Available via license: CC BY
Content may be subject to copyright.
Gates Open Research
Open Peer Review
Anyreportsandresponsesorcommentsonthe
articlecanbefoundattheendofthearticle.
METHODARTICLE
CUBES: A practical toolkit to measure enablers and
barriers to behavior for effective intervention design [version 2;
peer review: 2 approved, 1 approved with reservations]
ElisabethEngl ,SemaK.Sgaier 1-3
SurgoFoundation,Washington,DistrictofColumbia,20001,USA
DepartmentofGlobalHealth&Population,HarvardT.H.ChanSchoolofPublicHealth,Boston,MA,USA
DepartmentofGlobalHealth,UniversityofWashington,Seattle,Washington,USA
Abstract
Apressinggoalinglobaldevelopmentandothersectorsisoftento
understandwhatdrivespeople’sbehaviors,andhowtoinfluencethem.Yet
designingbehaviorchangeinterventionsisoftenanunsystematicprocess,
hobbledbyinsufficientunderstandingofcontextualandperceptual
behavioraldriversandanarrowfocusonlimitedresearchmethodsto
assessthem.Weproposeatoolkit(CUBES)oftwosolutionstohelp
programsarriveatmoreeffectiveinterventions.First,weintroduceanovel
frameworkofbehavior,whichisapracticaltoolforprogramstostructure
potentialdriversandmatchcorrespondinginterventions.This
evidence-basedframeworkwasdevelopedthroughextensive
cross-sectoralliteratureresearchandrefinedthroughapplicationin
large-scaleglobaldevelopmentprograms.Second,weproposeasetof
descriptive,experimental,andsimulationapproachesthatcanenhance
andexpandthemethodscommonlyusedinglobaldevelopment.Sincenot
allmethodsareequallysuitedtocapturethedifferenttypesofdriversof
behavior,wepresentadecisionaidformethodselection.Werecommend
thatexistingcommonlyusedmethods,suchasobservationsandsurveys,
useCUBESasascaffoldandincorporatevalidatedmeasuresofspecific
typesofdriversinordertocomprehensivelytestallthepotential
componentsofatargetbehavior.Wealsorecommendunder-used
methodsfromsectorssuchasmarketresearch,experimentalpsychology,
anddecisionscience,whichprogramscanusetoextendtheirtoolkitand
testtheimportanceandimpactofkeyenablersandbarriers.TheCUBES
toolkitenablesprogramsacrosssectorstostreamlinetheprocessof
conceptualizing,designing,andoptimizinginterventions,andultimatelyto
changebehaviorsandachievetargetedoutcomes.
Keywords
Interventiondesign,implementationscience,behaviorchange,behavioral
drivers,behavioralmodels,researchmethods,globalhealth,global
development.
1 1-3
1
2
3
Reviewer Status
InvitedReviewers
version 2
published
06Jan2020
version 1
published
18Mar2019
123
report report report
,UnitedStatesAgencyHope Hempstone
forInternationalDevelopment(USAID),
Washington,USA
,UnitedStatesAgencyforAngela Brasington
InternationalDevelopment(USAID),
Washington,USA
1
,LondonSchoolofRashida A. Ferrand
HygieneandTropicalMedicine,London,UK
BiomedicalResearchandTrainingInstitute,
Harare,Zimbabwe
2
,AshokaUniversity,Sonipat,Neela A. Saldanha
India
3
18Mar2019, :886(First published: 3
)https://doi.org/10.12688/gatesopenres.12923.1
06Jan2020, :886(Latest published: 3
)https://doi.org/10.12688/gatesopenres.12923.2
v2
Page 1 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
SemaK.Sgaier( )Corresponding author: semasgaier@surgofoundation.org
:DataCuration,FormalAnalysis,Investigation,Methodology,Resources,Validation,Visualization,Writing–OriginalDraftAuthor roles: Engl E
Preparation,Writing–Review&Editing; :Conceptualization,FormalAnalysis,Investigation,Methodology,ProjectAdministration,Sgaier SK
Resources,Supervision,Visualization,Writing–Review&Editing
Nocompetinginterestsweredisclosed.Competing interests:
ThisworkwasfundedbytheBillandMelindaGatesFoundationandtheSurgoFoundation.Grant information:
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
©2020EnglEandSgaierSK.Thisisanopenaccessarticledistributedunderthetermsofthe ,Copyright: CreativeCommonsAttributionLicense
whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited.
EnglEandSgaierSK.How to cite this article: CUBES: A practical toolkit to measure enablers and barriers to behavior for effective
GatesOpenResearch2020, :886(intervention design [version 2; peer review: 2 approved, 1 approved with reservations] 3
)https://doi.org/10.12688/gatesopenres.12923.2
18Mar2019, :886( )First published: 3 https://doi.org/10.12688/gatesopenres.12923.1
Page 2 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Introduction
Interventions that aim to shift what people do, and the choices
they make, are a major focus of global programs and policy.
With aggressive targets set by the United Nation’s Sustainable
Development Goals and limited funding for global develop-
ment, programs must leverage limited resources effectively and
efficientlyi. To achieve global development outcomes, such as
reducing maternal mortality or the number of HIV infections,
a plethora of diverse interventions have been designed and
implemented across the world that ultimately all aim to shift
behavior. Examples include increasing the demand for polio
vaccinations in India through mobilization activities1, develop-
ing checklists for nurses in Uttar Pradesh, India, to increase
adherence to labor and delivery guidelines2, and creating
incentives to drive voluntary male circumcision for HIV prevention
in Kenya3.
Not all these examples have resulted in successful and sustain-
able behavior change at the levels needed to have the desired
impact on development outcomes. Creating lasting change
through successful interventions is hard. First, it requires a
thorough understanding of why a target behavior is currently not
occurring in a given context. Designing and evaluating inter-
ventions in the field without a thorough understanding of the
underlying drivers of the target behavior in question can be an
inefficient use of time and resources4. Second, since people’s
choices and behaviors are influenced by many factors, a single
intervention is often insufficient to drive change. For instance,
a failure of malaria net uptake may be rooted in a lack of
availability or accessibility. But this external context is just
one part of the picture: beliefs on the part of the end-user, for
example about the benefits and risks of the nets, along with many
other factors, may also influence decision-making. All these
factors must be addressed through a well thought-through set
of complementary interventions (an intervention portfolio) to
significantly improve the usage of bed nets. Third, not all
people are the same. Sub-groups of people within the target
population can be differentiated by the varying drivers behind
their behavior, necessitating a different set of interventions to be
targeted at each subgroup5,6.
The right levers to focus on, and the portfolio of targeted
interventions to scale, are often far from obvious. Programs
need a holistic and practical behavioral framework that accounts
for and structures all types of barriers and enablers of behavior.
Many models of behavioral enablers and barriers exist, but
most focus either on systemic drivers, without addressing how
individuals can be motivated to respond to such changes7–9, or on
people’s beliefs, personality characteristics or cognitive biases,
neglecting context10–19. Several approaches, such as the COM-B
behavior change wheel, the Fogg model, and MINDSPACE,
focus most strongly on the appropriate types of interventions to
change behavior20–22. In global development, several organiza-
tions such as PSI, Johns Hopkins’ Center for Communication
Programs, and FHI360 have created behavioral models incor-
porating various subsets of drivers. Being application-focused,
they usually place great focus on incorporating guidelines on
implementation design and monitoring, communication, and
advocacy, or on providing a rich compendium of intervention
optionsii. However, no current behavior change framework helps
programs select the right research tool to assess an intervention’s
components. Programs need a repertoire of validated methods
that will help them assess distinct enablers and barriers of
behavior change in the field, because not all methods are effective
at measuring each type of enabler and barrier, and only a limited
set of methods are used in the development sector today.
In this paper, we provide programs with a practical two-part
toolkit to help programs design an effective portfolio of
interventions. We call the toolkit CUBES: to Change behavior,
Understand Barriers, Enablers, and Stages of change. First, we
present an evidence-based framework for understanding behav-
ior. The framework synthesizes stages of change, contextual
and perceptual drivers (which can act as enablers or barriers),
and layers of influencers, using evidence from multiple
sectors. The components were first articulated and applied in
the voluntary medical male circumcision program23, and we
later refined the framework through a thorough evaluation of
existing behavioral models, and by testing its applicability
and practicality in several large-scale development programs.
Second, to help programs generate actionable insights into the
components of the framework in their own context, we recom-
mend a set of research methods from various sectors and detail
their strengths and weaknesses, expanding the methodological
toolkit of qualitative interviews and quantitative surveys
that many programs use by default. We support this with a
decision tree to aid the choice of research method according to the
practitioners’ specific development program and context.
Real-world programs aim to drive change in complex dynamic
systems of people, places, and information channels, and
iUnited Nations. About the sustainable development goals. Available from:
https://www.un.org/sustainabledevelopment/sustainable-development-goals/.
Amendments from Version 1
Addressing the reviewers’ comments, we have amended this
article to further clarify and contextualize selected constructs
and research methods used in the behavioral toolkit, discuss
approaches from development organizations, and comment
further on the utility and validity of the toolkit.
Any further responses from the reviewers can be found at the
end of the article
REVISED
iiJohns Hopkins University Center for Communication Programs. JHU/CCP
Pathways models. Available from:
https://www.thecompassforsbc.org/sbcc-tools/jhuccp-pathways-models;
Population Services International. PSI behavior change framework “Bubbles”:
proposed revision. Available from:
https://www.psi.org/publication/psi-behavior-change-framework-revision/;
FHI360. SBCC framework. Available from:
https://www.c-changeprogram.org/resources/sbcc-framework-0
Page 3 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
CUBES can be applied at any level in the system. Ultimately, we
encourage practitioners use the toolkit to:
1) Understand determinants of behavior: barriers and
enablers, both perceptual and contextual.
2) Design idea-channel interventions that address barriers
and leverage enablers.
3) Design to all levels of change – individual, family,
society, and systems.
To illustrate the usability of the proposed approaches, we present a
case study showing how the CUBES framework and the methods
toolkit were applied in a large-scale program for voluntary
medical male circumcision in Africa. While we have developed
this approach through the lens of our programs in global devel-
opment, its principles can be applied to any behavior change
context.
Methods
Constructing a best-practice framework of behavior
Grounding intervention design in a comprehensive and action-
able behavioral framework is important. Many such models
exist, with varying levels of evidence supporting their compo-
nents. We surveyed models that were a) most influential and
b) had an evidence base confirming that they predict behavior,
rather than a comprehensive survey of all models in existence. We
defined ‘influential’ as having been used to guide behavioral-
intervention design across sectors, including health behaviors,
and ‘evidence-based’ as the existence of original research evalu-
ating the predictive power of individual components of a model
(such as ‘perceived severity of risk’24 or ‘conscientiousness’12)
on behavior.
We began with a list of models known to the authors, then
surveyed key approaches cited in these models, then searched
PubMed and Google Scholar for the following terms: ‘behavior’
AND (model OR framework OR drivers OR barriers OR facili-
tators OR enablers), and focused on the first five pages of
results in the search engines. The main search was performed
between October and December 2016, with additional targeted
searches until December 2018. We identified 17 models fitting
the criteria, each focusing on a different set of behavioral
drivers (Table 1). The drivers for which we found experimental
evidence of moderate to high predictiveness on behavior
were then placed into a framework. We subjected the drivers
included in the framework to critical review by four experts in
behavioral science, health psychology, and the development
sector, focusing on the drivers’ comprehensiveness and applica-
bility to global development. Finally, we applied the framework
to design research in our own large-scale programs (see sample
case study in this paper) to test for actionability and to further
refine its components.
Assessing and developing a curated set of research
methods to measure drivers
To help programs select appropriate research tools to capture
enablers and barriers to behavior, we surveyed methods
used across disciplines, using literature research and expert
conversations. Unless otherwise mentioned below, we then
applied and tested these methods in various combinations in our
own large-scale development programs to assess each method’s
feasibility, strengths and weaknesses in insight generation. In
Zambia and Zimbabwe, we investigated voluntary medical male
circumcision4,5,23, and in different areas of India we conducted
programs on household behaviors relating to maternal and child
health (family planning, antenatal care, institutional delivery,
postnatal care), tuberculosis care-seeking behaviors, and health-
care provider and front-line worker behaviors within medical
facilities and communities (unpublished reports).
Results
Cubes: a practical framework of behavior
Following our review of influential models of behavior, we
distilled their most evidence-based components into a practical
behavioral framework that programs can use to evaluate
existing evidence, conduct research to close evidence gaps, and
ultimately design interventions to match barriers to behavior
(Figure 1). The CUBES framework articulates three critical
components of behavior change. First, the path toward a target
behavior consists of a series of distinct stages. Second, the
progression through each of these stages is influenced by a set
of contextual and perceptual drivers. Third, these barriers and
enablers may be transmitted to the individual, reinforced, or
weakened through influencers (such as friends, family, or com-
munity members), either directly or through media channels.
Below we outline each component and the contributions of the
behavioral models surveyed (Figure 1).
Stages of the behavioral change process
Both contextual and perceptual drivers influence whether
individuals possess the knowledge needed for behavior change,
intend to act, or are already acting. These drivers either hinder or
facilitate progression along the stages of change.
In global development, many intervention programs focus on
enhancing awareness (passive knowledge) and skills (active
knowledge). However, it is possible to be aware of an option,
and even have the skills to do something, without intending
to take advantage of it. For example, in Uttar Pradesh, India,
increasing nurses’ skills did not always increase their practices
accordingly25. Clearly, knowledge does not equal action. There-
fore, understanding where people are on the pathway to behavior,
and why they are not moving forward, is key to designing
interventions that can move people toward action. This is an
essential first step for program designers to orient themselves
when evaluating behavior.
The Transtheoretical Model13 proposed a series of relatively
rigid stages: precontemplation (being unaware), contemplation,
preparation, action, maintenance, and termination, and adds a
set of interventions (‘processes of change’) to help an individual
progress from one stage to the next. However, the efficacy of
assigning individuals to very detailed stages has been called
into question in several systematic reviews26–28. In line with the
Health Action Process Approach16, the CUBES framework there-
fore divides the behavioral change process more simply into
three stages of knowledge (ending with the necessary aware-
ness or skills to engage in a behavior), intention (which can also
Page 4 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Table 1. Behavioral models surveyed, and their main advantages and limitations.
Models of behavior surveyed
Origin sector Main advantage Main limitation
Main focus: perceptual drivers
Health Belief Model Psychology, public
health
Very widely used, wealth of data demonstrating that
components explain some variance of behavior.
Neglects factors other than beliefs (biases, emotions,
habits) and context/environment.
