Clinical science is a process for continually revising ideas about what is best for us should we become ill. It challenges older ideas by comparing old to new tests or treatments. The fuel of science is competing ideas, the grist of science is comparison. Clinical science has four main components; a) measurement, b) options, or alternative ideas, c) study design, and d) communication. All four have been on display for the COVID-19 pandemic. Mistakes have been made in every component.
Measurement.
If we cannot measure something accurately, there is poor clinical science. The first questions to ask of any report on any malady is what clinical events are being measured and how accurately are they being measured. The main events being measured during COVID-19 are the diagnosis, and outcomes. The two diagnosis measures are of present illness (COVID-19 test) and past illness (antibody). The outcome (clinical event that results from having COVID-19) primarily includes death.
How good are these measures? The diagnostic measures are not good. Both the test for present and past disease are inaccurate. By inaccurate I mean that some people with COVID are missed by the test, while some who are well are falsely identified. This measurement error undermines clinical science and leaves the public confused.
What about the clinical outcome of death? Death is an accurate measure, but, the issue for COVID-19 is whether the death is due to that infection. Unfortunately, death due to COVID-19 is likely an inaccurate measure. How could this be? For one, not all patients who die have a COVID-19 test, many of the deaths are “presumptive”. Second, other potential diseases that can cause death, like influenza, are not being measured concurrently. Measuring the cause of death is a precarious scientific measurement and often inaccurate for many clinical maladies.
Options.
Clinical science tests new against old ideas. The ideas being compared are called, options. If you have coronary artery disease, for example, options for your care might be medical versus surgical therapy. Options must be as clearly defined as outcomes, and, if we want to know which option is best, a study must assure that all people receive the same options. For example, if medical therapy is being evaluated we must know what medicine is being given, it’s dose, and how long it will be given. There must be specificity to the description of the option. If the options are unclear, or undefined and variable, clinical science cannot determine which is best. Suppose, for example, some patients take only 50% of the medicines while others take 100%? If that occurs, studying the medicine option is useless.
Unfortunately, many of the proposed treatment options for COVID are poorly defined and haphazardly carried out. For example, staying home as an option is unclearly being delivered and followed. Some people are staying home, some are at work, or school. What about masks? Some wear masks, some don’t. Some states are opening now, some are not. Options are inconsistent, so clinical science is compromised. This leaves us unsure of the best ways to approach an epidemic like COVID-19. A strategic, specific plan for options is not being done, even yet, 4-5 months into the epidemic.
Study Design.
The next most important items in clinical science are how measures are ascertained, and in whom are they measured. Besides the critical requirement of accurate measurement, the measures must be obtained in ways that allow us to appropriately compare options for care and then extrapolate to others not in a clinical trial.
In clinical science, the design of a study is “how” measures are taken. Two terms are important to know, “Random sampling” and “Randomization”. Random sampling is required to learn prevalences of diseases and outcomes, and randomization is required to test one option versus another to see if we can alter the prevalences of diseases or their outcomes.
Random sampling means that people are chosen for study based on a coin-flip process. Hence, random sampling assures that the sample of people being studied, for example, tested for COVID-19, will all have had an equal chance of being studied. For example, if I had the phone number of every person in the country, I could randomly call some of them and ask them to be tested for COVID-19. Random sampling provides the most accurate estimate of the number and percent (prevalence) of people with COVID-19.
Presently, this is not how the numbers for COVID-19 are being measured. Who is/is not tested is not being done randomly. Only about 15% of the population of the United States has had a test and we do not know the processes used to decide who does/does not get tested. Hence, we are compromising our understanding of the burden of COVID-19.
Randomization, is the process for choosing who gets what option once we are ready to examine what options might be best, for example, to reduce the transmission of COVID-19. Randomization assures that every person in a study has an equal chance of getting the options. For example, presently there are over 125 agents proposed to treat COVID-19. Testing these agents will require randomized trials. Randomization balances personal and clinical factors that may influence the results of a study. With those factors balanced, studies can assess the independent contribution of a studied option.
Communication.
There is presently poor reporting on the COVID-19 epidemic because quality clinical science principles are not being used. This epidemic has exposed, in my view, a disorganized approach to the clinical science of epidemics. While determining accurate numbers for the prevalence of COVID-19 and it’s burden is paramount, it is important to realize that clinical science is not a decision-making technology; it is a number generator. Decision-making is a process of judgement informed by numbers, but decision-making is hampered by poor clinical science.
