| | | |

Demystifying Medical Research.

"Owl," said Rabbit shortly, "you and I have brains. The others have fluff. If there is any thinking to be done in this Forest--and when I say thinking I mean thinking--you and I must do it."
From a conversation between two great intellectuals in The House At Pooh Corner.

It can be a little intimidating at first when you start looking at medical research, but there are a number of good reasons to go straight to the source:

• Popular media reports do not give you the full story.
• Results are taken out of the context of the existing information base.
• Reports often focus on one aspect of the study, which may or may not have been the object of the study.
• Media reports often do not follow up the story when additional research is completed.
• Popular media only covers what they perceive as important or dramatic.
• Popular media often reports results without discussion from analysts who are knowledgeable and open-minded.
• Doctors often don't have the time (and some of them, the inclination) to read research.
• Your Aunt Sally has limited expertise.

It's not too hard to delve into the medical literature yourself, you just need a few tools:

• Access to a medical library at a large library, hospital or a college that has a medical/nursing program.
• A good medical dictionary to help you with unfamiliar terms (there will be lots of them).
• A primer on reading research reports.

I can't help you with the first two items; you'll need to find them on your own, but I can help you understand what it is that you're reading and whether it really means anything.

Anatomy of a Research Paper.

A research paper usually starts out with an abstract, which is a summary of the results.

• Read the abstract to find out if the article contains information that is useful to you; if it doesn't you can quickly eliminate the paper without having to read the whole thing.
• There are publications at medical libraries that are simply compiled abstracts. You can collect the references and then look up the whole report.
• Don't just read abstracts! They don't give you all the information. It's sort of like trying to write a book report from Cliff Notes--you get the basics but you miss all the fine detail.

The next part of the paper usually is a history (which sometimes is rather long) giving you a synopsis of current thought in the field and results of previous research. It also tells why the researchers felt the study was necessary.

Then we go on to the methodology section, which details how the study was designed and what the researchers attempted to accomplish. This can be a very confusing section, as here is where you'll find all the statistics-related terminology, but it's important to look at to see if this was a well-designed study. This reflects on whether the results were valid or reliable, which we'll cover in a bit.

Then there is a results section, often with lots of handy tables of information. Here again, read the text! All the information cannot possibly fit in a table. This is the section with the most medical terminology, so have that dictionary handy!

The next section is a discussion or comment section, where the researchers discuss the results and where they feel additional research is needed. There's almost always more research needed!

Finally, there is a bibliography which lists other studies that the researchers consulted.

[top]

Important Factors in Research.

How was the research done?.

• validity--is a measure of how well the study is designed to find what it is looking for. If you design a study to measure the weight of elephants, but the design calls for the use of a measuring stick, then the results are not valid.
• good reliability means that someone else could do a similiarly designed study and come out with comparable results. An example of good reliability is with studies on continuous electronic fetal monitoring in labor. All studies showed no benefit to babies but did show an increase in cesarean sections.

Types of Research Studies.

• Retrospective studies look at things that have already happened.
  • Researchers look back over medical records or interview people and compare care or circumstances before the outcome with the outcome itself.
  • The researchers try to match the populations that experienced the different outcomes, that is they try to make them as similiar as possible by selecting participants from the same age, socio-economic status, or other factors.
• Qualitative studies collect information that reflects people's opinion or satisfaction with the study item.
  • This information is often gained by use of questionaires and interviews.
• Meta-analyses are studies that look at a number of independent studies on the same topic to decide if the results of these studies can be generalized to be meaningful to larger groups of people.
• Quasi-experimental are studies that are almost true experimental studies (see below) but don't quite make it.
  • The researcher couldn't meet all of the criteria for the true experimental design, for instance using a convenience sample instead of a random sample.
  • The limitations are explained in the methodology section of the paper.

The final type gets a section all to itself:

True Experimental Design.

The true experimental design has specific characteristics:

• the subjects are representative, that means that if you are studying a pregnancy condition you look at pregnant women, not elderly men;
• research is done to determine current findings in the field, or a test study is done to collect preliminary data;
• it features a control group, which is a group that isn't exposed to the experimental variable (ie. they take a placebo pill instead of the medicine);
• subjects are randomly assigned to the control group or the experimental group (this may be done by assigning every other person to the experimental group or even by a coin toss);
• lastly, it contains an observation of the results; that is, you don't give the medicine and then not stick around to see what happens.

Some types of true experiments are:

• randomized controlled trials (RCT) which is the basic form of the experimental study; and:
• double-blind studies, in which neither the researchers nor the subjects know whether the subjects are in the experimental or control group.

