Epidemiology Research Design

Content:Thanks Lee for giving such a concise overview of the recommendations of the task force. This entire week I have been hearing debates about the U.S. Preventive Services Task Force (USPSTF) recommendations and stirred up the hornet’s nest. Every debate I have been hearing makes me understand the political implications of the recommendations and the brave effort by the taskforce to actually state the findings genuinely. Obviously they do not recommend that women over 40 should not be taking mammograms but only that they make a more informed choice about this rather than being herded to the X-ray room once they hit 40. It is really interesting to hear the other side trying to demonize the taskforce by stating that the taskforce has passed the death knell on mammograms as a tool for effective screening, which is definitely not the case. It is surprising to see how people can take statements out of context and stir up such a racket.
Dr Pettiti’s lecture was a very good over view of epidemiological methods. Understanding design concerns is the first step towards designing a study. I think that having a firm understanding of the various designs (case control, cohort, ecological, randomized control trial) and the biases and confounding factors underlying each of these is the first step towards your dissertation/thesis. Dr Pettiti talked about the Type I and Type II errors and the importance of avoiding Type II errors in research. I would like to add that Type I error or alpha (significance level of study) is also important to consider. We traditionally consider 0.05 as the magical number for setting this error and I have learnt all through my Masters that this is a big mistake. Every study needs to consider the scope of the study, the population being impacted and the outcome measures of the study before setting the alpha level at 0.05. Thus a p=0.05 is not a universal alpha and researchers need to consider other aspects before setting their alpha. The same applied to Type II error or beta (also known as power of the study) and there is nothing magical about 80% (0.8).
Another aspect of every research design is defining the measurement errors, biases and confounding factors beforehand. This is very critical to your study and there are textbook chapters on these. These biases obviously affect the internal and external validity of the study. I highly recommend going through a simple epi text to understand these as these concepts are not limited to epi research but encompass any kind of research study.
I found a very simple and easy online text and hope everyone likes it.
http://www.bmj.com/epidem/epid.html

Posted by Sheetal Shetty