# Lesson 12: Writing Up Research

*Companion-podcast transcript, Sarah and Kiffer*

---

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

**Kiffer:** And I'm Kiffer. This is the last episode for Research Methods in Health Sciences, and it covers lesson twelve, on writing up research.

**Sarah:** It feels fitting that we end with writing. Why does it come last?

**Kiffer:** Because it is the step that turns a study into something other people can use. Every earlier lesson produced a piece of a study. The report is where those pieces meet. Until somebody else can read what you did and what you found, the study is unfinished in any practical sense.

**Sarah:** And we're doing it with Cedar Valley again.

**Kiffer:** We are. The Cedar Valley Social Connection Study is our fictional mixed-methods study of loneliness among adults aged sixty-five and older. The team, led by Doctor Maya Hart, has one thousand six hundred completed surveys, and three hundred and ninety-two of those people, which is twenty-four point five percent, were classed as lonely. The team has also run twenty-four interviews and four focus groups.

**Sarah:** Where do they start?

**Kiffer:** They start with the structure. Most empirical reports in the health sciences follow a structure called IMRaD, which stands for Introduction, Methods, Results and Discussion. Many journals call the introduction the background, so that's the word we use in the lesson.

**Sarah:** And each part has a job.

**Kiffer:** Each part answers one question for the reader. The background answers why the study was needed and what it asked. The methods answer what was done, with whom, and how the data were analyzed. The results answer what was found. The discussion answers what the findings mean, how far they can be trusted, and what should follow.

**Sarah:** You describe it as an hourglass.

**Kiffer:** The background starts broad, with a problem that matters to a lot of people, and narrows to one question. The methods and results stay narrow because they're about one study. Then the discussion widens again, back to the literature and to practice. The shape tells you where a sentence belongs. A general claim about loneliness goes at the top of the background or the end of the discussion, and a percentage from your own survey goes in the results.

**Sarah:** Let's start with the background, then. What is it actually for?

**Kiffer:** It's an argument. Its job is to persuade the reader that your question was worth asking. A background that just lists facts about a topic leaves the reader wondering why the study exists, while one that moves step by step from a problem to a question makes everything after it easy to follow.

**Sarah:** Is there a model for that argument?

**Kiffer:** The best known one comes from a linguist named John Swales, who studied a large number of research article introductions. He found that most of them make three moves, and he called the pattern the Create a Research Space model, or CARS for short. First, you establish a territory by showing that the topic matters. Second, you establish a niche by showing that something about it is unknown or contested. Third, you occupy the niche by saying what your study did to fill it.

**Sarah:** What does that look like for Cedar Valley?

**Kiffer:** The territory is loneliness and health. Meta-analyses by Julianne Holt-Lunstad and colleagues found that weak social relationships, loneliness and social isolation are associated with earlier death, and a study of older adults in the United States linked chronic loneliness to more physician visits. The niche is local. The team didn't find any published estimate of loneliness for Cedar Valley, and local planners don't know whether lonely older adults use emergency departments more often than others. The study occupies that niche.

**Sarah:** Let me push on something. When I was a student, my backgrounds were basically a list. This study found this, and that study found that.

**Kiffer:** That's the most common problem I see. We call it a study-by-study summary. Each sentence reports one paper, and the reader is left to work out how the papers relate to each other and to your question.

**Sarah:** And the alternative is synthesis.

**Kiffer:** Right. A synthesis organizes the literature by idea instead of by paper. Each paragraph opens with a claim, gives the evidence for it from several sources, and ends by connecting the claim back to the study. So instead of three sentences about three studies, you'd write that loneliness matters for health and for the health care system, and then bring in all three studies as evidence for that one claim.

**Sarah:** How do you get from a pile of papers to that kind of paragraph?

**Kiffer:** A synthesis matrix helps. It's a simple table in which each row is a source and each column is a feature you care about, such as the population, the design, the relevant finding and how you plan to use it. When it's filled in, you read down the columns, and the patterns become your paragraphs.

**Sarah:** Let's talk about the gap. You call it the hinge.

**Kiffer:** It's where the writer turns from what others have done to what this study does. It helps to know what kind of gap you're stating. A population gap means a question has been studied elsewhere but not in your group or setting. A question gap means a relationship hasn't been examined. A conflict gap means studies disagree, and a methods gap means earlier studies used a measure or design that limits what they can show. A perspective gap means a group's experience hasn't been described in its own words, which is often what a qualitative strand fills.

