Opinion: Quality data on physical activity is missing in action from public health surveillance
If “That which is measured improves. That which is measured and reported improves exponentially” is true, then public health might suffer from a data void, a void that confounds our efforts to remedy the entrenched physical inactivity status quo.
To put the above in context, many in public health already consider physical inactivity the neglected ‘Cinderella’ risk factor for chronic disease prevention. In a society more focused on treatment than prevention, there’s an acknowledgement that if you could convert physical activity to pill-form, it would be a miracle drug.
Physical inactivity is a gargantuan problem. According to the National Health Interview Survey, despite the health benefits, fewer than half of U.S. adults met the minimal guidelines for aerobic activity and almost one-third of adults were physically inactive in 2015 (table at end). To fully meet the 2008 Physical Activity Guidelines, adults must engage in at least the equivalent of 150 minutes/week of moderate-intensity aerobic activity and in muscle-strengthening activities on 2 or more days/week.
As a public health researcher looking for new solutions to our tragic, persistent, and expensive chronic disease burden, I thought it might be helpful to conduct a cost effectiveness analysis (CEA) for investment in urban bike shares. Benefits from bike share programs are multifaceted and fun, and I am interested in modeling physical activity changes as the result of transportation mode shifts to bicycling from other modes (walking, bus, subway, private car, etc). Specifically, I want to model purported — and forecast future — population health benefits from an increased number of people achieving the above Guidelines as the result of NYC’s bike share, Citi Bike.
To model changes in physical activity, I started looking for cross-sectional/baseline and continuous surveillance system data for NYC’s 2,167 census tracts. Census tracts are a pretty standard unit of measurement, and I thought census tracts were particularly important in light of NYC’s great diversity and inequity, a city known for high residential segregation.
The benefit of small area epidemiologic data to develop and implement effective and targeted prevention activities, identify emerging health problems, and establish and monitor key health objectives is well established. For example, RWJF supports a Healthy Communities initiative with a focus on how ‘local data, financing and policy and practice changes, coupled with multi-sector community leadership and capacity building, can change community environments in ways that make it easier for people to be healthier.’
The marginal benefits of exercise may in fact be greatest for the most sedentary individuals; thus, public health value may be found with an increased understanding of the physical activity distribution and local data driven, site-specific models.
My data search was also informed by the paper, Physical Activity Associated with Public Transport Use — A Review and Modelling of Potential Benefits. These Australian researchers’ variable of interest — self-reported minutes of physical activity per week— allowed them to model population health benefits from 8, 16, and 24 minutes of added physical activity and also, percent uptake of physical activity by those insufficiently active. Their model’s findings below:
Conservatively, if only 20% of inactive people in NSW, Australia, walked for 16 minutes more each week, across the state there would be 6.97% more adults meeting public health recommendations for physical activity, which has significant public health implications. As very few public health interventions increase population physical activity by anything like this amount, this would represent a significant improvement.
Fabulous, right? Especially the part about the significant improvement?
Inspired by the above, I sought data that measured physical activity or inactivity consistent with my bike share cost effectiveness analysis objectives — to identify an optimal exercise dose for reducing health risk in the general public, customize an exercise prescription for a neighborhood, measure compliance with physical activity guidelines and establish dose–response relationships with my study outcomes.
Nationally, NHANES provides data for adults (leisure-time, transportation, and household activities) and children (leisure-time activities) and in 2003, accelerometry data were obtained in addition to self-report. NHANES surveys about 5,000 individuals each year, whereas the Behavioral Risk Factor Surveillance System (BRFSS) completes more than 400,000 adult interviews each year.
Since I was seeking small area epidemiologic data, I turned to physical activity, as self-reported, available from the NYC Department of Health and Mental Hygiene’s annual telephone, Community Health Survey and the Physical Activity and Transit Survey in 2010/2011, but this data set did not cover census tracts, only boroughs and neighborhoods.
More helpful was the CDC’s 500 Cities Project: Local Data for Better Health, Chronic Data Portal, launched in 2015 and funded by RWFJ. So far, the data is available at the census tract level for just one year, 2014, but 2015 data is expected later this year. Of their 27 measures selected from the BRFSS, only one question pertains to physical activity and the 500 Cities Project has no plans to add additional measures without additional resources and/or public interest (their one question happens to be the same question employed by NYC DOHMH’s survey). The question asks about ‘leisure time physical activity’ in an arguably over-simplified ‘yes’ or ‘no’ format:
“During the past month, other than your regular job, did you participate in any physical activities or exercises such as running, calisthenics, golf, gardening, or walking for exercise?”
