In this video, get introduced to the 2x2 table as a framework for applying study designs and understanding the different measures of association.
- [Instructor] Congratulations, you made it to chapter three where we dive into talking about one of the main concepts in the field of epidemiology, the two by two table. So far in this corse you have learned that you need a hypothesis that defines an exposure, which is a proposed cause and an outcome that supposedly gets caused by the exposure. The whole study is about determining how exposure and outcome are related. And how you do that depends on the study design you select. So the two by two table helps guide you as to how you will look at that measure of association depending on your study design.
So, I have to emphasize again that the two by two table is completely conceptual. It's like a guide to help, it's not literal. It makes a lot of assumptions. One of them is that you have to imagine both your exposure and outcome are binary. Think about a continuous exposure, like household income and a continuous outcome, like money spent per week on cafe lattes. We have to pretend for the sake of the two by two table philosophy that both of these are binary. Okay, great, now I'm craving a cafe latte.
Feeling artistic? Pull out a pice of paper and a pen. You know, those analog instruments they used to use back in the early 2000s. Draw the shape I have on the screen. This is your blank two by two table. Remember, we are trying to relate exposure with outcome. In this old fashioned diagram we will used the term 'disease' instead of 'outcome'. So the next step in your drawing is to put the E+ to the left of top row of boxes and the D+ over the first column. These ancient inscriptions indicate that what you are going to put in the box where they intersect is the number of people in your study who have both the exposure and also had the disease, or outcome.
So naturally, we put E- and D- in their respective places next. Then, we label the cell. A for the first cell we define, that double-whammy of both having the exposure and the outcome. You'll see cell D is the double-whammy, the other one. Those have neither the exposure nor the disease. A and D are then called concordance cells because they agree. B and C cells are discordant because in B the participants have the exposure and not the outcome and in C it's vice versa.
We can even define the marginals using these letters. The row totals are known as A+B and C+D, respectively. And the column totals are A+C and B+D. Finally, the grand total is, you guessed it, A+B+C+D. Now that we have that filled in, let me introduce you to the term we use called relative risk. What relative risk refers to is the risk of one group having a certain status relative to another group. So here's an example.
The risk of the outcome in the exposed compared to the unexposed would basically compare what's going on in the top row, the exposed, compared to what's going on in the bottom row, with the unexposed. Whereas assessing the relative risk of the exposure in those with the outcome, compared to those without the outcome would deal with what's going on in the left column compared to the right column. Relative risk we are talking about from this table is actually another term for measure of association.
So as a generic approach to assessing relative risk we make ratios of one group to another group. If the ratio is greater than 1, the group we put on the top of the ratio has more risk relative to one on the bottom. And that relative risk can go as high as, well, theoretically, infinity. But, if in the ratio, downstairs is bigger than upstairs, your ratio is going to be less than one. In fact, it can go really low, approaching zero. But, if in the end the relative risk is close to one, well, there's basically no relationship between the exposure and outcome.
But if you find there is no relationship between the exposure and outcome, don't be sad! Turn your frown upside down. I've gotten a relative risk close to one before, and I admit, it's disappointing. But I cope by just spilling my guts in the write-up and telling the truth. Which is, I was wrong about my hypothesis. And then moving onto the next study.
- What is epidemiology?
- Study design overview: descriptive, analytic, cross-sectional, and case control
- Planning a study
- Planning the analytic data set
- Analytic data set requirements