**Sarah:** And Cedar Valley has two of those.

**Kiffer:** It has a local population gap for the survey question, and a perspective gap for the interviews with people who moved to a smaller town.

**Sarah:** Here's a question I suspect students have. How do you know a gap is real? What if somebody somewhere has already done the study?

**Kiffer:** That's exactly the right worry. A gap statement can only claim what your reading supports. If you write that no study has ever examined something, you're making a claim that only a systematic search could back up, and systematic searching is taught in another course in the series. In a short report, you write something you can defend, such as "we did not find a published estimate of loneliness for this region." That claim is honest and defensible.

**Sarah:** How does the background end?

**Kiffer:** It ends with a one-sentence aim and the questions, written in the same words the methods and results will use later, so a reader can follow each question all the way through. Cedar Valley has two. The first asks whether lonely older adults in the region are more likely than others to visit an emergency department in the following twelve months. The second asks how older adults who live alone experience social connection after moving to a smaller town.

**Sarah:** Let's move to the methods section. What makes a good one?

**Kiffer:** Think about two readers. One wants to judge whether your findings can be trusted. The other wants to repeat your study or compare it with another one. Both need specific detail: the instrument, the number invited and the number who took part, the cut-off used to classify a score, and the analytic approach. It's written in the past tense under subheadings, commonly design, setting, participants, measures, data collection, analysis and ethics, and most of it records decisions students made in lessons six to eleven.

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

**Kiffer:** Partner clinics mailed invitations to a stratified random sample of five thousand two hundred and fifty adults aged sixty-five and older, with an online survey first and a paper copy later for people who had not answered. One thousand six hundred completed it, about thirty-two percent of those whose invitations were delivered. Of those, one thousand three hundred and twelve, or eighty-two percent, agreed to linkage with administrative health records.

**Sarah:** And how did they measure loneliness?

**Kiffer:** They used the three-item version of the University of California, Los Angeles Loneliness Scale. It asks how often you feel you lack companionship, feel left out, and feel isolated from others. Each item is scored from one to three, so the total runs from three to nine, and the team classed people scoring six or higher as lonely. That definition has to be in the methods, or a reader has no idea what the word lonely means in the results.

**Sarah:** What about the qualitative strand?

**Kiffer:** The methods report that two trained interviewers did the interviews: the graduate research assistant, a woman in her twenties from a large city, and the community research associate, a woman in her sixties who grew up in North Bench. Interviews lasted forty-five to ninety minutes, followed a guide piloted with the advisory group, and were recorded and transcribed. Two team members coded the first three transcripts independently and revised the codebook, and then four team members shared the coding in a framework matrix and developed themes, which the advisory group reviewed.

**Sarah:** Now, reporting guidelines. I think this might be new for a lot of students.

**Kiffer:** It probably is. A reporting guideline is a checklist of the information that a report of a particular kind of study should contain. The statistician Douglas Altman made that argument forcefully, and he helped to found the Enhancing the Quality and Transparency of Health Research Network, which everyone calls EQUATOR. It keeps a searchable library of reporting guidelines.

**Sarah:** How do you actually use one?

**Kiffer:** While you're collecting data, the checklist tells you what you'll need to record, like how many people declined or how long each interview lasted. While you're writing, you work through it item by item and note where each item is reported. Many journals ask authors to submit the completed checklist with the manuscript.

**Sarah:** Does a completed checklist mean the study is good?

**Kiffer:** It means the study is fully reported. Whether the design was strong is a separate judgement, which another course in the series teaches. The authors of the main observational guideline say that it was written to guide reporting and should be kept separate from quality assessment.

**Sarah:** Tell me about that guideline.

**Kiffer:** It's STROBE, which stands for Strengthening the Reporting of Observational Studies in Epidemiology. It has twenty-two items for cohort, case-control and cross-sectional studies, arranged in the order of a report. Its methods items ask for the design, setting, participants, variables, data sources and measurement, bias, study size, how quantitative variables were handled, and the statistical methods.

