Walk into a shop or restaurant ... and cough. Everybody will notice — and many will rapidly move away from you. Cough in a hospital waiting room and you will find a cotton swab in your nose as health workers rush you into isolation.
Five months ago, there were more than 100 distinct infectious diseases that might have explained that cough. Right now, there seems to be only one disease: COVID-19. Has influenza disappeared from the planet? How about tuberculosis? Whooping cough? The common cold?
The diagnosis of disease can be summed up in a simple mathematical formula. When somebody develops a cough and a fever, doctors ask themselves three questions: 1) What diseases can cause these signs and symptoms? 2) How common are the diseases themselves? and 3) How often will each individual disease produce cough and fever? All that remains at this point is to make a mental list of these numbers and rank them. Diseases that are common — and that commonly produce these symptoms — rise to the top of the list.
The mathematical basis for generating a ranked differential diagnosis list based on statistical probabilities is known as Bayesian analysis.
Over the past 3 decades, our team of researchers and infectious disease specialists have developed and expanded a massive set of computer spreadsheets, which follow the actual numbers involved. We use these data to generate a ranked list of diagnoses for every infectious disease scenario. For example, if a group of American adults is experiencing severe illness, cough and fever, the most likely diagnosis is COVID-19. In this scenario, an infectious disease database platform ranks COVID-19 “number 1,” with a statistical likelihood of 87%. However, the program also lists an additional 76 infectious diseases. (Influenza is the second most likely diagnosis, with a probability of 8%.)
If the statistical likelihood for COVID-19 in this case is 87%, then the possibility for another diagnosis is 13%. Failure to diagnose — or even consider — some of these diseases will waste time and resources and might even endanger the patient. When clinicians are challenged by a list of over 70 diseases compatible with a clinical scenario, they will say to themselves, “Of course, I should also look for that condition” or, perhaps, “What is that disease, and why does it appear on my list?”
In order to further explore and focus the diagnosis list, clinicians will include additional signs, symptoms and laboratory tests, as well as relevant details concerning exposure (foods, animals, foreign travel). Even the dates of travel or exposure can be used to improve the specificity of this list.
But what if we are dealing with a group of adults in Australia with identical clinical findings? In this case, influenza is most likely (45% probability) and COVID-19 ranks second (22% probability). Similarly, Ebola and monkeypox would be twice as likely as COVID-19 — if this were a group of identical patients in the Democratic Republic of the Congo.
For the past 5 months, thousands of physicians worldwide have been confronted by waves of people with fever, cough and difficulty breathing. Inevitably, the “art and science” of diagnosis will be degraded by a certainty that the next patient who enters my clinic will also have COVID-19. But every patient is unique, and we must continue to discount alternative diseases — even at the height of this global pandemic.
