Rethinking Behavioral Approaches to Compliment Biological Advances to Understand the Etiology,Prevention,and Treatment of Childhood Obesity

Twelve-Week Solutions

Childhood obesity has been a persistent and global public health issue over the past 30 years. ⁵ The etiology of obesity is driven by a complex web of social, behavioral, environmental, and genetic factors. Given the persistence of the problem and the complexity of the etiology, we ask whether it is rational to expect childhood obesity to be successfully addressed with a 12-week solution? Many behavioral interventions are delivered over a 3-month time frame (origin unknown). There is ample evidence of these short-term approaches leading to behavior change during the intervention, but, once the intervention ends, behaviors and any weight-related changes often regress to preintervention levels. Altering obesogenic behaviors for a short time frame is insufficient for triggering longer term behavior change, which would be necessary to alter current weight status and long-term weight trajectory. This giveth and taketh away perspective, where interventions provide support for short-term behavior change but then remove that support, fails to recognize the complex interactions between social, genetic, environmental, and behavioral determinants of obesity. Further, adopting this approach does not address the challenges associated with altering behaviors that may be reinforced or continuously undermined outside the intervention delivery setting that is highly obesogenic.^6,7 For these reasons, we contend that 12-week interventions are incompatible with our understanding of human behavior, and the pervasive and persistent practice of treating this public health problem with a 12-week solution may be contributing to the inability to halt the childhood obesity epidemic.

Intervening on Interventions

When interventions are delivered over longer durations, they often target settings where children already engage in the healthiest behaviors. These include schools, child care centers, afterschool and summer programs, and organized sports. The rationale for targeting these settings is based upon the ability to capitalize on the reach of these settings. The majority of children attend school regularly, 14% attend afterschool programs, ⁸ over 19% attend summer camps, ⁸ and 37% play on a sports team regularly. ⁹ Of these settings, school has been the primary location for delivering obesity-related interventions. ⁴ However, a robust evidence-base indicates when children are in these settings they engage in more activity, reduce their screen time, have more consistent sleep schedules, and eat a more healthful diet compared to when they are not in these settings.^10–13 For instance, studies show that school, in the absence of a formal obesity intervention operating under routine practice, has a weight prevention effect (i.e., children gain minimal or no weight) for children classified as healthy weight and a treatment effect (i.e., children lose weight) for children classified as overweight or obese.^14,15 Moreover, making playgrounds more conducive to activity, or increasing activity in physical education (PE) and sports fails to recognize that these settings are where children are the most active throughout their day. Likewise, attempts to improve the foods/beverages served and consumed in these settings may result in negligible effects given there is limited room to make changes because the food environment is highly regulated by state and federal agencies for caloric and nutrient content. For these reasons, we believe the majority of behavioral childhood obesity interventions that target such settings are simply intervening on what is already one of the most effective childhood obesity interventions, and thus unlikely to be the most effective use of resources on a population level.

Public school provides universal access to a structured environment that, in most instances, is health-promoting. In contrast, not all children are afforded the access to programs outside of the school setting. Many millions of children are unable to regularly access afterschool programs, summer day camps, or participate in organized sports.^8,16 The reasons for this are multifaceted, but the fee-for-service model upon which these types of programs operate is a major contributor, especially for children from low-income households. ⁸ Instead of pouring resources into increasing the health promotion of settings where few children have access, providing access to those who currently are unable to access the health benefits that these programs provide may produce greater public health benefits.

Beyond the Individual

Historically, public health accomplishments were realized through mandatory, macro-level policy-driven participation. In the childhood obesity prevention and treatment field, there is an overreliance upon behavioral theories that focus almost exclusively upon psychosocial individual motivational behavior change processes and/or the need for greater education (e.g., knowledge, skills) about the choices one should make.^17,18 Approaching interventions through this lens results in solutions designed to help individuals to optimally respond (make choices) within environments where few healthy alternatives may be present and ignores the fact that an individual's behaviors are highly influenced by environmental cues to engage in less healthful behaviors. For individuals with means, individually focused behavior change processes may be sufficient. ¹⁹ Yet for high risk, vulnerable populations (e.g., children from low-income households), asking them to make better choices within environments that offer few “good” choices is disconnected from the larger macro-level factors and social determinants of health experienced by children and parents in these settings. ²⁰

