# Lesson 2: Developing a Research Question

*Companion-podcast transcript, Sarah and Kiffer*

---

**Sarah:** Welcome back to Office Hours. I'm Sarah.

**Kiffer:** And I'm Kiffer. This week we are talking about the research question, which is the sentence that every later decision in a study depends on.

**Sarah:** Last week we went from a topic to a problem to a question. So what is left to do with the question?

**Kiffer:** There is quite a lot left to do. In Lesson one we sorted questions into four kinds: descriptive, explanatory, predictive and exploratory. That tells you what sort of question you have. It does not yet tell a research team whom to recruit, what to measure or what to compare. This lesson gives the question that precision, and then it asks whether the question deserves a study at all.

**Sarah:** And the lesson has four sections, right?

**Kiffer:** It does. Section one covers PECO and PICO, which are formats for questions that compare groups. Section two covers qualitative questions, mainly with a format called SPIDER. Section three judges questions with the FINER criteria, and section four turns a chosen question into aims, objectives and hypotheses and shows how to find key papers.

**Sarah:** Before we start, remind listeners about our running example.

**Kiffer:** We use the Cedar Valley Social Connection Study, which is fictional. A small research team at a British Columbia university, led by a researcher named Dr. Maya Hart, is working with the Cedar Valley Health Authority. It is a mostly rural region of about two hundred and ten thousand people, and about forty-six thousand of them are aged sixty-five and older.

**Sarah:** And the team starts with a worry.

**Kiffer:** Clinic staff and community organizations told the team that older adults in Cedar Valley seem lonely. That is a good starting point, but nobody can study it as it stands.

**Sarah:** Why not? It sounds clear enough to me.

**Kiffer:** Try turning it into a question: does loneliness affect older adults' health? That question does not say which older adults, whether everyone over sixty-five in Canada or the residents of one region. It does not say how you will decide who counts as lonely, or whom the lonely group is compared with. And health means a hundred things, from blood pressure to hospital stays to how people rate their own wellbeing. Each of those gaps is a decision that someone will make eventually, and a structured format forces those decisions at the start, before the study drifts.

**Sarah:** Where do these formats come from?

**Kiffer:** The best known comes from evidence-based medicine. In nineteen ninety-five, Richardson and colleagues described what they called the well-built clinical question, with four parts: the patient or problem, the intervention, the comparison and the outcome. Those four parts became known as PICO. Researchers in environmental and occupational health adapted it for exposures that nobody assigns, like air pollution or shift work. Morgan and colleagues described PECO in twenty eighteen. The letters stand for population, exposure, comparator and outcome.

**Sarah:** Let's go through them with Cedar Valley. Population first.

**Kiffer:** The population is the group the question is about, defined by person, place and time. For Cedar Valley, it is adults aged sixty-five and older living in the health authority region. Lesson seven refines that into target, source and study populations.

**Sarah:** What about the exposure?

**Kiffer:** An exposure is a characteristic, behaviour, condition or circumstance that people have or experience, and that the researcher observes without assigning it. The team's exposure is loneliness. And here is the important part: they had to say how loneliness would be defined. They used the three-item University of California, Los Angeles Loneliness Scale, which scores from three to nine, and they counted a score of six or higher as lonely.

**Sarah:** And the comparator?

**Kiffer:** The comparator is the group the exposed group is compared with. Here it is the one thousand two hundred and eight respondents who scored three to five on the same scale, compared with the three hundred and ninety-two who scored six or higher. And the outcome is the number of emergency department visits in the twelve months after the survey, taken from administrative records that are linked to survey responses with participants' consent. Put together, the question reads: among adults aged sixty-five and older living in the Cedar Valley Health Authority region, is loneliness, defined as a score of six or higher on the three-item scale, compared with a score of three to five, associated with the number of emergency department visits in the twelve months after the survey?

**Sarah:** I noticed you said associated with. Why not just ask whether loneliness causes emergency visits?

**Kiffer:** Because the team cannot assign loneliness to anyone. They observe people who are already lonely and compare them with people who are not. Lonely older adults may also be older, poorer, sicker or more likely to live alone, and any of those differences could explain a link with emergency visits. An association is a statistical link between two variables that may or may not reflect cause and effect.