MINDSPACE Checklist Public policy
(interdisciplinary
influences)
Concrete, practical checklist of evidence-based
techniques to effect change across many sectors.
Focuses almost entirely on unconscious processes and
corresponding nudges.
Integrative Model of Behavioral
Prediction/Reasoned Action
Approach/Theory of Reasoned
Action
Psychology Differentiates between different kinds of beliefs. Context/environment is only accounted for superficially.
Does not elaborate on how beliefs are formed; neglects
intention–action gap (focus on intention, but intentions do
not equal actions) and unconscious processes
(e.g. biases).
Transtheoretical Model
(Stages of Change)
Psychology Change-as-process over time is unique component. Evidence for six clearly delineated stages of change is
weak.
Health Action Approach Psychology Stages of change extended to repeat behaviors. No recognition of biases or contextual factors.
Self-determination Theory Psychology Differentiates between extrinsic and intrinsic motivations
and names drivers for intrinsic motivation.
Focused on only one aspect of decision-making: ignores
all non-motivational individual and systemic factors.
OCEAN model of Personality Psychology Trait-based models of personality reliably explain part of
the variance in (health) behaviors.
Factors only account for part of an individual’s personality,
which in turn only accounts for parts of their behavior.
Personality has limited predictive power for a specific
behavior, but rather for patterns of behavior.
Theoretical Domains Framework Psychology Validated and extensive list of barriers and facilitators. Biases and personality mostly absent.
COM-B (‘capability’,
‘opportunity’, ‘motivation’ and
‘behavior’)
Psychology Emerging from the Theoretical Domains Framework,
the first model to link different intervention and policy
categories to behavioral drivers in a systematic and
parsimonious way.
Limited dimensions of drivers of behavior makes the model
easy to understand, but it does not provide much detail.
Fogg Behavior Model Psychology Similar to COM-B: behavior is understood as a mixture of
motivation, ability, and prompts. Uniquely, strong focus on
characteristics of contextual cues that are most effective in
shifting behaviors.
Model’s view of motivation and ability is simplistic.
Expected Utility Theory and
Prospect Theory
Behavioral economics Gives insight into appraisal process of a decision. Accounts for a small subset of drivers of behavior.
Collection of cognitive biases
and heuristics
Behavioral economics,
psychology,
neuroscience
Insight into ‘automatic’ and unconscious drivers of
behavior.
Accounts for only one aspect of decision-making.
Evo-Eco Approach Evolutionary biology,
neuroscience
Evolutionary aspects of behavior and embodiment given
due importance (e.g. disgust as a primal emotional
reaction).
Views behavior as largely caused by automatic/habitual
processes.
Main focus: contextual drivers
Social-Ecological Model Psychology Shows the dynamic ways that different strata of the social
sphere influence each other.
Does not account for perceptual drivers of behavior.
Social Cognitive Theory Psychology Shows how social influence can mediate some perceptual
drivers.
Focuses most on self-efficacy, little emphasis on context.
Practice Theory Sociology, anthropology Focuses on environmental constraints on behavior. Neglects individuals, focus on theoretical level rather than
testing components’ explanatory value.
Diffusion of Innovations Theory Communication studies/
sociology
Clear guidance on techniques to reach different segments
of a population to adopt a novel behavior.
Segments individuals in a specific way (how receptive
they are to an innovation), does not account for other
environmental and cognitive factors driving decision-making.
Page 5 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Figure 1. The CUBES behavioral framework. Contextual and perceptual drivers combine to act as enablers and barriers along an
individual’s path from knowledge – encompassing awareness and skills – to intention (or motivation to act towards a goal) and action and
beyond. Layers of influencers can affect these drivers and reach an individual through various channels.
be understood as a plan of action towards a specific goal29), and
action. While the path to action and beyond can be understood
as a sequence, people can also move back and forth between
already-reached stages.
Almost no models of behavior focus on repeat behaviors and
habits, despite the importance of sustained change in real-life
contexts. Different drivers become more or less important for
repetition and habit. The intention to repeat a behavior becomes
more likely when two factors converge: a positive evaluation
of the previous experience (‘experienced utility’), and a revised
self-efficacy in comparison to what was expected16,30. However,
merely repeating a behavior does not create habits. The
creation of a habit requires the development of automaticity (the
behavior is performed with low awareness, control, attention,
or intention) and an association of a behavioral response with
contextual cues and an experience of reward31–33. Some behavioral
drivers that can be targeted for single actions, such as intentions,
goals, or beliefs, are much less important to the formation of a
habit16,31,33. For example, delivering a child in a health facility
is a one-time behavior, for which intention, goals, and beliefs
matter greatly, but exclusive breastfeeding after the child is born
is close to a habit, which after initiation does not require a woman
to form the intention from scratch every single time. Instead of
targeting beliefs and intentions, then, restructuring of environ-
mental cues and conscious inhibition of unwanted habits may
have greater success in creating lasting habits34,35.
Contextual drivers of behavior
Behavior emerges out of a complex system of interactions
between individuals and the systems they act in. The Social-
Ecological Model introduces the concept of ecosystems, which
examines the dynamic ways that different layers of the social
sphere influence each other8,36,37. Adapting the Social-Ecological
Model8,36,37, social norms and customs influence individuals in
an ecosystem in several layers. These norms may be explicit,
but they can also emerge from an implicit layer of what
Practice Theory defines as ‘shared cultural knowledge’ that
expresses itself as routines and habits7,38. Social norms are a
construct that can only exist on the level outside the individual:
through collective behavior and ‘shared knowledge’, norms
describe a set of practices of what other people do (descriptive
norms), or prescribe what people should do (prescriptive norms).
Both of these may influence attitudes and behavior39. Unlike
individual-level beliefs, norms usually imply some consequence
to the individual should they deviate from the norm, such as
disapproval40.
Structural factors are further contextual drivers that may shape
and constrain perception and action. These aspects are usually
not fleshed out in behavioral models. However, unless these
constraints are removed, interventions acting on perceptual
drivers will not allow for the target behavior to occur. We
propose differentiating between infrastructure, policies and laws,
and systems and processes, all of which vary strongly with their
Page 6 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
respective context. For example, infrastructure drivers may
include availability and condition of roads leading to a health
facility, equipment to perform a test, or seeds for farmers to use.
Policies and laws constraining behavior can exist on several
levels, from national laws to facility-level guidelines. Systems
and processes could mean supervision, training, feedback, or
incentive systems under which healthcare providers operate in
a facility, or the tools teachers have available to plan lessons and
receive feedback on their tuition. Finally, demographic factors
such as age or education level act as contextual constraints on
behavior.
Perceptual drivers of behavior
Multiple models recognize that behavior is also shaped by
perceptual drivers that, together with contextual drivers, combine
to make a target behavior more or less likely.
Biases and heuristics. Most aspects of an individual’s cognition
and behavior are influenced by ‘automatic’, often unconscious,
mental shortcuts or rules of thumb (heuristics) that can bias
decision-making. Many biases can be difficult to change, but
knowing about them makes it possible to construct environ-
ments where the best decisions are the easiest. A typical example
is setting the default option of a pension program to ‘opt-out’
instead of ‘opt-in’: staying in a pension program after auto-
enrollment is easier than making the effort to join in the first
place41,42. Examples of heuristics and biases are the optimism,
confirmation, and availability biases, anchoring-and-adjustment,
hyperbolic discounting heuristics, and the status quo and repre-
sentative biases17,18,41–47. MINDSPACE, an influential checklist
for behavior change developed in part by UK government
agencies20,48, is an example of using biases (such as a bias to
choose the default option) as tools to design interventions.
Beliefs. Beliefs are formed by learned experience, differentiating
them from biases (which humans share to varying degrees as
a result of how our brains evolved). In the Integrative Model
of Behavioral Prediction/Reasoned Action Approach, beliefs
about what outcome can be expected from a behavior (called
‘attitudes’ in that model), normative beliefs (how others will
judge a behavior), and beliefs about the extent of one’s control
over the behavior (self-efficacy) all influence intention, which
is seen as the main driver of behavior19,49,50. Experiments have
shown that some beliefs predict behavior better than others. For
example, perceived control (self-efficacy) and beliefs about
a behavior’s outcome are better predictors of behavior than
normative beliefs51,52. Indeed, self-efficacy beliefs have emerged
as a strong predictor of behavior across other models, such as
Social Cognitive Theory, the Health Belief Model, and the
Transtheoretical Model. The strong influence of self-efficacy on
behavior has been shown experimentally in many studies53–57.
Outcome expectations are another example of a belief. An
individual appraises a potential behavior by weighing perceived
costs and perceived benefits, which may be emotional (‘How
will the outcome of the behavior make me feel?’), social (‘How
will others judge me?’), or functional (‘How does this help or
hurt me?’). Outcome expectations, together with the perceived
severity of an outcome, the perceived susceptibility to that risk,
and self-efficacy, are central to one of the most widely-used
models of behavior, the Health Belief Model14,24,58. Evidence
from several systematic reviews shows that increasing perceived
benefits and decreasing perceived costs to a behavior will
be most likely to cause an individual to engage in the target
behavior10,24,59,60. Beliefs around (professional or social) self-
identity may also be predictive of behavior. For example,
environmental self-identity, or seeing oneself as ‘a person who
acts environmentally-friendly’, is related to several environ-
mental behaviors61. However, empirical research on identity and
behavior is still emerging.
Emotion. The experience of emotion (affect) also drives
behavior. Affect arguably colors all perception and powerfully
shapes decision-making20. While affect is used as a distinct
driver of behavior in the MINDSPACE checklist20, the COM-B
framework22 includes it as a rapid, automatic component of
motivation. One example of targeting affect to drive motivation
to engage in a new behavior can be seen in an intervention
promoting soap use in Ghana: education around the benefits of
soap did little to drive up its popularity, but emphasizing the
feeling of disgust from ‘dirty hands’ resulted in significantly
increased soap use62. This emphasis on automatic emotional
responses, such as disgust or a desire to conform with others,
is a key component of the Evo-Eco behavioral model63.
Personality. Personality traits are not often included in models
of behavior—not even in the Theoretical Domains Framework,
which includes the perhaps most comprehensive list of
barriers and facilitators64 —but they can strongly influence an
individual’s propensity to engage in and maintain health
behaviors65,66. Current applications of personality models to
behavior prediction focus on aggregates of behavior, or
behavioral patterns of behavior such as going to check-ups and
regular physical activity (preventive health behaviors). Such
models often do not attempt to predict single instances of
behavior, which has been shown to be much less reliable65.
Currently, the dominant personality model with a large evidence
base behind it is the so-called Five-Factor Model12,66–68 with
the five broad traits of openness to experience, conscientious-
ness, extraversion, agreeableness, and neuroticism (‘OCEAN’).
Conscientiousness appears to be an especially strong predictor
of behavior patterns, such as sticking to preventive health
routines65,66.
Influencers and channels
Following from the Social-Ecological Model8,36,37, influencers
surround an individual in layers. Family and friends or peer
groups are the closest layer to the individual. The next layer
usually consists of relationships in the community, for example
in workplaces, schools, and neighborhoods. The most distant
layer is the larger social context. Influencers can reach
individuals either directly or at scale via various media
channels, which is important information to determine how
to deliver interventions. For example, female self-help groups
can serve as a channel for rural women in India to reinforce or
change social norms relating to a certain target behavior. For
Page 7 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
an intervention to work, the content, the type of influencer,
and the channel through which they reach individuals must be
identified as relevant to the target individuals. Finally, social
influence is often not intentional as a self-help group might be,
but less explicit influence may not be any less powerful.
Interaction between the ‘building blocks’ of CUBES
Depending on the behavior, all the drivers mentioned above
will be relevant to varying degrees to any one individual, and
their combination will result in a larger or smaller tendency
to act. The elements of CUBES influence each other deeply.
For instance, punitive supervision (systems and processes) by
medical-officers-in-charge (influencers) in a health facility
might lead nurses on the receiving end to experience high
anxiety (emotions), and to beliefs that trying hard will not
result in any benefit to them (outcome expectation belief). It is
important for programs to disentangle these components, even
when they influence each other, because this determines what
kind of intervention is best placed at what level.
Therefore, the enablers and barriers in the different ‘building
blocks’ of CUBES can be seen as a checklist for programs that
they can utilize to design effective interventions.
Models of behavior that simplify to the point of ‘motivation’ or
‘ability’ (such as COM-B or the Fogg model) are simpler, but
also less actionable for programs on the ground. For example,
if a study finds a lack of motivation to go to the doctor despite
having symptoms indicative of tuberculosis, this alone is not
actionable by an intervention. Instead, programs need to
know where potential patients are on the knowledge–intention
spectrum (they have clearly not yet taken action), whether
they perceive their symptoms and the disease as a danger
to their and others’ health (risk perceptions), whether they
think going for a check-up will actually alleviate symptoms
(outcome expectations), whether they feel able to skip work and
other responsibilities to attend an appointment (self-efficacy),
whether appropriate facilities are even available, affordable,
and accessible (structural factors), and whether there is stigma
involved in seeking care (social norms). All these components
would feed into the concepts of ‘motivation’ or ‘ability’, but
require very different interventions to effect change.
Once CUBES has been used to understand and categorize
existing evidence, evidence gaps can be closed in a focused way
with primary research. Not all components of CUBES are best
captured and intervened on in the same way. In the following
sections, we introduce a method mix designed to identify a
comprehensive set of drivers.
Methods of measuring drivers
Qualitative interviews, focus groups, and quantitative surveys are
some of the most common methods of insight generation in the
global development. These strategies complement each other:
qualitative methods are best suited to exploration and capturing
nuances, whereas quantitative methods are indispensable for
discovering patterns and weighing the relative importance of
different drivers, which is essential for developing interventions
that address the barriers that matter. Here, we propose two
overarching considerations for programs to add value to the
methods used.
First, existing methods can be improved by using the CUBES
framework as a checklist against survey or discussion-guide
items, to check whether a comprehensive set of enablers,
barriers, influencers, and stages is captured. Too often, methods
such as quantitative surveys remain at the level of measuring prac-
tices (or ‘what’ data) and demographics, at the expense of the
‘why’, or perceptual and contextual drivers of the target behaviors.
Second, programs could benefit from expanding their own toolkit
of methods by selecting approaches to investigating behavior
from a variety of sectors. The right method will depend on what
type of data needs to be captured, and whether the purpose of
research is exploration or testing specific hypotheses. A method
mix can also help counteract any weaknesses of individual
methods.
