A new open-access guide published in Chiropractic & Manual Therapies offers researchers a step-by-step method for writing one of the most notoriously difficult parts of a scientific paper: the Discussion section. Written by Charlotte Leboeuf-Yde and Claire Gudex, the article is the last in a series of three that walk authors through structuring the Background, Methods, Results and now Discussion sections of clinical and epidemiological studies. Its central argument is that a good Discussion is not a free-form essay but a carefully planned text built around six recurring elements, each with a distinct purpose.
The six elements, as the authors define them, are a Summary of Findings, an Explanation of Findings, a Comparison of the results with those from other studies, a Critique of the authors’ own work, Perspectives, and a Conclusion. The Discussion usually opens with the summary and closes with the conclusion, but the four middle items can be arranged in different ways. Explanations and comparisons are often combined into a single passage, and the perspectives can appear in several locations. The authors even suggest that writers who struggle with structure should temporarily give each element its own heading while drafting, then remove the headings if the target journal will not accept them.
The Summary of Findings is where most Discussions should begin, and the guide is refreshingly strict about what belongs there. It should answer each research objective in simple, complete sentences, using the same key words for the study population and variables that appeared earlier in the paper. Crucially, it should not repeat detailed results, exact estimates, confidence intervals or p-values, unless the study was designed to produce a numerical result such as a prevalence estimate. Nor should it cherry-pick: the authors warn against reporting only the positive or only the negative findings, insisting that disappointing results be included alongside favourable ones. The summary reports results; it does not explain them.
The Explanation of Findings is the place to deploy the expertise the researchers accumulated while doing the work. This is where possible reasons for associations found or not found, mechanisms of action, technical aspects of the study process, and potential links or pathways belong. The key discipline here is avoiding repetition: the Background section covered what the authors knew and thought before the study, while the Discussion covers what they know after collecting and analysing the data. Anything that simply restates pre-study knowledge should be cut. The authors also note that published Discussions often grow too long precisely because writers stray from their research objectives, recycle Background material, or over-describe other people’s studies.
For the comparison with previous research, the guide offers what may be its most elegant trick: reuse the research gaps identified when the study was planned. If the study filled a ‘jigsaw puzzle’ gap, the Discussion should refer only to the studies that are relevant to completing that picture, with phrases such as ‘these findings fit well with’ earlier work. If previous studies contradicted each other in a ‘ping-pong’ pattern, the writer can simply state which studies the new results agree with, list the ones they contradict, and then attempt to explain the disagreement, citing typical culprits such as differences in study design, population, sample size, definitions, cut-off points, statistical approach or research quality. If the gap was a true ‘hole’ with no prior studies, or a clinical myth with no scientific literature, the authors provide template phrasings for those situations too. They also reassure readers that studies published after their own project began can legitimately be cited in the Discussion, since the Background should not introduce new information.
The Critique of own work section receives perhaps the most candid treatment. The authors argue that honesty is strategically as well as ethically sound: methodological weaknesses acknowledged by the authors are generally accepted and ‘forgiven’ by readers, whereas flaws uncovered by the reader make the authors appear untrustworthy. They suggest a systematic method: revisit the Methods plan, stop at each heading, and ask what data-quality or bias issues an insightful reader might detect. Each issue should then be classified as a real weakness with consequences, a real weakness with unknown consequences, or a minor or irrelevant weakness, and described accordingly. Strengths may be mentioned, but only those beyond what the study design would normally deliver; randomization in a randomized controlled trial does not count, though having enrolled enough subjects to permit subgroup analyses might. The choice of heading matters too, ranging from the broad ‘Methodological considerations’ to the narrower ‘Study limitations’.
Perspectives, the fifth element, push the findings out into the wider world. The guide divides them into three types: research perspectives, which should offer specific recommendations rather than the vague and unhelpful phrase ‘more research is needed’; clinical perspectives, covering knowledge that could change the behaviour or attitudes of health professionals and patients; and public health perspectives, such as implications for vaccination programmes, health policies, resource allocation or population-level guidelines. Notably, the authors point out that negative results can be especially valuable here, because they can redirect research, stop irrelevant clinical procedures or change useless policies. For validation or pilot studies conducted as preliminary steps toward a main study, the perspective may legitimately concern only the researchers’ own needs.
The Conclusion, the authors suggest, is best written immediately after the Summary of Findings so the two do not become near-identical. It should consider the wider picture and the overall aim of the study rather than the detailed objectives, contain few if any statistical details, avoid abbreviations, and be brief and memorable enough to be reused in the Abstract. The reasoning is pragmatic: many readers read only the title and the conclusions of a paper, so those parts must carry the essential message on their own.
The guide also tackles artificial intelligence, a topic that has transformed academic writing since the series began. Generative AI can be a legitimate aid for starting the writing process, generating ideas and checking grammar and flow, but the authors warn that it can make serious errors, including citing sources that do not exist or fabricating information, a phenomenon known as AI hallucination. They insist that generative AI must be used alongside human intelligence, because responsibility for an article’s content rests with the human author, not the model. Their recommended workflow is to create one’s own structure and content first, thinking through the specific needs of the study, and only then ask AI for feedback, a sequence designed to preserve both ownership of the manuscript and the writer’s critical thinking skills. They also note that in September 2025 the International Association of Scientific, Technical and Medical Publishers published a classification framework for reporting how authors use generative AI in academic writing, and that any AI use, whether at the planning, data collection, analysis or manuscript-preparation stage, should be described in the paper.
For researchers who want to sharpen their skills without leaning on AI, the article points to two free resources: the Medical Writing Checklist developed at the University of Southern Denmark, available as a phone app and covering sentence length, unnecessary words, abbreviations, apostrophes and commas, and the Academic Phrasebank from the University of Manchester, a categorized collection of scientific English phrases organized by article section and rhetorical function. The overarching message of the guide is that the Discussion is where authors demonstrate their critical analysis, and that this requires planning anchored in well-formulated research objectives. The final tip is a change of perspective: instead of writing for oneself, write for the reader, asking what they need to know and how to make the most important pieces of information easy to find. Done well, the authors conclude, even a long and information-heavy Discussion can remain a genuinely interesting read.
Subject of Research: A practical guide to structuring the Discussion section of clinical and epidemiological research articles
Article Title: Practical guide to structuring the Discussion section in clinical and epidemiological studies
Article References: Leboeuf-Yde, C., & Gudex, C. (2026). Practical guide to structuring the Discussion section in clinical and epidemiological studies. Chiropractic & Manual Therapies, 34(1), Article 45. https://doi.org/10.1186/s12998-026-00645-z
Image Credits: AI Generated
DOI: 10.1186/s12998-026-00645-z
Keywords: scientific writing, Discussion section, clinical research, epidemiology, research methodology, academic publishing, research objectives, peer review, artificial intelligence, study limitations, manuscript structure, science communication
News Source: Phoebe Ingram. (October 29, 2026). How to Write a Discussion Section That Readers Actually Trust. Scienmag.