If the caregiver knows what group the subject is in they may unknowingly (or knowingly!) alter the care they give and therefore bias the results. An example is a study that was done to see if a treatment for gestational diabetes effected the cesarean rate. The problem was that the doctors delivering the babies knew whether the mother was in the experimental or control group and if the doctor believed the treatment was effective, he was less quick to do a cesarean, thereby decreasing the cesearean rate for that group.

Study Variables.

The things research studies are interested in are called variables. Variables is the term used for both the thing that is suspected of causing a particular result, and the result itself. Say you're wanting to study how light periods of different length effect the germination of a certain type of seed. You make sure all other factors (moisture, type of seed, growing medium, temperature) are all the same. So the variables are the length of light (in this case the independent variable because it is altered on purpose) and the rate of germination (the dependent variable since it depends on the amount of light).

[top]

Choosing Participants.

There are several ways researchers select participants for studies:

• random assignment (the creme de la creme of selection processes), as I mentioned above;
• random sampling--used when the subjects aren't divided into control or experimental groups, this is like telephone polling--the larger the sample the more valid the results;
• convenience sampling--the subjects are convenient to the researcher, for instance her patients (this can create a bias--more on this later).

Probability and Significance.

If you toss a coin an endless number of times, statistically you have a chance of it coming up heads 30 times in a row. How do you know that the results of a study weren't arrived at by sheer luck (or lack thereof)? Researchers have ways of determining the probability that results were a product of chance.

The most common way (and I admit the only one I really understand) is satistical significance. You'll see this referred to as the "p value" with a number, such as "p=.01". If you convert the decimal to a percentage, you find that p=1%. This means that the researchers figured out that there was a 1% chance that the data was the result of some sort of fluke in the study. Results are considered statistically significant if the p value is smaller than or equal to .05, or 5%. The way they figure this out is by looking at the raw data and seeing the range among each of the groups. The more tightly packed is the data within a group, the more likely the values are to be significant.

[top]

Establishing Cause.

Okay, so the research has established a link between a particular factor and a result. Does this mean that factor caused the result? Not necessarily. The correlation coefficient describes the relationship of the sets of data to each other, whether an increase in the values in one set is associated with a change in the other set.

• The correlation is assigned a number from +1.00 to -1.00.
• +1.00 indicates a step for step increase in the values.
• 0.00 indicates no relationship between the variables
• -1.00 indicates that as one value moved one step up, the value for the other variable moved one step down.
• The closer the correlation coefficient is to +1.00 or -1.00, the stronger the association between the variables is.

Now, just because the correlation coefficient shows a strong relationship it doesn't mean that one variable caused the other variable. For instance, there is a strong correlation between one's age in years and the value of inflation. Did the fact that you had a birthday effect the rate of inflation? No, of course not. That's why separate studies are done to establish a correlation, then to discover if there is a causal relationship.

Confound It!.

Now that you have all this faith in the protocols used by researchers to ensure accurate results, please excuse me while I deflate your bubble somewhat. These are the things that drive researchers mad because they can mess up a perfectly good study:

• confounding factors;
• selection bias;
• human tendency to alter behaviour when being observed;
• reporting bias.

These terms are used to describe factors that can invalidate and alter the results of a study. There may be errors in the way subjects were assigned to the groups, or differences in how the protocol was carried out by various participants (making comparison impossible), or factors may have come up in the course of the study which were not considered when the study was designed.

Say you're studying those germinating seeds again, and someone comes along and turns off a light when it was supposed to be on. That's called a confounding factor--something messed with the variable. How do you tell what the results mean if you don't even know what you started with?

Then there is selection bias--when the group assignments were not random or aren't typical of the population that the findings are applied to. For example, a study that came to the conclusion that gestational diabetes was very dangerous to babies because they found more stillbirths in women with the condition. But there was a selection bias; the women were selected for the study because they had previously suffered a stillbirth and were at higher risk.

Two other factors that may effect results are:

• the tendency of people to act differently because they are part of a study (doctors may alter their routine to encourage the desired result);
• the difficulty in a medical setting of limiting additional procedures that fall outside the study design (usually these cases are excluded from the study).

Last but not least is reporting bias. If you conducted a study that totally contradicted what you an your corportate sponsor expected to find, would you report the findings? Some people don't.

Well, that's my primer. I hope it helps. I'm a total number klutz myself and I figure if I can understand these papers, then anyone can! Normally I'd have a list of links for you to follow to learn more, but my extensive searching found not one site! If you come across one, I'd sure appreciate it if you'd let me know at rugratz3@geocities.com.

Bibliography.

• ICEA. 1994. Teacher Certification Program Study Guides.
• Rathus, S. 1990. Psychology. pp. 689-701.

| | | |

You are here:
http://www.oocities.org/HotSprings/9235/research.html