Hidden Epidemics

Country/area: United States

Organisation: Columbia Journalism Investigations and Center for Public Integrity with McClatchy’s The State and The News & Observer, California Health Report, Centro de Periodismo Investigativo and InvestigateWest

Organisation size: Big

Publication date: 25 Aug 2020

Credit: See short biography(ies) section

Project description:

For decades, scientists warned that climate change would harm our health, spurring an unprecedented rise in deadly heat, infectious disease and disaster-related trauma. But the U.S. public health system, hampered by underfunding and political resistance, is ill-prepared for this crisis, as documented in our 16-month investigation. Americans are suffering the consequences.


Columbia Journalism Investigations and the Center for Public Integrity anchored the series, collaborating with newsrooms around the country. In addition to the organizations that reported or edited submitted stories, we partnered with City Limits, IowaWatch, The Island Packet, The Lens, The Mendocino Voice and Side Effects Public Media.

Impact reached:

Stories in the “Hidden Epidemics” series ran in or were cited by at least 30 national and local publications, including The Guardian, The Nation, Mother Jones and the Texas Tribune. Covering Climate Now, which carried the series for republication, named one of our pieces a “CAN’T-MISS [STORY] FROM 2020.” The Center for Cooperative Media listed the series’ second installment on the mental toll of climate disasters as one of the top 10 collaborative journalism projects of the year.

CJI and Public Integrity reporters discussed our findings on a half-dozen radio and TV programs, including Public Radio International’s The Takeaway. We hosted a virtual community event on the mental health toll of repeated hurricanes, floods and wildfires that attracted nearly 200 participants and sparked a public debate over important issues such as U.S. public-health preparedness and climate justice. On the academic front, CJI reporters presented our findings at the Baltimore Neighborhood Indicators Alliance’s annual Data Week. 

Local editors and reporters who worked with us saw great interest in these stories from readers; our Vibrio story was the most-read environmental piece at the McClatchy-owned South Carolinian flagship The State over the past two years. But perhaps the most important statement of impact comes from one veteran media watchdog in North Carolina, who wrote, of that Vibrio story, also published in The State’s sister paper The News & Observer, “I thought as the alert from N&O arrived on my screen, ‘This is a fearmongering bit of click bait and here I am clicking on it.’ Reading to the end, my appreciation deepened and deepened as the stories and technical realities unfolded. This is a rare and helpful example of public interest journalism.”

Techniques/technologies used:

Data analyses for the series showed us where to focus geographically (heat deaths, vibriosis cases), which newsrooms to recruit as partners (regions with repeat climate disasters and higher shares of social vulnerability), how to identify climate’s current impacts on people’s lives and more. Data was key. But we aimed to slide it into stories in the manner of ducks swimming: If we did it right, the hard work would be virtually invisible.


Data that powered an indoor heat deaths interactive — one that shows temperature, air-conditioning status and other details about each case in an Arizona county — was processed using Python/Pandas and visualized using d3.js. For all the other heat-related analyses, as well as the environmental data for the Vibrio story, we used RStudio. Illustrator was used to refine figures made with ggplot.  


For the Vibrio story, we used dynamic tables on Google Spreadsheets to see trends in time and geography and identify the most dangerous species, the riskiest seasons, the most common ways of infections and the places with higher incidence in state and county level across the nation.

We maintained and analyzed datasets for our mental health stories using Microsoft Excel, with an occasional assist from Microsoft Access. We used Tabula to collect some datasets from PDFs, Dataminr to webscrape others and OpenRefine to clean the data. That allowed us to show that student homelessness spikes after major disasters, among other findings.

We used JotForm as a platform for a survey of disaster survivors, putting its skip-logic options to work. Other data work included partnering with a Columbia University epidemiologist on a model to estimate the number of Arizona heat deaths that are primarily due to warming temperatures, and tapping buoys and satellite data to show water temperature changes over time.

What was the hardest part of this project?

Obtaining data and documents through public-record requests was a slog. For example, two agencies denied the existence of data tracking progress in the federal Crisis Counseling Program for disaster survivors, even after we learned that counselors must record this information. Only after we found blank forms and formal memos confirming responsibility for maintaining the data did we get it.

We also wrangled with incomplete datasets. Salinity data used in the Vibrio story contained many missing values and dates as well as values suspiciously outside of normal ranges. We had to drop our plans for a broad analysis and use a case study approach: illustrating one weather event and coinciding vibriosis case, relying on one salinity sensor.

To analyze warming in the Chesapeake Bay, we needed long-term water temperature data. Data from buoys was mostly limited to the last 10 years. Scientists pointed us to data derived from a mix of satellites and buoys, but there were accuracy questions. Ultimately, we found one buoy with a longer time record that we could compare with the satellite data. The buoy data had similar inconsistencies as the salinity data, with changing formats between years. But after cleaning it up and interpolating between missing data points, we were able to use the buoy as a check for the satellite data and we determined the satellite data was solid. 

To create the indoor heat death database used in our reporting and in a graphic, we pulled information manually out of 130 autopsy records. It was both tedious and emotionally demanding. Some physical descriptions were gruesome, and most records contained snippets of circumstances surrounding the death; all the people we read about died alone, and many of the records suggested lonely, isolated lives.

What can others learn from this project?

Climate change is an all-of-the-above story. It affects everyone, and just about every journalist can find climate stories on their beat. The datasets we tapped show that reach: student homelessness, a federal social vulnerability index, mental-health services, water salinity, death records, disaster declarations, 911 calls and more. 

So that’s one lesson: Look widely for data when reporting a climate story. Climate change is creating measurable havoc, and showing that human impact with data is powerful and necessary.

Another is the power of collaboration. By joining forces with 10 regional newsrooms in areas hit by disasters, we were able to leverage everyone’s reach for an extensive survey answered by more than 200 disaster survivors and mental-health professionals. In that way, we created data to fill a gap and found sources we never would have known about otherwise.

Designing and managing that survey was a months-long affair. Outreach ranged from postcards to Facebook ads to three interns who spent weeks reaching out to groups around the country. So, word to the wise: Embark on surveys only if you’ve got the time to invest.

One useful aspect of this partnership was that newsrooms could do a little (share the survey) or a lot, depending on their interest and capacity. We shared our disasters reporting and data with partners; some co-published our story, some localized it, some produced their own pieces, some collaborated with our interns on stories. Being flexible rather than prescriptive can work out better for everyone.

And the final lesson: It’s possible to connect local, human stories to big datasets — even during a pandemic when travel is restricted or impossible. We talked to hundreds of people, most of them by phone.

Project links: