Unvaccinated Covid Patients Push Hospital Systems Past the Brink

Country/area: United States

Organisation: Bloomberg

Organisation size: Big

Publication date: 15/12/2021

Credit: Drew Armstrong, writer; David Ingold, data and graphics; Paul Murray, data and graphics


Drew Armstrong is the senior editor for health care at Bloomberg News, where he has helped lead global coverage of the pandemic over the last two years. David Ingold is the Data Desk Editor at Bloomberg News, where he focuses on stories that combine code and statistics. Paul Murray is a graphics developer; he was previously at The New York Times and studied Data Visualization at the University of Illinois, Chicago.

Project description:

Vaccination in the U.S. is often framed as an individual choice. Bloomberg used a huge government hospital dataset to show how places with low vaccination rates affect their more-vaccinated neighbors. 


The team used local-level vaccine and hospital data to identify regions whose health systems had been strained by Covid. An analysis of thousands of hospitals put the spotlight on Kentucky, which had just been through a tragic surge of infections, hospitalizations and deaths. 


The story used data analysis to target on-the-ground reporting and showed how the U.S.’s patches of vaccine vulnerability can overwhelm state-wide health networks.

Impact reached:

Bloomberg’s data journalism and health team’s have spent the last year tracking how vaccines are being administered, and assessing their impact on the Covid-19 pandemic. While government datasets can give snapshots of what’s happening around the country, we took the data a step further to explore more complex relationships between vaccination rates, hospital utilization, health-system collapse and medical-staff stress and burnout. 


What resulted was a meld of data journalism and personal storytelling that was read by almost a million people. Online, health workers expressed gratitude that the horrors they’d seen inside hospitals for the last two years were finally getting shared more broadly. The story also served as a warning for what other hospitals around the country were about to go through as a fall wave of cases expanded. 

Techniques/technologies used:

Bloomberg used two large and challenging datasets: Daily county-level vaccination rates published by the U.S. Centers for Disease Control and Prevention, and weekly hospital utilization reports published by the U.S. Department of Health and Human Services. 


The goal was to try and identify counties that met two criteria: high vaccination rates compared to their neighbors, and hospitals that hit capacity after facilities filled up in surrounding, less-vaccinated counties.



To accomplish this, data for all 3,000+ U.S. counties was brought into R. A series of scripts was run to identify places that met those specific rules, and the team identified 25 counties for follow-up. The team began calling hospitals and public-health officials to get a clearer picture of what had happened on the ground in those places. 


The analysis and follow-up led the team to Lexington, Kentucky. Lexington contains several large health systems and is highly vaccinated, but is surrounded by rural counties with low vaccination rates and more limited health care resources. More than 30 in-person and phone interviews confirmed the pattern: A wave of infections that started in less-vaccinated Appalachia swamped hospitals there, then spilled over and filled ICUs in Lexington, and then gridlocked the state’s health-care system.


To tell the story to readers, Bloomberg returned to the same datasets and built a visualization showing the fall 2021 outbreak’s rapid progress. Additional hospital data fed graphics showing how the wave of Covid had filled their beds and stretched staff past capacity.


The article used Svelte and D3.js to present an animated map of Covid-19 cases in Kentucky, as well as several charts showing how hospital bed usage changed over the course of several months of the outbreak. Exploratory charts were made in R with ggplot2 and adapted for the web.

What was the hardest part of this project?

Often data is used as a supporting component to an idea that’s already in progress. A source provides a tip and a reporter hunts down the data to confirm it. For this story, we did the opposite: We asked what would happen if you wrote a script to identify places most impacted by low-vaccination neighbors. We used that idea to narrow thousands of leads down to a handful of places we thought might have the most compelling stories to tell. 


Throughout the U.S. Covid outbreak, other news organizations have told compelling stories about individual hospitals or doctors. Bloomberg’s effort was unique in that it treated the broader system as a story, using data to illustrate the complex and hard-to-see relationships that keep regional health systems running – and that can fail in times of pressure. 


That data helped us precisely target our reporting on the ground. It identified hospitals such as Saint Claire in Morehead, Kentucky, which had been so overwhelmed by patients that one nurse told us about putting somebody in a body bag every day for two months. It led us to the University of Kentucky, which typically accepts more inbound patient transfers than any other hospital in the U.S. – but had to strictly limit those handoffs as the viral surge overwhelmed the state. Our on-the-ground reporting was made much more efficient, and much richer, by arriving with a deep understanding of how the fall wave of Covid had crashed into the state’s hospitals. 

What can others learn from this project?

The U.S. has expanded the amount of Covid-19 data available to the public over the last year, which is a welcome development. But many of the tools necessary to analyze the pandemic’s impact and to tell stories about it still have to be built from scratch. 


Bloomberg’s teams have worked with Covid-19 data for more than a year, in large part through our widely-read vaccine tracker. By making Covid-19 data a story in and of itself, we created deep expertise inside the newsroom that helped us identify new questions and stories that other outlets hadn’t attempted.


It also helped us identify pitfalls. Much of the U.S.’s Covid-19 data is reported from state and local health departments and compiled by the federal government. That’s created significant potential for data errors that can mislead journalists. Bloomberg’s team was able to spot several of those issues – such as an apparent change in the definition of how some Kentucky hospitals were reporting capacity – and to keep that problem from showing up in the story. 


Any journalist working with similar datasets should view building relationships with primary and secondary sources of such data as a core part of the job. Those sources – in states, county health departments, hospitals and the U.S. government – have been an invaluable source of tips, fact-checking and idea-testing. 

Project links: