Comparisons of coronavirus infections to the mundane seasonal \u2018flu have invited widespread mockery. In fact, the comparison is oddly and usefully apt. Estimates of infection and death rates \u2013for both \u2013 require heroic, unverifiable assumptions. And how we deal with flu outbreaks provide useful lessons for coping with what the World Health Organization has declared a global pandemic.<\/p>\n
Like many, I never see a doctor when I believe I\u2019ve gotten the flu. Most patients who do see physicians are \u201cclinically\u201d diagnosed, in a rough and ready way. The US Center for Disease Control (CDC) recommends<\/a> laboratory tests only for hospitalized or high risk patients. And, according to the CDC, \u201cproper interpretation\u201d of tests requires consideration of numerous factors such as whether the specimen is taken from the upper or lower respiratory tract, accuracy of the test used as compared to a \u201cgold standard\u201d test, and the prevalence of the flu in the patient population.<\/p>\n These interpretations (like my personal guesses and physicians\u2019 clinical diagnoses) are fallible: laboratory testing isn\u2019t like dipping litmus paper into a beaker.<\/p>\n Attributing deaths to the \u2018flu is also a fallible judgment call, not the unambiguous result of a foolproof test.<\/p>\n How many are diagnosed, whether by themselves, clinically by physicians or nurses, or through lab tests, is also a guesstimate. I keep my self-diagnosis to myself and professional results aren\u2019t entered into a national registry or data base.<\/p>\n <\/p>\n With the new coronavirus, the problem of testing mistakes arising from errors in the underlying science, the design and construction of instruments, and, in the interpretation of results are even more acute. We cannot even objectively assess the magnitude of the errors \u2013 how could we reliably test the accuracy of a new test for a new disease? And as with new vaccines and drugs, turning promising laboratory research into practically useful tests requires extended trial and error; the recent Theranos scam should warn us about trusting miraculous leaps. Training clinicians to operate and interpret new tests is another huge practical problem as the history of mammography shows<\/a>.<\/p>\n But, without reliable tests of known accuracy and comprehensive, well-maintained registries, epidemiologists cannot escape classic \u2018garbage-in garbage out\u2019 problems of modeling the severity and progression of infections.<\/p>\n