**Sarah:** So a causal question is off the table forever?

**Kiffer:** No, and that is a good push. Lesson three shows how to draw a causal diagram that identifies the other variables the team needs to measure, and Health Sciences three forty-one develops the reasoning for moving from association toward a causal interpretation.

**Sarah:** Now, when would someone use PICO instead of PECO?

**Kiffer:** I give students a simple test. Could a researcher, in principle and ethically, decide who receives the factor? If yes, it is an intervention and you have a PICO question. If people bring the factor with them, or circumstances assign it, it is an exposure and you have a PECO question. Nobody can make people lonely, so loneliness is an exposure. The health authority has asked the team about a possible future study of a community connector service, where a trained worker links lonely patients with local groups. That service could be assigned to some patients and withheld from others, so a question about it would be a PICO question, comparing referral with usual care.

**Sarah:** What about descriptive questions? They don't really compare anything.

**Kiffer:** Right, and you should not force a comparator into them. What proportion of older adults in the survey scored six or higher? That question has a population and an outcome and nothing else. A partly filled format is fine for a descriptive question.

**Sarah:** Give me the practical advice for writing one of these.

**Kiffer:** Work in a fixed order. Start from your problem statement. Name the outcome first, because it is what you are trying to explain, and decide how and when you will measure it. Then name the exposure or intervention and how people will be classified, and name the comparator explicitly. Then define the population by person, place and time, and write it all as one sentence ending in a question mark.

**Sarah:** What are the common mistakes?

**Kiffer:** Drafts often have a vague outcome, like worse health, or no comparator at all. Some are compound questions with two exposures and two outcomes, which is really four questions. Some use causal wording for an observed exposure. And some have a population that is either too broad, like all Canadians, or so narrow that the study would end up with a handful of people.

**Sarah:** Let's move to section two. Why do qualitative questions need a different format?

**Kiffer:** Because they ask different things. Qualitative questions ask how or what. How do people experience a situation? What does a service mean to them? How does a process unfold? The answers come as words and accounts, usually with no intervention, no comparison group and no single outcome to count.

**Sarah:** What happened when the Cedar Valley team tried PICO anyway?

**Kiffer:** Their advisory group of six older adults had talked a lot about moving. Several had moved from Cedar City to smaller towns after retiring or after a spouse died, and they described that move as the point where their social lives changed. The team's first attempt in PICO form asked whether moving to a smaller town, compared with staying in Cedar City, reduces social connection among older adults living alone.

**Sarah:** What is wrong with that version?

**Kiffer:** There are several problems. Moving is treated as an intervention, although nobody assigns it. Reduces social connection presumes the direction of the answer before anyone has been interviewed. And the question says nothing about the kind of data the team plans to collect. The format pushed a question about experience into the shape of a trial.

**Sarah:** So what does a good qualitative question look like?

**Kiffer:** Creswell and Poth recommend one broad central question supported by a few sub-questions. The central question usually starts with how or what and uses open verbs such as explore, describe or understand. It names the people and the setting, and it focuses on one experience or process. It avoids words that imply measurement or cause, such as effect, impact or to what extent. And it avoids wording that assumes the answer.

**Sarah:** Now tell me about SPIDER.

**Kiffer:** SPIDER was developed by Cooke, Smith and Booth in twenty twelve, originally as a tool for searching for qualitative studies when doing evidence syntheses. They kept the idea of a short list of parts and changed the parts to fit qualitative work. The letters stand for sample, phenomenon of interest, design, evaluation and research type. Sample replaces population because qualitative studies choose small groups for the depth of their experience, without aiming to represent a population statistically.

**Sarah:** Walk me through the Cedar Valley version.

**Kiffer:** The sample is adults aged sixty-five and older who live alone and who moved to Riverside, North Bench or Kestrel Lake in the past five years. The phenomenon of interest is the experience of social connection after the move, including relationships kept, lost and formed, and a sense of belonging. The design is twenty-four semi-structured interviews. The evaluation element is the kind of findings they want, which is how participants describe their relationships and what helped or hindered them. The research type is qualitative, as one strand of a mixed-methods study. The short version of the question is: how do adults aged sixty-five and older who live alone experience social connection after moving to a smaller Cedar Valley community? Three sub-questions then break it into parts.