**Sarah:** Did the Cedar Valley draft pass?

**Kiffer:** It mostly did. The team found two gaps. They hadn't said anything about bias, so they added a sentence about the reminder postcard, the replacement questionnaire and the final contact used to reduce non-response. They also hadn't explained the study size, so they added a sentence saying that they selected five thousand two hundred and fifty people because they expected a response of about thirty percent. And because Cedar Valley linked survey answers to administrative records, the team also used an extension of STROBE for routinely collected health data, called RECORD, which asks how the linkage was done and how its quality was checked.

**Sarah:** What about the interviews and focus groups?

**Kiffer:** There are two main options. The Consolidated Criteria for Reporting Qualitative Research, known as COREQ, is a thirty-two item checklist for interview and focus group studies, with three domains: the research team and reflexivity, the study design, and the analysis and findings. The Standards for Reporting Qualitative Research, known as SRQR, has twenty-one items and applies to qualitative research of any design. COREQ suits a study like Cedar Valley's that rests on interviews and focus groups.

**Sarah:** And a mixed-methods study needs both strands covered.

**Kiffer:** It needs both strands, plus the integration. There's a short guideline called Good Reporting of A Mixed Methods Study, or GRAMMS, from Alicia O'Cathain and colleagues. Its six items ask you to justify using mixed methods, describe the design and each strand, and explain where and how the strands were integrated.

**Sarah:** Let's get to results. What belongs there?

**Kiffer:** The results report what was found, in the order of the questions, without interpretation. The section usually opens with the flow of participants, which for Cedar Valley is five thousand two hundred and fifty selected, one thousand six hundred responding, and one thousand three hundred and three linked. Then it describes the participants, reports the quantitative findings, presents the qualitative themes, and closes with the integrated findings.

**Sarah:** And tables do a lot of the work.

**Kiffer:** Tables and text divide it. The table holds the full set of numbers or themes in a form readers can scan and check. The text tells the reader what to notice, cites the table by number and quotes only the key figures.

**Sarah:** The reading walks through the anatomy of a table.

**Kiffer:** In the American Psychological Association style, which everyone calls APA, every table has the same parts. The table number sits above it in bold, and the title follows on the next line in italics and title case. The column headings name each column and give the sample size, and the left-hand column holds the row labels. Horizontal lines separate the headings from the body and close the table, with no vertical lines. Underneath, a note that begins with the word note in italics explains every abbreviation, so the table can stand on its own.

**Sarah:** Let's talk about Table One. Why is it always called that?

**Kiffer:** It's because the table is so standard. In most quantitative reports the first table describes the participants, and researchers simply call it Table One. The Cedar Valley version has columns for all one thousand six hundred respondents, for the three hundred and ninety-two classed as lonely, and for the one thousand two hundred and eight who were not.

**Sarah:** And there are some design decisions in there.

**Kiffer:** There are several. Categorical variables are shown as counts with column percentages, so each column describes one group. Age is roughly symmetric, so it gets a mean and standard deviation, while the number of chronic conditions is skewed, so it gets a median and interquartile range. Living alone takes a single row, because the other category is implied. The eleven people who didn't answer the self-rated health question get their own row for missing values, so the percentages are honest about the denominator.

**Sarah:** Are there any significance tests?

**Kiffer:** There are none, because the table is there to describe. The explanation paper that accompanies STROBE advises against significance tests in descriptive tables, and measures of association come in later courses.

**Sarah:** Then comes the cross-tabulation.

**Kiffer:** A cross-tabulation shows how the categories of one variable are spread across the categories of another. Cedar Valley puts loneliness status in the rows and emergency department visits in the columns, using the linked respondents. In the lonely group, eighty-two of two hundred and ninety-three had at least one visit in the following year, which is twenty-eight percent. Among everyone else, one hundred and ninety-two of one thousand and ten did, which is nineteen percent.

**Sarah:** Why row percentages?

**Kiffer:** The question is whether lonely people were more likely to have a visit. With the exposure in the rows, row percentages give the share of each group with the outcome, which answers that directly. Column percentages would tell you that, of the two hundred and seventy-four people who had a visit, about three in ten were lonely. That's true, and it answers a different question, about the make-up of the people who visited.

**Sarah:** And the results sentence stops at the numbers.

**Kiffer:** It stops there. Whether loneliness itself explains the difference, or whether lonely people are simply in poorer health, is a question for the discussion, and the adjusted analyses that address it are taught in later courses.

