Introduction

HSCI 130 introduced the people, institutions and landmark studies of public health. Each of those studies began when someone noticed a problem, decided which part of it could be studied, wrote a question and chose a way to answer it. HSCI 207 is about that work. It teaches the practical steps of health research, and it follows one fictional study through each of those steps.

1.1 What Research Is

Canada’s three federal research agencies (the Canadian Institutes of Health Research, the Natural Sciences and Engineering Research Council and the Social Sciences and Humanities Research Council) share an ethics policy, the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans, known as TCPS 2. It defines research as “an undertaking intended to extend knowledge through a disciplined inquiry and/or systematic investigation.” The definition has two parts. Research aims to produce knowledge that is useful beyond the immediate situation, and it is systematic: the researcher decides in advance what to observe and how, records those decisions, and applies them consistently so that others can judge the work and repeat it.

Many activities in health care use similar methods for a different purpose. A clinic that audits its wait times to improve scheduling collects and analyzes data for internal use. TCPS 2 states that quality assurance and quality improvement studies, program evaluation activities and performance reviews do not require research ethics board review when they are used exclusively for assessment, management or improvement purposes (Article 2.5). The boundary can be difficult to draw, and Lesson 5 returns to it.

Health research is carried out by universities, health authorities, ministries, clinicians, community organizations and First Nations health organizations, and the skills in this course apply in all of these settings.

1.2 The Running Case: The Cedar Valley Social Connection Study

Each lesson uses one invented study as its main worked example, so that you can see each step of the research process applied to the same project.

The two terms at the centre of the study have distinct meanings. Loneliness is the distressing feeling that one’s social relationships are fewer or less satisfying than one would like (Perlman & Peplau, 1981). Social isolation is the objective state of having few social contacts or relationships. A person can live alone without feeling lonely, and a person can feel lonely in a busy household. Lesson 7 returns to this distinction when it turns concepts into measures.

1.3 The Research Cycle

Textbooks draw the research process in different ways, and most share the same core stages. This course uses an eight-stage cycle (Figure 1.1). A team identifies a topic and a problem, reviews what is already known, writes a research question, chooses an approach and a design, plans the ethics, sampling and measurement, collects and analyzes the data, and interprets and shares the findings. The findings then raise new questions, and the cycle begins again.

The cycle is drawn as a circle because research is cumulative, with each study suggesting new questions, and because the stages are connected in both directions. A team that pilots its survey may find that older adults misread a question and return to the planning stage. A team that cannot obtain the data it needs may have to narrow its question. Researchers expect these loops and record the changes they make, because readers need to know which decisions were made in advance and which were made in response to what the team found.

How the course follows the cycle

The twelve lessons of HSCI 207 follow the cycle in order, as the table below shows. Two activities run through every stage. The first is engagement with the people who have an interest in a study, whom this series calls interest holders (Lesson 4). The second is ethics, which begins at the planning stage and shapes every later decision about recruitment, consent, data and reporting.

Other courses in the series take several stages further: systematic searching in HSCI 241, study designs in HSCI 230, sampling theory and questionnaire writing in HSCI 341, qualitative analysis in HSCI 841, and statistical modelling in HSCI 410. HSCI 207 gives you a working version of each stage for those courses to build on.

1.4 Topic, Problem and Question

New researchers often begin with a topic and try to move straight to collecting data. The steps most often skipped are the move from a topic to a problem and the move from a problem to a question. The cards below define the three terms, along with the aims that Lesson 2 derives from a question.

Where research problems come from

Research problems arise from several sources, and most studies draw on more than one. The panels below describe five common sources, each with a Cedar Valley example.

Writing a problem statement

A problem statement is a short paragraph, usually 100 to 200 words, that explains what the problem is, why it matters and what is not yet known. It persuades a reader that the question that follows is worth answering. A useful template has four parts.

The first sentence names the population and the issue, and the second says why it matters. The third states what is known and the gap, which contains three distinct unknowns, each of which could become a research question. The fourth names who needs the answer and for what decision.

1.5 From Problem to Question

A single problem can generate many research questions. The Cedar Valley statement points to at least three: how common loneliness is in the region, whether loneliness relates to emergency department use, and how older adults experience connection after a move. Lesson 2 gives you tools for choosing among them. The task here is to recognize what makes a question workable.