**Sarah:** The lesson mentions some other formats too.

**Kiffer:** Three of them. SPICE, proposed by Booth, stands for setting, perspective, intervention or interest, comparison and evaluation. It suits questions about services where the setting and a particular group's viewpoint are central. The Cedar Valley team used it for a question about how clinic staff experience asking older patients about loneliness during routine visits. PEO, which stands for population, exposure and outcome, appears in many nursing and health sciences textbooks for qualitative questions about people who share a circumstance, such as family caregivers whose parent lives alone in another community. Then PCC, which stands for population, concept and context, is the format the Joanna Briggs Institute recommends for scoping review questions. That one describes a literature review, so it lives in Health Sciences two forty-one.

**Sarah:** How do students choose among all of these?

**Kiffer:** Ask what the answer will look like. If the answer is numbers that compare groups, it is PECO or PICO. If it is numbers that describe one group, a population and an outcome are enough. If it is an account of experiences or meanings in words, start with SPIDER, and consider SPICE when the setting and a group's viewpoint drive the question.

**Sarah:** The Cedar Valley study has both kinds of question. How do they fit together?

**Kiffer:** Many mixed-methods researchers write a third question that integrates the two strands. The team wrote: in what ways do the experiences of older adults who live alone help explain the association between loneliness and emergency department use found in the survey? Lesson six covers the mixed-methods designs that decide when and how the two strands come together.

**Sarah:** What are the most common mistakes students make with qualitative questions?

**Kiffer:** The most common is slipping back into measurement language. Asking about the impact of moving on loneliness implies an effect that interviews cannot estimate. Another is the yes or no question, like do older adults feel lonely after moving, which invites a one-word answer. A third is the loaded question. If you ask what barriers prevent older adults from using the Riverside seniors' centre, you have assumed that barriers exist and that they explain why people stay away.

**Sarah:** How would you fix that last one?

**Kiffer:** Ask how older adults in Riverside decide whether to take part in activities at the centre. Then participants tell you what matters, which might be barriers or might be other preferences.

**Sarah:** Let's go to section three. The team now has well-structured questions. Isn't that enough?

**Kiffer:** It is a necessary step, but a well-structured question can still be a poor choice. It might need more participants than anyone could recruit. It might have been answered convincingly already. It might expose people to risks that no ethics board would accept. Judging questions before you commit to one finds those problems while the question is still cheap to change.

**Sarah:** And FINER is the tool for that.

**Kiffer:** FINER comes from Hulley and colleagues, in their textbook Designing Clinical Research. A good question, they argued, is feasible, interesting, novel, ethical and relevant. They wrote it for clinical research, and it travels well to public health, health services and qualitative work.

**Sarah:** Start with feasible.

**Kiffer:** Feasible means the team can actually answer the question with the people, skills, time, money and data it can get. Hulley and colleagues list participants, expertise, time, money and a manageable scope, and in public health we add access to data. For Cedar Valley, one thousand six hundred surveys are plenty for the comparison, but the linked emergency department records need participant consent and an approved data access request through Population Data B C. Lesson nine covers that process.

**Sarah:** Interesting sounds a little soft. Interesting to whom?

**Kiffer:** That is a fair question. It should interest the researcher, because research takes longer than planned and you need a question you care about to keep going. It should interest other researchers. And it should matter to interest holders, meaning the people affected and the organizations that serve them. The emergency department question came from clinic staff, and the health authority wants the answer to plan services.

**Sarah:** What does novel mean here?

**Kiffer:** Novel means the study adds something. Hulley and colleagues describe a novel study as one that confirms, refutes or extends earlier findings, or provides new ones. Novelty asks what your study would add. A careful study of a known association in a new population or setting can be novel, as can a study with better measures than earlier work.

**Sarah:** And ethical and relevant?

**Kiffer:** Ethical means the study could be done in a way that a research ethics board would approve. In Canada, those boards apply the Tri-Council Policy Statement, which Lesson five introduces. At the question stage you ask whether the risks can be justified, whether consent and privacy can be secured, and whether community agreements are needed, like the one the Cedar Valley team is developing with its First Nations health partner. Relevant means someone could use the answer, whether in practice, in policy, in future research, or for a community priority.

