Identifying Methodological Flaws in Studies
- May 22, 2019
- MCAT Blog, MCAT Chemistry, MCAT Organic, MCAT Physics
One of the things that really surprises students about the MCAT is the heavy focus on how experiments are set up and ran. This can really catch a student off-guard, as most undergrad courses don’t really emphasize this. But from the point of view of the AAMC, it’s obvious that they want to test this. In order to be a good doctor, you need to be able to understand the quickly changing research that is coming from our more experimental-minded colleagues. It’s incredibly important for doctors to know when a certain treatment is likely to help, and when it may make matters worse.
Take for example two factory workers, one that works on metal and one that works on wood. A certain drug may seem like it makes sense to give to both, but it may actually be a mistake to give that medication to the woodworker. Perhaps the drug shows some side effects for the respiratory system and the woodworker is much more likely to be affected due to the sawdust that he is constantly breathing in.
The AAMC wants medical students that are able to understand data, and perhaps more importantly, understand when data was collected improperly. There are many things that can throw off a study, and the MCAT loves to test these ideas. It can be quite tricky at times as many of the terms the MCAT uses may overlap! Also adding to the difficulty is understanding that some of these issues can be fatal to a study, and other issues are just an unavoidable part of research. A top MCAT prep course should devote some time to review these concepts. Let’s talk about some of the most common terms that the MCAT throws around:
Bias: Researcher, selection, attrition
- Researcher bias occurs when the scientist performing the data skews the data (perhaps even subconsciously!) because they are looking for a certain result. This is a difficult thing to combat if the researchers know what people are in the experimental vs the control group. The solution? A double-blind study of course! Now the people and the researchers both do not know what group people are in. No way to skew the results that way!
- Selection bias is also combatted by performing a double-blind study. Let’s say you are setting up a study testing your new weight loss pill. You have two patients that want to join the study, but one of them seems like they REALLY want to lose weight. If you get to pick which person goes into which study, you’re likely to choose the more dedicated person to be in the experimental group. There’s a decent chance that will throw off your data. Basically, when researchers get to decide which group the test subjects go into, it has the opportunity to create selection bias.
- Attrition bias is a little different. Let’s stick with the weight loss pill analogy to really understand this one. Imagine that you are creating a study for your weight loss pill, but you want to watch what happens to people over the course of 2 years. So you set up your control and experimental group, and you ask them to come in every day for 2 years. Naturally, some people are going to drop out of the study over that time period. The problem? People who are losing weight and feeling healthier are much more likely to continue coming in once a week. If you had 100 people on the experimental drug and it only caused 20 people to lose weight, there is a good chance the other 80 may have bailed on the study. As a result, it looks like 100% of your experimental group lost weight, but really it was 20%.
Validity: Internal, external, and face
- Validity is a nice pass/fail measurement. Either something is valid, or it isn’t. An internally valid study is basically a study that is done well. To meet that criteria, you need to know that the samples you are observing and the conditions that you are measuring are collected in way that demonstrates causation. (If you are testing a blood pressure medication, ideally you want your control and experimental groups to start off similar and have some consistency in the way you take measurements)
- It is entirely possible that you set up a study well and it is internally valid, but it doesn’t apply to the population at large, which is the measure of external validity. For example, if you are studying a rare disease that only occurs in one person, any research you pull from that may not be externally valid.
- Face validity is a measure of “does this experiment sound right” on the surface. If I wanted to study how healthy people in a certain county are, counting the number of hospitals isn’t really a good way to measure that. It fails face validity.
- Another issue that plagues human studies is the Hawthorne effect. Basically, if a person knows that they are being watched, they will behave differently. If I tell a patient that I must weigh them every week for a year over the course of the study, the fact that they are constantly being weighed may make them more likely to exercise or skip that second donut. Having a robust control group can help, but since it is illegal and unethical to do medical research on humans without informing them, there really is no way to completely get rid of the Hawthorne effect.
These topics show up all the time on the MCAT (especially the psychology and sociology section). Understanding these ideas will not only help you with the MCAT, but it’s going make you a better doctor in the long run as well.
Search the Blog
Free ConsultationSchedule Now
MCAT Blog How to Review MCAT Full-Lengths
Free MCAT Practice AccountNeed great MCAT practice?
Get the most representative MCAT practice possible when you sign up for our free MCAT Account, which includes a half-length diagnostic exam and one of our full-length MCAT practice exams.Learn More