top of page


Guest post by MARZIEH GHIASI

MSU SciComm blog contest winner

I’ve always loved detective stories—mostly the classical figures operating out of London: Hercule Poirot, Sherlock Holmes etc. But detectives aren’t only in fiction. John Snow, a founder of epidemiology, was a real-life disease detective. Notably not the one with dragons, Snow trawled through the streets of mid-19th century London trying to trace the origins of a deadly cholera pandemic that was killing hundreds. Using clever methods, including interviews, mapping and even an early epidemiologic experiment, he isolated the outbreak to some water supply companies providing sewage-contaminated waters. One major contaminated source was a water pump, which Snow helped shut down. The pump itself, no longer in use, is still there to see on Broad Street in London.

John Snow helped bring the cholera outbreak in London under control decades before people understood that bacteria caused cholera and other disease. That has been the power of epidemiology for a century and a half: to understand patterns of health and disease and save lives even if we don’t understand the biology of a disease perfectly. However, since its 19th century origins, the field has come to encompass more than infectious diseases.


When I began my training in epidemiology, two things became quickly apparent. First, no one including my family could actually pronounce what I am doing, or they confused it with dermatology or entomology. Unfortunately, I don’t know much about the skin, nor do I have a great affinity for insects, unless they are carriers of disease. Second, while the field finds itself in headlines in times of global crisis such as the COVID-19 pandemic, few people actually know what modern epidemiologists do.

Think of COVID-19. Virologists are working around the clock to understand the mechanisms of how the SARS-CoV-2 injures cells. Clinicians are in the frontlines of healthcare reporting symptoms and treating patients. Public health officials are running campaigns to make sure everyone is informed about the need to maintain social distancing. So, what are the epidemiologists doing?

There are many complex questions of health and disease that we take for granted. Questions like ‘does taking a daily aspiring prevent heart attacks?,' ‘do warnings on cigarette packs reduce deaths from smoking?,' ‘who is at higher risk of dying from COVID-19?'—can’t exactly be solved in a laboratory because the human body is complex, and there are many variations between people, our economic and social environments.

Even in the early days, epidemiologists were looking for correlations Epidemiologists work closely with biostatisticians to understand the patterns of disease across population ‘variables’ such as age, sex, health status, social environment, economic status and so on. This has been described as a fifth dimension, where human health is measured over large populations, understood in real and imagined worlds, and described through probabilities. Similar to how meteorologists can’t predict the exact amount rain in any given backyard or on someone’s head, an epidemiologist can’t say what has happened or will happen to an individual person. But a meteorologist can, with some margin of certainty, describe average rain over a region and make a reasonable recommendation to carry umbrellas when traveling in an area. Similarly, epidemiologists can describe population health factors and make recommendations about health precautions or interventions for populations.

Today, epidemiologists in dozens of subfields are working to understand the COVID-19 pandemic. They count the number of people infected and dying (descriptive and spatial epidemiology), identify what puts people at higher risk of infection and death (observational epidemiology), predict the future trajectories of the pandemic (predictive epidemiology), and how different measures from national policies (social epidemiology), to face masks and drugs work to prevent death (interventional epidemiology).


The most important way that epidemiology is applied in medicine is to figure out whether a given drug or vaccine works in patients, and their side-effects. This has been very important when faced with new diseases such as COVID-19, where epidemiologists and clinicians, in close associations with basic scientists, biostatisticians, and others are working in synchrony to answer these questions.

But not all epidemiologic evidence is equal. For example, in the early phases of the pandemic with no cure in sight clinicians desperate for cures reported that they administered various medications on their patients in their hospitals—hydroxychloroquine, azithromycin, vitamin C, tocilizumab, and you name it-- which appeared to have improved the condition of their patients. Quickly, other clinicians refuted these findings, reporting that the drugs did not work and in fact harmed patients. Why the controversy? Many of these studies were what epidemiologists call ‘case series’ studies or ‘cross-sectional’ studies, the lowest level of evidence. It doesn’t mean that the drugs don’t work. It simply means that we don’t have real evidence one way or the other, because these studies have a great risk of bias.

Nevertheless, the promise of ‘cures’ have attracted headlines. As people and countries have rushed to get access to unproven cures, many have argued that publicizing and trying these unproven drugs on patients even in desperate times, raises great concern about ethics and scientific integrity.

Hierarchy of study types in epidemiology

Another example of bias-prone studies are correlational studies. We’ve all heard at some point that ‘correlation is not causation’ and can reason that the decline of pirates and rise of global warming are probably unrelated. But it becomes a bit more tricky in studies of health. For example, a number of have looked at levels of tuberculosis vaccination across countries, a disease which most developed countries don’t vaccinate for anymore. These studies report that countries with more vaccination for TB have lower rates of COVID-19. However, because they are comparing two average measures over an entire country, they have a serious problems with bias when drawing the conclusion that these things are related. Another entertaining but real example of this is if we were to compare average chocolate consumption of a country and the number of Nobel prize winners.

