Thursday, July 31, 2008

Medical Writing: Reporting About Confounders

You have heard it in the radio or watched on TV. The news reporter is talking about the latest study that shows that video games are related to school violence. That's it, no more details about the study population, or the possible confounders. Just the conclusion of the study in sound-bite format.

Should you just accept the conclusion of that study at face value? If you're a savvy medical writer, you know that the conclusions in epidemiological studies may be affected by what is known as confounders. What are confounders?

Confounders (or “confounding variables”) are factors that are independently associated with a risk: they can make it appear as though two unrelated things are actually related. Epidemiology statistical tests can control for these factors, as long as the factors are measured during the study. Common confounders that are controlled for in epidemiology are age, gender, socioeconomic status, and lifestyle factors such a whether a participant is a smoker.

For example, suppose a researcher was interested in the relationship between gender and common heart attack symptoms. The researcher might see that women have a different symptom pattern than men and conclude that there are heart attack pathology differences between men and women. However, a closer look at the data might reveal that women were an average of 15 years older than men when they experienced their first heart attack. This finding would be important, because age has been associated with uncommon heart attack symptoms in past research studies. Therefore, it would be possible that the older age of the women, rather than a difference in disease pathology, was responsible for the differences in symptoms among men and women (i.e., age is a potential confounder).

When writing a paper or reporting about a study, make sure that the researcher has collected as much information on possible confounders as possible and has controlled for them. Otherwise, the conclusions of the study would not be valid.

Click here if you want to learn more about epidemiology and biostatistics.

No comments: