Overview

The NIH Health Behavior Theories Project aimed to identify core needs and activities required to advance the quality and use of health behavior theories (HBTs) and establish a foundation to support future progress. The project’s vision was to advance the conduct of health behavior research that is meaningfully guided by specific and predictive theories and contributes to a cumulative evidence base that ultimately advances efforts to promote healthy behaviors. Leveraging HBTs across the research translation continuum will enable the systematic development and testing of potent interventions, deep understanding of the mechanisms through and conditions under which they operate, and efficient implementation of interventions into widespread use.

To advance these aims, four subgroups, each co-led by an NIH representative and an academic expert, were created to discuss a core issue surrounding HBTs and lay the groundwork for future progress:
 

• Subgroup 1: What an HBT Should Do and How We Can Get There
• Subgroup 2: Advancing HBT Development and Application to Reduce Health Disparities
• Subgroup 3: Synthesizing Evidence to Advance HBTs
• Subgroup 4: Novel Approaches for Advancing Theory Development and Testing

In this one-day conference, HBT Project members came together to present on their work. The conference aimed to highlight the insights of each subgroup, engage attendees in dialogue, and foster discussion across the subgroups to identify core challenges, needs, and opportunities to advance HBTs.

Subgroup 1: What an HBT Should Do and How We Can Get There

Members: Alex Rothman (Co-Chair), Arielle Gillman (Co-Chair), Angela Bryan, Pedja Klasnja, Allecia Reid, Paschal Sheeran

Presentation Highlights

• A key challenge with the prevailing health behavior theories is that they remain static in their structure despite their use and the on-going accumulation of evidence regarding the relationships they specify. The subgroup argued that HBTs need to be reconceptualized as dynamic entities that evolve in response to new empirical findings
• Culture change will be needed to shift this understanding of HBTs. Currently, theories are used as tools to guide research.  Yet, the evidence that comes from this work is rarely applied to the theory – strengthening its arguments in the face of supporting evidence or modifying its arguments in the face of evidence that does not align with the theory.
• Theories do not address key features of relationships among constructs that they specify; the time course; the shape of relationships; the ease with which constructs can be changed, and, if they can be, what the most effective technique is for doing so. More precision and specificity in HBTs are needed.
• The field should agree on a common set of core constructs and resolve the overlap between constructs to move beyond the jangle fallacy. The malleability of a given construct should be identified and acknowledged in theories, and the assumption that relationships between constructs are linear should be addressed. The time course and magnitude of the relationships between constructs also should be identified, and the conditions that modify such relationships should be postulated.
• The subgroup presented a refined approach to graphically representing health behavior theories with goal of making salient the state of key features of the theory -- to make it easier to add or revised information regarding the theorized relationships, highlight what is not known, and identify areas where future research is needed. These changes require collective action, as theories are collectively owned by the community and therefore should be developed collectively.

Discussion

• One key idea is the nature of theories as communal properties. Data that inform the development of a theory are valuable regardless of whether they support or challenge that theory, because even negative findings help to evolve a theory. Increasing precision may generate more negative findings, but over time should enable researchers to utilize theories more purposefully. Current theories have broad scopes of acceptance, which makes discerning consistent phenomena difficult.
• Scholars trying to guide theory development should prioritize focusing on developing explanations for a specific behavior rather than searching for a universal theory that can be applied to all behaviors.
• Approaching theories as entities that are managed by the research community would improve the translatability of core constructs among groups who use different vocabularies. For example, public health experts can be added to an interdisciplinary team to help identify contexts that affect theories and constructs. The ideal makeup of an interdisciplinary team will depend on the group’s goals, but both content and methodological expertise likely is required.
• A good theory explains what is already known, generates new predictions, and has the flexibility to evolve. By these criteria, many current theories are insufficient. Criteria for evaluating theories may be needed.
• Entities that fund interventional research want the evidence generated by that work to be used – both in guiding the design and application of behavioral interventions and in guiding the improved specification of health behavior theories. 