**Sarah:** Now, the lesson turns FINER into a scoring rubric. Is that how Hulley and colleagues use it?

**Kiffer:** They present it as a checklist. The scoring is our addition, so that judgments are written down and can be compared. Each criterion gets a one for serious concern, a two for some concern or a three for strong. Each score gets a one-sentence reason that cites evidence, like a number, a source or a planned safeguard. The total runs from five to fifteen. Then three decision rules apply, in order. If any criterion scores one, you revise the question to remove the problem and score it again, or you replace it if the problem cannot be fixed. If nothing scores one and the total is twelve to fifteen, the question is ready to carry forward. If nothing scores one and the total is ten or eleven, you strengthen at least one criterion that scored two before you carry it forward. Those thresholds are conventions of this lesson.

**Sarah:** Why does a single one matter so much? A total is a total.

**Kiffer:** Because one serious problem can make a study impossible, however strong it is in other ways, and adding up the other scores does not change that. The Cedar Valley team drafted four candidate questions, and one of them shows this. It asked does loneliness cause dementia among older adults in Cedar Valley? It scored three for interest and three for relevance, two for novelty and two for ethics, for a total of eleven. But it scored one for feasibility. Dementia develops over many years, and the study has twelve months of linked follow-up. Early memory changes can also lead people to withdraw socially, so a short study could not separate the two directions of influence. No rewording fixed that, so the team set the question aside for a future long-term cohort study.

**Sarah:** What about the others?

**Kiffer:** The emergency department question scored two for feasibility, because of the data access request, and two for novelty, because studies in other countries have examined loneliness and health care use among older adults. It scored three on the rest, for a total of thirteen, so it was ready. A question about social media use and loneliness in Kestrel Lake scored nine, with ones for feasibility and novelty. Only about seventy of the survey responses came from Kestrel Lake, the survey had no social media items, and the topic has been studied extensively. The team replaced it.

**Sarah:** How did the interview question score?

**Kiffer:** It scored fourteen. It scored two for ethics, because interviews about loss and isolation can cause distress, so the team planned a distress protocol and a list of local supports. It scored three for novelty, because the team found few qualitative studies of older adults who had moved to smaller communities in rural British Columbia.

**Sarah:** How did the team arrive at those scores? Did Dr. Hart just decide?

**Kiffer:** No. Dr. Hart, the graduate research assistant and the community research associate each scored the questions independently and then met to agree. That matters, because FINER scores are judgments. Two careful people can score the same question differently, and the value of the rubric is in the reasons it makes them write down.

**Sarah:** What errors do you see most when people score questions?

**Kiffer:** Treating novel as never studied is the big one. Students give novelty a one because they found any earlier study, or a three because they did not look. Another is scoring interest only from their own point of view. A third is judging feasibility for the wrong study. Feasibility refers to the planned study, carried out by the team that would actually do it, such as a small research team with a modest budget and about two years.

**Sarah:** Let's move to section four. Once a team has its question, what comes next?

**Kiffer:** A question says what the study wants to know. A protocol, a grant application or an ethics application also has to say what the study will do. Researchers use three further statements for that: the aim, the objectives and, for some questions, the hypotheses.

**Sarah:** Define them for me.

**Kiffer:** The research aim is one sentence stating the overall purpose of the study. Objectives are the specific steps that, taken together, achieve the aim, and each one usually corresponds to some data collection or analysis. A hypothesis is a testable prediction about the relationship between variables, written for a quantitative explanatory question before the data are analyzed.

**Sarah:** What is the Cedar Valley aim?

**Kiffer:** To understand how loneliness relates to health and health service use among adults aged sixty-five and older in the Cedar Valley Health Authority region. One aim covers the whole mixed-methods study, and it has four objectives. The first is descriptive: to estimate the proportion of survey respondents who score six or higher on the loneliness scale. The second serves the PECO question: to compare the number of emergency department visits in the twelve months after the survey between respondents who score six or higher and those who score three to five. The third explores, through interviews, how older adults living alone experience social connection after a move. The fourth uses four focus groups to describe how older adults, caregivers, clinic staff and community connectors view local supports.

**Sarah:** How do you know an objective is well written?

**Kiffer:** A reader should be able to tell from it what data the team will collect and what it will do with them. Many researchers use the smart criteria. Doran described them in nineteen eighty-one for management objectives as specific, measurable, assignable, realistic and time-related. Research methods texts now usually read them as specific, measurable, achievable, relevant and time-bound.

