Opinion

Defective data, risky results; Here’s how the Harvard COVID-19 model got it wrong

(Getty) Coronavirus around blood cells

Op-ed views and opinions expressed are solely those of the author.

As world governments face unprecedented medical, social, and economic fallout from the COVID-19 crisis, they turn to a group of specialists who rarely gain notoriety or acclaim outside their communities – virologists, immunologists, epidemiologists and a myriad of statisticians, mathematicians, and other academics.  The specialists are being called upon to step out of the mostly theoretical worlds of science and academia and lead in a real-world global crisis.  

The Harvard Global Health Institute (HGHI) issued one of the key models our nation’s leaders use to evaluate and counter COVID-19. To calculate COVID infection across the US population, HGHI utilized a widely circulated and often-referenced study by the Chinese Center for Disease Control and Prevention (CCDC), entitled “Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China.”  

The CCDC’s case study was extensive and included 72,314 individuals.  According to the study, patients experienced the following severity of reactions to COVID-19: 81 percent mild, 14 percent severe (e.g. hospitalization), and 5 percent critical (e.g. ICU/ventilators).  

Defective data, risky results

Extrapolating from this data set, the Harvard study purports to demonstrate the utilization of hospital beds across the entire country based on a 20 percent, 40 percent, and 60 percent “infection rate of the general population.”  They then teamed with ProPublica—an ostensibly non-partisan and non-profit news organization run by Paul Steiger, founded by billionaires Herb and Marion Sandler, and that has George Soros as a board member. After teaming up with Harvard, ProPublica created the following terrifying graphic.

 

Red X added to make clear the image is grossly misleading and to avoid promoting false and erroneous information

According to the graphic, the US appears to be headed for an almost certain catastrophe.  

There’s only one problem.  It’s completely wrong.  

The CCDC study that Harvard based their model on was flawed.  The CCDC authors clearly said as much; Harvard knew or should have known this.  Yet, Harvard’s model had based its number of “projected infected individuals” on a sample of patients of whom 100 percent displayed symptoms and of whom 100 percent tested positive for COVID-19.  This is not representative of a spectrum of COVID-19 cases.   

Based on available studies and research, a likely majority of people with coronavirus are asymptomatic.  Logic dictates that because the CCDC failed to account for asymptomatic individuals and based their statistics only on confirmed symptomatic cases who sought treatment, the model also fails to account for those who experienced symptoms and never sought medical treatment.  Failing to account for, or address, what likely comprises a majority of all infected individuals is completely indefensible.

To illustrate this flaw, consider New York State, which has garnered the majority of media attention.  HGHI’s worst-case scenario models a national infection rate of 60 percent of the general adult population.  Remember that the model is basing their 60 percent on patients who all displayed symptoms, all sought medical treatment, and were all diagnosed positive.    

However, for argument’s sake, 40 percent of New York State’s roughly 15 million adults yields approximately 6 million infected people.  Applying the CCDC’s numbers and percentages would result in more than 1.2 million people, or 20 percent, requiring hospitalization—of whom 268,000 or 5 percent would require ICU care.  However, as previously noted, the model is not accurate; At least half or roughly 600,000 should be asymptomatic or simply recover at home. The remaining 600,000 would be those who seek medical treatment, of whom 120,000 would eventually require hospitalization and 30,000 ICU care.  Again, this is the model’s worst-case scenario for which there is no precedence and these numbers would be spread over the life of the disease, 12-18 months.  

Simply put, the model’s outcome would not be as grim as Harvard and ProPublica’s graphic has predicted.  Harvard’s model grossly inflates the numbers and makes the situation appear more dire than it is realistic. To be clear, analysis of the flawed model does not discount that the coronavirus is a dynamic evolving situation. 

If Harvard’s and ProPublica’s intent was not to stoke fear, create panic and prod an economic collapse that led to the unemployment of approximately 17 million US workers, then their model and graphic will at best be called shoddy scholarship.  At worst, this is criminal negligence.

Powered by Topple

Robert Spalding

Comments

Latest Articles