Since the beginning of the COVID-19 pandemic, there have been numerous projections and predictive models released on how it will unfold, how many people will become infected and how many will die.
Earlier this week, on May 26, Gov. Asa Hutchinson shared two COVID-19 predictive models conducted by the College of Public Health at UAMS at his daily press conference.
The first predictive model was from March 25 and showed the number of projected cases versus the number of observed cases up until May 20. Based on the model, it was projected there would be nearly 8,000 cases on May 20. However, the state actually had 5,458 cases on that date.
The second predictive model was an updated version of the first one, but from May 21. It showed the new prediction curve based on updated data from March 25 to May 20. Based on the model, it is projected that the state will have approximately 8,500 cases by June 23.
“COVID-19 is a new virus. That is why you often see the ‘novel coronavirus.’ We still don’t know how it operates or how it really works,” Dr. Mark Williams, dean of the College of Public Health at the University of Arkansas for Medical Sciences, told Arkansas Money & Politics.
According to Williams, UAMS used COVID-19 data that the Arkansas Department of Health collects.
“The dataset is not publicly available, but it is very similar to what the governor shares daily in terms of number of caes, hospitalizations, deaths and such,” he said. “The models that we have developed are based on the data we collect.”
“Models are based on assumptions in which assumptions are based on data. A model is only good as the assumptions you put into it. As the pandemic continues, the models are improving and getting closer to accuracy. They are not meant to be perfect or exact. The intent is to show trends and directions.”
Williams compared the predictive model to a forecast in terms that both are “based on knowledge of what happens today in the given circumstances if nothing else changes.”
He said, “they are constantly updated weekly with new data. Given the data, the prediction curve changes. We projected to have 6,000 cases by June 1 last week and the state now has 6,830 on [May 28].”
Although the UAMS predictive modeling is not available publicly, Williams noted that it is an effort to learn more, provide information to the Arkansas Department of Health as well as make a contribution in helping better understand COVID-19.
“From an epidemiology standpoint, the projections are interesting. However, it is easy to get confused as to what the models are actually doing and showing,” he said. “They are easy to misinterpret, and even to politicize.”
When asked about the state’s efforts to minimize the spread of COVID-19 in March and April, Williams said “it is demonstrable in the change in the prediction curve on the models.”
Williams wanted to recognize the scientists at the College of Public Health who have been working hard “as incredible public servants” to help the state. They are dedicated to collecting the data, looking at assumptions and adapting the models to Arkansas, so that we can better understand the effects of the virus in the state.”
Image courtesy of UAMS College of Public Health