Asking a Research Question

Tareq Kheirbek, MD ScM FACS

What makes a good research question?

If you are curious about your surroundings you will ask questions. In medicine you SHOULD have a lot of questions as you progress through training or learn new skills. Most of these questions will have some sort of an answer out there. You should learn how to find these answers so you don’t waste your time repeating studies. Sometimes, the answers you find have limitations or drawbacks and this generate more questions. Once you find an interesting question that you believe have not be adequately answered, then you have a research question to work with. Your research question has to be:

  • Ethical
  • Interesting
  • Clear
  • Focused
  • Simple vs complex
  • Testable
  • Relevant (You’re answering this question to improve patient’s care, NOT to make your CV longer!)
  • Feasible/achievable/doable

Most of my clinical questions early on have been in a “why ” format. Then I started asking “what if” questions. An important advise I got from a seasoned trauma surgeon was “not to confuse work frustrations with research questions”. We tend to believe in one opinion strongly and as a result disagree with others. As humans, we then make every effort to prove the other wrong. We therefore propose a research question powered with emotional drive to prove the opinion we already have. This in fact depletes the research question of equipoise – which is an essential element in scientific research.

In order to come up with a good question think of a recent question where surgeons on the team disagreed on the next step or the prognosis, or where a surgeon couldn’t easily make a decision. Did that make you curious as to what is the best approach to the clinical problem? Does answering this question make the care of your patient better? Then, look up what is out there in regard to this question. Is there any publication on the topic? Any opinions out there that could guide you in how to formulate your question?

Now that you have a question and you have refined it, it is time to define its components correctly and specifically using PICO:

  • Population: Who are you studying? Do you need to make this narrow (less confounders, but less generalizability) or wide (more generalizibility, but many confounders to control for)?
  • Intervention: What is the thing your interesting in studying? This could be a medication, a surgical approach, a new protocol, a new technology, a demographic characteristics, etc.
  • Control: what are you comparing your intervention to? This is usually the standard of care that is already in place. It is something that you can get information about from the literature to help you anticipate the effect of the intervention. For example, if you are asking whether a new procedure (I) in a specific patient population (P) results in a shorter duration of mechanical ventilation (O) you should at least know whether the current standard of care (C) results in need for mechanical ventilation and how long does that typically last. This information will become available as you’re doing the literature search prior to designing your study.
  • Outcome: The outcome of interest should be simple and single. You should have one outcome only that you’re interested in. The power of your study will rely on that one single outcome of clinical interest. Dr Benoit gave a very interesting talk on using mortality as an outcome in clinical research. Mortality is an easy target, but it is not always relevant to patient’s care. Choose an outcome that will be physiologically related to the exposure (antibiotics –> less infection, physical therapy –> early discharge home, etc). If you have other outcomes in mind they are secondary outcomes. You should not rely on these outcomes to make a conclusion from your study. Secondary outcomes should only serve as hypothesis generators for future work. You are conducting this study to answer a question you developed a priori, not fish for random relationships you come across as you’re analyzing multiple outcomes and exposures!.

The four questions you should explore as you are answering your research question or hypothesis:

  1. Is there a relationship between the exposure and the outcome?
  2. Is the observed relationship between exposure and outcome casual?
  3. What concepts explain the exposure-outcome relationship?
  4. How representative is the exposure-outcome relationship across persons, settings, times?

Once you have completed all the above, now it is time to consider your data sources and your study design.

Below is a good chapter on conceiving the research question:

[embeddoc url=”http://kheirbek.com/wp-content/uploads/2019/11/HUlley-Conceiving-the-research-question-1.pdf” download=”all” viewer=”google”]