Choosing the right type of method for different stages of
research
Research methods can be divided into descriptive, experimental,
and simulated approaches; the last two are relatively under-used
in global development. Whether qualitative or quantitative,
descriptive methods such as interviews or observation aim to
describe and explain behavior without testing the effects of
manipulating variables systematically. While they can explore
‘how’ and ‘why’ questions (and are therefore often called ‘explora-
tory’), they do not have the ability to systematically relate the
effect of change to outcomes (‘confirmatory’). Experimental
approaches can be used to test hypotheses and find causal rela-
tionships by systematically varying variables and testing their
effects on outcomes. However, they lack the ability to survey
a broad spectrum of factors at once that exploratory methods
provide. Finally, simulated methods can assess cause–effect
relationships in a virtual environment when experimental field
methods are not possible or are too complex, but they rely on
many assumptions to construct the simulation.
Expanding the descriptive toolkit
Journey mapping. Developed and primarily used in market
research69,70, journey mapping systematically tracks people’s
experiences and interactions with a product, service, or life event
over time, as people form beliefs about the product or event
and make decisions, perhaps via influencers, to interact with or
avoid it. This method is especially well-suited to get a sense
of stages of change. Journey mapping can also help form
hypotheses about segments of customers who share distinct
characteristics in order to target them with bespoke messages
via different channels. In addition, it can be useful in generating
hypotheses about underlying behavioral drivers that can then be
tested further.
Journey mapping uses many different techniques to collect
data, including one-on-one qualitative interviews, focus groups,
ethnography, web analytics, customer reports via apps, and
(qualitative) network mapping69,71. Below, we show in a case
study how journey mapping was successfully integrated in a
program understanding decisions around voluntary medical male
circumcision5.
Page 8 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Observation. Observation is a versatile and routinely used tool
in global development to collect data about what people do, the
context that surrounds them, and how they interact with processes,
objects, or each other72. The spectrum of observation techniques
ranges from researchers interacting closely with communities
(participant observation), to covert or overt ‘natural observation’
without participation, to controlled observation, where proce-
dures are highly standardized. Measuring time and frequency of
practices of medical residents in hospitals73, or of nurse practices
in hospitals in India25, are typical examples of controlled
observation. Such time-and-motion studies observe the time
taken and actions of participants executing distinct components
of a process. Observation is also a tool to measure contextual
drivers such as infrastructure or processes. For example, facility
and infrastructure audits commonly combine observations with
interviews of key stakeholders to track characteristics such as
hospital staff coverage, equipment availability, or communication
tools74.
To get a holistic sense of contextual drivers, observation can
track more than behaviors or supplies. Instead, immersive
observations in the participants’ natural environment could be
structured to assess the set of contextual enablers and barriers
outlined in the CUBES framework. We call this approach
‘structured immersive observation’. For example, in a recent
observational study on nurses in healthcare facilities, we
measured key contextual dimensions as follows. First, we assessed
facility infrastructure available to nurses: whether equipment
and drugs required for routine tests were available at the time
of testing, staff coverage throughout the period of observation,
availability of beds, water, and electricity, and transport options
for patients to be referred. Second, we assessed systems and
practices, such as interactions with and feedback from other
staff, time spent with patients and tests performed, documenta-
tion systems and job aids, and training records. Third, we assessed
community norms by observing community interactions and
communications with the nurse.
Enhanced quantitative surveys. Quantitative surveys are a critical
tool to obtain insights on many enablers and barriers to behavior
simultaneously and at scale, and consequently are a mainstay of
global development research. Survey design is a broad field with a
large array of approaches. Here, we recommend three key
techniques that can enhance the design of quantitative surveys
to measure potential behavioral drivers. Survey questions
can be structured to account for as many components of
CUBES as possible. In sensitive contexts, surveys can also be
enhanced to counteract respondent biases. Finally, while not
within the scope of this article, programs would benefit from
leveraging quantitative data for insight beyond descriptive
analyses: for example, population segments can be found for
targeted intervention design5.
Driver-structured surveys. Programs can use the CUBES
framework as a checklist to assess whether a survey captures
the range of potential contextual and perceptual enablers and
barriers, influencers and channels, and stages of change relating
to behaviors of interest. We often find surveys only measure a
narrow set of drivers, which presents the opportunity to generate
a more holistic view of behaviors of interest and the system.
This can also be enhanced by adapting existing validated tools,
such as scales testing personality or self-efficacy (see below).
Standardized scales. A simple and high-yield modification to
quantitative surveys is the inclusion of previously validated and
standardized rating scales relating to CUBES perceptual drivers.
Standardized scales that test specific cognitive processes
include the Risk Propensity Scale75 and the ten-item General
Self-Efficacy Scale76. Types of emotions and their felt strength
have also been widely measured with graphical rating scales,
such as the Self-Assessment Manikin77. All these scales can be
flexibly adapted to specific contexts. For example, in a study
investigating women’s propensity to engage in breast cancer
prevention, self-efficacy was asked in two items: ‘the extent
to which participants were confident that they could conduct
breast self-exams every month; and when they conducted a
breast self-exam, how confident they were in their ability to
identify a “lump that needs medical attention.”’78.
Personality tests, widely used in the private sector, also use
standardized rating scales. Many studies show some predictive
value of the OCEAN model’s ‘Big Five’ personality traits on
health behaviors65, especially conscientiousness66. Questionnaire
designers can tap a large number of validated instruments to test
OCEAN components, such as the public-domain International
Personality Item Pool79.
All standardized scales have the advantage of using previously
validated instruments, and that individual differences can be
captured with relatively little effort. A limitation is that, like all
self-reports, such tests are susceptible to reporting bias, since
participants can deduce or guess socially desired responses
(those that the respondent thinks will make them appear in a
favorable light). Responses may therefore be compatible with the
participants’ sense of self rather than their actual behavior.
Informal confidential voting interviews and polling-booth
surveys. Methods that stress anonymity and confidentiality, such
as polling-booth surveys (PBS) and the Informal Confidential
Voting Interview (ICVI) approach, can be used to probe sensi-
tive topics, as they counteract social desirability bias. The ICVI
consists of a one-on-one interview followed by self-completion
methods80. Similarly, PBS collects feedback from a group of
people who respond anonymously and independently through a
ballot box. Comparison of one-on-one interviews with PBS81 and
with ICVI82 on sexual risk behavior in Indian men and women
demonstrated their value, as more risky behaviors were reported
with each method.
Standardized patients. Standardized patients (SPs) are people
trained to play the role of a patient with certain medical and
personal characteristics, who interact with healthcare providers
in a realistic setting. SPs are comparable to ‘mystery shoppers’
in consumer research83, in that the healthcare professional
does not know the patient is not real84. In other scenarios, both
parties know about the setup85. The SP method has been applied
Page 9 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
to investigate healthcare provider behavior in many contexts, such
as prescription practices of pharmacists83. It has also been used
to assess how doctors communicate with their patients, such
as how surgeons disclose medical errors to patients85. The SP
approach can compare expected with actual behaviors, and
analyze communication, such as the content of advice given
to patients84. While the method on its own is very well suited
to capture practices and contextual drivers such as infrastruc-
ture and processes (of patient interaction), other drivers, such
as beliefs or biases, are less accessible to investigation.
Social network analysis. Social network analysis (SNA) maps
relationships between people, organisms, groups, or devices86.
When analyzing behavior, SNA can be an excellent tool to
describe which influencers and which channels are most impor-
tant to transmit certain norms and information. SNA can be both
qualitative or quantitative. Data can be generated from surveys,
ethnography, or observation, or mined from existing resources
such as GPS coordinates or twitter messages87,88. Many field
studies have used SNA to focus on where in the system to inter-
vene. For example, in Uganda, researchers mapped the process
of obtaining a diagnosis for tuberculosis through provider and
patient networks, and the steps where delays were most common
could be identified89. Ultimately, SNA is a flexible and versatile
method, but specifically focuses on identifying centers of
influence.
Leveraging ‘passive’ datasets. As in the SNA example, insight
can also be generated from leveraging ‘passive’ datasets, gener-
ated for a different original purpose, without direct interaction
with or observation of respondents. Examples are information
obtained from GPS, satellites, and sensor systems, as well as
other databases. To investigate contextual drivers, satellite images
can map physical conditions of the built environment, which can
then be related to behaviors and drivers from other datasets90.
For some audiences, social media data can be an appropriate
source of aggregate estimates of positive or negative sentiment91.
More analog data sources can also help generate insight, as in
an analysis of Kenyan newspaper articles about voluntary medi-
cal male circumcision, which provided insight on the types of
risks that were presented to readers92.
Assessing decision drivers through ‘in vitro’ experiments
We propose that programs can benefit from ‘in vitro’ experimental
methods before testing specific interventions in lengthy trials
in the field. Experimental methods that track the decisions
participants make in laboratory-like conditions serve several
purposes: programs can systematically change and test enablers
and barriers to behavior, predict behaviors in response to specific
interventions, determine those features of a service or product
that are most likely to align with the customers, and forecast
the market size of a product or features based on predicted
behaviors. All these factors narrow down the potential character-
istics of an intervention to be tested in the field, and ultimately
make the design of effective interventions more probable.
Discrete choice experiments. Discrete choice experiments
(DCE), extensively used in market research, uncover preferences
and value attribution from the choices that participants make,
rather than from the participant disclosing them. They are a
powerful tool to predict behaviors ‘in vitro’, assess which
features of a product or message are most important to the
customer, and to forecast market shares of products. DCE have
been shown to be predictive of health behaviors93. For feature
selection, participants are typically shown multiple iterations of
sets of products with varying features. In each trial, the participant
picks one option. For example, a discrete choice experiment in
South Africa evaluated which characteristics of HIV prevention
products, such as the method of use, or the protection against
diseases other than HIV, would be most valued by participants94.
From participant choices, a model can be built showing
which level and which combinations of a product’s features
predominantly drive decisions, where the tipping points of
certain preferences lie, and forecasting product market share.
DCEs can also be used to test various ‘what-if’ scenarios, and
results can then be used as a funnel to select the most promis-
ing attributes for a field intervention. Simulated test market-
ing is a related concept, in which the consumer is asked to
make choices in a realistic environment, with similar systematic
manipulation of test variables.
Decision games. Purely quantitative DCE approaches are mostly
used to cycle through permutations of features and types of
products or interventions. However, a related ‘decision game’
method mixing quantitative with qualitative elements can help
investigate which behavioral drivers most influence choices
made by participants. In a recent study with healthcare providers
in Uttar Pradesh, India, we used such a group ‘decision game’:
participants were given a set of scenarios, each with a set of
response options, and were asked to choose the option they thought
other participants would select (unpublished reports). Response
options coded for different behavioral drivers, and participants
were later qualitatively probed on their choices. For example,
nurses were asked which nurse in a scenario was likely to
be most stressed: the one working in an understaffed facility
(coding for infrastructure-staffing), the one dealing with demands
from patient family members (influencers-community), or the
one facing constant scrutiny and accountability from supervisors
(systems and processes – supervision).
Implicit attitude tests. Few research methods are suitable to
measure enablers and barriers that respondents cannot or do
not want to report. Over the last two decades, experimental
psychology has developed a battery of experimental approaches
to measure ‘implicit’ biases, or biases that are inaccessible to
conscious awareness and self-report, but nevertheless influence
behavior. The underlying concept of implicit attitude tests is
that our brains perform unconscious evaluations of concepts,
people, and objects, which have arisen from past experiences
and cannot be measured by explicit questioning.
The most widely-used implicit attitude test is the Implicit
Association Test (IAT). This test is based on the concept that
participants can perform a task more quickly when they see two
concepts as related than when they do not associate them with
each other95. IATs have been used to measure a plethora of social
Page 10 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
iii Project Implicit Research Group. Project Implicit 2011. Available from: https://
implicit.harvard.edu/implicit/ethics.html.
stereotypes, such as gender and racial biasesiii, for instance in
the Democratic Republic of Congo96. In market research, the
IAT has been used to gauge consumer attitudes toward differ-
ent products97. The predictive value of implicit attitude tests on
behavior is still under debate98,99. For this reason, and because
IATs can only test a small set of associations within a test
that requires training participants, we only recommend implicit
tests when a behavior is likely to be influenced by a specific,
deep-seated bias that respondents are unlikely to report.
Using simulations to model ‘what-if’ scenarios
Simulations have the unique advantage that complex ‘what-if ’
scenarios can be explored at the push of a button, which can be
used to supplement and inform data collection or to optimize
interventions. Through the construction of ‘virtual worlds’,
mathematical models can simulate the impact of implementing
certain interventions, or of targeting interventions to specific
sub-groups. They can also generate hypotheses on what a likely
driver for behavior might be. As an example, agent-based
models have been used to simulate the large-scale effects that
emerge from the actions of many single agents, such as the
spread of disease or of social norms and beliefs100,101. Similarly,
Bayesian cognitive mapping builds probabilistic models of
the likelihood that agents make certain decisions102. However,
simulations are only as good as the model and assumptions that
underlie them. The relevant and correct starting parameters
must be chosen with caution, which includes a degree of
subjectivity103, and generalizations from a model based on specific
assumptions are limited. Table 2 summarizes the approaches
discussed, and their strengths and weaknesses.
Choosing the right method at the right time for the right
purpose
The goal of any program is to implement successful interventions
in the field. Figure 2 depicts the process of setting the research
agenda, generating insights, and designing and optimizing
interventions that programs can use, depending on the knowledge
level at the start of the research process. First, when programs
define a target behavior to be changed, they can evaluate exist-
ing evidence against the components of the CUBES framework.
This can be done either from existing literature or analyzed from
datasets, or both. To directly choose an intervention that works,
programs must already have narrowed down specific drivers to
intervene on. This may be the case in a data-rich environment;
in other cases, political or resource constraints limit what is
testable in the field. In such cases, primary research may not be
required or appropriate. On the other end of the spectrum, a
program might know what it wants to change—for example, to
increase the uptake of modern methods of contraception—but
have little systematic knowledge of the types of drivers that may be
involved. In this case, exploratory research is warranted (‘insight
generation’ in Figure 2). We have found that a quantitative sur-
vey often provides the best practical balance between assessing
many components of CUBES at scale. Either before a survey (to
inform its design) and/or after (to dive deeper into specific
findings), specific descriptive or experimental methods offer
particular strengths assessing specific CUBES components and
can supplement a survey (Table 2).
Descriptive methods can be qualitative or quantitative, and many
can take both forms. To choose a qualitative or quantitative
focus, programs can consider whether the freedom to explore
limited aspects in depth is most important (which means a
greater focus on qualitative research), or representativeness
at scale and the relative likely impact of each driver (which
points to quantitative methods). Often, programs use prelimi-
nary qualitative research to the inform the design of quantitative
research, especially in field settings where quantitative research
is conducted in person and therefore is expensive and time-
consuming, whereas small-sample qualitative research is less
resource-intensive. The freer structure of qualitative research
also typically allows for follow-up questions and clarification.
