4. Research Questions & Hypothesis

The true sign of intelligence is not knowledge but imagination’ (Albert Einstein)

Einstein highlighted in one simple sentence that human intelligence is not what you know, but it is having the ability to be creative and to find a solution by making connections between ideas or objects.  For those who are new to research methodologies this brief piece is intended to further guide the reader in how to formulate a research question and/or a hypothesis.

As discussed in the previous newsletter (3. Literature Reviews), having performed the necessary background reading and literature review, the PICO model (Population, Intervention, Control/Comparator, Outcome), is a simple way of focusing and narrowing the research question so that it stays clinically relevant.  When choosing a topic to research, I would strongly advise that you either choose a population or an intervention that you are drawn to, and/or look at what expertise is available to you within your group and department.  If you are not sure on what topic to choose, an article in BMC Research Methodology by Munn et al (Munn et al., 2018) has a wealth of information on the different methods used in the health sciences.

Answering the PICO model will make sure the research question remains clear. The population and who the study is relevant to is first defined.  This is followed by a brief definition of the intervention that is typically what your hypothesis or main variable of interest will be.  For example, patients with cerebral small vessel disease (CSVD) would be the population, and whether a new vasoactive compound could improve cerebral circulation would be the intervention.  As a comparator, this can easily be a healthy population. However matching factors such as age, sex and/or disease can often provide better impact and clinical relevance to the study outcomes. For example, comparing the effectiveness of the drug in different severity of CSVD, or in an older cohort of patients with known cardiovascular disease but who do not have any history of stroke/TIA.  The outcomes and the main focus of the study will decide whether the research worked. The outcomes will be related to what is expected to change, or what is important to the participant and population being studied.  Being ambiguous about the factor that is being used to link the exposure and outcome is one of the most common errors in formulating a research question (Mayo et al., 2013).  Thus, when the question is poorly formulated, the design, analysis, sample size calculations, and presentation of results may not be optimal.

Depending on which model you follow there are several types of research questions that can be asked.  I will briefly highlight quantitative research questions that rely on the quantity of the data (for examples in qualitative research, where the data is interpreted by its quality see (Agee, 2009) and (Korstjens & Moser, 2017)).  Quantitative research questions are often grouped (and influenced) by the methodology and will explain a phenomenon by using statistics to analyse the numerical data.  They will either describe what is going on by using descriptive questions such as ‘how much’, explore the relationship between a variable (a factor, trait or condition etc.), or be comparative in saying ‘what is the difference’.  For suggestions on how to structure these research questions please see this online article: https://dissertation.laerd.com/how-to-structure-quantitative-research-questions-p2.php

The top 10s of priorities for research (https://www.jla.nihr.ac.uk/top-10-priorities) have been formulated for many areas of research.  With these questions having recently (in 2021) being prioritised for a whole range of vascular conditions (https://www.jla.nihr.ac.uk/priority-setting-partnerships/Vascular/vascular-conditions-top-10s.htm), I would first advise every study and funding proposal to consider how their research questions compare to this body of work.  If your research question does lie outside those that are outlined (as there are more than 10 questions to be answered in the world), make sure you clearly justify the purpose and the benefit to patients that your question will address.  The top three priorities for carotid disease are below and include an example for a question that is descriptive, relationship and comparative:

1. Can doctors accurately predict which people with carotid artery disease are most at risk of a stroke? - Descriptive
2. Is there an association between carotid disease and cognitive decline? - Relationship
3. What is the optimal management of patients with carotid disease using individualised risk benefit ratios? - Comparative

Proposing a hypothesis to a research question can be a little more challenging.  A hypothesis will predict a specific event and will need to be written and tested accordingly.  I know Christmas is still a long way away, however the comic strip below by phdcomics.com (https://phdcomics.com/comics/archive_print.php?comicid=1457) gives a nice example of inductive reasoning, a scientific conclusion that is based on a series of observations.

