The second session of the June 2021 workshop featured two presentations highlighting data-driven obesity solutions and innovative approaches in the contexts of education and health care. Panel and audience discussions about these presentations followed. Carlos Crespo, professor at Oregon Health and Science University and Portland State University School of Public Health and vice provost for undergraduate training in biomedical research at Portland State University, moderated the session.
In his introductory remarks, Crespo observed that although data play a critical role in science, they have limitations and do not work alone to change behavior. He cited emerging challenges to data use, such as the proliferation of artificial intelligence approaches based on profiles that could perpetuate intrinsic and extrinsic bias. He added that, although the field of precision medicine is also expanding, over- and underrepresentation are seen in the racial, ethnic, and anthropometric groups that are included in many of the clinical trials and biospecimen banks used to produce precision medicine guidelines. Crespo urged transparency about such limitations when leveraging data in systems-level solutions to reduce the prevalence of obesity.
INNOVATION IN EDUCATION: PHYSICAL ACTIVITY ACROSS THE CURRICULUM
Joseph E. Donnelly, professor of medicine and director of the Division of Physical Activity and Weight Management, Department of Internal Medicine at the University of Kansas Medical Center, presented findings from a 15-year series of studies focused on increasing physical activity across school curricula.
Donnelly began by listing several reasons to explore the school setting as a venue for increasing physical activity. First and foremost is that most school-age children are in these settings for 9–10 months of the year. State mandates and school mission statements encompass a variety of health outcomes, including physical activity, Donnelly continued, and research conducted in the past one to two decades has begun to explore potential relationships between physical activity and academic achievement. He stressed that the physical infrastructure of schools provides a protected environment where children can play under the supervision of an educated workforce
(i.e., teachers) that is generally respected for its authority, at least among elementary students. This workforce is already salaried to educate students, Donnelly pointed out, and physical activity can therefore be integrated at no additional (salary) cost. Finally, he noted that schools are entrenched in society and may be better positioned than other, less permanent settings to continue an intervention beyond a research study’s funding period.
Participation in schools is historically sedentary, Donnelly maintained, with students often experiencing long bus rides and what he called the traditional teaching paradigm of “sit down and be quiet,” which he said discourages movement during the off-task periods of the school day. He contended that in most cases, dedicated recess and physical education periods fail to provide adequate energy expenditure to protect against adiposity or promote fitness, a failure he attributed to challenges with equipment, space, and lack of teacher encouragement and guidance.
Donnelly presented a theoretical model of the connections among physical activity, improved health, and academic achievement (Figure 7-1), which he said can help build support for integrating physical activity into the school day. This model led to the development of an intervention called Physical Activity Across the Curriculum (PAAC), which Donnelly’s group explored in a 3-year randomized controlled trial of physical activity and academic achievement for students in second and third grades.
Donnelly explained that the premise of PAAC was to increase physical activity by using classroom teachers to integrate it into existing lessons. He clarified that this did not imply a decrease in academic instruction time and that the physical activity was not intended to be delivered as a break in the academic agenda. The intervention’s primary aim was to reduce increases in BMI, he recounted, and its secondary aims were to (1) determine associations between physically active lessons and academic achievement, and (2) describe time on task, a variable expected to be associated with academic achievement.
SOURCE: Presented by Joseph E. Donnelly, June 22, 2021; Donnelly and Lambourne, 2011. Reprinted with permission of Elsevier.
Turning to describe the conceptual framework for PAAC, Donnelly explained that it was a minimal intervention in that it did not set out to diffuse change throughout the school environment. He added that other intervention characteristics, such as the promise of enhanced learning with no additional cost or teacher preparation time, were designed to appeal to school administrators and teachers. It was important that PAAC be easily perpetuated and replicated in schools, Donnelly continued, and for it to be fun for teachers and students. Most students follow teacher directions to participate in classroom lessons, he pointed out, in contrast with recess, where they can choose to engage in sedentary behavior for the duration of the period.
Donnelly next described the PAAC approach, which was to integrate 10-minute periods of physical activity into a variety of academic lessons for a total of 90 minutes across the week, ideally distributed across morning and afternoon instruction times each day. This guidance quickly deteriorated, he admitted, into simply encouraging teachers to integrate the 90 minutes/week of physical activity whenever and however it was best for them and their classroom dynamics.
Greater levels of physical activity were observed in PAAC intervention versus control schools, Donnelly reported, based on direct observation of student behavior in response to teacher instructions using the SOFIT (System for Observing Fitness Instruction Time) tool. Average SOFIT-measured physical activity levels across intervention schools corresponded to such energy-expending movements as walking, hopping, and leaping, which Donnelly asserted could induce fitness and perhaps improve academic achievement. Although PAAC did not have an effect on students’ time spent on task, he pointed to the preponderance of evidence suggesting that classroom-based physical activity does increase time spent on task. This outcome is particularly important to teachers, he noted, who tend to be concerned that integrating physical activity into classroom lessons will distract students from learning.