However, we recommend that this order not be followed by
default, but rather examined on a case-by-case basis. In our
experience, qualitative research can sometimes divert resources
from quantitative research, which can be wasteful if results
from a smaller sample cannot be generalized. Instead, quali-
tative and quantitative research can also be run in parallel,
with a complementary focus on different drivers; pockets of
qualitative components can be mixed into qualitative research;
or qualitative back-checks can be conducted after quantitative
research.
In global development, many programs tend to focus on
descriptive methods for insight generation, followed by field
implementation. In the field, randomized controlled trials tend
to be seen as the gold standard for assessing the effectiveness
of an intervention, even if they are not always employed in
practice. In addition to preliminary evidence synthesis and using
a flexible methods toolkit for specific deep-dives, we argue that
this approach misses a key step, namely ‘in vitro’ experimental
methods. These methods can narrow down the many potential
hypotheses emerging from exploratory research, so that only the
enablers and barriers likely to be most impactful are ultimately
tested in the field. This optimization can be used to choose
between different types of interventions (such as monetary incen-
tives versus more information on risks and benefits), as well as
the components of a specific interventions (such as the magnitude
of incentives likely to be most effective). If rich descriptive data
is already available, and so a limited set of specific hypotheses
around limited drivers can be formed from the outset, programs
can directly skip to this step. This step can also be done in par-
allel with exploratory research, if evidence is strong on specific
drivers but weak on others and a holistic picture is desired. As
detailed above, purely simulation-based methods can be of use here
to model the effect of many different changes. As a result of this
step, field testing will be based on much stronger evidence.
Applications of the cubes toolkit: case study
Designing interventions to increase the uptake of voluntary
medical male circumcision (VMMC)
In Figure 3, we briefly outline the methodological choices made
to investigate enablers and barriers to uptake of voluntary medi-
cal male circumcision (VMMC) in Zambia and Zimbabwe.
Page 11 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Table 2. Overview of enhanced and novel insight generation methods as part of the CUBES toolkit.
Method Primary insight gain Most testable CUBES
components
Method type Advantages Disadvantages
Descriptive
Journey mapping Tracking experiences and influencers
over time
Stages of change,
beliefs, emotions,
influencers and channels
Qualitative Mapping experience, influencers
drivers over time
Self-report
Observation
▪ Time-and-motion
▪ Infrastructure audits
▪ ‘Structured immersive
observation’ (SIO)
Systematically tracking practices/
duration (time and motion), infrastructure
and supplies (audits), and CUBES-
structured contextual drivers (SIO)
Contextual drivers: structural,
systems and processes;
behaviors observed
Quantitative Behaviors and contextual drivers
and barriers can be measured
in their natural environment, in a
standardized and replicable way
Observed participants and
researchers are prone to behavioral or
recording biases, respectively.
Enhanced surveys
▪ Driver-structured surveys Using CUBES as checklist aids
systematic capture of enablers, barriers,
influencers, and stages of change
All Quantitative Holistic overview of all potential
drivers possible in one dataset
per respondent
Not all drivers are equally well
captured by self-report
▪ Informal confidential
voting interview (ICVI),
polling-booth surveys
Adding anonymized components
encourages responses on sensitive issues
Social norms, beliefs Quantitative,
ICVI also
qualitative
Greater disclosure on sensitive
issues
Yes/no response format leaves no
room to explore; anonymous data can
only be analyzed in aggregate
▪ Standardized scales Testing perceptual drivers with
validated, standardized tools (e.g. self-
efficacy, risk propensity, personality)
Beliefs, personality Quantitative Ready-made aids to assessing
perceptual drivers and barriers
Prone to self-report bias
Standardized patients Tracking behaviors, context, and
interactions through simulated
‘patients’ with a set of standardized
characteristics
Contextual drivers:
structural, systems and
processes; behaviors
observed
Quantitative,
qualitative
components
Standardization allows for
comparability, realistic setting
and covert data collection for
realism
On its own, is mostly limited to ‘what’
data and cannot explore drivers for
practices.
Social network analysis Revealing direction and strength of
relationships in a system
Influencers, social norms Qualitative or
quantitative
Versatile (qualitative or quantitative),
useable for networks of any size
and type, unique method of
identifying influential targets for
potential intervention
Network modeling can only investigate
limited drivers and barriers in one
network.
Leveraging ‘passive’
datasets
Generating insights from sensor, mobile
phone, satellite, GPS, social media,
and other databases, with no direct
customer interaction
Different, depending on
dataset
Quantitative Large-scale existing datasets can
be tapped and integrated with
other research methods, ‘bird’s-
eye’ view of context possible
Passive nature means no opportunity
to probe; existing datasets may not
focus on key customer groups
‘In vitro’ experimental
Discrete choice
experiments
Participants make repeated choices
between a set of options whose
attributes are systematically varied, in
order to uncover which attributes are
most important
All, least useful for
biases
Quantitative Quick to develop, test, and
analyze. Participants do not have
to explain ‘reasons why’, which
are inferred from choices
Correlation of hypothetical with real-
world choices is difficult to predict.
Providing response options that clearly
represent distinct drivers and barriers
is not trivial
Decision games Gamified, social experiment version of
a discrete choice experiment
All, least useful for biases Quantitative
and/or qualitative
Gamification increases
engagement, asking about what
other participants select instead
of own choices circumvents
some respondent biases
Same as discrete choice experiments;
qualitative approach is difficult to
interpret
Implicit attitude tests Using reaction time in response to
tasks and other measurements to
determine whether participant sees
concepts as related or not
Biases Quantitative Unique method to assess strong
biases inaccessible to self-report
or observation
Method not well tested in low-resource
settings, correlation of output and
behavior not obvious, each test can only
test a limited number of associations
Simulated
‘What-if’ simulations Modeling simulated decision-making
or outcomes in response to changing
parameters in complex systems
All can be simulated n/a Unlimited permutations of
changes (‘what-if scenarios’) in a
complex system can be modelled
Any model will only be as good as the input
data (which does require field-level input),
highly specialized skills required
Page 12 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Figure 2. Decision aid for choosing the right research approach at the right time, for the right purpose.
VMMC is a highly cost-effective intervention for preventing
HIV acquisition that is being scaled up in eastern and southern
Africa5,104. The achievement of the program’s ambitious targets
necessitated shifting the behavior of many men in the commu-
nity who either did not consider circumcision, or if they did, did
not take action. Therefore, the program needed to understand
the multiple interacting factors that facilitate or inhibit men’s
decision to get circumcised, and to test interventions to address
those factors. A synthesis of previous studies revealed a variety
of existing insights on many behavioral drivers, such as concerns
around pain, complications, or cost, as well as patterns of
influencers105. Analysis of a large quantitative survey further
showed a relatively small awareness-intention gap, but a large
drop from intention to action106 A total of 64% of men intended
to get circumcised, but only 11% did. However, a holistic view
assessing the prevalence of each of these, and other, drivers and
their relative strength was lacking, and most research was either
small-scale or qualitative5. The existing studies could not
answer why there was a strong intention-to-action gap. Also,
the research did not examine or reveal the heterogeneity among
men—the fact that a given enabler or barrier may be important
to one man, while not as relevant to another.
(BUIFSBWBJMBCMFFWJEFODFGSPNMJUFSBUVSF
BOENBQBHBJOTU$6#&4DPNQPOFOUT
"OBMZ[FBWBJMBCMFEBUBTFUTUPHFOFSBUFOPWFM
JOTJHIUTGPS$6#&4DPNQPOFOUT
*OUFHSBUFJOTJHIUTBHBJOTUFBDI$6#&4DPNQPOFOU
*EFOUJGZ$6#&4FWJEFODFHBQT
*OTJHIU
HFOFSBUJPO
%FTJHO
PQUJNJ[F
&WJEFODF
TZOUIFTJT
/PUFOPVHIJOGPSNBUJPOPSBIPMJTUJDQJDUVSFPGESJWFSTEFTJSFE
%FTJHOJOUFSWFOUJPOT
*EFOUJGZUBSHFUCFIBWJPSUPDIBOHF
'JFMEUFTU
*UFSBUF
.BUDINFUIPETNJYUP$6#&4ESJWFST
A*OWJUSPFYQFSJNFOUT
RVBOUJUBUJWF
A8IBUJGTJNVMBUJPO
RVBOUJUBUJWF
%BUBDPMMFDUJPO
EFTDSJQUJWFBOEPSFYQFSJNFOUBM
RVBMJUBUJWFBOEPSRVBOUJUBUJWF
4QFDJGJDESJWFSTUPJOUFSWFOFPOJEFOUJGJFE
Page 13 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Figure 3. Process of evidence evaluation, insight generation, and intervention design and optimization in a VMMC program.
A broad set of CUBES drivers therefore needed to be captured
at scale to assess the relative importance of each driver in a
single dataset. However, as the stages of change appeared to
be of primary importance, journey mapping and qualitative
decision games were first used to understand the stages of
change and associated beliefs and influencers at each stage in
more detail5. In summary, qualitative research pointed to a much
more nuanced picture. Men develop positive as well as negative
beliefs, influenced by individuals around them, as they move
through the various stages of change. These competing beliefs
and associated emotions move men towards or away from the
decision of getting circumcised in distinct stages of change. For
-JUFSBUVSFSFWFBMFEMBSHFTFUPGQPUFOUJBMESJWFSTJODMVEJOH
TPDJBMJOGMVFODFQBJOBOEPUIFSQFSDFQUJPOTCVUNPTU
TUVEJFTBSFTNBMMTDBMF
/PTZTUFNBUJDQJDUVSFPGESJWFSTBWBJMBCMFJOTJOHMFEBUBTFU
%BUBDPMMFDUJPO
2VBMJUBUJWF
&WJEFODF
TZOUIFTJT
A*ODSFBTFVQUBLFPG7..$BNPOHNFOJO;BNCJBBOE;JNCBCXF
*OTJHIU
HFOFSBUJPO %FTJHO
PQUJNJ[F
2VBOUJUBUJWF
'JFMEUFTU
r
r
4FHNFOUTQFDJGJD
NFTTBHJOH
3FDPNNFOESJHIUEFWJDF
UPSJHIUTFHNFOU
*UFSBUF
4JNVMBUFEDIPJDF
FYQFSJNFOUTUP
GPSFDBTUEFWJDFVQUBLF
3FQSFTFOUBUJWFTVSWFZ
"MM$6#&4DPNQPOFOUT
4FHNFOUBUJPOPGTBNQMFJOUPTVCHSPVQT
CBTFEPOESJWFSTPGCFIBWJPS
*OTJHIUTJOGPSNEFTJHOPG
A*OWJUSPFYQFSJNFOUT %FTJHOJOUFSWFOUJPOT
+PVSOFZNBQQJOH
4UBHFTPGDIBOHF
CFMJFGTFNPUJPOT
JOGMVFODFSTBOE
DIBOOFMT
4PDJBMOPSNT
CFMJFGTQFSTPOBMJUZ
FNPUJPOT
JOGMVFODFST
%FDJTJPOHBNFT
4QFDJGJDESJWFSTUPJOUFSWFOFPOJEFOUJGJFE
IFSFCZTFHNFOUPGUIFQPQVMBUJPO
*OUFHSBUFEJOTJHIUTBHBJOTUFBDI$6#&4DPNQPOFOU
A8IBUBSFNFOTLOPXOESJWFSTBOEJOGMVFODFSTPG7..$VQUBLFBMPOHUIFTUBHFTPGDIBOHF
"OBMZTJTPGQSFWJPVTSFHJPOTQFDJGJDEBUBTFU
TIPXFETUSPOHJOUFOUJPOBDUJPOHBQT
7PMVOUBSZNFEJDBMNBMFDJSDVNDJTJPOJEFOUJGZUBSHFUCFIBWJPSUPDIBOHF
Page 14 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
example, beliefs in the early stages include “VMMC protects
myself and my partner from STIs” (positive) and “the procedure
is painful”. As men move to consider undergoing the procedure,
there emerges a strong conflict between the positive beliefs and
negative ones, such as circumcision threatening self-identity,
leading to distrust between the man and his female partner, and
the perceived long healing time. The conflict between emotions
such as shame, distrust, and fear and perceived potential benefits
move men into a state of cognitive dissonance and stall them
from taking-action. However, it was also clear that not all men
held all beliefs equally strongly, or were equally subject to the
same type and strength of influencers. These insights informed
the design of a large-scale quantitative survey investigating
CUBES components comprehensively. Having this dataset
available then allowed us to segment men on the enablers and
barriers to VMMC, so that messaging interventions could be
designed targeting the most important drivers for each segment.
For instance, among the six segments found in Zimbabwe,
Embarrassed Rejecters had mostly negative beliefs about
VMMC, as well as fears and concerns regarding the procedure,
and had little social support5. However, they did not lack
the knowledge that strongly characterized another segment,
Neophytes. Accordingly, segment-specific interventions, including
messaging through front-line workers or media campaigns, could
be developed. In addition, the roll-out of circumcision devices
as an intervention to improve the uptake of circumcision was an
important consideration for programs. However, it was unclear
what the potential demand and market share of these devices
would be. The demand for different devices to carry out the
VMMC procedure was forecast using simulated test marketing, a
technique related to discrete choice experiments, so that the right
devices could be marketed with the right message to the right
people104. Interventions designed based on this research are cur-
rently being piloted at national scale in Zambia and Zimbabwe.
Discussion
Effective interventions to drive key outcomes are sorely needed
in global development and many other sectors. In this paper, we
aim to help programs arrive at an effective portfolio of interven-
tions in two ways.
First, to effectively design interventions that change target
behaviors, we introduce a novel and practical framework of
behavior, CUBES, to help programs categorize and understand
the barriers and enablers, influencers and stages of behavior
change (Figure 1 and Table 2). CUBES synthesizes widely
validated evidence across psychology, behavioral economics,
market research, and sociology7,8,11–14,16,19,20,22,35,41,48,51,56,76, and
presents its building blocks with a view to actionability
by programs. CUBES provides a checklist for programs to
systematically assess what is already known about drivers
of a target behavior, where novel research is most needed in
order to design actionable levers of change, and, after closing
evidence gaps, where interventions could focus.