Furthermore, the reliance on individual behavior change models of obesity are not only ineffective but also harmful. ²¹ The subtext of interventions that emphasize individual level behavior change implies that the root causes of obesity lie within an individual's immediate control. This perspective reinforces societal views of obesity as a personal failure of character or willpower. ²² Behavioral science's nearly exclusive reliance on the individual-level model of behavior change reinforces the societal perspective of obesity as a personal failure and serves to alienate both individuals and communities (particularly low income and minority communities) and increases inequities in populations already disproportionately affected by health disparities.^23,24

There have been many attempts to develop comprehensive interventions that focus on changes in policies, systems, and environments, and individual behavior change. Many of these interventions, however, have relied upon strategies within sectors that have demonstrated limited impact on behavioral or weight changes by themselves (e.g., homework activities, point-of-decision signage for stair usage).^18,25 Although some approaches target multiple settings, they are still comprised of largely individual-focused behavioral change approaches within those settings.^26,27 The prevailing wisdom is that combining approaches that have demonstrated limited short-term effectiveness will somehow result in long-term effectiveness through some unknown synergistic means to result in reductions in childhood obesity. Unsurprisingly, this has not been realized.

Combining many limited-effective approaches does not yield a new and more effective approach; rather it simply yields an equally ineffective, more costly approach. When policies, systems, and environmental changes are employed, they have largely been surface-level within a setting and have failed to address the root causes of childhood obesity. For example, providing posters to promote behavioral goals, highlighting healthy choices in vending machines, parenting classes with grocery store field trips, or giving low-income families smaller plates and glasses to decrease portion size consumption do not address key barriers to engaging in more health-promoting behaviors within highly obesogenic environments.^28,29 Thus, it is unlikely we will be able to goal set, self-monitor, or prompt our way to better health in environments that are highly obesogenic.

Developing Scalable Interventions

The design, implementation, and evaluation of childhood obesity prevention or treatment interventions often consists of a multi-stage pipeline of studies that commonly begin with pilot/feasibility studies, followed by efficacy and effectiveness studies, and then a scaled version of an intervention to be tested for dissemination/implementation. ³⁰ Interventions following this developmental pipeline are typically evaluated at the early stages under optimal conditions, then transition to more “real-world,” pragmatic conditions for testing. ³¹ Unfortunately, optimal conditions rarely exist for addressing public health issues, and interventions tested under such conditions are likely to fail when they are tested under the actual conditions for which they were intended. Recently published reviews demonstrate numerous failures to translate results from pilot or efficacy studies to larger scale interventions. This pattern is observed when interventions are taken from a pilot to a larger, well-powered trial or attempts are made to scale a formerly “efficacious” intervention tested as a well-powered efficacy study.^32,33 Pilot or efficacy studies are often delivered by highly trained interventionists, in optimal settings and to receptive audiences. In contrast, when those same interventions are scaled, they are usually delivered by lesser qualified delivery agents, in more diverse settings, and to less receptive audiences. What's apparent is that the testing of public health interventions at any stage under optimal conditions is not based in reality and has a high likelihood of failure, yet such practice is a hallmark of behavioral science's approach to intervention development. ³⁴ This may be a function of the way the peer-review process operates, and funding decisions are made. Regardless of origin, evidence is emerging that such a process leads to exaggerated early discoveries that can lead to subsequent failed well-powered trials^32,33

Advancing Behavioral Monitoring

Over the past 30 years, behavioral science has moved away from self-reported measures and adopted, where possible, objective assessment of obesogenic behaviors. This was accomplished through the introduction in the 1990s of actigraphy devices to measure physical activity, sedentary behavior and sleep, and more recently the use of digital photography to measure dietary intake. Despite these improvements, the scientific field has largely failed to embrace more recent advances in digital health that incorporate devices with multiple channels of sensor input for the classification of behaviors. ³⁵

