Over the last decade, a body of public health research has shown that place — where you were born and where you live — is one of the most important determining factors of how healthy you are. It’s fascinating stuff, and it’s made for some splashy headlines proclaiming that “zip code” is more important than “genetic code.” But while the overlying point is correct, those headlines can reinforce a problematic trend in public health research: Zip codes were designed to help mail carriers be more efficient, not for finding meaningful health trends among communities. There are limitations on how they can be used, and knowing those limitations is key to producing valid research.
As the Flint water crisis was unfolding in 2015, after tens of thousands of residents had been unwittingly exposed to lead in their drinking water after the source was switched to one that wasn’t properly treated, state researchers examined the blood lead level tests of children in the area. Their goal was to determine whether the water switch had caused an increase in lead in the blood. They did so, however, by looking at all the blood lead tests conducted within the many zip codes that cover the city of Flint — zip codes that also include homes outside of city limits. As a result, the samples were diluted. People who didn’t live in the city and didn’t drink the city’s water, and therefore had no potential exposure, were included alongside people who faced the threat.
Dr. Mona Hanna-Attisha, a pediatrician at the local children’s hospital, was sure something was wrong, so she went to Richard Sadler, a geographer with Michigan State University who works primarily on food systems in Flint. He mapped the blood lead tests that Hanna-Attisha had access to, and quickly saw there was a problem: When you included only children who had been exposed to the city’s water, blood lead levels in children had gone up after the water source had changed. It is a problem Sadler believes most geographers would be quick to spot, Sadler said recently, noting that the state agency that did the research didn’t have anyone on staff with expertise in the field. It is a point that’s too often lost on public health researchers, however.
For big regional or national studies, zip codes can be very useful. For starters, most people know their address, meaning you can ask them for their zip code on a survey and expect accurate answers (imagine asking people what Census tract, areas created by the Census to delineate homogenous neighborhoods, they live in).
But when it comes to local research in communities, zip codes are rarely a good way to look for geographic differences, and can cloud whatever relationships a researcher might be looking for. For starters, they are big. Sadler earned his doctorate in Canada, where postal codes are tiny, often representing individual city blocks or small parts of a neighborhood. “Moving back to the U.S., it was one of the things that I knew would irritate me about research,” Sadler said.
Researchers looked at all the blood lead tests conducted within the many zip codes that cover the city of Flint — zip codes that also include homes outside of city limits. As a result, the samples were diluted. People who didn’t live in the city and didn’t drink the city’s water, and therefore had no potential exposure, were included alongside people who faced the threat.
They were also created to deliver mail based on the best routes for mail carriers, and as a result they don’t necessarily represent areas with shared characteristics. In fact, they don’t represent areas at all: a zip code is a group of addresses, not a geographical area. In the case of Flint, the city’s zip codes don’t capture neighborhoods built around the same time (which would likely have similar water pipes) or areas drinking from the same water source.
Zip codes also change over time, with several being added and retired on a near annual basis. In a foundational study conducted in the late 1990s on the problem of using zip codes in research, a group of Harvard researchers showed that more than 10 percent of colon cancer cases in Massachusetts that took place in 1990 were in zip codes that existed in 1999, but not ten years prior. When the researchers excluded the cases that didn’t live in consistent zip codes, it changed the findings. What’s more, comparing the work that used zip codes with data from Census blocks showed contrary relationships between colon cancer and wealth, among other things.
Still, since the late 1980s, zip codes have become frequent features of health research.
In fact, they have been so relied on that the U.S. Census came up with a way to turn them into geographical areas, to overcome the problem that zip codes aren’t a geographical area at all. Called ZCTAs (Zip Code Tabulation Areas), they work like this: Every Census block is examined to determine which zip code applied to the addresses contained within is the most common. These blocks are then aggregated into larger areas, creating ZCTAs.
That solves one the problem of geography, but not necessarily the problem of homogeneity — whether the ZCTA’s represent groups of people with shared characteristics.
One way Sadler has worked around it and still gotten accurate survey results is by asking what streets people live near. Some people will name the closest corner, while others provide the nearest big intersection. The answers aren’t perfect, but they give him a lot more information than he would have by placing everyone at the center of a zip code, he says, especially when evaluating neighborhood characteristics — things like access to grocery stores or parks. When it comes to people’s health, place matters, which is why it's important to make sure that the places we examine are meaningful.
[Photo by Brett Carlsen/Getty Images]