Fever screening is the practice of checking the temperature of everyone passing through a checkpoint, such as at an airport or public gathering, for fever. This can be done individually with regular thermometers, or on entire crowds with infrared cameras. Fever screening has been widely deployed against COVID-19, especially in China, since fever is its most common and often chronologically first symptom.
Fever screening applies an evolutionary pressure on pathogens to not cause fever, or to cause fever later relative to when it becomes transmissible. Pathogens mutate. In many diseases, this results in drug resistance; antibiotics, for example, lose effectiveness over time.
Drug resistance evolves slowly, because partial resistance is only a slight benefit to a pathogen which acquires it; most patients who receive antibiotics do not infect anyone after their treatment has started. Evasion of public-health screening procedures, on the other hand, is likely to evolve much faster, because any incremental improvement in a pathogen’s ability to evade screening will give it a large advantage.
Evidence This Is Already Happening
I searched for papers recording the percentage of patients which had a fever in Google Scholar and in the citations on this page. For each paper, I extracted the percentage of patients reported to have a fever, and the date range of cases covered by the data. If the paper did not state the date range of cases, I wrote down the paper’s publication date instead. I excluded papers which didn’t report a percentage of patients with fever, or had n<20.
I found seven studies, listed below sorted by end date.
Dec 16-Jan 2: 98% Jan 1-20: 83% Jan 1-28: 98.6% Jan 6-31: 58.3% Pub. Feb 18: 78.2% Jan 21-Feb 15: 85.7% Before Feb 20: 87.9%
This is suggestive of a negative correlation between study date and percent of patients with fever. This is what you would expect if COVID-19 were evolving to evade fever screening; however, it could also be explained by later studies finding earlier and less-severe cases. This is weak evidence compared to the what could obtained by someone with access to a good dataset; these papers come from different sample populations and different points in the disease progression, and don’t all specify what their cutoff temperature was for diagnosing fever. If someone has access to a suitable dataset, I ask that they do the analysis and publicly state whether they see a declining rate of fever.
Other Properties Will Likely Also Change
It is well known that most diseases evolve to become less severe over time, because patients with severe cases are easier to detect and will take greater precautions. However, this takes place on very slow time scales. Evading fever screening, on the other hand, involves greater selective pressure and so may happen on a faster time scale, possibly fast enough to significantly influence the shape of the pandemic this year.
I don’t know how fever-screening evasion relates to disease severity; I can see plausible mechanisms by which this would make it either increase or decrease.
Effectiveness of Fever-Screening Will Decline
COVID-19 Is Under Strong Selective Pressure
Fever screening is the practice of checking the temperature of everyone passing through a checkpoint, such as at an airport or public gathering, for fever. This can be done individually with regular thermometers, or on entire crowds with infrared cameras. Fever screening has been widely deployed against COVID-19, especially in China, since fever is its most common and often chronologically first symptom.
Fever screening applies an evolutionary pressure on pathogens to not cause fever, or to cause fever later relative to when it becomes transmissible. Pathogens mutate. In many diseases, this results in drug resistance; antibiotics, for example, lose effectiveness over time.
Drug resistance evolves slowly, because partial resistance is only a slight benefit to a pathogen which acquires it; most patients who receive antibiotics do not infect anyone after their treatment has started. Evasion of public-health screening procedures, on the other hand, is likely to evolve much faster, because any incremental improvement in a pathogen’s ability to evade screening will give it a large advantage.
Evidence This Is Already Happening
I searched for papers recording the percentage of patients which had a fever in Google Scholar and in the citations on this page. For each paper, I extracted the percentage of patients reported to have a fever, and the date range of cases covered by the data. If the paper did not state the date range of cases, I wrote down the paper’s publication date instead. I excluded papers which didn’t report a percentage of patients with fever, or had n<20.
I found seven studies, listed below sorted by end date.
Dec 16-Jan 2: 98%
Jan 1-20: 83%
Jan 1-28: 98.6%
Jan 6-31: 58.3%
Pub. Feb 18: 78.2%
Jan 21-Feb 15: 85.7%
Before Feb 20: 87.9%
This is suggestive of a negative correlation between study date and percent of patients with fever. This is what you would expect if COVID-19 were evolving to evade fever screening; however, it could also be explained by later studies finding earlier and less-severe cases. This is weak evidence compared to the what could obtained by someone with access to a good dataset; these papers come from different sample populations and different points in the disease progression, and don’t all specify what their cutoff temperature was for diagnosing fever. If someone has access to a suitable dataset, I ask that they do the analysis and publicly state whether they see a declining rate of fever.
Other Properties Will Likely Also Change
It is well known that most diseases evolve to become less severe over time, because patients with severe cases are easier to detect and will take greater precautions. However, this takes place on very slow time scales. Evading fever screening, on the other hand, involves greater selective pressure and so may happen on a faster time scale, possibly fast enough to significantly influence the shape of the pandemic this year.
I don’t know how fever-screening evasion relates to disease severity; I can see plausible mechanisms by which this would make it either increase or decrease.