Challenges of Harnessing the Informatics Landscape to Promote Health Behavior Change
1. Challenges of Harnessing the
Informatics Landscape to Promote
Health Behavior Change
David B. Abrams, PhD
Executive Director, The Schroeder Institute for Tobacco Research
and Policy Studies
The Johns Hopkins Bloomberg School of Public Health
Georgetown University Medical Center
KEYNOTE PRESENTED AT THE AMERICAN ACADEMY OF HEALTH BEHAVIOR
AUSTIN, TEXAS
MARCH 19, 2012
2. Population Impact: The
Example of Tobacco
FDA
act
Source: Mendez, Warner. Tobacco control. Nicotine & Tobacco Research., August 11, 2010.
3. Revisit Goal of
Population Impact
Impact = Reach x Efficacy
Efficiency: Continuous optimization of quality of
evidence-based intervention
delivery at scale, cost-effectively
RE-AIM: multi-level integration
SOURCES: (1) Abrams et al. (1996). Integrating individual and public health perspectives
for treatment of tobacco: A combined stepped care matching model. Annals of Beh
Med,18,290-304. (2) Glasgow, Green, Klesges, Abrams et al. (2006). External validity: we
need to do more. Ann Behav Med,31(2),105-108.
4.
5. Back in 2005…
• Internet adoption in US: from 15% in 1995 to 75% in 2006
– More than 70 million adults go online each day
• ~ 80% of Internet users have searched online for health
information at some point in their lives (Pew, 2005)
BUT…
• In spite of a surge of technologic capability, research and
evaluation methodologies have not kept pace with rapid
evolution & proliferation of communication technologies
• Nor has the dissemination of effective eHealth interventions
achieved the level of penetration one might have hoped, given
the number of people who now access the Internet
Source: Atienza, Hesse, Abrams, Rimer, et al. Critical Issues in eHealth Research. Am J
Prev Med. 2007 May; 32(5 Suppl): S71–S74.
6. 5+ Years Later: Where
Are We Now?
Crounse commentary (2007):
“Even though robust communication and collaboration
solutions exist to speed scientific discovery and the
delivery of care, all too often our methodology falls
back on that which we know and have always done
before… But we must not dig in our heels, resist
change, and continue to conduct business as we have
always done before just because it suits our comfort
level. Others around the world will not indulge in or
tolerate that luxury.”
Source: Crounse B. The newspaper, the wristwatch, and the clinician. Am J Prev Med.
2007 May;32(5 Suppl):S134.
7. Assumptions
1. The promise of informatics and technology to change
public health can be realized using traditional scientific
theories and methods (with perhaps only some fine
tuning)
2. Single level interventions delivered at scale (mass
customization) can change health behavior at the
population level and make a timely impact.
3. Integration across platforms in real time can overcome
barriers to reach, engagement, and efficient delivery of
behavior change interventions and their seamless
integration into delivery systems and policy
8. Assumption 1:
Traditional Science
The Individual Effectiveness to
Population Impact Chasm
Source: Abrams, D (1999). Transdisciplinary paradigms for tobacco research. Nicotine
& Tobacco Research, 1, S15.
9. A New Definition of
Translational Research
T1 T2 T3 T4
Potential Application Efficacy Effectiveness Population-Based
Basic Science Potential Evidence- Clinical Care Health of
Discovery Clinical Based or Community
Application Guidelines Intervention or Population
Basic Theoretical Efficacy Applied Public Health
Knowledge Knowledge Knowledge Knowledge Knowledge
Types • Phase 3 trials • Phase 4 clinical trials
• Phase 1, 2 trials •T3 type studies in community
of • Systematic reviews • Implementation
• Observational • Population / outcome studies
Research • Health services studies • Communication
• Cost-benefits, policy impact
• Observational studies • Dissemination
• Studies beyond clinical care
• Diffusion
• Systematic reviews
Sources: 1) Szilagyi P. 2010: From Research to Dissemination Implementation:
http://www.research-practice.org/presentations.aspx. 2) Khoury M, et al. Gen Med,
2007;9:665-674. 3) Glasgow et al., RE-AIM.
10. Assumption 2:
Single-level interventions
Outside
the skin
Under
the skin
12. Source: Lazer et al. (2009). Life in the network: the coming age of computational social
science. Science. 323(5915): 721–723.