A workable research question has five features. It names a population and a setting. It refers to things that can be observed or measured, or to experiences that people can describe. It can be answered with data the team can realistically obtain. It is narrow enough to be answered in one study and broad enough to matter to someone. Its answer is unknown in advance, which rules out questions settled by definition (“Do socially isolated people have fewer social contacts?”) and questions of value that data alone cannot answer (“Should the health authority fund more seniors’ centres?”).

The tabs below show three versions of a Cedar Valley question.

First drafts of research questions are rarely final. The Cedar Valley team will revise its questions as it reads key papers and hears from its advisory group, and it will record each revision with its reason.

Introduction

Research questions differ in what they ask for. Some ask how much or how many, some ask why, some ask who will experience an outcome, and some ask what an experience is like. The kind of question determines the kind of answer a study can give, the data it needs and the approach that suits it. A team that is clear about the kind of question it is asking is less likely to collect the wrong data or to claim more than its data support.

Several classifications are in use. Hernán, Hsu and Healy (2019) sort the tasks of quantitative data analysis into description, prediction and causal inference, and Shmueli (2010) showed that explaining and predicting are distinct goals that call for different analytic decisions. This course adds a fourth kind, exploratory questions, for situations in which the experiences or concepts themselves are not yet well understood. Other textbooks use further categories, such as comparative or relational questions, and most of these fit within the four used here (Figure 2.1).

2.1 Descriptive Questions

A descriptive question asks what is happening, to whom, where and when. It characterizes a population, condition or situation as it is, without asking why. Epidemiologists often describe health by person, place and time, and descriptive questions put that habit into practice: they ask how many people have a condition, how often an event occurs, how a characteristic is distributed across groups or communities, or how it has changed over time.

The Cedar Valley team’s first descriptive question asks: “What proportion of adults aged 65 and older in the Cedar Valley Health Authority region are lonely, defined as a score of 6 or higher on the three-item UCLA Loneliness Scale?” The scale asks how often a person feels that they lack companionship, feel left out and feel isolated from others. Each item is scored from 1 (hardly ever) to 3 (often), so totals range from 3 to 9 (Hughes et al., 2004). The answer to a question of this kind is a number. The box below uses the result of the team’s regional survey, which later lessons follow from design to analysis, to show what such an answer looks like.

This figure describes the people who answered the survey. Whether it also describes all 46,000 older adults in the region depends on how respondents were chosen and who chose to respond, which Lesson 7 takes up.

Descriptive questions are sometimes regarded as the simplest kind, yet much of public health depends on them. Planners cannot place services until they know where need is concentrated, a program cannot show change without a baseline, and differences between groups often reveal inequities that later studies try to explain. Descriptive studies are also a frequent source of the ideas that explanatory studies go on to test.

2.2 Explanatory Questions

An explanatory question asks why something happens or whether one factor affects another. In quantitative health research, explanatory questions are usually causal. They ask whether a change in one factor, called the exposure, would produce a change in another, called the outcome. An exposure is any characteristic, behaviour or condition whose effect is being studied, and an outcome is the health state or event that it might affect.

The Cedar Valley team’s explanatory question asks: “Among adults aged 65 and older in Cedar Valley, is loneliness associated with the number of emergency department visits in the following twelve months?” The question uses the word associated, and the team’s underlying interest is causal. Planners want to know whether reducing loneliness would reduce emergency department visits, because only then would loneliness programs be expected to ease pressure on emergency care. Lesson 2 shows how to structure this question with the PECO framework.

The distinction between association and causation is central to explanatory research. Two factors are associated when they occur together more often, or less often, than chance alone would produce. One factor causes another when changing the first would change the second. An association can arise for several reasons, and causation is only one of them. The panels below set out the main alternatives using the Cedar Valley question.

Answering an explanatory question therefore requires a comparison, for example between lonely and less lonely older adults, and a plan for ruling out the alternative explanations. Lesson 3 introduces the causal diagrams that make those alternatives visible, and HSCI 230 and HSCI 341 develop the study designs and reasoning that address them.

Explanation is also a goal of qualitative research. Interviews might show how giving up driving led one woman in North Bench to stop attending her church and then to feel cut off, describing a process that a survey can only infer. Section 3 returns to how the two approaches each contribute to explanation.

2.3 Predictive Questions

A predictive question asks who is likely to experience an outcome in the future, using information available now. The answer is a forecast for individuals or groups, often expressed as a risk or a score. Its quality is judged by its accuracy: how often it identifies people who go on to have the outcome, and how often it raises false alarms.