Second, not every type of driver is best measured in the same
way. We therefore curate a set of descriptive, experimental,
and simulation approaches across sectors, and advocate for a
method mix tailored to the gaps in knowledge in a given program
(Figure 2 and Table 2). Some approaches, such as different
types of observation and self-report, are already well-
established in global development, but using CUBES to struc-
ture the components of insight generation ensures that programs
can design tools in a systematic way, ultimately saving time and
money. For example, quantitative surveys would benefit from the
selective incorporation of validated scales to measure specific
drivers, or from ways to encourage participants to respond to
sensitive topics with greater fidelity. Other methods are well-
used in other sectors such as market research, experimental
psychology, and decision sciences, and programs could benefit
from them for specific purposes. For instance, discrete choice
experiments and decision games provide an experimental way to
systematically vary and identify key enablers and barriers
before testing interventions in the field. Implicit attitude test
can be considered as a complementary method to self-report
when strong biases are presumed to be at play, and simulations
modeling complex systems provide programs with a way to
estimate the importance and interaction of multiple drivers, as
well as test ‘what-if’ scenarios. We demonstrate the process of
choosing methods for specific purposes in a case study on
voluntary medical male circumcision uptake (Figure 3). Of
course, data collection in any program will also be influenced
by considerations about cost, time, and skill resourcing. There is
no hard-and-fast rule of how each method ranks on those three
parameters, as much depends on existing organizational and pro-
gram infrastructure. Nevertheless, we urge programs to estimate
these parameters before choosing a methodological path, and
to also consider trade-offs in investing upfront versus potential
time and cost savings in the intervention phase.
Once data has been collected, the CUBES framework can again
be used to structure findings and highlight the specific content
and potential targets of an intervention. Creating interventions
to fit the varied barriers to behavior is a challenge as well as an
opportunity for global health. For example, higher levels of con-
scientiousness have consistently been associated with higher
adherence to medication107 or the contraceptive pill108. Rather
than attempting to influence conscientiousness, an intervention
might consist of identifying those with lower conscientiousness
and targeting increased levels of support to this sub-population.
In other situations, drivers may be affected directly. For example,
using Facebook to alert college students that peer social norms
around drinking were lower than they thought changed drinking
behavior109. As an example of targeting a belief, enhancing
self-efficacy through encouragement on progress, attribution
of progress to participants’ own abilities, observation of others
carrying out the target behavior, and other strategies significantly
increased physical activity in older adults110. These examples
are by no means indicative of success in other contexts and
behaviors, and interventions all need to be piloted. However,
using a framework of behavior allows for the identification of
possibilities that may otherwise remain hidden, and conversely
narrow the options for choosing suitable intervention types.
Previously, Michie et al.111 provided a comprehensive overview of
93 behavior change techniques, from social comparison to incen-
tives, feedback on behavior, prompts, and goal setting111, as well
as a more high-level categorization of nine types intervention
Page 15 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
from education to training, incentivization, and environmen-
tal restructuring22. While identifying mechanisms of actions for
interventions remains a work in progress112, these categories can
serve as decision aids and an overview of options to programs.
The Fogg Model of behavior, as well as the MINDSPACE check-
list, also provide a useful classification of what characterizes
effective prompts that can increase motivation and ability20,21.
In many contexts, a single type of intervention may not be
enough. We previously showed that psycho-behavioral segmenta-
tion can be a powerful method for finely targeting interventions
beyond a one-size-fits-all approach5,6,113. In the voluntary
medical male circumcision program, we used quantitative survey
data to segment men on what drove them toward or away from
the procedure, and could therefore tailor interventions specifically
to each segment5. However, sound and actionable segmentation
can only be performed on large-scale quantitative datasets,
which is a consideration for the method mix chosen.
After designing interventions to match key barriers found,
whether at the population or segment level, we also recommend
optimizing and ‘funneling’ interventions from a large pool of
potential options down to a narrow set that can be thoroughly
evaluated in the field while maximizing the likelihood of success
(Figure 2). Discrete choice experiments can help programs choose
between types of interventions and their components at the
design stage. Even at the field test stage, factorial designs
could test sets of interventions more efficiently, such as differ-
ent magnitudes of monetary incentives, a distinct messaging
component in each, and a different channel through which they
are deployed. Recently, this approach has been refined in the
Multiphase Optimization Strategy (MOST) for determining the
best set of intervention components114. ‘In vitro’ experiments
have inherent limitations, as they are not fully replicative of
real-world contexts and behaviors that participants are asked to
engage in. However, experiments can link a large set of features
to an actual behavioral outcome in a controlled way. It is plau-
sible, but yet to be tested systematically, that the closer the ‘in
vitro’ behavior to its real-world counterpart, the more informa-
tive such experiments might be. For example, being asked to
pick among different physical products (such as contraceptive
packages) with different features is a task not far removed from
reaching for an actual product in a pharmacy. Asking commu-
nity health workers to pick among job options with varying
attributes such as salary, workload, and career progression115
might not be too different from workers weighing those consid-
erations when looking at a job ad. An experiment asking nurses
to evaluate and react to a hypothetical emergency, however,
might be much more distant from how the choice behavior would
play out in a real-life context.
CUBES and the methods toolkit proposed here have several
limitations. Frameworks of behavior in general have an ‘evaluation
problem’, as it is not feasible to directly compare the insights
generated de novo using many different behavioral frameworks
with very different components. The true test of time will lie
in whether programs judge CUBES to be useful in increasing
intervention fit, as well as effectiveness at reaching target
outcomes, as we have found in our own programs. So far, we
have used this ‘utility test’ in our own work, and to give speedy
feedback to other organization, in two ways: first, we have
used the framework to evaluate planned data collection for
comprehensiveness. Second, the framework has been used
for dimensionality reduction of expansive datasets (such as
household surveys), to enable more clarity in analysis. Future
testing should also include critical review of whether interven-
tion options expanded when CUBES was used, and ultimately
the change in impact that systematic intervention design yields.
A second limitation is that it is unlikely that any one program
will be able to draw on expertise for all method options equally,
and many techniques require specialized skills to design, field,
and analyze. This limitation can be somewhat mitigated by
bringing in expert resources, although programs may also face
difficulties gathering that network of expertise.
CUBES and permutations of its methods toolkit have now
been used in several large-scale programs, from investigating
healthcare provider behavior and household behaviors along
the maternal and neonatal healthcare pathway in Uttar Pradesh,
India, to understanding and influencing tuberculosis care-seeking
in South India (unpublished data) and voluntary medical male
circumcision in Africa5. We hope that linking the categoriza-
tion and measurement of enablers and barriers to behavior
will enable many more programs to design efficient and effec-
tive interventions that get results, and in turn iteratively refine
the approaches introduced here.
Data availability
All data underlying the results are available as part of the
article and no additional source data are required.
Acknowledgments
The authors thank Prof. Richard Thaler (University of Chicago),
Prof. Jessica Cohen (Harvard University), and Anabel Gomez
(AVAC) for their insightful comments on the CUBES behav-
ioral framework. We thank Steve Kretschmer, Hannah Kemp,
and all past and present Surgo Foundation team members and
partners for their contributions to the development of CUBES
(formerly called CUE). Finally, we would like to thank Yael
Caplan for contributions to the figures.
We thank the Surgo Foundation, who has kindly granted the
authors the permission and licence to use materials and information
owned by them, for this article.
Page 16 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
References
1. Weiss WM, Rahman MH, Solomon R, et al.: Outcomes of polio eradication
activities in Uttar Pradesh, India: the Social Mobilization Network (SM Net) and
Core Group Polio Project (CGPP). BMC Infect Dis. 2011; 11: 117.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
2. Semrau KEA, Hirschhorn LR, Marx Delaney M, et al.: Outcomes of a Coaching-
Based WHO Safe Childbirth Checklist Program in India. N Engl J Med. 2017;
377(24): 2313–24.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
3. Thirumurthy H, Masters SH, Rao S, et al.: Effect of providing conditional
economic compensation on uptake of voluntary medical male circumcision in
Kenya: a randomized clinical trial. JAMA. 2014; 312(7): 703–11.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
4. Sgaier SK, Reed JB, Sundaram M, et al.: Interventions to Drive Uptake of
Voluntary Medical Male Circumcision-A Collection of Impact Evaluation
Evidence. J Acquir Immune Defic Syndr. 2016; 72 Suppl 4: S257–S261.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
5. Sgaier SK, Eletskaya M, Engl E, et al.: A case study for a psychographic-
behavioral segmentation approach for targeted demand generation in
voluntary medical male circumcision. eLife. 2017; 6: pii: e25923.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
6. Sgaier SK, Engl E, Kretschmer S: Time to scale psycho-behavioral segmentation
in global development. Stanford University, and Stanford Center on Philanthropy
and Civil Society: Stanford Social Innovation Review. 2018.
Reference Source
7. Bourdieu P: Outline of a theory of practice . Cambridge University Press; 1977.
Publisher Full Text
8. Bronfenbrenner U: Ecological models of human development. International
Encyclopedia of Education. 1994; 3.
Reference Source
9. Rogers E: Diffusion of innovations. 5 ed. New York: The Free Press; 2003.
Reference Source
10. Carpenter CJ: A meta-analysis of the effectiveness of health belief model
variables in predicting behavior. Health Commun. 2010; 25(8): 661–9.
PubMed Abstract
|
Publisher Full Text
11. Kahneman D, Tversky A: Prospect theory: an analysis of decision under risk.
Econometrica. 1979; 47(2): 263–92.
Publisher Full Text
12. McCrae RR, John OP: An introduction to the five-factor model and its
applications. J Pers. 1992; 60(2): 175–215.
PubMed Abstract
|
Publisher Full Text
13. Prochaska JO, Velicer WF: The transtheoretical model of health behavior
change. Am J Health Promot. 1997; 12(1): 38–48.
PubMed Abstract
|
Publisher Full Text
14. Rosenstock IM, Strecher VJ, Becker MH: Social learning theory and the Health
Belief Model. Health Educ Q. 1988; 15(2): 175–83.
PubMed Abstract
|
Publisher Full Text
15. Ryan RM, Deci EL: Self-determination theory and the facilitation of intrinsic
motivation, social development, and well-being. Am Psychol. 2000; 55(1):
68–78.
PubMed Abstract
|
Publisher Full Text
16. Schwarzer R: Modeling health behavior change: how to predict and modify the
adoption and maintenance of health behaviors. Appl Psychol. 2008; 57(1): 1–29.
Publisher Full Text
17. Tversky A, Kahneman D: Judgment under uncertainty: heuristics and biases.
Utility, Probability, and Human Decision Making: Selected Proceedings of an
Interdisciplinary Research Conference, Rome, 3–6 September, 1973. 1974; 141–62.
Publisher Full Text
18. Tversky A, Kahneman D: Loss aversion in riskless choice: a reference-
dependent model. Q J Econ. 1991; 106(4): 1039–61.
Publisher Full Text
19. Yzer M: The integrative model of behavioral prediction as a tool for designing
health messages. Health communication message design theory and practice.
2012; 21–40.
Reference Source
20. Dolan P, Hallsworth M, Halpern D, et al.: Influencing behaviour: The mindspace
way. J Econ Psychol. 2012; 33(1): 264–77.
Publisher Full Text
21. Fogg B: A behavior model for persuasive design. Proceedings of the 4th
International Conference on Persuasive Technology. Claremont, California,
USA.1541999: ACM; 2009; 1–7.
Publisher Full Text
22. Michie S, van Stralen MM, West R: The behaviour change wheel: a new method
for characterising and designing behaviour change interventions. Implement
Sci. 2011; 6: 42.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
23. Sgaier SK, Baer J, Rutz DC, et al.: Toward a Systematic Approach to Generating
Demand for Voluntary Medical Male Circumcision: Insights and Results From
Field Studies. Glob Health Sci Pract. 2015; 3(2): 209–29.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
24. Janz NK, Becker MH: The Health Belief Model: a decade later. Health Educ Q.
1984; 11(1): 1–47.
PubMed Abstract
|
Publisher Full Text
25. India Health Action Trust: Rolling facility survey: results of two successive
rounds, 25 high priority districts, Uttar Pradesh. Lucknow: IHAT; 2017.
26. Riemsma RP, Pattenden J, Bridle C, et al.: Systematic review of the
effectiveness of stage based interventions to promote smoking cessation.
BMJ. 2003; 326(7400): 1175–7.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
27. Spencer L, Adams TB, Malone S, et al.: Applying the transtheoretical model
to exercise: a systematic and comprehensive review of the literature. Health
Promot Pract. 2006; 7(4): 428–43.
PubMed Abstract
|
Publisher Full Text
28. Salmela S, Poskiparta M, Kasila K, et al.: Transtheoretical model-based dietary
interventions in primary care: a review of the evidence in diabetes. Health Educ
Res. 2008; 24(2): 237–52.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
29. Ajzen I, Madden TJ: Prediction of goal-directed behavior: attitudes, intentions,
and perceived behavioral control. J Exp Soc Psychol. 1986; 22(5): 453–74.
Publisher Full Text
30. Kahneman D, Thaler RH: Anomalies: utility maximization and experienced
utility. J Econ Perspect. 2006; 20(1): 221–34.
Publisher Full Text
31. Neal DT, Wood W, Quinn JM: Habits—a repeat performance. Curr Dir Psychol
Sci. 2006; 15(4): 198–202.
Publisher Full Text
32. Verplanken B, Orbell S: Reflections on past behavior: a self-report index of
habit strength. J Appl Soc Psychol. 2003; 33(6): 1313–30.
Publisher Full Text
33. Wood W, Tam L, Witt MG: Changing circumstances, disrupting habits. J Pers
Soc Psychol. 2005; 88(6): 918–33.
PubMed Abstract
|
Publisher Full Text
34. Quinn JM, Pascoe A, Wood W, et al.: Can’t control yourself? Monitor those bad
habits. Pers Soc Psychol Bull. 2010; 36(4): 499–511.
PubMed Abstract
|
Publisher Full Text
35. Verplanken B, Wood W: Interventions to break and create consumer habits.
J public Policy Mark. 2006; 25(1): 90–103.
Publisher Full Text
36. Bronfenbrenner U: The ecology of human development: experiments by nature
and design. Harvard University Press; 1979.
Reference Source
37. Stokols D: Establishing and maintaining healthy environments. Toward a social
ecology of health promotion. Am Psychol. 1992; 47(1): 6–22.
PubMed Abstract
|
Publisher Full Text
38. Reckwitz A: Toward a theory of social practices: a development in culturalist
theorizing. Eur J Soc Theory. 2002; 5(2): 243–63.
Publisher Full Text
39. Smith JR, Terry DJ, Manstead AS, et al.: The attitude-behavior relationship in
consumer conduct: the role of norms, past behavior, and self-identity. J Soc
Psychol. 2008; 148(3): 311–33.
PubMed Abstract
|
Publisher Full Text
40. Brauer M, Chaurand N: Descriptive norms, prescriptive norms, and social
control: an intercultural comparison of people’s reactions to uncivil behaviors.