13. Iterative Continuous
Improvement
Dynamic model of research for multi-level impact:
Theory to mechanisms to practice to policy loop
14. Example:
Multiphase Optimization
Strategy (MOST)
• Collins, Murphy, Strecher. The
multiphase optimization strategy
(MOST) and the sequential
multiple assignment randomized
trial (SMART): new methods for
more potent eHealth
interventions. Am J Prev Med.
2007 May;32(5 Suppl):S112-8.
PMCID: PMC2062525.
• Collins et al. The Multiphase
Optimization Strategy for
Engineering Effective Tobacco
Use Interventions. Ann Behav
Med. 2011 Apr;41(2):208-26.
PMCID: PMC3053423.
15. From Gene Chip Arrays
To Population Arrays
Multi-level tailoring at:
• biological level
• individual level
• proximal socio-behavioral level
• community level
• population level
GENOMICS TO POPULOMICS
Source: Murray et al. (2006). Eight Americas: Investigating Mortality Disparities across
Races, Counties, and Race-Counties in the United States. PLoS Medicine: Vol 3,
15139, e260.
16. Illustrative Examples from
the Schroeder Institute
1. The iQUITT Study - Internet (Graham, PI)
2. Facebook (Cobb, PI)
3. POSSE (Kirchner, PI)
4. Adaptive designs in clinical trials (Niaura)
17. Assumptions
1. The promise of informatics and technology to change
public health can be realized using traditional scientific
theories and methods (with perhaps only some fine
tuning)
2. Single level interventions delivered at scale (mass
customization) can change health behavior at the
population level and make a timely impact.
3. Integration across platforms in real time can overcome
barriers to reach, engagement, and efficient delivery of
behavior change interventions and their seamless
integration into delivery systems and policy
18. Internet and Telephone Treatment for Smoking Cessation
Amanda L. Graham, PhD (PI)
National Cancer Institute
5 R01 CA104836
2004 – 2010
19. Initial Evaluation of
QuitNet
• Observational study in December 2002
• Total # surveyed = 1,501
• Responders: 25.6% (N=385)
20. Initial Evaluation of
QuitNet
Least conservative
ADHERENCE SAMPLE (N=223): 30.0%
– Respondents only
• Used site ≥ 2x (N=336): 13.1%
• Used site >1x (N=488): 9.8%
• Excluding bounced (N=892): 8.0%
INTENTION TO TREAT (N=1,024): 7.0%
– Counts all non-responders as smokers
Most conservative
21. 2005 participants
Recruited online
Randomized to
“real world”
Internet or phone
treatments
~ 70% follow-up
rates 3-18
months
Source: Graham AL, Bock BC, Cobb NK, Niaura R, Abrams DB. Characteristics of smokers reached
and recruited to an internet smoking cessation trial: a case of denominators. Nicotine Tob Res. 2006
Dec;8 Suppl 1:S43-8.
22. Control Condition
Static site designed
by research team
“look and feel” of
QuitNet
Extracted content
from QuitNet
No interactive
features
No online
community
26. Research Questions
1. Informed Consent: For low-risk, population-based studies
focused on dissemination and implementation research (i.e.,
evaluating interventions as they are used in the “real world”),
what is the appropriate and optimal level of informed
consent? How might informed consent be a barrier that
actually limits the reach and understanding of the target
population in fundamental ways?
2. Control/Comparison Group: What is the appropriate
control condition or comparison condition? Is one needed at
all? How can we move away from traditional RCTs and
consider SMART/adaptive designs, practical & comparative
efficacy trials, and other approaches?
28. Population Impact
Impact = Reach x Efficacy
Efficiency: Continuous optimization of quality of
evidence-based intervention
delivery at scale, cost-effectively
RE-AIM: multi-level integration
SOURCES: (1) Abrams et al. (1996). Integrating individual and public health perspectives
for treatment of tobacco: A combined stepped care matching model. Annals of Beh
Med,18,290-304. (2) Glasgow, Green, Klesges, Abrams et al. (2006). External validity: we
need to do more. Ann Behav Med,31(2),105-108.
30. IMPACT:
Secondary Analyses
• Of funnels and tunnels and rabbit holes…
• From community newspaper to Internet tx seekers…
• From 10+ million to 99,900 to 2,005…
• Who do we have here, who is NOT here, and how much
implementation dissemination, generalizability and
scalability do we REALLY have here?
• Oh (nearest and dearest) denominator wherefore art
thou?
