By Ana V. Diez Roux, MD, PhD, MPH

One of the many remarkable aspects of the times we are living through is the frequent discussion in the press and among the public of basic epidemiologic concepts that are usually relegated to abstract discussions in public health courses. Many mornings, sitting over my coffee, I have heard radio debates or read discussions in the paper on how best to measure the speed with which disease is developing (number of cases or number of cases per population?) or how deadly it is (cause-specific mortality or case-fatality?). More recently there have been intense discussions (sometimes a bit muddled) about what is meant by the accuracy of various tests: Is it best for a test to have high sensitivity (probability that someone with the disease tests positive?) or high specificity (probability that someone without the disease tests negative?) And how can it be that even a test with very high sensitivity and specificity can have a low positive predictive value, i.e., the probability that someone who tests positive actually has the disease is low? (This is one of my favorite epidemiologic concepts…if you do not know the answer check it out!)

And then of course the simulation models of various types, some statistical models used to predict into the future based on associations between variables observed in other places or in the past, others traditional infectious disease equation-based models (with flows and stocks across pools of susceptible, infected and immune), and even agent-based models that attempt to capture the behaviors and interactions of individuals in order to see under what condition disease rates increase or decrease, and how various interventions may affect incidence or deaths. Each type of model has its strengths and limitations, all hopefully based on data, but also (as every modeler knows) often heavily reliant on many assumptions on which we have (yes, even today many weeks into the pandemic, limited information).

As an epidemiologist at heart, I have been alternatively thrilled, amused, and bemused by seeing epidemiology (and epidemiologists!) in the limelight. On one hand discussions have illustrated the utility (and in my mind also beauty and elegance…if you will indulge me…) of very basic epidemiologic concepts, concepts I remember reading about in Lilienfeld’s Foundations of Epidemiology so many years ago on the top floor of Hampton House (the public heath library) at the then Johns Hopkins School of Hygiene and Public Health. On the other hand, the discussions have also illustrated, in an often shocking way, how ill-prepared we have been as a society to respond to the crisis we are facing: how we are still lacking important scientific information on the very basic epidemiologic parameters that we need to understand where we are, what we should do, and how we should monitor and evaluate our actions.

There are, of course, some good reasons for this. Sometimes the metrics we need to calculate (e.g., the case fatality rate) although deceptively simple can be difficult to accurately capture in the midst of a pandemic. But the challenges we are facing go beyond that. They are related fundamentally to two things. One is the lack of investment in public health infrastructure, even in the United States, one of the richest countries in the world and the one that spends the highest proportion of its GDP on health care. The second, as striking as the first, is the absence of a clear and coordinated national response in the United States to address the basic elements of epidemiologic surveillance in a way that is systematic and coordinated across the country.

Access to testing (as extensively discussed in the press) has been a major factor in our inability to fully characterize the pandemic, but another major factor has been the absence of coordinated systematic efforts to conduct studies that can provide compelling answers to questions like the proportion of the population that has been infected, the true incidence, the clinical manifestations and severity of infection (including the predictors of severity beyond age and chronic diseases), and the duration of immunity. These questions are critical to our ability to move beyond blanket lockdowns and design, implement, and monitor the most effective strategies for the future, not just in the United States but globally.

In the absence of a coordinated national response, multiple groups and individuals have attempted to fill the gaps as best they can publishing their work in pre-peer reviewed websites. Results of surveys and studies are beginning to emerge, but they can only offer partial and incomplete views, and sometimes can generate information that appears to be contradictory (even if sometimes understandable scientifically because of different populations studied and methods used), creating confusion among the public and providing opportunities for the information to be used in politically motivated ways.

Where does this leave us? The lack of basic epidemiologic data has been one of the most troublesome aspects of the pandemic. There are some signs that better data may be available in the future. For example, the CDC has launched some initiatives to collect relevant public health data. It is also true that what appears to be disjointed pieces of evidence emerging from the work of various groups, can often come together and coalesce into a few facts around which there is consensus. Another question is whether this is the best and most efficient way to gather information on basic epidemiologic parameters in the midst of a public health emergency.

Perhaps the greatest hope is that all these public discussions of incidence and prevalence and case-fatality rates and sensitivity and specificity will spur greater recognition of the importance of science for policy among the public, and will ultimately reinforce the fact that public health is par excellence a government responsibility critical to the health of all of us. On both counts unfortunately, the jury is still out.