US: South Dakota State University (SDSU) researchers are using the tools of spatial analysis to explore nationwide data for insights on what influences obesity.
“We can identify and map some of these regions or ‘hotspots’ of high and low obesity,” said associate professor Michael Wimberly of SDSU’s Geographic Information Science Center of Excellence. “Ultimately what we want to do is explain what some of the drivers are.”
The SDSU study set out to map spatial patterns of obesity and risk factors nationwide by using Behavioral Risk Factor Surveillance System (BRFSS) data from telephone surveys compiled annually by the Centers for Disease Control and Prevention. The BRFSS data includes self-reported height and weight, as well as respondents’ answers to questions about their levels of physical activity, and about fruit and vegetable consumption.
“The advantage of using BRFSS compared to a variety of other data sources is that we can get wall-to-wall national coverage. They actually do sampling in every county across the United States,” Wimberly said. “So we can map things, first of all, and we can also use various spatial statistics to test hypotheses about what the environmental correlates of obesity, physical activity, fruit and vegetable consumption are at a national level as opposed to other studies that have been more localised.”
SDSU’s preliminary analysis of data from 48 states showed that the probability of obesity increased with distance from supermarkets, while consumption of five or more servings of fruits and vegetables per day decreased. The research also showed clear differences between large metropolitan areas and sparsely populated rural areas.
“The geographic perspective opens up a unique window. Looking at maps, people relate very intuitively to the patterns and it really catalyses a lot of new thought, ideas and hypotheses. That is the power of what we refer to as ‘exploratory spatial data analysis,’ working with the data using statistical techniques that allow us to tease out real spatial trends from the underlying noise and using that as a method for hypothesis generation. We can also pull multiple sources of data together to actually test hypotheses about the underlying relationships.”