**Sarah:** Qualitative tables now. These seem harder to design.

**Kiffer:** They take some thought. Qualitative results are mostly written in text, organized by theme, and a summary table lets the reader see all the themes at once. The first thing to get right is the theme names. A good theme name states a meaning. "Transportation decides who you see" is a theme, while "transportation" on its own is only a topic.

**Sarah:** What are the Cedar Valley themes?

**Kiffer:** There are four, from the twenty-four interviews. Losing everyday encounters is about the routine, unplanned contacts that disappeared after a move. Transportation decides who you see is the second. Arriving as a newcomer is about small-town networks that feel hard to enter late in life. The cost of asking is about pride, and not wanting to be a burden, keeping people from saying they were lonely.

**Sarah:** And each has quotations.

**Kiffer:** Each theme has one or two short exemplar quotations from different participants, which in this fictional study are invented for teaching. Each one is labelled with a participant code, gender and decade of age, and nothing more. In a small town, a combination of the town, an age band and a gender can identify someone, so the team leaves the town names out.

**Sarah:** And then there's the joint display.

**Kiffer:** This is my favourite table in the lesson. A joint display puts the quantitative and qualitative findings side by side, one domain per row, and the last column records the meta-inference, which is what the two strands mean together. Michael Fetters and colleagues describe three kinds of fit. Confirmation is when the strands agree, expansion is when one adds something the other couldn't show, and discordance is when they disagree.

**Sarah:** Give me an example of each.

**Kiffer:** For confirmation, the survey shows that respondents who lived alone were more often lonely, and the interviews describe losing everyday encounters after moving alone. For expansion, the linked data show that lonely respondents visited the emergency department more often, and clinic staff in a focus group described patients going there at night because they had nobody to call. That suggests a pathway that administrative records could never show.

**Sarah:** And discordance?

**Kiffer:** Of the three hundred and ninety-two lonely respondents, more than half said on the survey that they would take part in a weekly social program. In the interviews, though, several participants said they would avoid any program described as being for lonely people, because the label felt embarrassing.

**Sarah:** So which is right?

**Kiffer:** They may both be right. Ticking yes on a survey costs nothing, while walking into a room labelled for lonely people costs a great deal, so stated interest may overestimate attendance. The way a program is described may also change who comes. That's why discordance belongs in the report. It often points straight to the next question.

**Sarah:** Which brings us to the discussion.

**Kiffer:** Most discussions make five moves. They restate the main findings as answers to the questions, compare them with earlier research, offer explanations including alternative ones, set out strengths and limitations, and close with implications and a conclusion. The discussion introduces no new results, and every comparison with earlier work needs a citation.

**Sarah:** How does the Cedar Valley discussion start?

**Kiffer:** It starts by answering the questions directly: about one in four older adults was lonely, and lonely respondents were more likely to visit an emergency department. It then compares that with the earlier study on physician visits, and raises the obvious alternative explanation, which is that poorer health among lonely respondents could account for part of the difference. Because the comparison is unadjusted, it is read as an association, and adjusted analyses are planned.

**Sarah:** That leads into limitations. I find students either skip them or write very vague ones.

**Kiffer:** The vague ones are the most common. "The sample size was small." "Correlation does not equal causation." Those sentences tell the reader very little. A useful limitation has four parts. It names the problem, explains how the problem could have affected the findings, says what the team did about it, and states what it means for how the findings should be read.

**Sarah:** Give me the Cedar Valley version.

**Kiffer:** Take non-response. About thirty-two percent of the people whose invitations were delivered completed the survey. Older adults who are most isolated may be less likely to answer a mailed invitation, so the prevalence of twenty-four point five percent may be an underestimate. The team sent a reminder postcard, a replacement paper questionnaire and a final contact, and it plans to compare the age and gender of respondents with regional population figures. Every part of that is specific.

**Sarah:** You also talk about hedging.

**Kiffer:** Hedging means choosing words that match the strength of your evidence. You'd write that loneliness was associated with emergency department visits, rather than that it causes them. You'd write that several participants who had moved described feeling like newcomers. And you'd keep any recommendation proportionate to what the study can show.

**Sarah:** Now let's turn to citations, and the seventh edition of the APA style.

**Kiffer:** The course uses the seventh edition of the Publication Manual of the American Psychological Association. It's an author and date system. Each citation in the text gives the author's surname and the year, and it points to a full entry in the reference list.