Don’t get me wrong. Associations are important and epidemiologist love to find them where they can! But the ultimate goal is to infer causality, that is to determine what things actually cause changes in human health so you can intervene. And this can only be done when you identify and eliminate biases as much as possible.


So what is this ‘bias’? Bias is anything that impacts the ‘trueness’ of our results and it is everywhere. Epidemiologists have named dozens, categorized them, and there are whole subfields in epidemiology dedicated to studying biases. Some biases are due to shortfalls in tools we have to measure things. For example, think about measuring your weight on a scale—your measurement is as good as the scale is, and if the scale is not very good, then your weight will fluctuate.

Other sources of bias may be more nefarious. For example, think about the COVID-19 drug studies described above.

Maybe in one study the clinicians only selected patients who were healthy enough to be able to take a certain drug in, and left others out if they were sick or had died (an example of selection bias). Maybe in another study, patients who were on a drug of interest were treated differently, or more attentively, than they would have been otherwise (an example of observation bias). Maybe one group of patients who did not take the drug were old and frail, compared with patients who took the drug but were also healthier and more likely to survive because of their age and health condition (an example of confounding bias).


For epidemiologists, higher quality evidence comes from ‘case-control’ and ‘cohort’ studies. These types of studies formed the bulk of the evidence we have today that smoking causes cancer. In the COVID-19 pandemic, case-control studies have tried to determine what factors put groups at higher risk. For example, here they group those who died from COVID-19 (cases) and compare them to people who developed mild disease (controls). They make sure groups are comparable, a process called matching by age or gender or location. Then they go backward in time and compare what people were exposed to (did they smoke?), what health state were they in (did they have heart disease?) and so on. Cohort studies, on the other hand, go forward in time looking at specific groups with specific exposures or health states, such as workers in factories or women who are pregnant, and follow them to see how many and what groups develop COVID-19. Of course, these studies have their biases too.

The highest quality evidence left for epidemiologists are randomized controlled trials (RCT). As of July 2020, there were over 1400 clinical trials registered in the US for COVID-19, many of them RCTs. They are like experiments. Imagine two worlds, one in which a person takes a medication, in the other the exact same person doesn’t. One of these worlds can never exist. So epidemiologists, limited by not being able to create two worlds (that’s physicists, right?), create two groups of people that are extremely similar. One group gets the treatment, and one doesn’t. Based on that, we can determine if a drug or vaccine actually works. These are not easy to do, and we can’t always do them—but they are the very best tool we have.

Will wearing a mask protect me from COVID-19? Should I take face-to-face university courses? Should my child go to school?

In the uncertainty of the past few months many of us are trying to make decisions about our health as individuals and society, and weighing risks and benefits. Naturally, friends and family and the public ask these questions from their nearest epidemiologist, socially-distanced of course. But a common joke, or as the New York Times calls it ‘informal motto’, among epidemiologists is that the correct answer is always “Well, it depends”.

Epidemiology is not truly equipped to answer questions of health and disease at the level of the individual. Decisions at the individual level require input from health care providers or public health specialists that know the personal context, such as health history and so on, who can appropriately synthesize and translate the best proven evidence in science. Epidemiologists can’t say with full certainty that a mask will definitely personally protect a person at all times. But evidence so far has shown that if many people wear masks, the probability of the transmission of the COVID-19 on average in the population will be reduced, which by extension protects every person.

What epidemiologists can’t answer is should I personally take face-to-face courses? Well, it depends on many population and individual variables: for example, is the disease outbreak controlled locally, does the university have a concrete safety plan, are there any personal or family health concerns and so on. What epidemiologists can answer is can universities offer face-to-face courses in a specific community, with specific population-level safety measures, without increasing the probability of an outbreak.

What to do with that information lies in the hands of public health officials, governance structures and administrators, and of course individuals in the system. Unfortunately, in the COVID-19 pandemic, this has been a source of frustration for many epidemiologists who have felt that their science has been vilified or ignored at times. But the science of epidemiology has never been a smooth ride. It was founded and solidified in the furnace of some of the world’s deadliest pandemics. Though we have come to encompass more— epidemiologists still play an important role in the face of pandemics. We work closely with other fields, including but not limited to biostatisticians, basic scientists, clinicians, public health experts—and on occasion even dermatologists and entomologists—to act as disease detectives to help improve human health.

MARZIEH GHIASI is an MD/PhD epidemiology trainee at Michigan State University. Her current research is focused on the genetic epidemiology of gynecologic disease, with a background in airborne infectious disease transmission and environmental health.

Image credits:

Coronavirus dashboard: by Markus Spiske, public license

Outbreak response: CDC Public Health Image Library, public domain

Broad Street pump: by Toby Bradbury, creative commons 2.0

Masks: Creative Commons

Epidemiological curves: London Bureau of Hygiene and Tropical Diseases, public domain

Hierarchy of evidence: by Marzieh Ghiasi, creative commons CC2.0

145 views0 comments


bottom of page