Donnelly suggested that a university curriculum designed to equip future teachers with the skills to encourage classroom-based physical activity is the most low-cost, effective strategy for promoting nontraditional physical activity in schools. Such training is virtually nonexistent at the university level, he observed, but would help minimize the burden on teachers of integrating physical activity by providing guidance on how to incorporate simple movements into lessons. He argued that additional evidence linking physical activity and fitness with academic achievement would also help promote nontraditional physical activity in schools and support policy changes that could lead to more widespread dissemination of programs like PAAC.
Shifting back to the challenge of increasing physical activity in schools without decreasing academic instruction, Donnelly provided several suggestions. He described increasing children’s physically active time during physical
education class and recess as an obvious approach and then listed other ideas, including providing access to physical activity before and after school, promoting active travel (i.e., walking, biking, or other nonmotorized transport) to school, providing physically active lessons, and using physical activity as a classroom management or behavior tool. He expressed skepticism about the level of school support for integrating physical activity, referencing data from the 2016 School Health Policies and Practices Survey indicating that relatively few U.S. school districts require schools to provide regular classroom physical activity breaks (CDC, 2017). Around one-third to one-half of districts recommend such breaks, he noted, suggesting that the probability of policy change is low in the absence of a requirement.
In the final portion of his presentation, Donnelly raised a series of issues that have emerged from interventions aimed at integrating physical activity throughout the school day. One such issue is that limited evidence is available to inform whether teacher-led or outside vendor–led efforts are more effective in increasing classroom physical activity because most studies have taken the latter approach. Donnelly identified as a second issue that school settings prioritize academic learning, which unlike physical activity is tied to state standards and consequences related to funding and accreditation. A third issue, he continued, is that teachers cannot reasonably be expected to design, organize, and incorporate physical activity into lessons without adequate training. In that scenario, he asserted, school principals are unlikely to hold teachers accountable for delivering physical activity as intended, and teachers are unlikely to react favorably to the directive to add another daily task. He acknowledged that these issues may generate doubt as to whether schools are good settings for promoting physical activity and that certain variations in educational settings (e.g., open classrooms, frequent moving between classrooms) may not be conducive to pursuing this goal. In such cases, he suggested, alternative settings such as boys and girls clubs and local parks and recreation facilities may be more suitable, as they already maintain a culture of physical activity, follow structured schedules, and are subject to minimal if any academic governing entities that would shift the focus away from physical activity.
INNOVATION IN HEALTH CARE: USING DATA TO GUIDE PERSONALIZED, EVIDENCE-BASED CARE FOR OBESITY USING A CLINICAL DECISION SUPPORT SYSTEM
Patrick J. O’Connor, senior clinical investigator and codirector of the Center for Chronic Care Innovation at HealthPartners Institute, discussed the use of a data-driven approach to guide personalized and evidence-based care for people with obesity. He drew on his experience with a clinical decision support tool that was developed for patients with chronic health conditions.
O’Connor began by affirming that primary care settings recognize the importance of intervening on systems-level factors that contribute to obesity, but that they also pursue approaches to address obesity at the patient level. He agreed that prevention of obesity is critical, but pointed out that solutions are also needed for people who already have obesity. Care and clinical management of people with overweight and obesity are far from optimal (Fang et al., 2021), he observed, not only for obesity but also for other chronic conditions, such as type 2 diabetes, hypertension, and high cholesterol.
To provide context for his discussion of the clinical decision support tool, O’Connor reviewed various strategies used to treat overweight and obesity and their average effects on weight loss. Lifestyle changes and interventions are clearly indicated and important for any patient with a weight-related condition, he began, but he pointed out that the impact of these approaches on weight varies among individuals and in the best cases is about a 5 percent weight loss over 1 year, on average. Several FDA-approved medications have been approved for longer-term treatment among people with BMIs greater than 27 kg/m2, he continued, and can help patients achieve a 5–15 percent weight loss at 1 year. The proportion of eligible people who use these medications is quite small, he acknowledged, despite their potential to effect substantial decreases in weight when combined with lifestyle changes. A third level of treatment is metabolic bariatric surgery, which O’Connor reported is effective—leading to average weight losses of about 30 percent of weight at 1 year postsurgery among patients with BMIs greater than 35 kg/m2—but used infrequently. He added that many patients regain substantial amounts of weight in the years after bariatric surgery.