**Sarah:** Run through the basic forms.

**Kiffer:** A parenthetical citation puts the author and year in parentheses at the end of the clause. A narrative citation makes the author part of your sentence, with the year in parentheses after the name. With two authors, you join the names with an ampersand inside parentheses and with the word and in a narrative citation. With three or more authors, you use the first author's surname and et al. from the very first citation.

**Sarah:** And the reference list?

**Kiffer:** It starts on a new page under the heading References, bold and centred. Entries are double-spaced, alphabetical by the first author's surname, and have a hanging indent, which means the first line sits at the margin and the following lines are indented. Every citation in the text needs an entry, and every entry needs to be cited.

**Sarah:** Let's do the three source types, starting with a journal article.

**Kiffer:** The order is who, when, what and where. The authors come first, surname then initials, separated by commas, with an ampersand before the last one. Then the year in parentheses and a period. Then the article title in sentence case, which means capitals only for the first word, the first word after a colon, and proper nouns. Then the journal name in italics, a comma, the volume in italics, the issue in parentheses in regular type, a comma, the page range and a period. Finally comes the digital object identifier, written as a link with no period after it.

**Sarah:** What about reports?

**Kiffer:** A report entry gives the author, the year, and the title in italics and sentence case, with any report number in parentheses after the title. The publisher comes next, but only if it differs from the author. So a report written and published by the Cedar Valley Health Authority leaves the publisher out, while the National Academies report on social isolation lists the National Academies Press as its publisher.

**Sarah:** And web pages?

**Kiffer:** A web page entry gives the author, the date with the month and day if the page shows them, the page title in italics, the site name unless it's the same as the author, and the address. If there's no date, you use the abbreviation for no date, and if the page is designed to change, like an events listing, you add the date you retrieved it.

**Sarah:** Which brings us to Zotero. Why spend time on a tool?

**Kiffer:** About two thirds of students in a later course in this series told us they had never used a reference manager, and using one makes a real difference to accuracy. Zotero is free and open source and runs on Windows, Mac and Linux.

**Sarah:** Walk me through getting started.

**Kiffer:** Install the Zotero desktop program and the Zotero Connector, a browser extension that saves the details of the article you're viewing into your library. Make a collection for each report. Add sources with the connector, or by pasting a digital object identifier into the add by identifier button. Then check every entry, which is the step people skip.

**Sarah:** What goes wrong?

**Kiffer:** The item type can be wrong. Titles are often stored in title case, and Zotero can't reliably tell which words are proper nouns, so you should store them in sentence case. A group author, like the Cedar Valley Health Authority, needs to be entered in single-field mode, or Zotero will split it into a first and last name.

**Sarah:** Then you write.

**Kiffer:** Then you write. Zotero installs a plugin for Word and LibreOffice, and the connector adds a Zotero menu to Google Docs. You choose the seventh edition APA style, insert citations as you go, and add the reference list at the end. If you find an error, you fix it in the library and refresh the document. Zotero can't check that a source supports the sentence you cited it for, so that responsibility stays with the writer.

**Sarah:** Let's finish by putting the report together. Where does a writer start?

**Kiffer:** With the sources and the first two sections. A short report runs to about two thousand words, with a background, methods, results, discussion, references, and at least one quantitative and one qualitative table. Set up a Zotero collection, add the sources from the reading and check each entry. Draft the background as a funnel that ends with the question, and draft the methods under the subheadings from section two, checking them against the STROBE methods items and the COREQ domains.

**Sarah:** And the tables?

**Kiffer:** Build the Table One and the coding results into one quantitative and one qualitative table in APA format, each with a number, a title and a note, and mention each in the results text. Then write a discussion with the five moves and at least two specific limitations. Use the word budget in the lesson, roughly four hundred words for the background and five hundred for each of the other parts, to keep things in proportion.

**Sarah:** Any last advice, since this is the final episode?

**Kiffer:** Read only the first sentence of each paragraph in your draft. Together, those sentences should tell the whole argument, and if they do, the report will be easy to follow. And remember that every part of a report records decisions made earlier in the study, so writers are often better prepared than they feel.

**Sarah:** That's a good place to end. Thanks, Kiffer.

**Kiffer:** Thank you, Sarah, and thanks to everyone who has listened this term. Good luck with your writing.