Subgroup 2: Advancing HBT Development and Application to Reduce Health Disparities

Members: Sydney O'Connor (Co-Chair), Maria Fernandez, (Co-Chair), Monica Webb Hooper, Darrell Hudson, Deborah Linares, Lorraine Halinka Malcoe, Rena Pasick, Rachel Shelton, Mary-Louise Millett

Presentation Highlights

• Health disparities are largely preventable health differences that adversely affect populations who experience greater challenges to optimal health. Health behaviors are shaped by a complex interplay of influences across multiple domains; these levels vary across population and location and compose the context in which health behaviors occur.
• HBTs have limited contextual integration, narrow development bias, limited theory adaptation, and methodological constraints. The subgroup aimed to explore the extent to which health disparities and broader context have been considered in the development and application of HBTs to date and identify future avenues to refine HBTs to reduce health disparities.
• Current HBTs overlook context and were developed primarily through work that focused on groups not experiencing health disparities, limiting their relevance and effectiveness in diverse populations. The research community has an opportunity to adapt or create HBTs to ensure they reflect the lived realities of people experiencing health disparities. Refinement of existing HBTs should prioritize empirical examination of applicability across populations and settings.
• Broader frameworks can be integrated into HBTs to support better understanding of behavior. Such frameworks should focus on contextual or socioecological levels.
• In addition to adapting HBTs, applying HBTs also can be enhanced to better consider contextual factors. Frameworks that explicitly account for social, environmental, and temporal contexts can be integrated to improve the relevance and impact of interventions.
• Systems thinking and co-development of interventions can make future HBTs more applicable for diverse populations. Complementary frameworks can be integrated to capture multilevel influences on health behavior.
• Crosscutting methodological approaches, including mixed methods, participatory and community-engaged methods, and systems thinking, can be leveraged to go beyond the deductive, quantitative methods that have generally been used to develop HBTs within narrow populations.

Discussion

• Training can be improved to help students recognize the interconnectedness and context of HBTs. Incorporating many models is critical to making theories relevant to many populations, and trainees should be guided to think about multilevel perspectives on HBTs, embed theories in context, and consider lived experiences. Interdisciplinary coursework also can be encouraged, and students can be encouraged to participate in partnerships and mentored research.
• Social determinants of health often are foundational in disparities, and some are modifiable. Context is not static—circumstances and policies may change, which can affect health. Quantifying constructs is difficult, but researchers can focus on incorporating theories that show the relationship among changing contextual factors and health behaviors.
• Ensuring interventions are appropriate for the communities that need them without creating new interventions for each population and context is difficult. Implementation and scalability of narrow interventions must be considered, as well as the question of whether communities will use them.

Subgroup 3: Synthesizing Evidence to Advance HBTs

Members: Elise Rice (Co-Chair), Katie Witkiewitz (Co-Chair), Carmela Alcantara, Talea Cornelius, Martin Hagger, Emily Hennessy, Bryan Kutner, John Sakaluk

Presentation Highlights

• Conducting syntheses (e.g., systematic reviews; meta-analyses) can illuminate the stability of theoretical models across a body of primary studies, test whether constructs proposed by an HBT act as mediators across studies that can explain an outcome, identify moderators (i.e., conditions that strengthen or weaken an observed effect), evaluate predictions made by new or emerging theories, and establish benchmarks for effects.
• Syntheses may be affected by bias in primary studies; heterogeneity in populations, control conditions, intervention components, measurement, and environment; and lack of transparency in reporting. Determining to what extent primary studies are examining the same pathways and measuring the same constructs is difficult.
• Primary research studies should adopt a rigorous study design that enables the testing of mechanisms; clearly describes the theoretical pathways by which the purported mechanisms of the intervention on outcomes are proposed to operate; clearly defines the mechanism and uses appropriate measures of variables and constructs; and reports the methods and results of the data analysis techniques used.
• Longitudinal and experimental studies with factorial designs may be able to address specificity needs. By carefully operationalizing constructs, selecting appropriate measures, and including moderators in power calculations, primary studies can provide many opportunities for synthesis.
• Additional considerations for enabling synthesis include using standardized terms, describing the specific time points at which the measure is used, and justifying decisions at each step.

Discussion

• Many current systematic reviews and meta-analyses take a confirmatory approach rather than asking whether theoretical hypotheses are borne out across studies. Specificity in a meta-analysis can help produce results that can be used to develop or adapt theories.
• To facilitate conversation between the findings generated by synthesis and the theories, synthesis findings must be translated, and researchers should partner with communities to discuss outcomes that are important to them. The subgroup’s paper is intended to be a call to action.
• A meta-analysis can only synthesize data included in the primary study; although more complex questions may be relevant, a meta-analysis may only be able to assess whether an intervention is effective.
• The field is beginning to provide more incentives for publishing unexpected results, and ideally the field will become more transparent over time.
• Bias in primary studies makes interpreting data from underrepresented groups difficult for analysts undertaking syntheses. Individual participant data may be helpful but is often difficult to access. A rigorous study design should assess the behavioral pathway of interest rather than using race or ethnicity as a proxy for that pathway.
• Emerging technologies, such as artificial intelligence, could be useful, but improving primary studies is more important. Standardized reporting and ontologies, which are necessary for both artificial intelligence and general improvements, should be the focus. Journals can contribute to this effort with such tools as standardized checklists.