Eur J Soc Psychol. 2010; 40: 490–9.
Publisher Full Text
41. Ariely D: Predictably irrational: the hidden forces that shape our decisions. 1st
ed. New York, NY: Harper; 2008; 280.
Reference Source
42. Thaler RH, Sunstein CR: Nudge: improving decisions about health, wealth,
and happiness. Rev and expanded ed. New York: Penguin Books; 2009; 312.
Reference Source
43. Brannon LA, Carson KL: The representativeness heuristic: influence on nurses’
decision making. Appl Nurs Res. 2003; 16(3): 201–4.
PubMed Abstract
|
Publisher Full Text
44. Bränström R, Brandberg Y: Health risk perception, optimistic bias, and personal
satisfaction. Am J Health Behav. 2010; 34(2): 197–205.
PubMed Abstract
|
Publisher Full Text
45. Laibson D: Golden eggs and hyperbolic discounting. Q J Econ. 1997; 112(2):
443–78.
Publisher Full Text
46. Poses RM, Anthony M: Availability, wishful thinking, and physicians’ diagnostic
judgments for patients with suspected bacteremia. Med Decis Making. 1991;
11(3): 159–68.
PubMed Abstract
|
Publisher Full Text
47. Salkeld G, Ryan M, Short L: The veil of experience: do consumers prefer what
they know best? Health Econ. 2000; 9(3): 267–70.
PubMed Abstract
|
Publisher Full Text
48. Halpern D, King D, Vlaev I, et al.: MINDSPACE: Influencing behaviour through
Page 17 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
public policy. Institute for Government. 2010.
Reference Source
49. Ajzen I: The theory of planned behavior. Organ Behav Hum Dec. 1991; 50(2):
179–211.
Publisher Full Text
50. Gold GJ: Review of predicting and changing behavior: the reasoned action
approach. J Soc Psychol. 2011; 151(3): 382–5.
Publisher Full Text
51. Armitage CJ, Conner M: Efficacy of the Theory of Planned Behaviour: a meta-
analytic review. Br J Soc Psychol. 2001; 40(Pt 4): 471–99.
PubMed Abstract
|
Publisher Full Text
52. Godin G, Kok G: The theory of planned behavior: a review of its applications to
health-related behaviors. Am J Health Promot. 1996; 11(2): 87–98.
PubMed Abstract
|
Publisher Full Text
53. Brassington GS, Atienza AA, Perczek RE, et al.: Intervention-related cognitive
versus social mediators of exercise adherence in the elderly. Am J Prev Med.
2002; 23(2 Suppl): 80–6.
PubMed Abstract
|
Publisher Full Text
54. Devine-Wright P, Devine-Wright H, Fleming P: Situational influences upon
children’s beliefs about global warming and energy. Environ Educ Res. 2004;
10(4): 493–506.
Publisher Full Text
55. Rimal RN: Perceived risk and self-efficacy as motivators: understanding
individuals’ long-term use of health information. J Commun. 2001; 51(4): 633–54.
Publisher Full Text
56. Bandura A: Health promotion by social cognitive means. Health Educ Behav.
2004; 31(2): 143–64.
PubMed Abstract
|
Publisher Full Text
57. McAlister AL, Perry CL, Parcel GS: How individuals, environments, and health
behaviors interact. Health Behavior. 2008; 169–188.
Reference Source
58. Rosenstock IM: Why people use health services. Milbank Mem Fund Q. 1966;
44(3:) Suppl: 94–127.
PubMed Abstract
59. Harrison JA, Mullen PD, Green LW: A meta-analysis of studies of the Health
Belief Model with adults. Health Educ Res. 1992; 7(1): 107–16.
PubMed Abstract
|
Publisher Full Text
60. Moreau C, Bouyer J, Goulard H, et al.: The remaining barriers to the use of
emergency contraception: perception of pregnancy risk by women undergoing
induced abortions. Contraception. 2005; 71(3): 202–7.
PubMed Abstract
|
Publisher Full Text
61. van der Werff E, Steg L, Keizer K: The value of environmental self-identity:
The relationship between biospheric values, environmental self-identity and
environmental preferences, intentions and behaviour. J Environ Psychol. 2013;
34: 55–63.
Publisher Full Text
62. Curtis VA, Garbrah-Aidoo N, Scott B: Ethics in public health research: masters
of marketing: bringing private sector skills to public health partnerships. Am J
Public Health. 2007; 97(4): 634–41.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
63. Aunger R, Curtis V: The evo–eco approach to behaviour change. In: Gibson MA,
Lawson DW, editors. Applied evolutionary anthropology. New York, NY: Springer
New York; 2014; 271–95.
Publisher Full Text
64. Cane J, O’Connor D, Michie S: Validation of the theoretical domains framework
for use in behaviour change and implementation research. Implement Sci.
2012; 7: 37.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
65. Booth-Kewley S, Vickers RR Jr: Associations between major domains of
personality and health behavior. J Pers. 1994; 62(3): 281–98.
PubMed Abstract
|
Publisher Full Text
66. Soto CJ, Jackson JJ: Five-Factor Model of Personality. In Dunn DS, editor.
Bibliographies in Psychology. New York, NY: Oxford University Press; 2013.
Publisher Full Text
67. McCrae RR: The place of the FFM in personality psychology. Psychol Inq. 2010;
21(1): 57–64.
Publisher Full Text
68. Digman JM: Personality structure: Emergence of the five-factor model. Annu
Rev Psychol. 1990; 41: 417–40.
Publisher Full Text
69. Følstad A, Kvale K, Halvorsrud R: Customer journey measures - State of the art
research and best practices. Norwegian Research Council. 2013; Report A 24488.
Reference Source
70. Powell T: Going beyond surveys for B2B research. Quirk’s Marketing Research
Review. 2016; 4.
Reference Source
71. Macdonald EK, Wilson H: Immersive market research. Warc Best Practice. 2011.
Reference Source
72. Kawulich BB: Participant observation as a data collection method. Forum
Qualitative Sozialforschung. 2005; 6(2).
Publisher Full Text
73. Finkler SA, Knickman JR, Hendrickson G, et al.: A comparison of work-sampling
and time-and-motion techniques for studies in health services research. Health
Serv Res. 1993; 28(5): 577–97.
PubMed Abstract
|
Free Full Text
74. Manasyan A, Saleem S, Koso-Thomas M, et al.: Assessment of Obstetric and
Neonatal Health Services in Developing Country Health Facilities. Am J
Perinatol. 2013; 30(9): 787–94.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
75. Meertens RM, Lion R: Measuring an individual’s tendency to take risks: The risk
propensity scale. J Appl Soc Psychol. 2008; 38(6): 1506–20.
Publisher Full Text
76. Schwarzer R, Bässler J, Kwiatek P, et al.: The assessment of optimistic self-
beliefs: comparison of the German, Spanish, and Chinese versions of the
general self-efficacy scale. Appl Psychol. 1997; 46(1): 69–88.
Publisher Full Text
77. Bradley MM, Lang PJ: Measuring emotion: the Self-Assessment Manikin and
the Semantic Differential. J Behav Ther Exp Psychiatry. 1994; 25(1): 49–59.
PubMed Abstract
|
Publisher Full Text
78. Rimal RN, Juon HS: Use of the risk perception attitude framework for
promoting breast cancer prevention. J Appl Soc Psychol. 2010; 40(2): 287–310.
Publisher Full Text
79. Goldberg LR, Johnson JA, Eber HW, et al.: The international personality item
pool and the future of public-domain personality measures. J Res Pers. 2006;
40: 84–96.
Publisher Full Text
80. Gregson S, Zhuwau T, Ndlovu J, et al.: Methods to reduce social desirability
bias in sex surveys in low-development settings: experience in Zimbabwe. Sex
Transm Dis. 2002; 29(10): 568–75.
PubMed Abstract
|
Publisher Full Text
81. Lowndes CM, Jayachandran AA, Banandur P, et al.: Polling booth surveys: a
novel approach for reducing social desirability bias in HIV-related behavioural
surveys in resource-poor settings. AIDS Behav. 2012; 16(4): 1054–62.
PubMed Abstract
|
Publisher Full Text
82. Gregson S, Mushati P, White PJ, et al.: Informal confidential voting interview
methods and temporal changes in reported sexual risk behaviour for HIV
transmission in sub-Saharan Africa. Sex Transm Infect. 2004; 80 Suppl 2:
ii36–ii42.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
83. Xu T, de Almeida Neto AC, Moles RJ: A systematic review of simulated-patient
methods used in community pharmacy to assess the provision of non-
prescription medicines. Int J Pharm Pract. 2012; 20(5): 307–19.
PubMed Abstract
|
Publisher Full Text
84. Satyanarayana S, Kwan A, Daniels B, et al.: Use of standardised patients to
assess antibiotic dispensing for tuberculosis by pharmacies in urban India: a
cross-sectional study. Lancet Infect Dis. 2016; 16(11): 1261–8.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
85. Chan DK, Gallagher TH, Reznick R, et al.: How surgeons disclose medical errors
to patients: a study using standardized patients. Surgery. 2005; 138(5): 851–8.
PubMed Abstract
|
Publisher Full Text
86. Lerman K, Yan X, Wu XZ: The “Majority Illusion” in Social Networks. PLoS One.
2016; 11(2): e0147617.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
87. Schiffer E: Net-Map – Influence mapping of social networks. Manual. Sunbelt
Conference of the International Network of Social Network Analysis. 2007.
Reference Source
88. Schiffer E, Hauck J, Andreini M, et al.: Net-map: Influence network mapping–
theory. Small reservoirs toolkit CGIAR challenge program on Water for Food. 2002.
Reference Source
89. Sekandi JN, Zalwango S, Martinez L, et al: Four Degrees of Separation: Social
Contacts and Health Providers Influence the Steps to Final Diagnosis of Active
Tuberculosis Patients in Urban Uganda. BMC Infect Dis. 2015; 15: 361.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
90. Maharana A, Nsoesie EO: Use of Deep Learning to Examine the Association of
the Built Environment With Prevalence of Neighborhood Adult Obesity. JAMA
Netw Open. 2018; 1(4): e181535.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
91. Patodkar VN, Sheikh IR: Twitter as a corpus for sentiment analysis and opinion
mining. IJARCCE. 2016; 5(12): 320–2.
Publisher Full Text
92. Muzyka CN, Thompson LH, Bombak AE, et al.: A Kenyan newspaper analysis of
the limitations of voluntary medical male circumcision and the importance of
sustained condom use. BMC Public Health. 2012; 12: 465.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
93. Quaife M, Terris-Prestholt F, Di Tanna GL, et al.: How well do discrete choice
experiments predict health choices? A systematic review and meta-analysis of
external validity. Eur J Health Econ. 2018; 19(8): 1053–1066.
PubMed Abstract
|
Publisher Full Text
94. Quaife M, Eakle R, Cabrera Escobar MA, et al.: Divergent Preferences for HIV
Prevention: A Discrete Choice Experiment for Multipurpose HIV Prevention
Products in South Africa. Med Decis Making. 2018; 38(1): 120–33.
PubMed Abstract
|
Publisher Full Text
95. Greenwald AG, McGhee DE, Schwartz JL: Measuring individual differences in
implicit cognition: the implicit association test. J Pers Soc Psychol. 1998; 74(6):
1464–80.
PubMed Abstract
|
Publisher Full Text
Page 18 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
96. Lowes S, Nunn N, Robinson JA, et al.: Understanding Ethnic Identity in Africa:
Evidence from the Implicit Association Test (IAT). Am Econ Rev. 2015; 105(5):
340–5.
Publisher Full Text
97. Perkins A, Forehand M, Greenwald A, et al.: Measuring the nonconscious. CP
Haugtvedt, P Herr, FR Kardes (2008) Handbook of consumer psychology. 2008;
461–475.
Reference Source
98. Sheeran P, Gollwitzer PM, Bargh JA: Nonconscious processes and health.
Health Psychol. 2013; 32(5): 460–73.
PubMed Abstract
|
Publisher Full Text
99. Oswald FL, Mitchell G, Blanton H, et al.: Predicting ethnic and racial
discrimination: a meta-analysis of IAT criterion studies. J Pers Soc Psychol.
2013; 105(2): 171–92.
PubMed Abstract
|
Publisher Full Text
100. Friedkin NE, Proskurnikov AV, Tempo R, et al.: Network science on belief system
dynamics under logic constraints. Science. 2016; 354(6310): 321–6.
PubMed Abstract
|
Publisher Full Text
101. Rand W, Rust RT: Agent-based modeling in marketing: Guidelines for rigor. Int
J Res Mark. 2011; 28(3): 181–93.
Publisher Full Text
102. Pope AJ, Gimblett R: Linking Bayesian and agent-based models to simulate
complex social-ecological systems in semi-arid regions. Front Environ Sci.
2015; 3: 55.
Publisher Full Text
103. Lee JS, Filatova T, Ligmann-Zielinska A, et al.: The Complexities of Agent-Based
Modeling Output Analysis. J Artif Soc Soc Simul. 2015; 18(4): 4.
Publisher Full Text
104. Fram F, Church F, Sundaram M, et al.: Employing Demand-Based Volumetric
Forecasting to Identify Potential for and Roles of Devices in Scale-Up of
Medical Male Circumcision in Zambia and Zimbabwe. J Acquir Immune Defic
Syndr. 2016; 72 Suppl 1: S83–9.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
105. Djimeu Wouabe E: Scoping report on interventions for increasing the demand
for voluntary medical male circumcision. Washington, D.C.: International
Initiative for Impact Evaluation; 2013.
Reference Source
106. Population Services International Research Metrics: Voluntary medical male
circumcision TRaC study among men and women aged 15–49 years in
Zimbabwe. Washington, D.C.: PSI Social Marketing Research Series, 2013.
107. Molloy GJ, O’Carroll RE, Ferguson E: Conscientiousness and medication
adherence: a meta-analysis. Ann Behav Med. 2014; 47(1): 92–101.
PubMed Abstract
|
Publisher Full Text
108. Leahy D, Treacy K, Molloy GJ: Conscientiousness and adherence to the oral
contraceptive pill: A prospective study. Psychol Health. 2015; 30(11): 1346–60.
PubMed Abstract
|
Publisher Full Text
109. Ridout B, Campbell A: Using Facebook to deliver a social norm intervention to
reduce problem drinking at university. Drug Alcohol Rev. 2014; 33(6): 667–73.
PubMed Abstract
|
Publisher Full Text
110. Allison MJ, Keller C: Self-efficacy intervention effect on physical activity in
older adults. West J Nurs Res. 2004; 26(1): 31–46; discussion 47–58.
