Movember Data Dive: Physical Activity Behavior in the US
Howdy! As you all know from my blog last week, I have decided to put my mustache to good use this November and raise money for the Movember Foundation. The Movember Foundation is a global charity that aims to increase awareness and funds for men’s health issues, with a focus on 4 key areas: prostate cancer, testicular cancer, poor mental health, and physical inactivity.
Physical inactivity, which is the 3rd leading risk factor for global mortality, is the newest focus area for the Movember Foundation. I have decided to jump in on this initiative by doing what I do best: analyzing data with my partners-in-crime, Elasticsearch and Kibana. And in order to help raise awareness around physical inactivity, I decided to look into the 2013 Behavioral Risk Factor Surveillance System (BRFSS) data from the Centers for Disease Control & Prevention (CDC).
Every year, the CDC conducts approximately 500,000 telephone surveys to collect data on a variety of personal health-related topics, such as nutrition, drinking habits, physical activity and health history. With this data set, I investigated macro patterns of physical activity and exercise in the United States. Next week, we will dive into relations between personal behavior and health risks such as diabetes, hypertension, and cardiovascular conditions.
The big picture!
The 2013 survey included ~491,000 respondents, with 40% men and 60% women. Given the Movember focus on men’s health, we will only be looking at the male respondent data in the rest of the analysis here.
It is heartening to see that almost 75% of the male respondents had at least some exercise in the last 30 days. On the flip side, only about 52% met their recommended aerobic guidelines and an even lower 30% met their recommended muscle strength guidelines. So, there’s definitely some room for improvement there.
Conducting large scale surveys is tricky because you want to make sure that survey design and execution do not pollute or bias the results in any way. For example, in this survey you see a slight impact of interview month on the results. The graph below shows that people interviewed in summer months showed slightly better aerobic fitness levels. People tend to be more active during summer, and what we are observing here is likely the effect of recency bias in the survey.
Walk on America!
Walking is by far the most popular physical activity, with close to 60% of the male respondents listing it as their primary physical activity. Running, weightlifting, and gardening figure in the 2nd, 3rd and 4th spot. Given my woody roots, I was psyched to see Gardening and Yard Work in the top 10.
Interestingly, if you look at how much time people spend on their primary physical activity per week, the picture changes a bit as expected. Popular fitness activities like walking, running and weightlifting figure pretty low on the duration metric. Outdoors activities like farming, fishing, skiing, and hunting rise to the top of the chart.
Yoga and Pilates are great, but ...
Relating the exercise type to the recommended goals, you can see that activities like yoga, weightlifting, and pilates are great for muscle strengthening, but don’t help much with meeting recommended aerobic guidelines. Activities like boxing, rock climbing, wrestling, calisthenics, and aerobics provide a more balanced cardio + strength workout.
Buck up, Texas!
The BRFSS survey measures personal behavior and health risk patterns at the state level, and this information can be then used by local authorities to design and deploy targeted health initiatives and disease prevention programs.
Let’s continue exploring the big picture from a geospatial angle. Montana residents takes the top spot as the most active state, with an average 650 minutes of total physical activity per week. Texas is at the bottom of the rung with total weekly activity of 368 minutes per respondent.
Exercise and Demographics
Moving on to demographic factors, it’s interesting, but not surprising, to see how the type of exercise varies with age. You can see that as people age, they gravitate towards activities with lower MET values, i.e lower energy cost activities like gardening and walking. However, it is surprising that on an estimated workout intensity scale, a greater fraction of people in the older age brackets are getting a vigorous workout compared to younger respondents. This seemed counter-intuitive and made me wonder if the formula used to estimate the intensity of a workout was somehow inversely dependent on age.
Exploring along other demographic dimensions like education and income level, you can see there is a strong correlation between fitness and both these demographic factors. Educated people and people from higher income brackets show a higher likelihood of meeting their recommended fitness guidelines. However, income and education are typically highly correlated, and it is quite likely that we are not looking at independent effects here.
Stay tuned for more!
That’s it in this edition of Movember Data Dive. Next week, we will investigate eating and drinking patterns, as well as the relations between health risks and personal behavior. We will also be publishing the example code to the examples repo soon. Be sure to check back at the Elastic blog for more.