Our eight-month investigation revealed how today’s heat illnesses connect to decades-old discrimination. We focused on heat-related illnesses in Arizona and Florida, two states that have seen sharp spikes in summertime temperatures over the last century. With difficult-to-access federal data on emergency room visits and hospitalizations, we identified ZIP codes with higher rates of heat-related illnesses. In both states, those areas are low-income neighborhoods with a long history of racial segregation. Our reporting found that residents are neglected by local governments failing to robustly prepare for and mitigate the increasing heat.
Our story had practical impacts on the communities we reported on.
Until we reached out to local health officials in Fort Pierce, Florida, they were unaware that heat was a particular issue in their community — even though the state health department tracks heat-related illness rates by ZIP code. After we presented our results, county health officials told us that they would work with the community to address the issue. Health department officials also took the problem to the city manager, whose spokesperson said the city would work to collect more data and create a program focused on heat-related illnesses to better protect the well-being of its citizens. Local grassroots organizations and nonprofits in the area also got back to us, saying they would include the subject in their projects and discussions.
In Arizona, Vitalyst Health Foundation said our story prompted them to propose a new 2022 budget category dedicated to climate change and urban heat.
The piece was distributed to Covering Climate Now and the Florida Climate Reporting Network, and picked up or cited by local, national and even international publications. We also directly partnered with two local newspapers, sharing our analysis of heat-hospitalization records that journalists can rarely access; the Arizona Republic and the Orlando Sentinel both published their own pieces (here and here) in addition to ours.
PreventionWeb, managed by the UN Office for Disaster Risk Reduction, shared and highlighted our story on its website. And we discussed our findings on a radio program at WUSF, the NPR station in Tampa, Florida, and KCSB News in Santa Barbara also covered our story.
Our data findings will also be included in a book entitled Weather and Climate Extremes in a Warming World: Changes, Causes, Consequences and Countermeasures, published by Elsevier and the Royal Meteorological Society.
What’s unique about the main analysis underlying this story is that it uses real patient data. Many of the heat-related health analyses in both the media and scientific literature use vulnerability indices to estimate where people are most at risk for heat illness. For example, such studies would use a combination of temperature and demographic data to identify areas where people are more likely to be at risk. We believe we are the first to do statewide analyses using real patient data to identify the ZIP codes where the rates of heat illness are actually highest.
The data had to be applied for, with details about methods to be used and the steps we would take to protect privacy. Our first attempt was not approved because we didn’t spell out the privacy-related efforts enough, but we adjusted our application and then received approval. We have only seen data from this agency used in one other journalistic story, and it was a different dataset with less privacy restrictions. So we didn’t have much precedent in terms of journalistic usage to guide us through the process.
We used RStudio, QGIS and SPSS for various steps of the project. The next answer details how we used them.
What was the hardest part of this project?
The data was challenging. First: It was mailed on CD-ROMS in a legacy format used by the U.S. government. The only instructions for opening the data were for paid software that we didn’t have access to. Just unzipping the files took four different software attempts. Eventually we used a free trial version of SPSS to open the files and then export as .asc and .loc files that we could use with RStudio to then parse the data and finally convert to more accessible datasets. The trial and error involved in this portion also took a lot of time because the files were several GBs each, so processing time was significant.
During the analysis in QGIS, we hit difficulties such as ZIP codes not matching up between the health dataset and our census data. The health dataset uses ZIP codes as reported by patients, some of which are P.O. boxes not found in census data. So we had to investigate those ZIP codes individually and combine them with a ZIP code that existed in our census data. There was also a privacy restriction that required at least two reporting hospitals for any grouping of data that we were making public. So we had to first determine how many reporting hospitals were represented in each ZIP code. For the Arizona Republic reporting, this privacy restriction was problematic for wanting to report numbers for one high-rate ZIP code in particular that only had one reporting hospital, so we had to find a way to creatively combine that ZIP code with another for reporting purposes, without sacrificing information quality.
What can others learn from this project?
Look for data that isn’t getting the attention it should – and explain how decisions made decades ago continue to impact people today. It’s that melding of past and present, data and individual experiences, that gave the story its power.
Another lesson that’s especially relevant to climate reporting: Remember that the people who are being impacted aren’t necessarily making the connection. Many residents we interviewed didn’t recognize the dangers of the rising temperatures. Both in Florida and Arizona, we encountered many people who grew up in an already hot environment, didn’t realize temperatures were worsening and weren’t thinking about the health impacts of heat. In Florida, it was also very hard to get a response from public authorities and agencies about the problem because they hadn’t identified it as one. It took a lot of persistence and trying alternative ways to gain access to sources willing to comment.