**Sarah:** Now hypotheses. Students often think every study needs one.

**Kiffer:** That is a common misunderstanding. Hypotheses belong to quantitative explanatory objectives. Researchers usually write a pair. The null hypothesis says there is no association or no difference. The alternative hypothesis says there is one. An alternative is directional if it predicts which group will have the higher value, and non-directional if it only predicts a difference.

**Sarah:** What did the Cedar Valley team write?

**Kiffer:** The null hypothesis says that the mean number of emergency department visits in the twelve months after the survey does not differ between respondents who score six or higher and those who score three to five. The alternative is directional: respondents who score six or higher will have a higher mean number of visits.

**Sarah:** The lesson insists hypotheses are written before analysis. Why does timing matter so much?

**Kiffer:** Because of something Kerr named in nineteen ninety-eight: harking, which stands for hypothesizing after the results are known. It means noticing a pattern in the data and then presenting it as if you had predicted it, so chance findings look like confirmed predictions. Writing the hypothesis into the protocol, which Lesson six covers, gives you a dated record of what you predicted.

**Sarah:** So you can never analyze anything you did not plan?

**Kiffer:** You can, and good researchers often do. You simply report those analyses as exploratory, so readers know they were not predicted.

**Sarah:** What about the qualitative strand? Does it get a hypothesis?

**Kiffer:** No. Objective three asks how older adults experience social connection. A prediction about what they will say would work against the open listening that interviews need. Descriptive objectives also skip hypotheses, because they estimate a quantity without comparing groups.

**Sarah:** The last part of the lesson is about key papers. How many does a researcher need?

**Kiffer:** Three to five. These are the published studies or reports most directly useful for your question. You need them to judge novelty, to borrow good measures and to explain why your question matters. Systematic searching, with documented strategies across databases and formal screening, is taught in Health Sciences two forty-one.

**Sarah:** What kinds of papers should be in the set?

**Kiffer:** A good set covers several roles. A recent review or meta-analysis tells you what is known. The Cedar Valley team used a meta-analysis by Holt-Lunstad and colleagues from twenty fifteen, which found that loneliness, social isolation and living alone were each associated with higher mortality. For a report, the team used the twenty twenty National Academies of Sciences, Engineering, and Medicine report on social isolation and loneliness in older adults. A third role is a study of a similar question somewhere else, and the team used a study by Gerst-Emerson and Jayawardhana of loneliness and health care use among older adults in the United States. That paper is why the team scored novelty two. Then comes a measurement paper, which for Cedar Valley was the paper by Hughes and colleagues describing the three-item loneliness scale. For a qualitative question, you also want a qualitative study of the same or a similar experience.

**Sarah:** How should a student actually search?

**Kiffer:** Use the parts of your structured question. Take two or three key terms, list a few synonyms for each, and combine them in simple searches in Google Scholar, PubMed or the library catalogue. Read titles first, then abstracts, and keep only papers that address your population, exposure or outcome directly. Look for a recent review first, because its reference list leads you to the main studies.

**Sarah:** What is citation chasing?

**Kiffer:** Backward citation chasing means reading the reference list of a useful paper to find the earlier work it relies on. Forward citation chasing means following the cited by link in Google Scholar to find later papers that cite it.

**Sarah:** What should students do with the papers once they have them?

**Kiffer:** Record each one in a small table with its citation, its role and its contribution, and keep the full reference for Zotero, which Lesson twelve covers. Also note any other variables the papers link to your exposure or outcome, such as age, chronic illness, income or living alone, because Lesson three uses them to build a causal web.

**Sarah:** Let's finish by putting it together. What does the Cedar Valley team have at the end of this lesson?

**Kiffer:** A linked set of statements for each strand. For the quantitative strand, there is the PECO question with each element labelled, its FINER scores with a reason for each, a total of thirteen and a decision that it is ready, the aim and objective two, and a null and a directional alternative hypothesis. Then there is a table of four key papers with their roles. The interactive FINER scorer in section three builds the same kind of summary for any question. For the qualitative strand, the set holds a SPIDER question, its scores, an aim and objectives, and key papers that include at least one qualitative study, with no hypothesis.

**Sarah:** Where do the causal web and the diagram come in?

**Kiffer:** Those come in the next lesson. Lesson three takes the key papers and turns them into a causal web and a first directed acyclic graph.

**Sarah:** That gives a clear picture of how a question takes shape. Thanks, Kiffer.

**Kiffer:** Thanks, Sarah. See everyone next week for causal webs and diagrams.