Common Terms and Concepts in Epidemiology

Okay, so over the last few weeks, we looked at incidence and prevalence, and we also looked at how prevalence relates to screening tests. Hopefully, you now know why prevalence of a disease under control can continue to rise (because people are surviving). You also know why incidence will continue to rise even if you have the exact same number of new cases year after year (because the number of people at risk declines). And you hopefully know why you can’t go screening everyone in the population at random for diseases using screening tests that are less than perfect (because you’ll have a low prevalence and high number of false-positives).

For today’s lesson, I decided to keep it a little bit simpler and just go through some common terms used in epidemiology and define them with simple examples. So let’s do that after the jump…

If she hits ten three-pointers in a row, is it statistically significant?

MORBIDITY AND MORTALITY
These are two concepts that sometimes get mixed up. Morbidity relates to being sick or afflicted with a condition. Incidence and prevalence are both measures of morbidity. That is, they measure how many people are sick with a disease or suffering from a condition over time (per year) or at a given time (today).

Another measure of morbidity is the attack rate (AR). The AR answers the question: Of all of those exposed, how many became sick? This measure comes in handy when the exposures are many, like in a food-borne outbreak. Take a look at this example:

As you can see, plenty of people ate the different foods (or combination of foods). But those who ate the lettuce and the tomato had the largest ARs. That gives us a clue as to what made people sick at that imaginary party. We’ll look closer at AR and how to take it into account in an outbreak investigation when we look into outbreak investigations.

On the other hand, mortality measures how many people die from the disease or condition. It answers the question: Of all who were infected or suffering from the condition, how many died? It is often expressed as a proportion and often called a “death rate.”

Death rates can help us determine how bad something is, of course. For example, in the early days of the H1N1 pandemic, none of us knew how bad it was going to be. As the weeks went by and we started to count deaths out of total infections, we knew that it was going to be as bad as any other flu season… Except that, of course, it came just after a flu season. For the purpose of epidemiology of diseases, we look at the death rate of those who have contracted a disease. We then compare that to the “crude” death rate (the death rate from all causes) to know if a disease is worse than, you know, living.

STATISTICAL SIGNIFICANCE
One of the things that is touted a lot when talking about a study is the “statistical significance” of the findings. “Is it statistically significant?” is one of the first questions you should ask yourself when looking at a study. Of course, there are other questions, but we’ll address them in our final class, where we review a study.

But what is “statistical significance”? Let’s do another mental exercise. Let’s imagine that we have a bag with 100 balls inside. Of those 100 balls, ten are red and the rest are blue. Simple math will tell you that you have a 10 in 100 (or 1 in 10) chance of drawing a red ball at random. So you draw one ball, and it’s blue. You draw another, and it’s blue. Then you draw another, and it’s blue… What if you draw 20 balls and they’re all blue? What are the odds of that?

Likewise, what if you draw five red balls in a row? What are the odds of that? Gut intuition would tell you that drawing so many red balls in a row is kind of weird because the total number of red balls is so low. (In fact, my $1 calculator didn’t have enough digits to calculate the odds of that. How to calculate that is for a class on statistics, not epidemiology.) Drawing 20 blue balls in a row is not as “weird” because, you reason, there are plenty of blue balls in the bag. (Insert joke about a little town in Pennsylvania here.) But how do you know it's common to draw 20 in a row?

GUT INSTINCT
Like I’ve said before, there is no such thing and “gut epidemiology.” Sure, we can take educated guesses based on our gut, but those guesses are less likely to get you in trouble and more likely to be correct if they are indeed “educated.” What we need to do to avoid getting into trouble and to look at our results objectively is to run our findings by a statistical test. Those statistical tests will tell us the likelihood of seeing the observed results by chance alone. A ton of factors go into that likelihood. Things like the sample size, randomization of subjects, etcetera.

DOUBLE-BLIND
In study design, when we say that a study is “double-blind” (not double-bling, as my phone auto-corrected), what we are saying is that neither the subjects nor the researchers know who is in the treatment/intervention group and who is in the control group. It is hoped that this will reduce bias in the study by eliminating any pre-conceived notions about the treatment or intervention. It may come as a shock to you that researchers would be biased. They are. We’re all human.

IN CONCLUSION
Of course, these are not all the epidemiological terms that you’ll need as a budding epidemiologist, but they’re a start. Let me know what other terms you have questions about, and I’ll try to address them later in the “night school.”

Powered by ScribeFire.