The Cedar Valley team could ask: “Can information recorded at a routine primary care visit, such as age, living alone, recent bereavement, hearing loss and a loneliness score, identify which adults aged 65 and older will visit an emergency department in the next twelve months?” A clinic could use the answer to offer extra follow-up to patients at high risk.

Predictive and explanatory questions are easily confused because both concern the relationship between factors and an outcome. The difference lies in what the answer is used for. A predictor needs only to be informative, and it need not be a cause. The number of emergency department visits a person made last year is likely to be among the strongest predictors of visits next year, yet no intervention could change last year’s count. The reverse error is more common: treating a predictor as though it were a cause. If hearing loss predicts emergency visits, that finding alone does not show that providing hearing aids would reduce them, because hearing loss may mark people who are older or in poorer health (Shmueli, 2010). Hernán and colleagues (2019) describe causal questions as questions about what would happen under different actions, which is a different task from forecasting what will happen if nothing changes.

2.4 Exploratory Questions

An exploratory question asks what an experience, process or situation is like and how it works, when too little is known to say in advance what should be measured. Exploratory questions are open. They often begin with how or what and use words such as experience, meaning or understand. Their answers take the form of concepts, themes or detailed descriptions, and they are usually pursued with qualitative methods such as interviews, focus groups and observation.

The Cedar Valley team’s exploratory question asks: “How do adults aged 65 and older who live alone experience social connection after moving to a smaller town such as Riverside or Kestrel Lake?” The team does not know in advance what matters most to people in this situation. Transport, churches, neighbours, volunteer roles, telephone and video contact with family, or the experience of being a newcomer might all play a part. An exploratory study can identify these factors, and its findings can later shape the questions in a survey or the design of a program. Lesson 10 develops the interviews and focus groups the team will use.

Exploratory work also occurs in quantitative research. An analyst who examines a large dataset for patterns without a prior hypothesis is conducting exploratory analysis, which HSCI 410 develops in depth. In both approaches, exploratory findings suggest ideas that later studies test or measure.

Students often confuse the words exploratory and explanatory because they look alike. An exploratory question opens up a topic that is poorly understood. An explanatory question tests or accounts for a relationship between factors that have already been identified.

2.5 Telling the Kinds Apart

The table summarizes the four kinds of question with their usual signal words. Signal words are a guide only, because the same word can serve different purposes, and the test is always what the team intends to find out.

Matching wording to intent

Many weak research questions are weak because their wording does not match what the team plans to do. The tabs show three common mismatches and a revision of each.

Check your understanding

Classify each question below as descriptive, explanatory, predictive or exploratory before opening the panel.

Each kind of question calls for a different kind of evidence. Section 3 matches the four kinds to quantitative, qualitative and mixed-methods approaches.

Introduction

Once a team knows what kind of question it is asking, it needs a strategy for answering it. An approach is the broad strategy a study uses to answer its question, defined mainly by whether it collects and analyzes numbers, words and observations, or both. Health research uses three approaches: quantitative, qualitative and mixed methods.

Four related terms are often confused, and Figure 3.1 sets them out from the most general to the most specific. A paradigm is a set of assumptions about what counts as knowledge and how it can be obtained. An approach is the broad strategy. A design is the specific plan for who is studied, when, and what comparisons are made, such as a cross-sectional survey, a cohort study or a qualitative interview study. A method is a technique for collecting or analyzing data, such as a questionnaire, a review of medical records or a semi-structured interview. This lesson concerns approaches. Lesson 6 and HSCI 230 cover designs, and Lessons 7 to 11 cover methods.

3.1 Quantitative Research

Quantitative research collects data in the form of numbers, or converts observations into numbers, and analyzes them with statistical methods. Its concepts are defined before data collection begins and are measured in the same way for every participant, usually with standardized instruments such as the three-item UCLA Loneliness Scale. Samples are often large and, where possible, are chosen so that they represent a defined population. The analysis estimates quantities such as proportions, averages and differences between groups, and it states how much uncertainty surrounds each estimate. The aim is usually to generalize from the sample to the population it came from.

These features give quantitative research its strengths. It produces precise estimates that can be compared across groups, places and times, it can test whether a relationship is larger than chance would produce, and its results can be combined across studies. The same features set its limits. A survey can measure only what its designers thought to ask, and a number can say how often something happens without saying what it means to the people involved.

