcome Expectations
Goals
Points
Effect
Time
Points
Goals
Time*Points
Time*Goals
Points*Goals
Time*Points*Goals
F
3.043
.347
.396
.054
.672
.195
1.536
𝜼𝟐
.03
.00
.00
.00
.01
.00
.01
P
.08
.56
.53
.82
.41
.66
.22
*adjusted for gender and income
Analyses
Multiple imputation used for missing
values
Windsorized when necessary
Aim 1 and Aim 2: Repeated-measures
factorial ANOVA
◦ Linear regression to identify covariates
◦ Points & goals entered as fixed factors
◦ Bonferroni-corrected posthoc analyses
Aim 3: Hierarchical linear modeling
◦
◦
◦
◦
Forward stepping
Model fit assessed via -2LL, AIC, BIC
Predictors retained at p ≤ .10
Significance at p ≤ .05
Aim 3: Feedback coded by theme
◦ Descriptive statistics
Results: App Usage Un-transformed
Points
Effect
Intercept
Time
Points
Goals
𝑩
6.904
-.168
1.875
1.909
Goals
𝑺𝑬
.78
.06
.90
.90
t
8.89
-2.76
2.08
2.12
P
<.01
.01
.04
.04
Results: App Usage Transformed
Points
Effect
Intercept
Time
Points
Goals
Goals
𝑩
2.72
-.04
.38
.36
𝑺𝑬
.13
.01
.15
.15
t
21.40
-4.78
2.61
2.48
P
<.01
<.01
.01
.02