A good study starts with a clearly defined question

Abstract

A clearly defined research question consists of (where appropriate) identification of study participants, intervention (or exposure) along with a comparator, and outcomes. Research questions should be defined in advance to allow identification of the appropriate study design, and they should be documented in a protocol before commencing the study. An important, but often overlooked, aspect of medical research is the research question. It is a statement that identifies the issue(s) to be studied. For example, are premature babies more at risk of developing respiratory illness than term babies? It should be posed by the researcher before designing the study to ensure that the study design and the data collected are appropriate and sufficient to answer the question of interest. The research question is a clear, concise and focused question. For example, if interest was in reducing the incidence of post-partum haemorrhage (Weeks, BJOG 2015;122:202-10), then the question might be, 'Does the prescribing of oxytocin reduce blood loss following delivery?' Having a clearly defined research question helps investigators to focus their research and allows them to work towards supporting or rejecting a specific hypothesis. Once the research question has been identified, other components of the study can be identified in a logical manner. Ideally the research question should determine the best design for the study. For the oxytocin example described above, a randomised control trial would be considered the reference standard (Bellad et al., BJOG 2012;119:975-86). However, if the research question had been, 'Do women who experience a post-partum haemorrhage have a perceived negative birth experience?', then a qualitative study would be more appropriate (Hin ton et al., BJOG 2014;121(Suppl. 4): 20-23). As a first step towards identifying a research question, a literature search can provide information on what is already known on a subject, the types of studies already undertaken and what questions previous research has left uncertain. A good research question often starts with a general observation and then is iteratively refined into a well-formulated question, using the PICO (Participants, Intervention, Control , Outcomes) criteria. For example, the initial question might be 'Is starting aspirin at 12 weeks of gestation in women at risk of pre-eclampsia a useful treatment?' and the refined question might be 'Does starting aspirin in women at risk of pre-eclampsia at 12 weeks of gestation reduce the risk of preterm delivery when compared with usual care?' So although the initial question contains information on the intervention (aspirin) and the participants of interest (pregnant women at risk of developing pre-eclampsia), in the refined question extra information is provided both on the comparator group (usual care) and the outcome (preterm delivery) (Villa, BJOG 2013;120:64-74). In summary, obtaining robust results starts with the definition of the research question. Above all else the question needs to be of clinical interest. If the research question is well posed then the design of the study to answer the question will often fall easily into place. With prospective documentation in a protocol, this will then lead to suitable data collection that will answer the research question on study completion.