Formulating a research question

I was a junior researcher, and MSc postgraduate student, wanting to write a dissertation's systematic review (SR). I learned that the first step in writing any piece of research, including SR, starts with finding a well-formulated research question^1,2. As a healthcare worker in the neurosurgery field interested in decompressive craniectomy's efficacy in increased intracranial pressure (ICP), I chose to ask a question related to this. Having formulated a focused question I then found nine published systematic reviews on this topic in PubMed and a handful of others in PROSPERO. I became frustrated, faced with a challenging question: 'How can I formulate a new and novel research question that has not been previously asked (in the ongoing systematic reviews) or answered (in published systematic reviews) and suits and supports your area of interest?'

Background

It has been argued that the first example of published SR was in 1753 by James Lind³, and in 1904 Karl Pearson published what we consider the first statistical synthesis of evidence⁴. However, this type of research did not take a recognisable place in healthcare research until 1993 with the foundation of the Cochrane Collaboration^5,6. Since then, SRs have increased exponentially. Moher estimated a decade ago that 2,500 SRs were published yearly⁷, and this rate has increased to 25,000 SRs each year⁸, which means more than 2000 reviews each month on average. So, what has been published yearly is being published monthly, and we are approaching daily publications. In recent times, finding an untouchable area of healthcare research without SR is like finding a needle in a haystack. This fact is not restricted to SRs but applied to other types of research. For example, when searching in PubMed with the term "COVID-19", we find since the outbreak of the COVID-19 pandemic until the end of October 2022, more than 308000 COVID-19-related articles and 6380 systematic review hits, i.e. on average more than 6 SRs daily. This situation leaves little room for postgraduate students to formulate a new and genuine question for a systematic review. Thus, I will discuss the solution I have adopted to overcome this dilemma.

The Nature of the Question in SR

According to the Oxford dictionary, a question is "A sentence worded or expressed to elicit information." or "A doubt about the truth or validity of something."⁹, and in Wikipedia, the question is: "an utterance which typically functions as a request for information, which is expected to be provided in the form of an answer."¹⁰. So the question and the answer/information are in a close relationship, and the existence of the last supposes the existence of the former and vice versa. In SRs, this relationship is more obvious because every SR aims to address a question that can be answered through prior information existing in primary research². And according to FINER criteria, it should be possible for the SR's question to be addressed through the existing knowledge^2,11. Thus, we can say that all SR questions are generated exclusively through our prior knowledge.

The Generation of an SR Question

The prior knowledge could be identified in various sources or what we can call the stakeholders of the research/SR question¹².  Stakeholders in healthcare range from researchers who ask the question themselves, other researchers in the field, clinicians in the same area of interest, committees/bodies or research groups (scientific query), patient/patients with a condition related to the research field, and lastly postgraduate students interested in the same topic. By examining prior knowledge and the needs of these stakeholders, a primary question can be generated. For example, a woman with a family history of breast cancer (stakeholder), who knows that genetic factors are important predisposing factors and that there are possible diagnostic ways to detect this condition (prior knowledge) early, she could be interested in the efficacy of screening in her case (primary question). This question could also be raised through a breast cancer advocacy group (stakeholder) working alongside researchers on their joint topic. To explore the possibility that this question could have been raised or answered previously, the stakeholders examine the prior knowledge individually and universally through a comprehensive search of the relevant literature. This search yields one of two possibilities: either the primary question has been previously asked and answered; thus, this part of knowledge will move from the universal level to be integrated into the induvial one of the stakeholders, and the cycle would start again, or the search would identify a knowledge gap giving the possibility to start formulating and refining the primary question (Figure 1).

Figure 1: The cycle of identifying the knowledge gap

Refining and formulating the question

Having identified a knowledge gap, the primary question could be feasible, interesting, novel, ethical and relevant^2,11,12. Hence, a process of refining and well-formulating the question initiates. Richardson et al. suggested that a well-formulated research question should contain all four parts of its anatomy mentioned by Oxman and Sackett in their Users' guides to the medical literature^13,14. The PICO-Format illustrates this anatomy: P: population, I: intervention, C: control, and O: outcome. This format could be extended in the case of SR to PICO-TSS, where: T: time, S: study design, and S: setting (e.g. primary healthcare, secondary, etc..)¹⁵. However, applying this format to the primary question will yield several variations of PICO-TSS questions. Appendix I shows the possible variations for our breast cancer question.