O’Connor then elaborated on the impact of bariatric surgery on coronary artery disease and mortality. Among people with BMIs greater than 35 kg/m2 and type 2 diabetes, he reported, the surgery is associated with coronary artery disease event rates (assessed 7 years postsurgery) that are substantially lower (2.3 percent) than the rates for matched patients not having undergone the surgery (4.2 percent) (Fisher et al., 2018). In another study of outcomes in surgical patients and matched controls, the percentage of people who died 12 years after surgery was around 20 percent for surgical patients and 30 percent for controls (Arterburn et al., 2015).
Despite these positive outcomes for surgical patients on average, O’Connor emphasized the tremendous magnitude of individual variation in benefits from bariatric surgery. To illustrate this point, he referred to an analysis that predicted a gain of 6 quality-adjusted life years in a 40-year-old female with a BMI of 40 kg/m2 and newly diagnosed type 2 diabetes that did not require exogenous insulin, but a potential small loss of quality-adjusted life years in a 68-year-old male with long-standing type 2 diabetes and poor glucose control, other comorbidities, and a BMI of 55 kg/m2 (Arterburn et al., 2015).
O’Connor explained that research using data from hundreds of thousands of patients has identified factors that can help predict the benefits in individual patients. These factors include age, sex, baseline BMI, presurgical hemoglobin A1c level, presence of comorbidities, length of time since diabetes diagnosis, and use of insulin to control diabetes.
O’Connor transitioned to describing the data-driven, primary care–based clinical decision support tool his team developed to improve the quality of chronic disease care in adult patients with type 2 diabetes and BMIs greater than 25 kg/m2. The web-based tool is linked to the patient’s electronic medical record (EMR) and uses algorithms (based on evolving clinical guidelines and FDA actions) to process the patient’s EMR data and any self-reported lifestyle data in real time to identify appropriate weight-loss options for that individual. O’Connor added that the next step for this tool is to promote a shared decision-making process by communicating the benefits and risks of each appropriate weight loss option to both the patient and the primary care clinician.
This nonproprietary clinical decision support tool has been in use for about 10 years, O’Connor recounted, and now serves around 3 million patients in 12 medical groups across 10 states. It has been used primarily for management of blood pressure, cholesterol, glucose, and smoking, he relayed, and will next be evaluated in relation to obesity decision support. Specifically, he elaborated, his team will evaluate the effect of the tool on weight trajectories, FDA-approved weight-loss medication starts and metabolic bariatric surgical referrals, and patient-reported shared decision making (i.e., conversations about weight) and intent to lose weight.
O’Connor then listed the questions typically asked about a given treatment by patients being treated for obesity, which revolve around anticipated amount of weight loss and how long the loss will persist, whether it will eradicate their diabetes and for how long, and whether they will be able to stop taking any of their medications. They also ask about the risks of medications or surgery; insurance coverage; and how the treatment will impact their risks for longer-term outcomes, such as quality of life, longevity, heart attack, and stroke.
According to O’Connor, a concise format is the best way to present clinicians with information useful for shared decision making. He observed that primary care clinicians are generally rushed and unaware of the potential (let alone patient-specific) benefits and risks of bariatric surgery and medications to treat obesity. To illustrate this observation, he pointed out that current diabetes guidelines suggest that all patients with diabetes and certain levels of obesity consider bariatric surgery, but nothing in the guidelines suggests the characteristics of patients who are likely to have substantial versus small benefits from the surgery. Accurate estimation of benefits for a given patient will be difficult without a decision support tool, O’Connor maintained, perhaps with the exception of bariatricians
(clinicians specializing in bariatric surgery), who work in weight-loss centers and not typically in primary care.
O’Connor stressed the importance of tailoring the information communicated to patients according to their numeracy and health literacy levels and considering their personal and cultural perspectives on eating, weight, and treatments. Many patients strongly resist suggestions for surgery or additional medications, he noted, and weight is often an emotionally loaded issue. He then showed screenshots of the clinical decision support tool’s graphic interface for patients, pointing out the use of colored symbols to prioritize the health parameters (e.g., weight, blood pressure, cholesterol, or glucose) and/or behaviors (e.g., tobacco use) for which improvements are expected to yield the greatest benefit for an individual patient. The relative benefits of weight loss and improvements in other clinical domains depend on what type of weight-loss treatment is under consideration, O’Connor explained, and he emphasized that optimal decision support tools address and compare benefits across multiple clinical domains. The tool has an additional tab for clinicians, he added, with more information about the estimated risks and benefits of specific options.
O’Connor provided several summary points with regard to the use of clinical decision support tools to manage overweight and obesity, which he predicted could promote uptake of effective weight management strategies. Directing tools to both patients and providers is key, he began, so that both parties are informed about the estimated potential individual-level benefits and risks of various weight management strategies and can engage in shared decision making. For adults with type 2 diabetes and obesity, O’Connor proposed that framing weight management options as treatments for diabetes (versus obesity) could motivate more serious consideration of such options by some clinicians and patients. He suggested that the impact of a decision support tool on quality of care could be maximized if combined with other strategies, such as the use of gamification, incentives, or active outreach to patients with registry-based case management.
