Inappropriate study design

From The Embassy of Good Science
Revision as of 16:56, 26 March 2021 by Benjamin.Benzon (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Inappropriate study design

What is this about?

This short article will talk about the experimental or study design, its logic and its appropriate use as well as its misuse.

Why is this important?

Addressing any question in empirical sciences implies that there will be some kind of study, which needs to be designed in such fashion that it answers the question in the most unambiguous way. Scientific questions can usually fall into one of these categories: identification of influencing factors (etiology and risk factors), predicting power of those factors (prognosis and prediction), effects of interventions into those factors. To answer these questions, factors that are measured or manipulated must correspond to questions that are asked. Meaning, your outcome measure has to be relevant for your question. Preferably, it should answer your question directly. The design of your study also has to fit the outcome measure. Furthermore, if there are some factors that we are unaware of or others that are known to us but we cannot measure them, then we have to use randomization to reduce their possible influence on things that we want to measure. If you do not address things that are listed above when planning research, you can simply get wrong answers to your questions . [1]

  1. Glasziou P, Heneghan C A spotter’s guide to study designs BMJ Evidence-Based Medicine 2009;14:37-38.

For whom is this important?

What are the best practices?

Observational studies, such as cohort or case – control studies, are sometimes overinterpreted in terms of cause-effect relationship. Correlation between a factor and an outcome does not necessarily mean causation. When it comes to experimental studies, sometimes randomization is not possible due to ethical reasons which should be taken in account when interpreting results of such studies. [1] Sometimes outcome measures do not correspond completely to questions asked in the study i.e. they are only indirectly connected.

All of this is usually addressed in research methodology classes. When planning, doing and reporting research, you can always rely on appropriate EQUATOR reporting guidelines to make sure you have everything accounted for.

  1. Colli A, Pagliaro L, Duca P. The ethical problem of randomization. Intern Emerg Med. 2014 Oct;9(7):799-804.

Other information

Virtues & Values
Good Practices & Misconduct