Tareq Kheirbek, MD ScM FACS
Choosing the design of your study depends on when you observe your patients, whether you follow them forward in the future, or collect data on them from the past, and whether you’re interested in comparing exposures or outcomes.
In the illustration above, individuals there is a group of individuals without any exposure or outcome. Eventually, some of them gets exposed to an intervention (red) and others don’t. Some of the exposed individuals will develop an outcome (circle) and some of the unexposed will also develop the outcome. The research question is whether there is a relationship between exposure and outcome. We are attempting to investigate whether in this group of individuals (and eventually in other similar groups in repeated experiments), the outcome occurs more frequently in the exposed individuals than in the unexposed. We are also going to explore if this different in rates of outcome is significant enough to occur more than random.
Prospective Cohort Study:
If you start observing the group of individuals prior to exposure and you follow them prospectively as you are collecting information that would reveal whether an outcome has developed then you are conducting a prospective cohort study. This is a very powerful form of research studies, arguably better than a prospective cohort study! This can be lengthy if the disease process is slow. It also requires patient’s consent to be followed prospectively and their data to be collected. It also risks loss of patients’ follow up.
Prospective Controlled Trial:
If you’re conducting a study where you follow patient’s prospectively to see if a certain exposure is associated with a specific outcome, BUT you are controlling the kind of exposure and the timing of it, then you’re conducting a prospective controlled trial. While this has been the gold standard for evidence-based medicine for a long time, its biggest drawback is limited generalizability and narrow interpretation. Of course, the consent process here is more complicated since you are “experimenting” on patients and generally approaching this with no knowledge whether the exposure is going to work or not, or whether it is harmful or not.
Retrospective Cohort Study:
If you obtain data on a cohort of patients who were exposed to an intervention/ disease/ environmental variable at some point in the past and you are exploring whether that exposure resulted in an outcome then you’re conducting a retrospective cohort study. You are still interested in the exposure and following the patient’s course from exposure forward. You are just collecting the data retrospectively. Most of the clinical research is conducted this way. There are many biases associated with this design, especially information bias. You are going to deal with a lot of missing data and you cannot correct that because you are collecting data from the past, so what was not measured or collected is not there. Of course there are statistical ways to deal with missing data, but that’s for another day to discuss
Case Control Study:
This is similar to retrospective cohort study from a time schedule standpoint. You are still collecting data retrospectively on your cohort of patients, but here you are interested in the outcome. You are comparing those who had the outcome to those who did not, then you are trying to see if a specific exposure in the past was associated with this outcome.
Confused?
The vast majority of your studies are going to be retrospective cohort studies or case control studies. So which one is which? If you are comparing a specific exposure and want to see if an outcome developed as a result of this exposure this a retrospective cohort study. This is used when the exposure is relatively rare. That way you have enough patients in the intervention group for your study, and therefore you should have enough rate of outcome to evaluate. If you are comparing patients with a specific outcome to others who do not have the outcome and want to see if an exposure is linked to this outcome then you are conducting a case control study. This is done when your outcome of interest is rare. That way you can focus on getting enough patients with the outcome to evaluate the relationship to exposure of interest.
Examples:
If you are interested in evaluating whether ventilator support after laparoscopic cholecystectomy is associated with increased rates of wound dehiscence, you would realize that ventilator support is not common after lap chole and therefore you should conduct a retrospective (or prospective, but this study would take a loooong time!!) cohort study. That way when you define your exposure group you’d find enough people who got intubated and compare them to those who did not and look for rates of dehiscence.
On the other hand, if you are interested in rates of mortality after a lipoma resection and whether performing the procedure in ASC versus Davol OR was a factor, you would recognize that death after lipoma is very rare and therefore you want to have enough patients with the outcome (cases) and then have a control group of those who did not die, then retrospectively compare where the surgery took place.
Surveys and Cross-sectional Studies:
In these studies you are collecting the information on “exposure” and “outcome” at the same time. You cannot determine whether the exposure preceded the outcome. You are basically determining the relationship between two variables at a single timepoint. For example, you ask your patients in the surgery clinic whether they feel depressed that morning and whether they took their pain medications prior to presenting to the clinic. You might find a relationship between the two, but you CANNOT conclude that not taking pain medications resulted in depression or that being depressed resulted in not taking pain medication. There is simply a relationship. That is it! Nothing casual!