Subgroup 4: Novel Approaches for Advancing Theory Development and Testing

Members: Angela Pfammatter (Co-Chair), Audie Atienza (Co-Chair), Marissa Burgermaster, David Conroy, Kayla de la Haye, Genevieve Dunton

Presentation Highlights

• Innovation can be a transformative catalyst for theory generation, testing, and refinement. The abundance of data produced by recent technologies can prompt new ways of thinking about HBT development. Better dissemination of HBTs can foster improvements in the theories and their use, and policy actions can help improve connections between HBT development and implementation.
• Temporal specificity of hypothesized relationships has received limited attention. The timescale, lag in changes, and temporal precedence can be considered.
• Frameworks for characterizing situational and contextual specificity, which are high-dimensional phenomena with significant individual variation, are lacking.
• HBT testing and refinement can be advanced using the science of behavior change. Targets should be identified and validated, and methods for interventions to engage targets should be developed. Computational modeling approaches can help expose assumptions, and novel experimental approaches for optimization have been used to develop innovative interventional approaches.
• Team science and participatory approaches can offer different ways to think about theory development.
• Dissemination can be bolstered by capitalizing on ontologies, measures, and data currently available in public repositories.
• HBT development should be emphasized in training programs.
• Policy can be an important tool for guiding improvements in compliance and harmonization. The significant amount of data available can be used as a basis for learning health care and policy systems to develop ways to triage risk and phenotype characteristics.

Discussion

• Computational artificial intelligence and machine learning approaches can be used to distinguish between nomothetic and idiographic models of health behavior; significant amounts of data are required, but heterogeneity may become identifiable. HBT researchers and computational scientists can collaborate to synthesize methods and concepts in a way that those in either field can understand. Although artificial intelligence tools are powerful, human creativity is a more critical need, and any work in this area must be inspired by actual lived experiences rather than determined empirically.
• More intentionality around data capture is needed to incorporate the rich context necessary to account for health disparities. Large data-gathering companies may have data with sufficient detail, but significant amounts of cleaning likely would be required to use these data for health research.
• Artificial intelligence and other new tools should be approached broadly and used to answer existing questions. New technologies can be used to adapt old theories, but the community also should not hesitate to retire theories that are no longer appropriate.

Overall Synthesis and Discussion of Future Directions

• The HBT field must collaborate to agree on terminology. Professional societies could provide a venue for the community to work on these problems. Terminology alignment will need to address the tension between distinctions made by researchers and those made by the general public.
• Adding context to HBTs could begin with specifying for whom and when a theory does and does not work. A model unique to a single person may not be useful, but a model that can be adjusted in certain conditions to overcome interference with the mechanism of action would be helpful. Theories with varying levels of specificity may be appropriate in different situations—more precision may be required to answer some questions, and more breadth may be helpful to answer others.
• HBTs should not be expected to work for everyone, given the complexity of individuals, and the role of the theory will vary in different situations depending on the scope of change desired. As change processes are modeled, researchers can extract the principles that support change, and context can help researchers identify underlying principles that may be missing from current theories.
• Trainees would benefit from being exposed to interdisciplinary research teams so they learn how to incorporate multilevel thinking into research.  
• Learning health systems and primary care networks may prove to be valuable partners for research. Chronic illnesses have a strong behavioral component, and health care systems collect significant amounts of data, so incorporating health behavior theories in this setting is a key opportunity.
• More work is needed to understand the mechanisms of behaviors and behavior change, which is a key focus of HBTs. Work also is needed to understand the processes that underlie these mechanisms. Neuroscientists can help uncover cognitive processes that underlie engagement in health behaviors, and behavioral geneticists can assess genetic predispositions for individual differences that relate to aspects of behavior related to health, such as impulsivity and risk aversion.
• Theories are important because they help researchers organize data in ways that may be reproducible; efforts to improve rigor and reproducibility may benefit from more precisely defining what makes a theory a “good” theory.
• In most training programs, HBTs are taught in a siloed manner. Although trainees need to understand the context related to HBTs, training researchers to have interdisciplinary expertise may result in a knowledge base that lacks depth. Training should be organized at a disciplinary expertise level, but trainees should be taught to collaborate across disciplines and recognize they cannot be experts in every area. Stronger training in methods is also needed, and training must be adapted to meet current and future needs to retain its appeal to students.