Blog

In this blog post, Gates Open Research chats to Dr. Abraham D. Flaxman (University of Washington) about his work in the global health and health metrics sciences.

Dr. Flaxman’s latest research uses data captured via Facebook to explore just how common COVID-19 is in healthcare workers versus non-healthcare workers. Keep reading to learn how researchers can use existing datasets to help answer important questions about public health and uncover the results of this new study.

Meet the author

Thanks so much for this invitation.  I am Abraham D. Flaxman, a mathematician at the Institute for Health Metrics and Evaluation at the University of Washington. I work as an Associate Professor of Global Health and Health Metrics Sciences.  My research focuses on data and methods for global health metrics. Over the last decade, I’ve worked closely with health workers to fill the gaps in our evidence base on the incidence, prevalence, and determinants of many different diseases, injuries, and risk factors. I’m always looking for new and better ways to use existing datasets to help answer my colleagues’ vexing questions in global public health.

What inspired you to investigate the incidence rates of COVID-19 in healthcare workers compared to the public?

It is a very personal story for me and one that takes me back to the first wave of the COVID pandemic in 2020.  In Washington State, we were hit hard and early by a COVID outbreak in long-term care facilities. And remember, this is before we knew if or when a vaccine would be available.  So in Seattle, we were taking social distancing and other non-pharmaceutical interventions very seriously.  Around this same time, one of my children had a medical mystery, and so my family was making multiple doctor visits to get it sorted out (everything is stable now!).  The public schools were closed, and our jobs had transitioned to remote work, so I wondered if the interactions we were having with the (excellent) healthcare workforce in Seattle were a substantial portion of our risk contracting COVID. 

I figured other people would have similar concerns. When I discussed the topic and existing evidence with my colleague and co-author Herbert Duber, who also works as an emergency medicine doctor, I discovered that healthcare workers were also keenly interested in this. Furthermore, these workers had some closely related questions:

“Would patients delay care because they were worried about COVID?”

“Were doctors exposing their families to the virus when they went home after a shift?”

Herbie and his colleagues knew from early evidence that healthcare workers had experienced higher incidence rates of COVID than the rest of the population. Still, they hypothesized that by the fall of 2020, this would no longer be the case. Healthcare workers knew how to keep safe and understood the importance of being safe. Once they had the resources to act on this, we hypothesized that they might have lower COVID incidence rates than non-healthcare workers.

Tell us more about your study design. What did you do, and why did you choose to use data from a Facebook survey?

In principle, it is straightforward to measure a relative incidence rate. But we wanted to do this immediately, and the straightforward approach from Epidemiology 101 is not fast. It is also not cheap and might take valuable time away from the healthcare workers in the middle of a pandemic! So, the straightforward approach was out.

I’ve always been drawn to the possibility of using existing datasets to answer urgent questions in health metrics. This Facebook survey we ended up using began as an effort led by a team at the CMU Delphi Group to collect regularly updated data on COVID symptoms. They partnered with Facebook to survey a large sample of Facebook users every day. In addition to asking about COVID symptoms, they expanded their questionnaire to collect occupational data and data on COVID testing. It was challenging to make sense of this evidence, though!

What was it about the data that made it so challenging?

The challenge was this: symptoms do not mean having COVID, but not having symptoms often means not getting tested for COVID.  So even though we had a lot of data, it would be easy to misinterpret.  The breakthrough that made our estimates possible was the CMU team’s question in their fourth update to their questionnaire: why were you tested for COVID?

We realized that by focusing our analysis on only the people tested because of a workplace requirement, we could develop with something that approximates a representative sample of healthcare workers and non-healthcare workers.

What did you find?

We found that, as hypothesized, by the end of October 2020, healthcare workers had a substantially lower incidence rate of COVID than non-healthcare workers.  Despite working on the frontlines of the epidemic, the healthcare workers were 20% to 32% less likely to be COVID-positive.  This was a reassuring update to the finding from earlier in the pandemic that found that, early on, healthcare workers were at higher risk of COVID. Additionally, these results fit with the theory that the doctors, nurses, and other healthcare staff know what it takes to keep safe.

Why did you choose to publish your research openly?

Considering the urgency of the topic, our primary interest in publication was to get this evidence out quickly to everyone who could benefit from it. However,  we also felt that peer-review was essential to ensure that we were not overlooking something about our interpretation of this data. We achieved all these goals by using an open research publishing platform with open peer review.

The peer review process was quite effective, and I would recommend it to others. Something about the process all being out in the open changes the whole process compared to a traditional anonymous peer review. I have used this approach before, so it was not entirely new for me. I want to thank the reviewers for participating in it, especially if they had not done it before.

Tell us, what’s next for you in this area of research?

I am still very concerned about healthcare workers’ challenges in the pandemic. Our work shows that, in the fall of 2020, healthcare workers had a lower COVID incidence rate than other workers. But this is not their only challenge.  I just read some new evidence about how healthcare workers are experiencing burnout, depression, and anxiety from the challenges of working in a pandemic for so long.  This is all part of the story when it comes to keeping workers safe.

Read the full Research Article today on Gates Open Research to dive deeper into Dr. Flaxman and colleagues’ study and findings. You can stay up to date with Dr. Flaxman’s latest work and insights by following him on Twitter.