The quantitative strand of the Cedar Valley study has three parts. A regional survey of adults aged 65 and older, which received 1,600 completed responses, measures loneliness with the three-item UCLA scale. With participants’ consent, survey responses are linked to administrative records of physician visits, hospital discharges and emergency department visits through Population Data BC, an organization that supports research access to linked administrative data in British Columbia. A review of 300 electronic medical record charts at six partner clinics adds clinical detail. Lessons 8 and 9 describe these methods.

3.2 Qualitative Research

Qualitative research collects data in the form of words, images and observations and analyzes them by interpreting their meaning. Its questions are open, and data collection is flexible: an interviewer follows up on what participants say, and the interview guide may change as the study proceeds. Samples are small and are chosen through purposive (judgement) sampling, which means selecting people because they can speak to the experience being studied, such as older adults who have moved to a new town. Analysis involves reading the data closely, labelling passages with codes and grouping codes into themes or categories. Because the researcher interprets the data, qualitative researchers practise reflexivity, the habit of examining how their own background and assumptions shape what they notice.

Qualitative research gives depth and context. It can show how people understand their situation, how processes unfold over time and which factors matter that researchers had not anticipated. It cannot estimate how common an experience is in a population, and its findings depend on careful interpretation. Qualitative studies are judged mainly by their credibility and by their transferability, meaning whether readers have enough detail about the setting and participants to decide whether the findings apply elsewhere.

The qualitative strand of the Cedar Valley study includes 24 semi-structured interviews with older adults living alone and four focus groups: two with older adults, one with family caregivers, and one with clinic staff and community connectors. Lesson 10 shows how the team prepares and runs these sessions, Lesson 11 introduces coding, and HSCI 841 develops qualitative analysis at graduate depth.

3.3 Paradigms at a Glance

Behind each approach lie assumptions about what counts as knowledge and how it can be obtained. Thomas Kuhn (1962) used the term paradigm for the shared assumptions, methods and model problems that guide a scientific community, and Guba and Lincoln (1994) applied it to the competing worldviews that underlie research approaches. HSCI 230 Lesson 1 examines these ideas in depth. For this course, you need enough to recognize the stance a study takes and to explain why a team chose its approach. The tabs describe four stances that you will meet in health research (Creswell & Creswell, 2018).

A study’s paradigm is often implicit, and many health researchers choose an approach mainly by the question they are asking, which is itself the pragmatic position. The essential test is that the approach fits the question and that the team can explain why.

3.4 Matching Questions to Approaches

The kind of question is the best guide to the approach. Figure 3.2 reduces the choice to one question about the form the answer should take, and the table that follows matches each kind of question from Section 2 to the approach that usually suits it. Lesson 6 extends this into a decision map from question type to design family.

3.5 Mixed-Methods Research

Mixed-methods research collects and analyzes both quantitative and qualitative data within a single study or program of research and deliberately integrates the two (Creswell & Plano Clark, 2018). Integration is the point at which the strands are brought together, for example when survey results are used to decide whom to interview, when interview findings are used to explain a survey result, or when the two sets of results are compared side by side (Fetters et al., 2013). Mixed-methods researchers treat integration as the defining feature of the approach, because a study that collects both kinds of data and reports them separately gains little from having both.

Greene, Caracelli and Graham (1989) proposed an early classification of the purposes for combining approaches, and later authors have extended it. A team may want a more complete picture than either approach can give alone, may need qualitative findings to explain a puzzling quantitative result, may use qualitative work to develop a survey or a program, or may want to see whether results from different methods agree, which is often called triangulation. Creswell and Plano Clark (2018) describe three core designs that differ in whether the strands run at the same time or one after the other, called convergent, explanatory sequential and exploratory sequential designs. Lesson 6 teaches these designs and how to choose among them.

Common misconceptions about approaches

Section 4 turns from planning a study to reading one.

Introduction

Every published study is the end of a long series of decisions about the question, the approach, the participants, the measures and the analysis. A research article is the authors’ account of those decisions and of what followed from them. In this course you will read studies for how they were done, because reading other researchers’ decisions is an efficient way to learn to make your own.

4.1 The Parts of a Research Article

Most empirical articles in the health sciences follow a structure known as IMRaD, for Introduction, Methods, Results and Discussion. A title and an abstract come first, and declarations and references come last. Each part records a particular set of decisions (Figure 4.1).