All these variations are possible SR questions ranging from a narrow specific question to a wide general one. The choice should not be arbitrary but systematic, preferably by a well-established hypothesis^2,16. And the relationship between the question and its hypothesis is bidirectional. It takes repeated steps of refining the question based on the supposed hypothesis to reach the ultimate question with the ultimately suitable hypothesis (Figure 2).

 Figure 2: The bidirectional question-hypothesis relationship                                                                                       

Starting from a hypothesis that a monthly self-examination of the breast for adult women >16 years reduces breast cancer mortality compared to women who don't examine will lead us to formulate a well hypothesis-driven PICO-TSS-question. This refining process could be integrated into the cycle of identifying the knowledge gap to produce a question-generation algorithm (Figure 3). Through the SR/research, we are questioning the hypothesis, which could fill the knowledge gap and add information to prior knowledge or would fail and lead to the same or other knowledge gaps producing infinity circles of question generation (Figure3).

Figure 3: The Infinity circles of question generation

Alternative Techniques

The previous algorithm produces what we call a FINER's question by careful application. However, this algorithm needs a lot of effort and time and sometimes fails despite extensive searching to identify the knowledge gap¹⁵. Alternatives are still available to help develop new and genuine SRs questions. Cummings et al. identified other sources for the research question: new ideas and techniques presented in conferences, creative imagination during informal conversations with colleagues or teaching, or working with a mentor who could be a fruitful source for developing the question^11,12. Moreover, many organisations provide lists of research questions: the James Lind Alliance (JLA) is one of the world-leading initiatives that provide the top 10s of priorities for research through workshops of a Priority Setting Partnership (PSP)¹⁷. However, these alternatives share a common pitfall: they might be oriented from the outside of the researcher's perspective and interests; this could affect the research process's motivation and relevance to the stakeholders' needs.

How I generated my SR question

As a postgraduate student and clinician (stakeholder), I was interested in decompressive craniectomy (DC) as a topic for my MSc dissertation. My prior knowledge (theoretical and practical) gave me uncertainty about this procedure's efficacy (primary question). Through searching (PubMed + PROSPERO), I examined the universal knowledge, where I found 13 published SRs and a handful of ongoing ones. After reviewing all research questions in these reviews, I identified uncertainty in a specific area of DC, namely the additional effect of DC on the neurological outcomes (in terms of GOS-E and mRS) in patients with ICP (Knowledge Gap). I then developed a PICO-TSS question based on my hypothesis that DC added to other less invasive therapies can enhance the neurological status of patients who have or are at risk of developing ICP. After multiple refinements of the PICO-TSS question and the hypothesis with thoughtful feedback from my supervisor, I finally reached my FINER's question (Figure 4).

Figure 4: The Generation of decompressive craniectomy question

Conclusion

Research question generation and well-formulation through identifying the knowledge gap and testing the suitability of a proposed hypothesis should be the main leading algorithm for finding a FINER's question for both SRs for postgraduates and almost all types of research and researchers. 

References               

Appendix I

Multiple variations for the formulation of a PICO-TSS-question for breast cancer screening

P: adult women/ or women older than 16/ or women between 16 and 50 years/ etc., all with a family history  of breast cancer

I: mammography/self-breast examination/physician examination/ ultrasound/ etc. as a screening test for breast cancer 

C: no screening/ usual screening (like self-examination) etc.  

O: breast cancer-specific mortality/or all-cause mortality/or quality of life/or free-progress survival/or side effects of breast cancer treatment etc., one or a combination of these as the primary outcome    

T: from the time of screening till mortality/or ten years after screening/or till progress etc.

S: randomised controlled trials/or observational studies/or any study design 

S: GP practices (primary healthcare setting)/or outpatient clinics/ or hospitalised patients for other reasons etc.

Appendix II

My Review question with the PICO-TSS form

Does decompressive craniectomy (DC) introduce any additional value to the neurological state of the patients who have or would develop increased intracranial pressure (ICP) syndrome compared with standard care alone?

P: patients (human, both adults and children) with different neurological/neurosurgical conditions that cause increased intracranial pressure, including:

1- Traumatic brain injury
2- Malignant middle cerebral artery infarction
3- Paediatric traumatic brain injury
4- Spontaneous intracerebral haemorrhage
5- Cerebellar infarction or bleeding
6- Acute subdural haematoma
7- Poor-grade aneurysmal subarachnoid haemorrhage