32. User Engagement &
Outcomes
Pilot study 2002:
• Use of any social support and
2-month continuous abstinence: OR = 4.03
• Intensity of website use and
2-month continuous abstinence: OR = 6.07
iQUITT Study 2011:
Compared to no treatment:
• 3+ logins were 1.9x more likely to quit (p < .05)
• 3+ calls were 2.4x more likely to quit (p < .01)
NOTE: to date we can’t explain the growth of the static minimal Internet comparison
(control) group
33. Engagement:
Social Networks & Cessation
NEXT STUDY
35. Assumptions
1. The promise of informatics and technology to change
public health can be realized using traditional scientific
theories and methods (with perhaps only some fine
tuning)
2. Single level interventions delivered at scale (mass
customization) can change health behavior at the
population level and make a timely impact.
3. Integration across platforms in real time can overcome
barriers to reach, engagement, and efficient delivery of
behavior change interventions and their seamless
integration into delivery systems and policy
36. Am J Public Health. 2010 Jul;100(7):1282-9.
J Med Internet Res. 2011 Dec 19;13(4):e119.
37. QuitNet By the
Numbers
• Website overview 2007
– 1.17 million unique visitors to the web site
– 76.45 million “page views”
– 123,927 unique registered users
– 160,000 active users
• Internal communications 2007
– 1.36 million internal email (“Qmail”) messages
– 815,070 forum posts, ~ equal numbers in “Clubs”
37
38. QuitNet Scope
• One of the 1st examples of large-scale, web-based therapeutic social network
• > 750,00 members – approx. 30-50K are active in any given month
• Growth rates of up to 22,000 members in a month.
39.
40. QuitNet Data
Applications
A: Longitudinal Social Network Analysis
– 5+ years of detailed network data
B: Content Analysis
– 10+ years of forum postings, chat logs, private
message history, blog posts, personal profiles and
testimonials.
C: Agent Based Modeling
– Recreation of QuitNet as a dynamic, synthetic
network that can be manipulated.
42. Example: Facebook
• 65 M users/month (US
alone)
– Covers over 50% of
people aged 15-24
• Age:
– 45% of the population
is over 25
– Over 35 population
doubling every 2
months
• Gender:
– Women are fastest
growing segment
43. Why Online Networks?
• For Interventions:
– Faster intervention development
– Better diffusion and dissemination
• For Evaluation:
– Faster recruitment
– Fewer barriers to enrollment
– Fewer barriers to follow-up
– Broader conceptualization of impact
46. “Impact 2.0”
• Traditional View:
Impact = Reach X Efficacy
• Network View:
Impact = (Initial Reach X R) X Effectiveness
Where R is the reproductive ratio or viral spread of
an intervention or behavior.
50. Example: Facebook
R01
• Nate Cobb, PI (2012 – 2015)
• Planned >12,000 participants
in factorial design
• Outcome is R - diffusion of
the application from one
member to another. Not
effect!
• Answers question of what
drives diffusion and spread?
• Entire process is automated
from enrollment to tracking of
diffusion.
52. Assumptions
1. The promise of informatics and technology to change
public health can be realized using traditional scientific
theories and methods (with perhaps only some fine
tuning)
2. Single level interventions delivered at scale (mass
customization) can change health behavior at the
population level and make a timely impact.
3. Integration across platforms in real time can overcome
barriers to reach, engagement, and efficient delivery of
behavior change interventions and their seamless
integration into delivery systems and policy
53. Ecological Momentary Tobacco Control
Thomas R. Kirchner, PhD (PI)
National Institute on Drug Abuse / DC Department
of Health
RC1 DA028710 / CDC CPPW Contract
2009 – 2012
62. Socio-economic
POST Variation
Average pack price: Newport
M = $7.75 block-group white
M = $7.29 block-group non-white
p = 0.004
Low pack price: All cigarette brands
M = $6.73
Average pack price: LCC
M = $3.71
Low cost LCCs more prevalent in
non-white block-groups
(2 = 4.31, p=0.04).
64. Relapse Dynamics
SOURCE: Kirchner et al. Relapse dynamics during smoking cessation: Recurrent
abstinence violation effects and lapse-relapse progression. J Abn Psych; 2012: 121(1).
65. SOURCE: Shiyko MP, Lanza ST,
Tan X, Li R, Shiffman S. Using the
Time-Varying Effect Model
(TVEM) to Examine Dynamic
Associations between Negative
Affect and Self Confidence on
Smoking Urges: Differences
between Successful Quitters and
Relapsers. Prev Sci. 2012 Jan 14.
[Epub ahead of print].