The resulting data is the prevalence of a ‘no’ answer to this question and is shown in the red map for NYC’s census tracts. The data appears reasonable; the darkest red areas are lower income census tracts associated with reduced opportunities for leisure time physical activity (this might perhaps be especially true for some of the leisure time examples included in the question, like gardening and golf).
This is the best data I’ve found to-date and I am somewhat frustrated because the limitations are many:
- First, self-reported physical activity is known to be highly problematic, particularly because people tend to compare themselves with peers when replying to physical activity questions. Research has shown differential reliability across subpopulations, both over and under reporting, and, understandably, confusion regarding levels of physical activity intensity
- Second, this data does not provide insight into compliance with public health’s exercise recommendations, 150 minutes / week
- Third, since the question only focuses on leisure time physical activity, this data offers no insights into total physical activity or other domains, for example active transit or work-related manual labor, housework
- Fourth, there is no trend data available; public health researchers should know if a neighborhood is engaging in more-and-more or less-and-less physical activity year-over-year
Seems to me there are new data collection opportunities though, perhaps more so today than ever before! Thanks to digital health innovations, some data collection methods could even help us leave behind the misadventures of the telephone. Already, private companies like Fitbit have massive amounts of data, but Fitbit owners are a self-selected sample. Early adopters of such tools are typically the healthy, wealthy, and worried; such a sample would not meet public health’s mission and population heath surveillance responsibility.
Also, helpfully and excitingly, the prices for digital health tools / wearable devices capable of measuring heartbeats are falling (heart rate monitoring in conjunction with an accelerometer improves accuracy). Such tools could easily improve physical activity measurement, and by doing so, increase our understanding of physical activity across geographic-small areas, and in conjunction with other data, a variety of built environments, and socioeconomic and health statuses, etc. Using wearable devices in research may also advance cardiovascular epidemiology’s understanding of benefits from all sorts of lifestyle activities. Accordingly, I am spreading the word about the CDC’s Healthy Behavior Data Challenge!
Before writing this post, I reached out to some folks responsible for these public health surveys and other public health surveillance data. They noted surveys such as the BRFSS require large samples and are extremely expensive. To be sure, there are ongoing efforts to improve the BRFSS.
Countering the cost and difficulties is the fact that physical activity is essential to health. In an epidemiology of health, physical activity is a basic building block with which to remedy our entrenched chronic disease burden, alongside smoking and obesity Fried, L. (2016)[i] . I would propose, although not discussed herein, that we have much better data on smoking and obesity prevalence and trends than physical activity.
Could public health not create better, more tailored, interventions — and advocate more effectively for their implementation — if practitioners had higher quality, robust, data on physical activity?
As with any public health challenge, practitioners need to understand the prevalence and distribution of physical activity as well as the extant barriers to — and opportunities for — physical activity in a given city, neighborhood, or census tract. How can public health write a ‘prescription for physical activity’ if the basic surveillance data does not exist.
I still like my idea to evaluate bike shares as a “prescription for physical activity.” Guess before I do the bike share CEA though, I may need to model a CEA for a proposed investment in higher quality, robust, physical activity data.
This might begin with simple big picture questions: What’s the cost of the BRFSS survey annually? And what’s the cost of physical inactivity? This CDC website estimates 11.1% of total health care expenditures (about $117 billion per year) were associated with inadequate levels of physical activity. Then we must also layer on the societal costs of early mortality and morbidity.
Any recommendations? Feedback? Disagreements? Thoughts on the importance of surveillance data to drive national and local action?
Thanks for reading my musings.
[i] Fried, Linda P. Dean of the Mailman School of Public Health and DeLamar, Professor of Public Health Practice, Senior Vice President, Columbia University Medical Center, Mailman School of Public Health. The Future of Epidemiology: An epidemiology of health. Department of Epidemiology Seminar Series, March 11, 2016.
Note: while the American Communities Survey includes commute to work data, covering bicycling, please note the margins of error are substantial, thus the reliability concerning.