PubMed Abstract
|
Publisher Full Text
111. Michie S, Richardson M, Johnston M, et al.: The behavior change technique
taxonomy (v1) of 93 hierarchically clustered techniques: building an
international consensus for the reporting of behavior change interventions.
Ann Behav Med. 2013; 46(1): 81–95.
PubMed Abstract
|
Publisher Full Text
112. Michie S, West R, Sheals K, et al.: Evaluating the effectiveness of behavior
change techniques in health-related behavior: a scoping review of methods
used. Transl Behav Med. 2018; 8(2): 212–24.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
113. Engl E, Smittenaar P, Sgaier SK: Identifying population segments for effective
intervention design and targeting using unsupervised machine learning: an
end-to-end guide [version 2; peer review: 2 approved]. Gates Open Res. 2019;
3: 1503.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
114. Collins LM, Baker TB, Mermelstein RJ, et al.: The multiphase optimization
strategy for engineering effective tobacco use interventions. Ann Behav Med.
2011; 41(2): 208–26.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
115. Abdel-All M, Angell B, Jan S, et al.: What do community health workers want?
Findings of a discrete choice experiment among Accredited Social Health
Activists (ASHAs) in India. BMJ Glob Health. 2019; 4(3): e001509.
PubMed Abstract
|
Publisher Full Text
|
Free Full Text
Page 19 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
Open Peer Review
Current Peer Review Status:
Version 1
29 April 2019Reviewer Report
https://doi.org/10.21956/gatesopenres.14025.r27004
© 2019 Saldanha N.ThisisanopenaccesspeerreviewreportdistributedunderthetermsoftheCreativeCommons
,whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalAttributionLicense
workisproperlycited.
Neela A. Saldanha
CentreforSocialandBehaviourChange(CSBC),AshokaUniversity,Sonipat,Haryana,India
Thisisanimportantandusefularticleandprovidesapracticalguidetocreateeffectivebehaviorchange
programs.
Therationaleforthistoolkitisclearlystated-behaviorchangeinterventionsoftendonotworkorarenot
sustainable.Weneedtomovefroma"single-intervention"silverbulletapproachtoamorestrategic
programapproachcomprisingmultipleinterventions.Butinthefaceofaudienceandcontext
complexities,multipleresearchtoolsanddisciplines,wheredowebegin?Thisarticleprovidesthat
answer,bothintheformofapracticalbehaviorchangeframeworkandappropriateresearchmethodsto
addresseachpartoftheframework.
Thebehaviorchangeframeworktheauthorsoutlineisextremelyuseful,aspractitionersstruggleto
cobbletogethermultipleframeworks,noneofwhichcomprehensivelyaddressesallbarriers.Inparticular,
Iappreciatedtheauthorsexplicitlycallingout inthemodel.Mostcontextual barriers & enablers
behaviorchangeframeworksdonotreferencethesefactorsclearlyandyetweknowfromabodyofwork
onbehavioralsciencethatsuchconstraintscanbeamajordriverofbehaviorandespeciallyofthe
intent-actiongap.Havingbothperceptualandcontextualbarriersandenablersinthemodelmakesita
completeframework.
Second,inmakingthedistinctionbetween orrepeatedbehaviorsandone-timebehaviors,thehabits
authorsmakethemodelmoreclearandeffective.Giventhathabitformationhasrarelybeenafocusin
muchofdevelopmentworkmakesitevenmoreimportanttoexplicitlyconsiderwhetherthebehaviorwe
arelookingtosolveforisaone-timebehaviororformingofahabit.
Third,thecuratedlistofresearch theauthorsoutlineisofrealvaluetopractitionerswhoneedtechniques
moreofthesekindsofliststoenablethemtoexpertlynavigatetheresearchworld.IknowIwillcertainly
turntothislistwhendesigningresearch.
Finally,Imustcommendtheauthorsforstronglycallingformore"in-vitro" .Rightnow,experimentation
theonlyevidence-drivenworkisthroughmulti-yearRCTsthatareexpensiveandnotflexibleandareonly
inplaceafterresourceshavealreadybeenexpendedondesigningandpilotingtheprogram.Havinga
Page 20 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
1.
2.
3.
4.
5.
inplaceafterresourceshavealreadybeenexpendedondesigningandpilotingtheprogram.Havinga
waytorapidlytestelementsoftheprogramatthedesignstagecanensurewearenotwastingvaluable
resourcesup-frontandcancoursecorrectearlyenough.
Asapractitioner,Ihaveafewsuggestionstoincreasetheusefulnessandlikelihoodofuptakeofthe
frameworkbypractitioners:
Make it easy: Iwouldurgetheauthorstosimplifytheframeworkanddrivetocognitiveeaseso
thatpractitionerscaneasilyadoptit.Forexample,theauthorsmayweighthebenefitsofincluding
the"StagesofChange"partofthemodelagainstthecomplexityitintroducesforimplementation.
Thiscouldcertainlybeamorein-depthversionofthebasicmodelbutinthefirstinstance,defining
thebehaviorwellandthenunderstandingperceptualandcontextualbarriersandenablerstothat
behaviorwouldgoalongwaytodesigningeffectiveprograms.
Itwouldalsobehelpfultogivepractitionerssomerulesofthumbthatareactionable.Forexample,I
wouldrecastthenarrativeoftheframeworkintheformofthreesimplerules:
1)Understanddeterminantsofbehavior:barriersandenablers,bothperceptualandcontextual.
2)Designidea-channelinterventionsthataddressbarriersandleverageenablers.
3)Designtoalllevelsofchange–individual,family,society.
Consider adding the construct of goals & identity to the perceptual barriers and enablers:
Allhumanbehaviorisgoaldriven.Goalsdriveattentionandvalueandarethereforecriticalto
motivationandestablishingintent.Atthesametime,currentresearchonthefieldprioritizes
knowledgeandattitudesratherthangoals.Indoingso,weoftenconcluderesearchwithlittleidea
ofthisimportantdeterminantofmotivation-people'sgoalsandhowtheintendedbehaviorscan
helpthemachievetheirgoals.Theauthorsmentiontheconstructofgoalsseveraltimesinthe
articlebutshouldconsiderincludingthisconstructintheframework.Similarlyafocusonidentity,
particularlycriticalinhabitformationwouldaddtotherichnessoftheperceptualbarriersand
enablers.
Clarify the role of social norms: Itwasnotintuitivelycleartomewhysocialnormswhichare
stronglyrootedinpsychologyarepartofcontextualfactorsthathasothernon-psychological
componentslikelaws&infrastructure.
Create more specifics for each construct:Theoverallarticledoesagoodjobofclearly
describingeachconstruct.Itwouldbehelpfulasajobaidtopresentaspecificchecklistforeach
construct.Forexample,thedifferenttypesofbeliefsonecoulduncoverinresearch.Practitioners
oftendon’tknowallthetheoreticalunderpinningsofaconstructandmayusetheirown
interpretation.Thiswouldresultovertimeintheconstructsbeinginterpreteddifferentlyand
thereforereducethescopeforcross-learningacrossprograms.
Create more detail and specificity on the design process:Thearticlehasalotofemphasison
theresearchprocesstogenerateinsight.Thereisamuchshortersectiononeffectivebehavioral
design,thoughthisisoftenthestagewhenpractitionersfailtoconvertrichinsightintoeffective
ideas.Similartothedetailedmannerinwhichtheresearchtechniqueshavebeenoutlined,itwould
beusefultooutlinethedesignprocesswithappropriatechecklists.Forexamplehowtoprioritize
ideas,howtoeffectivelyprototypeprioritizedideas,waystorapidlytestideasrangingfrom
experimentstorapidqualitativefeedbackloopsandothersuchstepsintheprocess.Withinthis
designprocess,thereneedstobeaclearunderstandingoftheprocesstodesignmassmedia
communicationasoftenmassmedia&nonmassmediainterventionsarecreatedseparately
Page 21 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
5.
communicationasoftenmassmedia&nonmassmediainterventionsarecreatedseparately
ratherthanaspartofoneintegratedprogram.
Overall,thisisaverypracticalandcleararticle.IhopethesuggestionsIhavelaidoutareusefultothe
authorsandthatthisworkcanleadtomoresuchbestpracticeswhichcanhelpinterventionistsdesign
effectiveprograms.
Is the rationale for developing the new method (or application) clearly explained?
Yes
Is the description of the method technically sound?
Partly
Are sufficient details provided to allow replication of the method development and its use by
others?
Partly
If any results are presented, are all the source data underlying the results available to ensure full
reproducibility?
Partly
Are the conclusions about the method and its performance adequately supported by the
findings presented in the article?
Partly
Nocompetinginterestsweredisclosed.Competing Interests:
Reviewer Expertise:Behaviorchange;behaviorscience;insights;behavioraldesign
I confirm that I have read this submission and believe that I have an appropriate level of
expertise to confirm that it is of an acceptable scientific standard.
AuthorResponse18Dec2019
,UniversityofWashington,Seattle,Seattle,USASema Sgaier
Wethankthereviewerforthepositiveevaluationofanddetailedcommentsonourwork.We
appreciatethecomments’focusonincreasingthelikelihoodoftheframework’suptakeby
practitioners,andhaveadaptedthearticleinthefollowingways:
1. Make it easy: I would urge the authors to simplify the framework and drive to cognitive
ease so that practitioners can easily adopt it. For example, the authors may weigh the
benefits of including the "Stages of Change" part of the model against the complexity it
introduces for implementation. This could certainly be a more in-depth version of the
basic model but in the first instance, defining the behavior well and then understanding
perceptual and contextual barriers and enablers to that behavior would go a long way to
designing effective programs.
It would also be helpful to give practitioners some rules of thumb that are actionable. For
example, I would recast the narrative of the framework in the form of three simple rules:
Page 22 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
1.
2.
3.
example, I would recast the narrative of the framework in the form of three simple rules:
1) Understand determinants of behavior: barriers and enablers, both perceptual and
contextual.
2) Design idea-channel interventions that address barriers and leverage enablers.
3) Design to all levels of change – individual, family, society.
Thebalancebetweensimplicityanddepthisindeedchallenging,andwehavefacedthisboth
internallyandwhenworkingwithpractitionersfromotherorganizations.Ultimately,webelievethat
inthismanuscript,thereisvalueinretainingcomplexityofthedriversatthecurrentlevel.For
example,thepresentStagesofChangehavealreadybeenreducedfrommorenuancedstagesin
theTranstheoreticalModel,andwebelievethatunderstandingbehaviorasajourneyfrom
awareness/skillstointentiontoactionandbeyondiskeytocreatingsuccessfulinterventionsthat
highlightthatawarenessisnotenough.However,wherewefullyagreewiththiscommentisthat
weneedtocreatebettertoolstohelppractitionersengagewiththetoolkit.Therefore,wehavenow
initiatedbuildingawebsitewherewewillprovidemorein-depthandanimatedwalk-throughsofthe
toolkit,inadditiontofurthercasestudies.
Wealsoappreciatetheactionablethreerulesofthumbre-framingthenarrativeofCUBES,and
haveincorporatedthemintotheIntroductioninaslightlymodifiedform:
Ultimately, we encourage practitioners use the toolkit to:
Understand determinants of behavior: barriers and enablers, both perceptual and
contextual.
Design idea-channel interventions that address barriers and leverage enablers.
Design to all levels of change – individual, family, society, and systems.
2. Consider adding the construct of goals & identity to the perceptual barriers and
enablers: All human behavior is goal driven. Goals drive attention and value and are
therefore critical to motivation and establishing intent. At the same time, current research
on the field prioritizes knowledge and attitudes rather than goals. In doing so, we often
conclude research with little idea of this important determinant of motivation - people's
goals and how the intended behaviors can help them achieve their goals. The authors
mention the construct of goals several times in the article but should consider including
this construct in the framework. Similarly a focus on identity, particularly critical in habit
formation would add to the richness of the perceptual barriers and enablers.
Thesearehighlyimportantpoints.Weagreewiththeimportanceofseeingbehavioras
goal-directed,buthavenotincludeditasaseparateconstruct,asitfoldsintotheconstructof
‘intention’.Whilegoalsandintentionsarenotequivalentconcepts,wehavenowclarifiedinthe
ResultssectionandthelegendofFigure1thatintentioncanmeanaplanofaction atowards
specificgoal(AjzenI,MaddenTJ.Predictionofgoal-directedbehavior:Attitudes,intentions,and
perceivedbehavioralcontrol.JournalofExperimentalSocialPsychology.1986;22:453–474):
The CUBES framework therefore divides the behavioral change process more simply into three
stages of knowledge (ending with the necessary awareness or skills to engage in a behavior),
intention (which can also be understood as a plan of action towards a specific goal), and action.
Forinterventiondesign,researchonspecificsofgoal-settinghasprovidedfascinatinginsights:for
instance,specificgoalsarebetterthanvagueencouragementsto‘doone’sbest’,goalsshouldbe
challengingyetachievable,andtime-proximalgoalsaremoreeffectivethangoalsfarinthefuture
(LockeEA,LathamGP.Buildingapracticallyusefultheoryofgoalsettingandtaskmotivation:A
35-yearodyssey.AmericanPsychologist.2002;57:705–717).Wethereforeagreethat
incorporatingtheconceptofgoalsintointentionsisveryuseful,although,aswementionbelow,in
Page 23 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
incorporatingtheconceptofgoalsintointentionsisveryuseful,although,aswementionbelow,in
thisarticlewecannotfocusonthemechanicsofinterventiondesign.
Wehadoriginallynotincorporatedtheconceptof‘identity’(suchasprofessionalorsocialidentity)
intotheCUBESframeworkfortworeasons:first,identitycanrefertoasetofbehaviors,personal
qualities,orinternalbeliefs(CaneJ,O’ConnorD,MichieS.Validationofthetheoreticaldomains
frameworkforuseinbehaviourchangeandimplementationresearch.ImplementationScience.
2012;7).Forinterventiondesign,identityisoftenthoughtofmoresimplyascorebeliefsabout
oneself.Second,itisexcitingthattherearenowseveralstudiesshowingthepredictivevalueofan
identityconceptonbehavior(forexample:vanderWerffE,StegL,KeizerK.Thevalueof
environmentalself-identity:Therelationshipbetweenbiosphericvalues,environmentalself-identity
andenvironmentalpreferences,intentionsandbehaviour.JournalofEnvironmentalPsychology.