With deductive reasoning, a single counter-instance shows a general claim to be false.  Karl Popper said that we can never prove something with certainty, we can only disprove with certainty (commonly referred to as Theory of Falsification, see (Popper, 2005)).  Thus, science ‘mainly’ seeks to refute conjectures, not to verify them (although see the demarcation problem and other historical/psychology theories for examples of when this theory does not hold true in other sciences (Nickles, 2013)).  Nevertheless, we cannot claim a general hypothesis is true by observing particular cases, but we can claim it is false by observing particular cases.  As scientists the aim of the game is to nullify a hypothesis that is based on the current scientific paradigm and propose an alternative hypothesis that will suggest an alternative theory to explain why the null hypothesis is not true.  This alternative theory will relate to the research question that we are interested in.  An example of a simple null (H₀) and alternative (H₁) hypothesis would be -

H₀ – drug X does not reduce pulsations in blood flow to the brain

H₁ – drug X does reduce pulsations in blood flow to the brain

It is now common in research articles only to state the alternative hypothesis.  But with these two statements significance testing can now be used to determine whether there is a difference between these populations (more on that in the statistics newsletter).  But it is now up to us to devise an experiment with suitable methodology that will test and exclude one of these hypotheses (see (Webb et al., 2021) for a suitable methodology that would test the hypothesis above).

The best way of understanding how a scientific hypothesis should be proposed and tested accordingly is to read a PhD thesis (or part of it at least). Focus on the statements proposed in the aims and hypothesis, prior to the methodology section.  You can download nearly any thesis within the U.K. that is open access from this website (https://ethos.bl.uk/Home.do;jsessionid=AC45FEB5C6A73BD97B93E81F1E0F80DC).  The introduction and background chapter will demonstrate all the relevant knowledge and theories that has led to why the research question and general hypothesis has been proposed.  Each subsequent chapter will then be focused in answering a specific set of questions related to the hypothesis.  The limitations and future directions chapter will offer insight on what can be done next.

Generating a research question is usually the easiest part of a study, it is much harder to obtain funding and complete a project.  However, there is no need to re-invent the wheel.  It will take time to put an idea in to practice but keep at it. And to finish off with Einstein, ‘The only reason for time is so that everything doesn't happen at once’.  Good luck.

Osian Llwyd

Vascular Studies Unit, Oxford University Hospitals

osian.llwyd@ouh.nhs.uk

References

Agee, J. (2009). Developing qualitative research questions: a reflective process. International Journal of Qualitative Studies in Education, 22(4), 431–447. https://doi.org/10.1080/09518390902736512

Korstjens, I., & Moser, A. (2017). Series: Practical guidance to qualitative research. Part 2: Context, research questions and designs. The European Journal of General Practice, 23(1), 274–279. https://doi.org/10.1080/13814788.2017.1375090

Mayo, N. E., Asano, M., & Barbic, S. P. (2013). When is a research question not a research question? Journal of Rehabilitation Medicine, 45(6), 513–518. https://doi.org/10.2340/16501977-1150

Munn, Z., Stern, C., Aromataris, E., Lockwood, C., & Jordan, Z. (2018). What kind of systematic review should i conduct? A proposed typology and guidance for systematic reviewers in the medical and health sciences. BMC Medical Research Methodology, 18(1), 1–9. https://doi.org/10.1186/s12874-017-0468-4

Nickles, T. (2013). The Problem of Demarcation. History and Future. Philosophy of Pseudoscience. Reconsidering the Demarcation Problem, Bauer 2001, 101–120.

Popper, K. (2005). The Logic of Scientific Discovery. In J. Shand (Ed.), Central Works of Philosophy Volume 4. Routledge. https://doi.org/10.4324/9780203994627

Webb, A., Werring, D., Dawson, J., Rothman, A., Lawson, A., & Wartolowska, K. (2021). Design of a randomised, double-blind, crossover, placebo-controlled trial of effects of sildenafil on cerebrovascular function in small vessel disease: Oxford haemodynamic adaptation to reduce pulsatility trial (OxHARP). European Stroke Journal. https://doi.org/10.1177/23969873211026698

Links

How to structure a research question https://dissertation.laerd.com/how-to-structure-quantitative-research-questions-p2.php

Top 10s of priorities for research https://www.jla.nihr.ac.uk/top-10-priorities

 Vascular conditions, Priority for research https://www.jla.nihr.ac.uk/priority-setting-partnerships/Vascular/vascular-conditions-top-10s.htm

PhD Comics https://phdcomics.com/comics/archive_print.php?comicid=1457

PhD Thesis web search  https://ethos.bl.uk/Home.do;jsessionid=AC45FEB5C6A73BD97B93E81F1E0F80DC).