This week brings more signals that the pediatric obesity epidemic may be in retreat. A new study using data from NHANES shows a much-celebrated 43% reduction in obesity prevalence among 2-5 year old children between 2004-2011. Before we jump to the why and how questions, we should exercise some caution in interpreting this trend estimate. First, the absolute change in prevalence is numerically quite modest, falling from 13.9% in 2003-4 to 8.4% in 2011-12 for an absolute change of 5.5 percentage points over 8 years. Second, the sampling variability surrounding this trend estimate is relatively large compared to the trend estimate itself, introducing considerable statistical uncertainty. The sample size of children in this age cohort in NHANES is not very large, and obesity prevalence is not very high in this cohort to begin with (a total of 91 kids in this subgroup were classified as obese in 2011-12), so we have some serious sampling error to contend with. Add in the distortions caused by the authors’ use of linear trend estimates, along with the problem of multiple comparisons that were not adjusted for in this analysis, and suddenly a p-value of 0.03 doesn’t seem all that definitive anymore. Of the 8 age cohorts examined in the study, these pre-K kids were the only group to exhibit a statistically significant reduction in obesity (but women over 60 years exhibited a significant increase in obesity).
Even if the headliner 43% reduction seems less impressive after taking a closer look at the numbers, the accumulating body of research suggests that finally we may have reached the downward-sloping face of the obesity epi curve, at least for young children. The next questions we need to ask are why and how. Of the myriad programs and policies used for obesity prevention and control across the U.S., which combination of approaches is working for whom, and under what circumstances? And readers of this blog surely want to know whether (and which) components of the public health delivery system – that constellation of governmental agencies and private actors that jointly implement public health programs and policies – are helping to generate these effects. By distinguishing the active ingredients from the instructive failures, we can rapidly and efficiently scale up what works in obesity prevention and limit future losses in quality and quantity of life, not to mention significant economic costs.
Fortunately, we have an accumulating body of research about the public health system’s roles in obesity prevention thanks to the growing field of public health services & systems research (PHSSR). For example, work from Frank Chaloupka’s group at UIC showed us that a decade ago much of the governmental public health infrastructure was not positioned to address the rapidly advancing obesity epidemic (recall my earlier posts on Chaloupka’s obesity and tobacco research). This group found that in 2003 less than half of the nation’s local health departments provided, supported, or advocated for obesity prevention programs targeted at youth. The CDC’s Xinzhi Zhang and colleagues used NACCHO Profile data to show that by 2005 this number had risen to 56%, although considerable geographic variation persisted in the availability of these programs. Better financed agencies were more likely to undertake obesity prevention activities, particularly those agencies that receive larger shares of their revenue from state government sources, which tend to be more flexible. Katie Stamatakis and colleagues at Washington University-St. Louis took a closer look at geographic variation in the NACCHO data, finding that the availability of obesity prevention activities in 2005 was not associated with the prevalence of obesity in the community. This finding suggests that any efforts to target obesity prevention funding and program activities to high-need areas are not working very well. Houbin Lou and his colleagues from CDC used NACCHO data from both 2005 and 2008 to better characterize the nature of obesity prevention activities supported by local health departments, finding that by 2008 nearly 70% of the nation’s local health departments were engaged in the most prevalent activity of disseminating information about obesity risks and prevention strategies to the public, health professionals, and policy decision-makers. The proportion of agencies that had no obesity prevention activities fell from 28% to 17% over the three year period.
What do we know about the impact of public health’s obesity activities? Here the evidence is still quite thin but encouraging. A paper published last year in the respected journal Health Services Research combined NACCHO data with county-level estimates of obesity prevalence from BRFSS to examine the connection between activities and outcomes. Using a careful econometric estimation approach, Adam Chen and colleagues from CDC found that local health departments that implemented obesity prevention programs in 2005 experienced significant reductions in obesity prevalence between 2004 and 2005, compared to agencies that did not implement the programs. Consistent with expectations, the reductions in obesity were larger among low-income population subgroups than for the populations as a whole, suggesting that health department-delivered programs may be particularly effective in reaching low-income populations and reducing socioeconomic disparities in obesity risks. This study also used one of my favorite tricks – a nonequivalent dependent variable approach – to show that the estimated effects could not be attributed to general temporal trends or simple forms of selection bias.
These population-level results are backed up by person-level results from a pragmatic randomized trial published last year in the journal Obesity. In this study by investigators at UNC’s CDC-funded Prevention Research Center, six North Carolina local health departments were trained to deliver a behavioral weight loss intervention tailored for low-income, mid-life women. The 16-week program produced significant weight loss among the women when compared to a delayed-intervention control group that received only general health education materials. All this at a cost of less than $350 per participant.
Intervention studies of specific obesity prevention policies and programs like the North Carolina trial are becoming more numerous, making it increasingly important to examine how multiple strategies interface and interact at the community and system level to impact health. Research on a wide variety of specific interventions continues to accumulate, including strategies for: promoting and supporting breastfeeding; reducing screen time; improving the food and physical activity environments of schools, child care settings and residential neighborhoods; expanding BMI screening and healthy lifestyle education in schools, child care, and health care settings; and changing menu labeling, food product pricing, and marketing practices in restaurants and retail outlets, worksites, schools, and community settings. While evidence supporting some of these strategies is already quite strong (like breastfeeding support), evidence for other strategies is preliminary, mixed or otherwise inconclusive. A missing element in many of these single-intervention studies is the ability to examine the characteristics of the surrounding public health delivery systems that may support or impede the obesity strategy’s reach, intensity, quality, efficiency, and its synergy with other prevention activities.
While the above studies are far from a comprehensive review of the evidence, it is clear that there is much more we need to learn about the public health strategies that work best to prevent and control obesity, and about the public health delivery system characteristics that best enable and support these strategies. For example, much of the existing work in PHSSR has focused heavily on governmental public health contributions to obesity prevention – an important but incomplete view of the strategies underway within the larger delivery system to tackle obesity. A large and diverse body of obesity prevention work is occurring through multi-sectoral initiatives and public-private ventures, including schools, worksites, churches, transportation programs, land use and zoning processes, and many other contributors. There is much to be learned from the careful study of these types of complex and multi-faceted strategies. Some of this research is already underway through the PHSSR Program supported by the Robert Wood Johnson Foundation, including results from the recently completed Multi-Network Practice and Outcome Variation (MPROVE) study that has been implemented in six diverse states through practice-based research networks (PBRNs). Stay tuned to this blog for updates on emerging findings related to obesity prevention and public health strategies.