2013;34:55–63;NigburD,LyonsE,UzzellD.Attitudes,norms,identityandenvironmental
behaviour:Usinganexpandedtheoryofplannedbehaviourtopredictparticipationinakerbside
recyclingprogramme.BritishJournalofSocialPsychology.2010;49:259–284).However,the
evidencebaseontheconcept’spredictivenessonbehaviorisstillinconsistent,andtoolimitedin
comparisontoothertypesofbeliefstorecommenditsinclusionintotheCUBESframeworkatthis
point.Wedo,however,nowmentiontheemergingevidenceinthemanuscript(Results):
Beliefs around (professional or social) self-identity may also be predictive of behavior. For
example, environmental self-identity, or seeing oneself as ‘a person who acts
environmentally-friendly’, is related to several environmental behaviors (van der Werff E, Steg L,
Keizer K. The value of environmental self-identity: The relationship between biospheric values,
environmental self-identity and environmental preferences, intentions and behaviour. Journal of
Environmental Psychology. 2013;34: 55–63). However, empirical research on identity and behavior
is still emerging.
3. Clarify the role of social norms: It was not intuitively clear to me why social norms
which are strongly rooted in psychology are part of contextual factors that has other
non-psychological components like laws & infrastructure.
Wethankthereviewerforbringingourattentiontothisimportantpoint.Wehavenowfurther
clarifiedourrationaleinthe‘Results–Contextualdriversofbehavior’section:
Social norms are a construct that can only exist on the level outside the individual: through
collective behavior and ‘shared knowledge’, norms describe a set of practices of what other people
do (descriptive norms), or prescribe what people should do (prescriptive norms). Both of these
may influence attitudes and behavior (Smith JR, Terry DJ, Manstead ASR, Louis WR, Kotterman
D, Wolfs J. The attitude–behavior relationship in consumer conduct: the role of norms, past
behavior, and self-identity. The Journal of Social Psychology. 2008;148: 311–334). Unlike
individual-level beliefs, norms usually imply some consequence to the individual should they
deviate from the norm, such as disapproval (Brauer M, Chaurand N. Descriptive norms,
prescriptive norms, and social control: an intercultural comparison of people’s reactions to uncivil
behaviors. Eur J Soc Psychol. 2010;40: 490–499).
4. Create more specifics for each construct: The overall article does a good job of clearly
describing each construct. It would be helpful as a job aid to present a specific checklist
for each construct. For example, the different types of beliefs one could uncover in
research. Practitioners often don’t know all the theoretical underpinnings of a construct
and may use their own interpretation. This would result over time in the constructs being
interpreted differently and therefore reduce the scope for cross-learning across
programs.
Page 24 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
programs.
Wefullyagreewiththereviewer’spointthatconstructsshouldbedefinedingreaterdepth,and
indeedtheyhavebythetheoreticalliteratureonwhichthistoolkithasbeenbuilt.Weseethemost
valueinbuildingoutadditionalguidanceintheformofexamplesmostlikelytobeusefulto
practitioners,suchasquestionsthathavebeenorcouldbeaskedinsurveystocaptureconstructs.
Thisisalonger-termprocess,andweaimtoincorporatethisguidanceinamoreinteractiveform
ontheupcomingwebplatform.
5. Create more detail and specificity on the design process: The article has a lot of
emphasis on the research process to generate insight. There is a much shorter section on
effective behavioral design, though this is often the stage when practitioners fail to
convert rich insight into effective ideas. Similar to the detailed manner in which the
research techniques have been outlined, it would be useful to outline the design process
with appropriate checklists. For example how to prioritize ideas, how to effectively
prototype prioritized ideas, ways to rapidly test ideas ranging from experiments to rapid
qualitative feedback loops and other such steps in the process. Within this design
process, there needs to be a clear understanding of the process to design mass media
communication as often mass media & non mass media interventions are created
separately rather than as part of one integrated program.
Inthisarticle,wehaveaimedtoprovideanoverviewofhowtosystematizepotentialdriversof
behaviorandmethodologiestogenerateinsightsaroundthem.Weagreethattheultimategoalis
toenableprogramstodesigninterventionsthataddresskeybarriersandleveragemainenablers.
However,thefocusofthisarticlecouldnotbeonthespecificsaroundinterventiondesign,for
severalreasons.Onereasonisthatreviewingtheevidenceontheinterventiontypesthatworkor
don’tindifferentcontextscannotbedonejusticeinasinglearticle.Forexample,onceaspecific
socialnormhasbeenidentifiedasabarriertoatargetbehavior,woulditworkbesttoleveragekey
influencerstochangethenorm,createmassmediaentertainmentshowingpeerrolemodels(asin
aninterventioninRwandaaroundthatusedasoapoperatomodelinter-groupharmony;StaubE,
PearlmanLA.Reducingintergroupprejudiceandconflict:Acommentary.JournalofPersonality
andSocialPsychology.2009),orsomethingelseentirely?Behavioralinterventiondesignisitsown
vastandongoingareaofresearchandtrialanderror,andherewecanonlyencourage
practitionerstodotheirownresearchonpotentiallyeffectiveapproaches,oncetheyhaveusedthe
CUBEStoolkittozeroinonthebarrierstosolvefor.However,wehavenowaddedsomeexamples
ofhowbehavioraldrivershavebeenorcouldbeaddressedtotheDiscussion(seeresponsetothe
firstcommentbyReviewer2).Second,fordetailonhowtoconductrapidprototypingofsolutions,
webelievethatpractitionersofhuman-centereddesign,andtheframeworksbydevelopment
partners(suchasJHU-CCP,PSI,andFHI360)raisedinthesecondpointbyReviewer1,already
providerichguidance.Incontrast,wethinkthismanuscriptwillbemostimpactfuliffocusedon
categorizingandmeasuringdriversofbehavior.
Nocompetinginterestsweredisclosed.Competing Interests:
01 April 2019Reviewer Report
https://doi.org/10.21956/gatesopenres.14025.r27007
Page 25 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
© 2019 Ferrand R.ThisisanopenaccesspeerreviewreportdistributedunderthetermsoftheCreativeCommons
,whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalAttributionLicense
workisproperlycited.
Rashida A. Ferrand
LondonSchoolofHygieneandTropicalMedicine,London,UK
BiomedicalResearchandTrainingInstitute,Harare,Zimbabwe
Thispaperdescribesanapproachtocomprehensivelyconsiderdriversofbehaviourchangetoinform
effectiveinterventiondesign.
Theauthorsproposeaframeworkthattakesintoaccountbothperceptualandcontextualfactorsaswell
asinfluencersofbehaviourbuildingonavailablemodels.Thisisarefreshing,muchneededapproachif
behaviouralinterventionsaretobeeffectiveandleadtoimpact.Iparticularlyliketheemphasison"repeat
andhabit"intheirframework,inthestagesofchange.
Issues:
Iamnotconvincedbysomeoftheapproaches(andtheiruse)thattheauthorspresentfor
informinginterventionsinglobalhealthanddevelopment.Whileparticularfactorsmaybe
important,howtheycanbemeasuredandmodifiedneedsfurtherthought.Forexample,the
authorsmentionuseofpersonalitytestsandmeasurementofperceptualbarriers;incorporating
thisininterventiondesignisnotthateasyparticularlyinpublichealthinterventions.Itisimportantto
consider thefindingscanbeusedininterventiondesign.how
Theauthorsneedtoconsiderthelimitationsof"invitro"experiments.Behaviour,astheauthors
pointout,iscomplexwithmultiplemediatorsandinvitroexperimentsmaybesubjecttothesame
limitationsastheapproachesusedtodate,particularlyifsomefactorsinthestagesofchangesare
notamenabletochange.Theissuesaroundeffectivenessversusefficacyarestillcriticalinthis
context.
Theauthorspresenttwoapproachestheyhaveused:1)inZimbabweandZambiatoscaleup
VMMCand2)inIndiaonTBcare-seekingandmaternalandchildcare.Theymentionthatthey
havefoundeffectivenessintheirownprogrammes.Disappointingly,thereisnoevidence
presentedoftheeffectivenessoftheapproachused-theapproachinZimbabwewaspublishedin
2016andtheauthorsstatethatitisbeingpilotednationallyinZimbabwein2019.Isthere
preliminarydataonuptakeetc.?AvailabledatafromDHSetc.infactsuggestthatthatVMMCrates
havefalleninthisregion.Similarly,theauthorspresentnoevidenceoftheeffectivenessofthe
frameworkandmethodstheypresent.Cantheauthorsdemonstrateevidenceoftheeffectiveness
oftheseapproaches?
Limitationsaboutthefeasibilityofusingtheseapproachesandtheirscientificvalidityandthe
feasibilityofthefindingsfromtheseapproachestoinformand"modify"thebarriersandenablers
needsmentionanddiscussion.Ultimatelythesearethefactorsthatwilldeterminewhetherthe
CUBESmodelisuseful.
1
2
Page 26 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
Is the rationale for developing the new method (or application) clearly explained?
Yes
Is the description of the method technically sound?
Yes
Are sufficient details provided to allow replication of the method development and its use by
others?
Partly
If any results are presented, are all the source data underlying the results available to ensure full
reproducibility?
Partly
Are the conclusions about the method and its performance adequately supported by the
findings presented in the article?
Partly
Nocompetinginterestsweredisclosed.Competing Interests:
Reviewer Expertise:Epidemiologyandpublichealth,implementationscience,behaviouralinterventions
I confirm that I have read this submission and believe that I have an appropriate level of
expertise to confirm that it is of an acceptable scientific standard, however I have significant
reservations, as outlined above.
AuthorResponse18Dec2019
,UniversityofWashington,Seattle,Seattle,USASema Sgaier
Wethankthereviewerforthedetailedandusefulcommentsonourarticle,andforhighlightingthat
thisapproachtoapproachingbehavioralinterventionsismuch-needed.Takingthedifferentpoints
offeedbackintoaccount,wehaveamendedthearticleinthefollowingways:
1. I am not convinced by some of the approaches (and their use) that the authors present
for informing interventions in global health and development. While particular factors may
be important, how they can be measured and modified needs further thought. For
example, the authors mention use of personality tests and measurement of perceptual
barriers; incorporating this in intervention design is not that easy particularly in public
health interventions. It is important to consider the findings can be used inhow
intervention design.
Wefullyagreethatmeasuringandcreatinginterventionsaroundamorecomprehensivesetof
behavioraldriversisnottrivial.Wethinktherearetwopartstothiscomment.Oneisonthe
researchmethodsthemselves.Inthisarticle,wehavealreadygivennuancedintentionsaround
theseapproaches:wedonotaimtoclaimthatallmethodsareequallyuseful,informative,and
validatedforallglobalhealthanddevelopmentusecases,orevenoverall.However,weoutlinea
Page 27 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
validatedforallglobalhealthanddevelopmentusecases,orevenoverall.However,weoutlinea
setofoptionsandchoices,allofwhichhavebeendeveloped,used,andvalidatedinothersectors.
Thesecondpartisonthendevelopinginterventionsthatfitthebroaderspectrumofbarriersfound,
whereweagreeaddressingsomebarriersmaybeharderthanothers,butperhapsthebigger
questionshouldbeabouthow itistoaddresssuchbarriers.Whilethismanuscriptdoesnotuseful
addressinterventiondesignitself,wementiononeapproachtocategorizingtypesofinterventions
bybehavioraldriversintheDiscussion(MichieS,RichardsonM,JohnstonM,AbrahamC,Francis
J,HardemanW, .Thebehaviorchangetechniquetaxonomy(v1)of93hierarchicallyclusteredet al
techniques:buildinganinternationalconsensusforthereportingofbehaviorchangeinterventions.
AnnalsofBehavioralMedicine.2013;46(1):81-95.).IntheDiscussion,wehavenowaddedmore
detailaroundwhatinterventionsaddressingperceptualdriverscouldlooklike:
Creating interventions to fit the varied barriers to behavior is a challenge as well as an opportunity
for global health. For example, higher levels of conscientiousness have consistently been
associated with higher adherence to medication (Molloy GJ, O’Carroll RE, Ferguson E.
Conscientiousness and medication adherence: a meta-analysis. Annals of Behavioral Medicine.
2014;47: 92–101) or the contraceptive pill (Leahy D, Treacy K, Molloy GJ. Conscientiousness and
adherence to the oral contraceptive pill: a prospective study. Psychology & Health. 2015;30:
1346–1360). Rather than attempting to influence conscientiousness, an intervention might consist
of identifying those with lower conscientiousness and targeting increased levels of support to this
sub-population. In other situations, drivers may be affected directly. For example, using Facebook
to alert college students that peer social norms around drinking were lower than they thought
changed drinking behavior (Ridout B, Campbell A. Using Facebook to deliver a social norm
intervention to reduce problem drinking at university: social norm intervention using Facebook.
Drug Alcohol Rev. 2014;33). As an example of targeting a belief, enhancing self-efficacy through
encouragement on progress, attribution of progress to participants’ own abilities, observation of
others carrying out the target behavior, and other strategies significantly increased physical activity
in older adults (Allison MJ, Keller C. Self-efficacy intervention effect on physical activity in older
adults. West J Nurs Res. 2004;26: 31–46). These examples are by no means indicative of success
in other contexts and behaviors, and interventions all need to be piloted. However, using a
framework of behavior allows for the identification of possibilities that may otherwise remain
hidden, and conversely narrow the options for choosing suitable intervention types.
2. The authors need to consider the limitations of "in vitro" experiments. Behaviour, as the
authors point out, is complex with multiple mediators and in vitro experiments may be
subject to the same limitations as the approaches used to date, particularly if some
factors in the stages of changes are not amenable to change. The issues around
effectiveness versus efficacy are still critical in this context.
Weagreewiththereviewerthat‘invitro’experimentshaveinherentlimitations,astheyarenotfully
replicativeofreal-worldbehaviorsintheirrichcontext.Westronglyarguethatthislimitationis
differentfromthoseofapproachesusedtodate.‘Invitro’experimentalapproachesprovide
additionaldataonwhichinterventionstoscaleup,asexperimentsuniquelylinkcontrolledinputsto
actualbehavioraloutcomes.Furthermore,alargersetofdriverscanbeincorporatedandtested
morerapidlyandonasmallersamplethaninafieldRCT.However,thereviewersraisean
excellentpointonthevalidityof‘invitro’approachesregardingbehaviorchange.Wehavenow
addedmorenuanced,butspeculative,guidanceonthekindsofbehaviorsthatmightespecially
benefitfrom‘invitro’testingtotheDiscussion:
Page 28 of 36
Gates Open Research 2020, 3:886 Last updated: 06 JAN 2020
Gates Open Research
‘In vitro’ experiments have inherent limitations, as they are not fully replicative of real-world
contexts and behaviors that participants are asked to engage in. However, experiments can link a
large set of features to an actual behavioral outcome in a controlled way. It is plausible, but