Collecting the data is rarely the hardest part of a health research project. The difficulty arrives afterwards, when a spreadsheet of numbers has to become an argument a reader can follow and trust.
Many projects stall at exactly this point. The analysis is sound and the sample is defensible, yet the write-up reads like a list of outputs rather than a piece of reasoning.
This guide covers that final stretch: organising your results, reporting them without interpretation, and building a discussion that earns its conclusions.
Start with the reporting guideline, not the blank page
Before writing a word, find the reporting guideline that matches your design. The EQUATOR Network maintains them for almost every study type in health research.
A randomised trial follows CONSORT. An observational study follows STROBE. A qualitative study follows COREQ. Each gives you a checklist of what must appear and where it belongs.
Using one from the start saves rewriting later. It also signals to assessors and reviewers that you understand the conventions of the field, which carries more weight than most students expect.
Separate what you found from what you think it means
The most common structural error is mixing results with discussion. The Results section reports. The Discussion interprets. Keeping them apart makes both far easier to write.
| Belongs in Results | Belongs in Discussion |
| Participant numbers and flow | Why recruitment fell short |
| Descriptive statistics | What the pattern suggests |
| Test statistics, confidence intervals | Whether the effect is clinically meaningful |
| Themes and supporting quotes | How themes relate to existing theory |
| Tables and figures | Limitations affecting interpretation |
A quick test: if a sentence contains “suggests”, “indicates” or “because”, it almost certainly belongs in the Discussion.
Writing the Results section
Open with participant flow. How many were approached, how many consented, how many completed, how many were included in the analysis. Reviewers look for this first.
Then present findings in the order your research questions were asked. That ordering gives the section a logical spine and stops it becoming a dump of every test you ran.
Report effect sizes and confidence intervals, not p values alone. A p value indicates whether an effect is likely real. It says nothing about whether the effect is large enough to matter clinically.
Use tables for numbers and prose for the story connecting them. Never repeat a full table in the text. Point to it, highlight the one or two figures that carry your argument, and move on.
A worked example
Suppose you surveyed 180 nursing students about sleep and perceived stress, and 142 completed the questionnaire.
Your Results paragraph might read: “Of 180 students approached, 142 completed the survey, a response rate of 79 percent. Mean sleep duration was 6.1 hours (SD 1.3). Perceived stress correlated negatively with sleep duration (r = -0.41, 95% CI -0.54 to -0.26).”
That is description only. There is no explanation and no speculation about cause. The explaining comes next, in its own section.
Writing the Discussion
A strong discussion usually moves through five stages, in this order.
Answer the question. Open with a direct statement of what you found, in plain language, without statistics. One or two sentences is enough.
Place it in the literature. Does your finding agree with previous work? Where it differs, say so, and offer a plausible reason: a different population, a different measure, a different time period.
Explain the mechanism. Why might this relationship exist? This is where you demonstrate that you understand the subject, not merely the statistics.
State what it means in practice. For a health project that usually means implications for care, policy or training. Be specific, and keep the claim proportionate to your evidence.
Address the limitations. This deserves its own treatment, because it is where most write-ups either over-apologise or quietly under-report.
A discussion is not a summary of your results. It is an argument about what they mean, built only on evidence you have already presented.
Limitations without self-sabotage
Students tend toward one of two mistakes. Either they hide limitations, or they list so many that the work begins to look worthless.
The useful approach is to name each limitation, state its likely direction of effect, and explain what you did to reduce it.
Rather than “the sample was small”, write: “the sample of 142 limited power to detect small effects, so the absence of a significant association in the subgroup analysis should not be read as evidence of no association”.
The second version demonstrates judgement. The first merely concedes a weakness and leaves the reader to decide how damaging it is.
Making tables and figures do real work
Every table and figure should answer a question the reader is already asking. If you cannot say which question, it probably belongs in an appendix.
Label axes fully, including units. Give each table a caption that can be understood without the surrounding text. Keep decimal places consistent, and do not report more precision than your measurement justifies.
Avoid presenting the same data twice. A figure and a table showing identical numbers is a common reason for reviewer comments on an otherwise sound manuscript.
Common write-up mistakes
| Mistake | Why it costs marks | Fix |
| Interpreting inside Results | Blurs evidence and argument | Move every “this suggests” to the Discussion |
| Reporting p values alone | Hides whether the effect matters | Add effect sizes and confidence intervals |
| Restating tables in prose | Wastes word count | Reference the table, highlight two figures |
| Vague limitations | Reads as evasive | Name the direction of effect |
| New results in the Discussion | Reviewers treat it as unplanned | Report everything in Results first |
Writing the abstract last
The abstract is written last, even though it appears first. It is a compressed version of a finished argument, and you cannot compress an argument you have not yet made.
Most health journals and institutions want a structured abstract: background, methods, results, conclusion. Check the word limit before drafting, because 250 words disappears quickly.
Include your headline number. An abstract that says “an association was found” tells a reader nothing. One that gives the effect size and confidence interval lets them judge whether to read on.
The conclusion sentence should match the conclusion in your Discussion exactly. Reviewers notice when an abstract claims more than the full text supports, and it is an easy thing to get wrong when both are written under time pressure.
Dealing with results you did not expect
A null result is still a result. If your primary analysis found nothing, report it plainly in Results and explain in the Discussion what would have been needed to detect an effect.
Resist the urge to run additional tests until something reaches significance. Reviewers and assessors recognise that pattern, and it undermines the analyses that were genuinely planned.
If you did run exploratory analyses, label them as exploratory. Honest framing of an unexpected finding is far stronger than an implausible explanation for a convenient one.
Final checks before submission
Read the Results section with the Discussion covered. It should stand alone as a clear account of what happened in your study.
Then read the Discussion with the Results covered. Every claim should be traceable to something you reported, rather than something you assumed along the way.
Check that each table and figure is cited in the text, numbered in order, and readable in isolation. Then work through your reporting guideline checklist line by line.
If the write-up is where your project keeps stalling, structured academic writing support can help turn a completed analysis into a coherent argument, particularly under deadline pressure.
Finally, confirm reference formatting against the ICMJE recommendations or whichever style your institution requires.
The short version
Data becomes a finding when it is reported clearly. A finding becomes a contribution when it is interpreted honestly.
Keep those two jobs apart, follow the reporting guideline for your design, and be precise about what your evidence can and cannot support. That is most of what separates a write-up that convinces from one that merely describes.