Research Summary by
Dan Sacks, Assistant Professor, Business Economics & Public Policy & Life Sciences Research Faculty Fellow, Kelley School of Business, Indiana University
How many people are currently infected with the virus that causes COVID-19? Information about current COVID-19 prevalence is critical for governments faced with decisions about gathering restrictions and school closures, for businesses deciding on staffing and safety precautions, and for individuals weighing how much to social distance. Unfortunately, even now – a full year since the start of the pandemic — we still do not have a clear sense of the current prevalence of COVID-19 in the general population in most parts of the country. The problem is that only a small fraction of the population is tested for COVID-19 on any given day. And the people who are tested tend to be either showing symptoms of COVID-19 or have had an encounter with another infected person. People who have mild or asymptomatic cases of COVID-19 are less likely to be tested but are still able to spread the disease. This selection bias in testing means that the number of confirmed COVID-19 cases reported on the dashboards maintained by state governments and newspapers likely substantially undercounts the true number of cases. Likewise, common metrics like the test positivity rate are probably overestimating the true prevalence of the disease.
In the face of these data limitations, the best we can do is place plausible lower and upper bounds on true prevalence. For example, in the first week of December, 2020, 0.5 percent of the population of Indiana tested positive for COVID, and 21 percent of the people tested were positive. This means we can be confident that COVID prevalence was between 0.5 and 21 percent. But that bound is very wide, making it hard to judge the severity of the epidemic in December and the risks associated with different activities.
In this research project, we investigated ways to tighten the bound COVID-19 prevalence by studying people who had been hospitalized. We focused on people hospitalized for non-COVID illnesses, such as pregnancy, cancer care, or car accidents. These non-COVID hospital patients are not at elevated risk of being infected with COVID-19. But because they are in the hospital, they are tested for COVID-19 at very high rates, 25 to 50 times as high as the general population. As a result, we obtain much tighter bounds on COVID prevalence for this group. For example, in the first week of December, 2020, COVID-19 prevalence was between 2.4 and 4.8 percent among people hospitalized for non-COVID reasons. Using non-COVID hospitalizations to measure population-wide COVID-19 prevalence is surely imperfect. The hospitalized population is not representative of the general population. It is older, for example, and it may be sicker or more susceptible for disease.
Despite these imperfections, we believe our approach is useful and feasible for states and health systems looking to measure COVID-19 prevalence. It is useful because even if the hospitalized population is non-representative, we can still learn something about COVID-19 prevalence. For example, if we think COVID-19 prevalence is higher in the hospitalized population, then the upper bound on COVID prevalence in the hospitalized population (4.8 percent in our example) is an upper bound for population prevalence, which could be as high as 21 percent without the hospital data. The approach is feasible because states already collect and report information on COVID testing and prevalence in the general population, and they report COVID-19 hospitalization rates. To report our prevalence bounds, states would only have to report COVID test and prevalence rates for a subset of hospitalizations, ones with diagnoses for non-COVID conditions.
This summary is based on the research paper ‘What can we learn about SARS-CoV-2 prevalence from testing and hospital data?’ The paper may be read here co-authored by Coady Wing, Associate Professor, Indiana University O’Neill School of Public Health; Nir Menachemi, Fairbanks Endowed Chair & Chair-Health Policy and Management department, Indiana University Richard M. Fairbanks School of Public Health; and Peter Embí, President & CEO, Regenstrief Institute, Inc.
Leave a Reply