Two parts deserve particular attention on a first reading. The final paragraph of the introduction usually states the research question or aim, and the discussion is where interpretation enters, so it needs the closest scrutiny. The abstract is a useful map, and it leaves out much of the detail that a judgement requires.

Qualitative and mixed-methods articles follow a similar structure with some differences. The methods section of a qualitative article usually describes the researchers’ position in relation to the topic, and a mixed-methods article reports both strands and explains how they were integrated. Reporting guidelines list what authors should report for each kind of study, including STROBE for observational studies (von Elm et al., 2007), COREQ for interview and focus group studies (Tong et al., 2007) and SRQR for qualitative research more generally (O’Brien et al., 2014). The guidelines double as checklists for readers, and Lesson 12 uses them to structure the writing of a report.

4.2 Reading for Decisions

Reading an article for its decisions is easier with a fixed template. The decision log below lists twelve decisions that almost every empirical study makes, where in the article each is usually recorded, and what to write down. Complete it in your own words, quoting the article only for the research question, so that you can later tell what the authors said from what you inferred.

Reading in passes

Experienced readers seldom read an article once from start to finish. Keshav (2007) described a three-pass method for reading research papers, and Greenhalgh (1997) advised readers of health research to establish what question a paper addresses and what type of study was done before turning to its results. The tabs adapt these ideas for this course.

Findings and interpretation

The most useful habit in reading research is to keep a study’s findings separate from its authors’ interpretation of them. Findings are what the data show, such as a proportion, a difference between groups or a set of themes. Interpretation is what the authors believe the findings mean and what they think should be done. Interpretation is a legitimate part of research, and it should stay within what the design can support. Three checks help. The first is whether the conclusion uses causal words, such as effect, increases or reduces, when the study observed only an association. The second is whether the conclusion extends to people or places that the study did not include. The third is whether the recommendations rest on the results or on assumptions that the study did not test.

4.3 Worked Example: Logging a Fictional Article

While drafting the problem statement, the Cedar Valley research assistant logs several articles. The abstract below is a fictional practice abstract written for this course, and its numbers are illustrative. It resembles the kind of study the team might find.

Rows 10 and 11 show where the article goes beyond its evidence. The findings show an association: lonely participants made more visits on average, and the difference remained after adjustment for four factors. The conclusion that loneliness increases emergency department use is a causal claim, and the adjustment may not have removed every alternative explanation, such as severity of illness. The recommendation goes further still, because it assumes both that programs reduce loneliness and that reducing loneliness reduces visits, and the study tested neither. A conclusion that stayed within the findings might read: “Loneliness was associated with more emergency department visits in the following year, and studies that test whether reducing loneliness reduces emergency department use are needed.”

For the Cedar Valley team, the article supports the problem statement, confirms a gap (it covers one city and no rural or First Nations communities), and prompts the team to plan how to reach older adults who rarely return surveys, which Lessons 7 and 8 address.

Bringing It All Together

This lesson introduced the research process that organizes HSCI 207. Research, as TCPS 2 defines it, is an undertaking intended to extend knowledge through disciplined inquiry or systematic investigation, and it moves through a cycle of eight stages from identifying a problem to sharing findings. The first and most often skipped steps are the move from a broad topic to a research problem, which combines a situation that matters with a gap in knowledge, and the move from that problem to a workable question that names a population, a setting and something that can be observed or described.

Research questions come in four kinds. Descriptive questions ask what is happening, explanatory questions ask why or whether one factor affects another, predictive questions ask who will experience an outcome, and exploratory questions ask what an experience or process is like. The kind of question points to the approach: quantitative research for counts, estimates and forecasts, qualitative research for experiences, meanings and processes, and mixed methods when a team needs both and integrates them. Paradigms describe the assumptions behind these approaches, and the essential test is that the approach fits the question.

A published article records the decisions its authors made, and a decision log makes those decisions visible. Reading in passes and keeping findings separate from interpretation are the habits that let you judge a study and learn from it. The fictional Cedar Valley Social Connection Study will carry these ideas through every lesson.

The final reflection asks you to apply the whole lesson to a health topic of your choice.

Final Knowledge Assessment

This glossary defines the terms, approaches and people introduced in Lesson 1; search it or hover over highlighted terms in the reading.