Skip to main content.

When investigating suspicious deaths, brace for some serious data work

Craft: Lessons From The Field

When investigating suspicious deaths, brace for some serious data work

Blog body

"What came back was hundreds of pages of heavily redacted records with no names attached," Corwin writes.
"What came back was hundreds of pages of heavily redacted records with no names attached," Corwin writes.

Tom Corwin and his colleague Sandy Hodson reported an investigative series for The Augusta Chronicle that looked at the deaths of 500 developmentally disabled patients who died in community care placements in 2013 while the under the auspices of the state of Georgia’s Department of Behavioral Health and Developmental Disabilities. The series detailed the extent to which Georgia is struggling to ensure such patients are properly cared for and kept safe. Here, Corwin shares how he obtained and analyzed the data behind the stories, and shares key lessons for reporters similarly investigating stories of suspicious deaths in their coverage area. — Ed.

This was probably the largest project I have attempted in terms of the number of deaths involved and the size of the open records requests we made and the records we received. That brings me to the first lesson that I learned.

1) Be prepared to reassess what you are doing and to be more realistic about what you are taking on. I was initially shocked when I learned of 82 unexpected deaths in 2013 among patients receiving care in the community under contract from the Georgia Department of Behavioral Health and Developmental Disabilities. It was a number buried deep inside a routine annual report the department puts out. I was already looking into the death of one of those patients, 12-year-old Christen Gordon, so we made an open-records request to get the state’s records on the other deaths. What came back was hundreds of pages of heavily redacted records with no names attached, making it difficult to tell what had actually happened with these patients. And it came with a bill for $440, which luckily I was able to avoid because of a technicality. It also turned out that it only included about 25 of the deaths, and it covered both behavioral health and the developmentally disabled. With only a six-month window before my fellowship deadline, and with the wise counsel of my advisor, I decided to focus on only the developmentally disabled, which still turned out to be hundreds of pages to sift through, again heavily redacted, and about 28 patient deaths to identify and investigate.

2) Get a database of your state’s deaths, and be prepared to work in cleaning it up. This turned out to be the key to identifying almost all of the patients we wanted to investigate. We made a request for the death records to the Georgia Department of Public Health and received back records on more than 76,000 deaths in 2013. But then the real work began, and you should expect to have the same problems if you get a similar database from a state agency. The biggest problem is the information is taken from county coroners and there does not appear to be a standard way of reporting.

For instance, one might list an address as 123 Brown St., and another might list it was 123 Brown Street. Many of the directional names, like South Main, might be listed as S. Main. This makes it almost impossible to do searches until all of the address listings are made uniform. In our case, that took almost a month of pretty dreary data work that consumed both work hours and nights and weekends. But in the end, you have a fantastic database that could lend itself to other projects. One of the projects the newspaper is considering is a look at gun deaths in Georgia.

3) Consider alternative sources to help identify relatives. Once we had almost all of the patients identified by matching up dates of death and addresses with our database of deaths, we began looking for relatives to talk to. In a very few cases, we were able to get autopsy records that listed a next of kin. And in some cases we could find an obituary. But we also turned to to find some of them, and we used Accurint to do address and phone number searches. Both Ancestry and Accurint require a subscription but they are worth it. Both can provide ages, for instance, which can help you determine if you are approaching the right person.

4) Understand that people may not want to be a part of your investigation. In our case, that was in fact the most common reaction. Many of the relatives we spoke to did not want to be involved in an investigation and did not want to discuss what had happened. My advisor, Martha Shirk, had done similar death investigations and tried to tell me that might happen. Some people do want to revisit these deaths. Still, it was disappointing and ultimately frustrating after all of the work we did to find these relatives. It is understandable that some might just want to be left alone and that they would prefer to think that their relative got good care. Some were just flat out indifferent. The sister of one patient, who had choked to death because he was left eating unsupervised while the staff member was in another room watching television, told me, “Yeah, he tried to do that to me, too.” And then she told me she didn’t want her brother’s death investigated.

5) Keep pushing the authorities to explain the numbers. One of the things we found out late in the investigation is that it wasn’t just 82 unexplained deaths in the community in 2013; it was 82 out of 500 community-based care deaths in 2013. And there were 141 unexplained deaths in 2014, out of 498 community deaths. These are numbers that were not included in any state reports we could find and were frankly unknown outside the department. But we only got them because we asked for them specifically, as we were trying to put the 82 deaths in context. And in the course of asking for that context, we also got the rules for how the state classifies deaths, which will vary from state to state, and discovered that the state was not asking for the autopsies that its own policy says should be done.

6) Realize that what you have done might feel unfinished. In my case, because I began with deaths in 2013, I knew there might be more deaths to investigate. I am in the process now of deciding how we might build on what has been published so far. I am evaluating how we might get the records for 2014, and whether there are any larger patterns going back through more years with some of these deaths. Even though it was a relief to finally publish the project, I have a sense that there is much more to say and explore on this subject.

I hope this is only the first chapter.

[Image by Wikileaks Mobile Information Collection Unit via Flickr.]


The nation's top infectious disease specialist will join us for a conversation with national health reporter Dan Diamond of The Washington Post. We’ll talk about the evolving threat posed by monkeypox, the current state of the COVID pandemic, and broader lessons on how we respond to emerging diseases. Sign-up here!

The USC Center for Health Journalism at the Annenberg School for Communication and Journalism is seeking two Engagement Editors and a social media consultant to join its team. Learn more about the positions and apply.


Follow Us



CHJ Icon