the psychology of mass-interpersonal behavioural change websites: a meta-analysis
DESCRIPTION
This paper presents a meta-analysis that investigates psychological design factors that can explain the efficacy of online behavioural change interventions. It makes a clear distinction between mass-media, interpersonal and mixed, mass-interpersonal communications. To this end, a model, called ‘the Communication-Based Influence Components Model’, is used to synthesize behavioural change and persuasion taxonomies.TRANSCRIPT
The Psychology of Mass-Interpersonal Behavioural Change
Websites: a meta-analysis
Brian Cugelman, Prof. Mike Thelwall, Prof. Phil Dawes
University of Wolverhampton Statistical Cybermetrics Research Group and the Wolverhampton Business School
http://cybermetrics.wlv.ac.uk
Medicine 2.0 Conference17-18 September 2009
Toronto, Canada
Overview1. Background and objectives2. Research challenges & solutions3. Meta-analysis4. Findings5. Conclusions
1. Background and Objectives
Examples of Online Interventions
• Don’t start smoking
• If you started, stop
• Exercise more
• Drink less alcohol
• Eat more good food
• Eat less bad food
Synthesis Research
1. Meta-analysis: positive results– Portnoy et al., 2008– Wantland et al., 2004
2. Systematic reviews: mixed and slightly positive– Norman et al. (2007) – Vandelanotte et al. (2007)
3. Real-world evaluation: unclear outcomes– Evers et al. (2003)– Doshi et al. (2003)– Lin and Hullman (2005)
Research Objectives
1. Assess the efficacy of online interventions appropriate for public campaigns
2. Identify psychological design factors
3. Investigate the role of adherence (dose)
2. Research Challenges &
Solutions
A. Prior Studies not Generalizable to Public Campaign
• Problem: Blend voluntary with mandatory behaviours (chronic disease management)
• Solution: More voluntary and common interventions
B: Ambiguous Online Communication Models
• Problem– Mass-Media (one-way)– Interpersonal (two-way)
• Solution: Mass-Interpersonal
(On
e-W
ay
) O
ne
-to
One
Impersonal
Many
Mass Media
(Tw
o-W
ay
) O
ne
-wit
h Interpersonal Mass Interpersonal
one-with-one
one-to-one
C: No Clear Design Guidelines on Online Behavioural Influence
• Problem: Too complex. Too simple. Not quite right.
• Solution: Communication Based Influence Components Model to integrate behavioural medicine and persuasion
Communication-Based Influence Components Model
SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
Framework to describe intervention psychology
Cugelman, B. Thelwall, M. Dawes, P (2009)
3. Meta-Analysis
Conducting the Meta-Analysis
• Searched five databases + grey literature
• Obtained 1,271 results
• Retrieved 95 full text studies
• Selected 31• Primary analysis: 30 interventions from 29
studies (N=17,524)
4. Findings
Effect Sizes
-0.4-0.3
-0.2-0.1
0.00.1
0.20.3
0.4
Survey Only (Waitlistor Placebo)
Website Print (Major)
Overall: d=.194, p=.000, k=30
d
Effect Size by Intervention Duration
-0.4-0.3
-0.2-0.10.0
0.10.2
0.30.40.5
0.60.7
One-time From 2 days to 1month
Beyond 1 to 4months
Beyond 4 to 7months
Beyond 7 to 13months
d
Dose: Three Variables
COR r=.37, p<.000, k=5
Intervention
Adherence
OutcomeEffect Size
StudyAdherence MR r=.481, p=.006, k=28
MR r=.455, p=.109, k=13COR r=.240, p<.000, k=9
COR: Correlation effect sizeMR: Meta-regression estimate
Relative Influence Components and Outcomes
876543210
Relative Behavioural Determinants (sum)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
Eff
ect
Siz
e (
d)
Print (Major)
Website
Survey Only (Waitlist or Placebo)
ControlMediaSimple
Media Channel SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
k% Across 30
Interventions
Website & Email 20 66.7%
Website 10 33.3%
Feedback Message SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
k% Across 30 Interventions
Tailoring 25 83.3%
Personalization 12 40.0%
Adaptation / Content matching 2 6.7%
Source Modifier SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
k% Across 30 Interventions
Attractiveness 5 16.7%
Similarity 3 10.0%
Credibility 1 3.3%
Source Encoding SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
k% Across 30 Interventions
Multiple Interactions 23 77%
Single Interaction 3 10%
Sequential Requests (Foot-in-the-door) 1 3%
Intervention Message SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
Top 5 of 40 Behavioural Change Techniques k% Across 30 Interventions
Provide information on consequences of behaviour in general 23 77%
Goal setting (behaviour) 21 70%
Provide feedback on performance 20 67%
Prompt self-monitoring of behaviour 19 63%
Provide instruction on how to perform the behaviour 18 60%
Audience Interpreter SourceInterpreter
InterventionMessage
AudienceInterpreter
FeedbackMessage
Media ChannelContext
Decode
EncodeDecode
Encode
Top 5 of 12 Behavioural Determinants k% Across 30 Interventions
Knowledge 30 100%
Motivation and goals (Intention) 26 87%
Social influences (Norms) 22 73%
Beliefs about consequences 21 70%
Skills 19 63%
5. Conclusions
Conclusions
1. Efficacy: Reasonable impact, and comparable to print, though more affordable with broad/rapid reach
2. Psychology: Most sites goal orientated, possible influence component correlation
– Communication Based Influence Components Model stood up across interventions
3. Dose: study adherence, intervention adherence and ES likely related. They may be explained by motivation
Thank you
University of Wolverhampton Statistical Cybermetrics Research Group and the Wolverhampton Business School
http://cybermetrics.wlv.ac.uk