Counting cases is not the way to measure the COVID-19 pandemic

For eight months, the U.S. government leaders have operated the machinery of the COVID-19 pandemic following case counts of positive laboratory tests. But, in the middle of a pandemic, case counts are not a representative indicator of infectious disease in a population. Without a measure that is proportional to infection, resistant to ever-change testing criteria and comparable across places and times, the public will not understand how the pandemic is progressing and government leaders will make the wrong decisions.

So far, according to the CDC, the United States has seen two waves of the COVID-19 pandemic, a larger one in April and a smaller one in July. Hospitalizations, which are roughly proportional to infections, peaked at 10 per week per 100,000 people during the week beginning April 11th fell then rose again to 8 per week per 100,000 people during the week beginning July 11th. Case counts, on the other hand, more than doubled from a weekly average of 31,000 during the mid-April peak to 67,000 at the end of July. From the top of one wave to another, hospitalizations fell while case counts rose — changes in medical visits for COVID-19 symptoms and deaths aligned with changes in hospitalization rates.

If case counts are the measure of pandemic progress, the public will understandably see a situation worsening when case counts rise even though the severe consequences may be declining. And decision-makers may naturally react with alarm, tightening restrictions unnecessarily. The measure of the pandemic itself becomes harmful. 

Public health fundamentals call for pandemic indicators to be proportional to the burden of disease in the population and to be insensitive to how public health measures infection. Both rates of hospitalizations and deaths can serve this purpose. Still, the shorter 12-day interval between infection and hospital admission makes the former the better choice for monitoring and public health action. In the U.S. today, few are denied hospital admission, admission criteria do not vary much from place to place, and everyone with fever, cough, or breathing difficulties and admitted to the hospital will be tested for the SARS-CoV-2 virus. CDC and HHS have been collecting some new hospital admissions data since March, but, so far, much of the information is not visible.

Because laboratory-based case counts are not proportional to either infection or severe disease, once a pandemic wave enters a growth phase — which happened in the U.S. by March — they are a poor tool for tracking infections even though they may retain value for individual medical care, for targeted public health disease investigations, and, when disease containment is possible, for contact tracing. During the H1N1 pandemic, the CDC abandoned state case reporting after a few months, explaining to the public, “as the 2009 H1N1 outbreak expanded, individual case counts become increasingly impractical and not representative of the true extent of the outbreak.” 

The National Pandemic Strategy calls for “transitioning from individual case reporting to severe disease and syndromic surveillance” during the acceleration phase of a pandemic. This is what happened during the Influenza H1N1 pandemic of 2009 when the CDC replaced the individual case reporting with reporting of H1N1 hospitalizations and deaths. After August 2009, CDC published monthly estimates of the total infections, hospitalizations, and deaths in each age group. 

Counting cases for pandemic monitoring is especially problematic given the characteristics of the SARS-COV-2 virus and the human factors that drive testing demand.

CDC defined a confirmed case of COVID-19 as a person who tests positive for the SARS-COV-2 virus regardless of symptoms or illness. Depending on age, up to 80 percent of people who test positive for the virus have no symptoms. A large, unreported fraction of U.S. cases are not sick at all. Conversely, a large and unknown fraction of healthy Americans have had exposure to the virus. Also, a new case found today may not be currently infectious. The SARS-COV-2 tests that detect fragments of viral RNA will return positive results for many weeks after contagiousness goes away.

Changing case counts reflects changing rules, access, and demand for testing. The CDC amended their guidance for who should get tested multiple times over the year, first prioritizing people with serious illness or vulnerabilities and later recommending testing for people without symptoms. Politicians habitually promise more tests. In the U.S. the number of weekly tests increased from 2 per thousand at the beginning of April to almost 10 per thousand in the middle of July. 

Counting cases, a lab-based case definition and ever-changing rates of testing affect the entire narrative of the pandemic. The Johns Hopkins Coronavirus Resource Center has been a primary repository of COVID-19 data for many in the media. The Johns Hopkins website describes pandemic trends and successes and failures largely in terms of new case counts. The Johns Hopkins’ authors do not explain the limitations of case counting, the biases that underlie these numbers, or the alternatives. 

Timely and comparable measures, which accurately represent disease burden, are critical for informing decisions on reopening society. The CDC’s Pandemic Intervals Framework, last updated in 2014, details when and how to respond at different points in a pandemic far worse that COVID-19. Public health actions should be proportionate to disease severity and levels of transmission. Increasing pandemic indicators should trigger more intense mitigation, while decreasing ones should allow relaxation. 

COVIDView, the CDC’s weekly summary of pandemic activity, already includes several of the better measures for pandemic monitoring, including some regional hospitalization data, but also emergency department visits and death rates. For example, the Aug. 28 summary reports continued declines in pandemic-related medical visits, hospitalizations and deaths with some regional variation. The summary of multiple measures paints a very different picture of the current COVID-19 pandemic than one based on case counts. Public health interests and journalists who represent the pandemic to the American people must use all of this available information, explaining how each measure corroborates another. 

Using multiple high-quality indicators together will allow decision-makers to act with more confidence and effectiveness. Given our nation’s public health expertise and experience, it remains unclear why many still promote counting cases to gauge the pandemic’s progress when our plans say otherwise, and much better measures exist. Our top priority today should be choosing more reliable pandemic indicators, like COVID-19 hospitalizations, medical visits, and deaths, that allow us to make data-driven decisions.

Rajiv Bhatia M.D., MPH, is an internist for the Veteran’s Administration and an affiliated assistant clinical professor of Medicine at Stanford University. He was a deputy health officer for San Francisco between 1998 and 2014.