68. Solutions & Future
Directions
Crounse commentary (2007):
“all too often our methodology falls back on that
which we know and have always done
before....But we must...not dig in our heels,
resist change and continue to conduct business
as we’ve always done so before just because it
suits our comfort level. Others around the world
will not indulge in or tolerate that luxury”
Source: Crounse B. The newspaper, the wristwatch, and the clinician. Am J Prev Med.
2007 May;32(5 Suppl):S134.
69.
70. Iterative Continuous
Improvement
Dynamic model of research for multi-level impact:
Theory to mechanisms to practice to policy loop
71. Assumptions
1. The promise of informatics and technology to change
public health can be realized using traditional scientific
theories and methods (with perhaps only some fine
tuning)
2. Single level interventions delivered at scale (mass
customization) can change health behavior at the
population level and make a timely impact.
3. Integration across platforms in real time can overcome
barriers to reach, engagement, and efficient delivery of
behavior change interventions and their seamless
integration into delivery systems and policy
72. Promises Promises…
Bio + behavioral + social + population - based sciences MAY
finally make the dream of efficient population behavior change
a reality if and only if:
• Rapid innovation across: platforms, modes, capacity in near
or in real time, will overcome prior barriers to:
– reach
– engagement
– utilization of efficient tailored behavior change
interventions
– and their seamless proximal and distal integration into
contexts (i.e. traditional and new -- social media, Internet,
community, low SES subgroups, health and public health
delivery systems and aligned policy at scale)
73. • “Today, the hurricane and earthquake do not pose the
greatest danger.
• It is the unanticipated effects of our own actions, effects
created by our inability to understand the complex
systems we have created and in which we are
embedded.
• Creating a healthy, sustainable future requires a
fundamental shift in the way we generate, learn from,
and act on evidence about the delayed and distal
effects of our technologies, policies, and institutions.”
Source: Sterman JD. Learning from evidence in a complex world. Am J Public Health.
2006 Mar;96(3):505-14. Epub 2006 Jan 31.
74. Embrace Complexity
• The world is complex, contextual, dynamic, multi-causal (causal
loops), multi-level, multiply determined…
– For every complex problem there is a simple solution….and it is
usually wrong
• Research designs, methods and measures should take this into
account and capitalize on advances in computer sciences,
technology, informatics, imaging, knowledge management,
networking and communications
• Vertical integration: cells to society across varying time units
(seconds to centuries)
• Solid basic behavioral and social and population science is needed
as a firm foundation to build systems within systems models
• Aligned incentives at every level of the system can change
populations
75. WE NEED EVIDENCE IN T2-T4 THAT…
IS MORE IS LESS
Contextual Isolated, de-contextualized
Practical, efficient Abstract, intensive
Robust, generalizable Singular (Setting, staff,
population)
Comparative Academic
Comprehensive Single outcome
Representative From ideal settings
75
76. www.re-aim.org EXTENDED CONSORT DIAGRAM
RE-AIM Issue Content Critical
Considerations
Total number potential settings
Settings Eligible Excluded by Investigator
n and % n, %, and reasons
ADOPTION
Setting and Agents Setting and Agents Other Characteristics
Who Participate Who Decline n and %
n and % n, %, and reasons
Of Adopters vs Non
Total Potential
Participants, n
Excluded by Investigator
REACH Individuals Eligible
N, %, and reasons
n and %
Individuals Enroll Individuals Characteristics
Not Contacted/
N and % Decline Other Of Enrolles vs.
N, %, and reasons N and % Decliners
Extent Tx Delivered Component A = XX% Extent Tx
IMPLEMENTATION By Different Agents Component B = YY% Delivered as
as in Protocol Etc.
Intended
Complete Tx Drop out of TX
Characteristics
N,%, and Reasons;
EFFICACY (n and % and
And Amount of change of Drop-outs vs
Amount of Change
(By Condition) (By Condition) Completers
MAINTENANCE Present at Follow-up Lost to Follow-up
Characteristics
(n and %) and Amount N, %, and Reasons
of Drop-outs vs.
a) Individual of Change or Relapse Amount of change or
(By Condition) Relapse (By Condition) Completers
Level
Characteristics
b) Setting of Settings that
Settings in which Program is Settings in which
Level Continued And/or Modified after Program not Continue vs
Research is Over Maintained Do Not
(n, %, and reasons) (n, %, and reasons)
*At each step, record qualitative and quantitative information and factors affecting each RE-AIM dimension and step in flowchart
77. The Challenge: If we have it all,
then will they really come?
• Impact = Efficacy x Reach /cost +
externalities
Not nearly as
much as we
could be!