Lastly, O’Connor highlighted several challenges to the clinical management of obesity. Many clinicians and patients underestimate the effectiveness and safety of FDA-approved medications and bariatric surgery for weight management, he observed, especially for patients with type 2 diabetes. If they had a more accurate view of the potential positive outcomes of these options, he suggested, many more people with obesity could benefit. At the same time, he acknowledged that communicating evidence-informed personalized estimates of the benefits and risks of weight management options is challenging because of the need to tailor such communications to the many factors that influence an individual patient’s capacity to understand information. Finally, O’Connor urged attention to ensuring that informatics-driven quality improvement strategies, including clinical decision support tools, improve health equity.
PANEL AND AUDIENCE DISCUSSION
Following the two speaker presentations, Crespo moderated a discussion with Donnelly, O’Connor, and workshop participants. The speakers responded to questions about opportunities to increase classroom-based physical activity, changes in participation in physical activity as schoolchildren age, intensive behavioral therapy for children who have obesity, prioritizing clinical actions in the presence of multiple comorbidities, and the availability of clinical decision support systems.
Opportunities to Increase Classroom-Based Physical Activity
Donnelly stated that a school’s culture is the most important factor in increasing classroom-based physical activity. A low-cost, efficient way to change the culture, he reiterated, is to include integration of physical activity into classroom lessons in the curriculum of teacher preparation programs across colleges and universities so that prospective teachers will expect to carry out this strategy and feel equipped to do so. Donnelly clarified that the goal of classroom-based physical activity is to expend energy, which can be accomplished through simple movements and instructions; complex motor tasks or sports-related movements are unnecessary. He acknowledged that teachers would likely need at least minimal guidance on how to adapt activities for students with special needs.
Changes in Participation in Physical Activity as Schoolchildren Age
Donnelly explained that as grade level increases, changes occur in the types of physical activity that are promoted and selected. He pointed out as an example that the typical approach with elementary school students tends to break down as children enter middle school, as they become self-conscious about doing silly movements that could cause them to sweat or could be difficult to perform in clothing or shoes that may have been selected with fashion rather than comfort in mind. Recreational and sports activities are typically promoted among older children, but Donnelly observed that data indicate a decrease in physical activity as children age despite the shift in activities offered.
Intensive Behavioral Therapy for Children Who Have Obesity
A participant asked how to advance the U.S. Preventive Services Task Force recommendation that children (older than 6 years of age) with obesity be offered or referred to intensive counseling and behavioral interventions to promote improvements in weight. O’Connor referenced efforts to engage
these children and their families in intensive behavioral therapy programs, noting a low completion rate despite free and relatively easy access to a carefully designed, sophisticated intervention. He noted that it is unclear whether the key barrier is families’ lack of time or parent–child conflicts, but he suggested that improving engagement in the intervention is an important step.
Prioritizing Clinical Actions in the Presence of Multiple Comorbidities
O’Connor addressed the question of how to prioritize clinical actions for patients who have obesity along with multiple comorbidities. He explained that algorithms can be designed so that a patient’s current weight, blood pressure, and cholesterol values can be input into the American College of Cardiology/American Heart Association’s risk calculator1 or other cardiovascular risk calculator (along with a few other key factors) to estimate a patient’s 10-year risk of heart disease or stroke as a reference point. The calculation can then be rerun to forecast the potential impact on that risk of decreasing certain values, and that information can be used to prioritize treatment options and discuss them with patients.
Availability of Clinical Decision Support Systems
O’Connor stated that the tool he had discussed is web based and scalable and is available to health care systems at a cost that covers such expenses as installing the tool at the recipient site. In terms of potential for the tool to interface with a learning health care system,2 he explained that if the tool communicates with the EMR and provides decision support at the point of care, the data can be deidentified and archived in a way that provides a roadmap for identifying care improvement opportunities. As an example, he said that data could be used to identify how many of a physician’s patients with diagnosed hypertension had been treated with blood pressure–lowering medications, how many had their blood pressure under control, and how many whose blood pressure was uncontrolled received a second drug. These kinds of data could also be compared among providers, he added, so as to target learning interventions to those with the most room for improvement. In addition, the data could be used to examine outcomes by specific patient characteristics to identify subgroups of patients who experience disparate outcomes.
___________________
1https://tools.acc.org/ASCVD-Risk-Estimator-Plus/#!/calculate/estimate (accessed January 12, 2021).
2https://nam.edu/programs/value-science-driven-health-care/learning-health-system-series (accessed April 21, 2022).
