The Nation’s Reported Covid-19 ‘Cases’ Are a Mix of Bad Math and Bad Reporting
If we want to use testing to effectively control this pandemic, this is what we need (now)
The U.S. Centers for Disease Control and Prevention (CDC) and many media outlets report the daily number of “new cases of Covid-19 infection in the United States.” They also regularly report infection rates—that is, the proportion of those tested who are positive. Such reporting occurs in other nations as well. But these statistics are misleading.
Testing for Covid-19 infection in the population serves three critical purposes in pandemic control:
- With the antigen test for Covid-19, the purpose is finding people who are currently infected and, if needed, recommending treatment or quarantine.
- Contact tracing, and then, if needed, recommending quarantine or treatment.
- With the antibody serology test, the goal is assessing the spread and monitoring trends of recent or past infection in the overall population to better understand the pandemic, ultimately to improve disease control.
Covid-19 testing numbers reported daily by the CDC and the media may be moderately effective for the first purpose, but they are not optimal for any of these purposes. Moreover, this number is not, as typically reported, “Covid-19 in the United States.” While reported numbers of Covid-19 deaths are a reasonable approximation of death counts in the population, the number of reported Covid-19 cases represents only the number of people who happen to get tested where testing happens to be available and that have shown to be positive. These “cases” are an unknown mixture of people who are asymptomatic, symptomatic, and those with Covid-19 who have been tested — again, where testing happens to be available. Estimating the number of cases in the population requires random, representative population sample testing. National counts and rates have been estimated to be more than 10 times what is reported in the media from available testing. Estimates of mortality rates associated with Covid-19 infection based on reported infection will be correspondingly exaggerated.
A basic problem with the daily reporting of Covid-19 “cases” is that we know little about who is getting tested and what population they represent. If most of those who get tested in one region are asymptomatic and simply want assurance that they are not infected, the proportion of “reported cases” will be low. If in another region most of those who get tested have symptoms, the proportion of “reported cases” will be high. In addition, access to testing is highly variable among states, by race, and by income, and has generally increased over time. Dividing one problematic number by another problematic number to estimate the Covid-19 infection rate (in the tested population) produces a problematic fraction — it’s bad math. The problem is compounded by comparison and addition of numbers across locations. Further, the reporting of infection rates to the CDC also varies by place, reporting institution, and time and misrepresents the results among those tested as if they were homogeneous indications for different regions of nation — that’s bad reporting.
The availability of testing
The availability and regulations regarding testing around the nation are far from standard. Rates of testing per population in states ranged from 206,000 per million residents in Oregon to five times as many — 1,093,000 per million—in Rhode Island (more than one test per person). Some testing sites require that the client have symptoms to be tested, while other sites do not. In some settings, tests are free; in others, there is a cost, though testing may be wholly or partially covered for those who are insured. Most states use federal funds to support testing, but some use additional state funds not available in other states. Comparison across testing locations is highly problematic.
Who gets tested
Given the great variability in who is getting tested, where, and why, the number tested is an unknown mixture of those who happen to get tested where testing happens to be available. A March survey found that more Democrats (74%) wanted testing than Republicans (52%), and Northeasterners were less interested in testing (52%) than Westerners (69%). A national survey in August found that rates of testing varied substantially by income (15.5% of those earning less than $25,000; 7.1% of those earning more than $150,000) and by race (7.1% of white people; 20.2% of Black people). It is unclear what it means to compare “cases” reported over time or location — a mix of oranges and apples, Red Delicious and Granny Smith.
How test results are reported
The Covid Tracking Project effectively monitors how Covid-19-related data is reported by states. The project grades states on the quality of reporting by evaluating data completeness. For example, less than optimal grades are given to states that do not provide hospital outcomes or infection data by race. As of October 27, 13 states still received grades of B or C; the remaining 38 (including Washington, D.C.) received an A or A+. States also report a lack of testing supplies and backlogs of unprocessed tests. Partly due to public health funding cuts, the means of reporting is also variable from place to place, as reported in the New York Times:
Health departments track the virus’s spread with a distinctly American patchwork: a reporting system in which some test results arrive via smooth data feeds but others come by phone, email, physical mail or fax, a technology retained because it complies with digital privacy standards for health information. These reports often come in duplicate, go to the wrong health department, or are missing crucial information such as a patient’s phone number or address.
The information collected and reported in this “system” of test data is of questionable validity and value for monitoring geographic differences or temporal trends. If we want to use a testing program to effectively control this pandemic, we need the following, and we need it now:
- A unified system of reporting across states and health departments, coordinated by the CDC. Standards for reporting, including a standardized reporting data form, are available. Electronic reporting to the CDC is also available.
- Consistent, affordable testing across the nation.
- Testing focused on people who believe they have been exposed or who are symptomatic, or separate testing for those neither exposed nor symptomatic.
- Contact tracing of people found to be positive for Covid-19 and a recommendation of quarantine of contacts found to be positive, or treatment if necessary.
- Random population seroprevalence studies for pandemic monitoring, epidemiological research, and control planning and evaluation.
Until we have greatly improved this system, daily reports of nationwide “Covid-19 cases” should be read with skepticism.
