While the public is eager for news about the novel coronavirus, the information can be confusing, outdated, and/or difficult to access. SPA Assistant Professor Nathan Favero has developed a new model to make more valid comparisons about the prevalence of the disease across states, or within a state over time.

Favero wrote an article entitled Adjusting confirmed COVID-19 case counts adjusting for testing volume, published on the website medRexiv, and launched the website COVID-19 Trendlines to share his research. All the code and data to run his analyses are available online through an open GitHub platform.

Rather than follow a traditional journal publishing route, Favero felt it was important to make his findings freely available immediately, with a preprint article and an open source platform. He plans to maintain a blog on the website to update information as it becomes available.

“Never before have I had to do research on a topic with the same sense of urgency,” said Favero, who leaned on other openly accessible papers for this project. “I felt a responsibility to make my findings publicly available, and [I] hope that others can build on it.”

The project addresses whether the recent national uptick in COVID-19 cases is mostly due to more testing or to actual increased spread, estimating that the latest surge is mostly due to spread of the disease. His research indicates a serious decline in prevalence throughout April and May, although the raw case counts, more widely reported, show more of a plateau during that time.
As media report on the novel coronavirus, comparisons based on confirmed case counts alone can be misleading. Trends in confirmed case counts only reflect the number of cases that have been detected. Therefore, an increase in the number of reported cases could mean that the number of infections has increased, or it could mean that states are now doing a better job of detecting whatever infections are out there––or both.
Looking at weekly test result records in each state, Favero adjusted confirmed cases based on the percentage of COVID-19 tests that came back positive in a regression analysis; this provides what he said is a clearer picture of the diseases’ progression in the U.S.  

“It’s complicated. My hope is to make things simpler and tell a story with one indicator and not have to navigate multiple measures,” Favero said. The new approach will make it easier to consider the true situation state-by-state for a more accurate assessment.

Favero’s work also examines why COVID-19 deaths have not increased alongside the spike in cases. Because of the switch from testing only those exhibiting serious symptoms to those with mild or no symptoms, Americans are detecting their infection earlier in the disease course. Therefore, he suggested, there will likely be a longer delay between detection and death.

Experts in public health have much of the same information, but Favero hopes his project will help the general public better understand the pace of the spread. It can also be valuable to policymakers making decisions about reopening plans to see what has and hasn’t been effective. Too often, politicians are “cherry picking” parts of the data to support their point of view; this model captures a more complete picture of the science, he said. This information––using a single measure––could empower the public to be more informed as they digest the news and politics around the virus, he says.

Favero conducted other COVID-19 research earlier this spring on attitudes toward social distancing during the pandemic.