Since the late 1990s, studies have shown that a combination of multiple channels of input, such as heart rate and accelerometry, produce more precise estimates of energy expenditure compared to either heart rate or accelerometry alone.^36–40 More recently, the combination of heart rate and accelerometry has proven more accurate in estimating sleep than relying upon actigraphy alone.^41,42 Unfortunately, such approaches have not been widely adopted, nor have the majority of manufacturers of scientific-grade motion devices attempted to incorporate additional input channels into a single device. This lack of advancement has allowed consumer-grade wearable devices—which now almost all include multi-channel capabilities capturing accelerations and heart rate via photoplethysmography—to gain market share for their use in behavioral research. ⁴³ The consumer market is not the only area making technological advances. The area of stretchable, wearable, miniaturized multi-channel devices is being rapidly developed by nanoengineers who have assembled prototypes that capture a variety of physiological input (e.g., heart rate, accelerations, skin temperature) within a single device the size of a postage stamp. ⁴⁴

One response of the behavioral science field to addressing the limitations of single channel information (i.e., accelerations) has been to employ more complex statistical modeling approaches for estimating energy expenditure and classifying activity levels. However, these approaches have provided no greater accuracy than traditional cutoff classification schemes. ⁴⁵ When attempts are made to incorporate multi-channel input to better understand behaviors, the solution has been to simply add more devices, with some researchers taping separate devices to different locations of the body as a means to improve estimates of 24-hour behaviors. ⁴⁶

Advancements in wearable devices, however, does not address issues with objectively capturing dietary or screen time behaviors. The ability to more accurately measure diet has eluded the behavioral science field for decades. The use of digital photography to measure consumption with photos taken before and after meals ⁴⁷ and e-buttons ⁴⁸ that capture images at prespecified intervals within a day exist but are not widely used and have their own inherent issues when deploying them with children. Advancements in measuring screen time have been fewer. Apps to monitor screen usage on smartphones and tablets exist, as well as embedding screen usage monitoring in software updates (e.g., iOS). Evidence from studies with adults show promise, ⁴⁹ but we are unaware of studies utilizing such technology with children.

Obesogenic Behaviors Are Dynamic

Widely accepted protocols for measuring obesogenic behaviors employ 7 or fewer days to assess “typical” levels of a behavior. This limited time frame presents numerous challenges for understanding highly complex and dynamic behaviors. Obesogenic behaviors are influenced by numerous contextual factors outside the control or even awareness of the individual. These include illness/injury, holidays/special events, ⁵⁰ weather/seasons,^51,52 lunar phases, ⁵³ and school. ¹³ The fallacy with current observation protocols is that four “complete” days of accelerometry or three 24-hour dietary recalls, adequately capture dynamic and ever-changing behaviors. Even when statistical indices provide evidence of the consistency of a behavior over short monitoring time frames (i.e.,<7 days), behavioral scientists need to ask whether, conceptually, this truly and satisfactorily represents a behavior. The assumption made here is that behavior by nature is stable and able to be captured and conveyed in a single metric. The idea that a few days can be extrapolated out across an entire calendar year to represent “typical” levels of a behavior is doubtful. Recognizing that behavior is dynamic, and measuring the implications of variability, cycles, and nonlinier dynamic relations between behaviors is essential.

More Than One Behavior

There is robust evidence linking children's obesogenic behaviors to overweight or obesity. Traditionally, obesogenic behaviors have been assessed and analyzed in isolation (i.e., physical activity and prevalence rates of overweight/obesity). This may largely be a function of how researchers are trained, with methodologies and measurement criteria shaped by a longstanding scientific paradigm that has called for narrowly focused investigations of a single obesogenic behavior in a given population and setting.

More recently, public health experts have called upon the behavioral sciences to consider a more integrated approach to better understand how obesogenic behaviors interact with one another within and across days. ⁵⁴ Regarded as a paradigm shift from current practice, exploring the “natural and intuitive integration” of the time children spend active, sedentary, and asleep, and their synergistic impact on childhood obesity has created multiple opportunities for future research. ⁵⁵ The potential of this relatively “new space” is complimented by advances in technology that allow for the measurement of multiple obesogenic behaviors from a single device, and incorporating integrative statistical techniques–that have largely been ignored since their introduction ⁵⁶ –such as compositional data analysis ⁵⁷ that can accommodate the assessment of patterns of co-dependent behaviors rather than individual “isolated” behaviors. ⁵⁸