How to Write a Research Article That Survives Desk Review
Sep 30, 2026
Most manuscripts that fail never reach a reviewer. They are returned by an editor, often within a day or two of submission, and usually for reasons that have little to do with the quality of the underlying science. That is the frustrating part: a solid study can be rejected because the abstract hid the result, the reporting guideline was ignored, or the paper was clearly built for a different journal.
Here is what actually happens in those first few minutes at the editor's desk, and how to write each part of an IMRaD paper so it survives the screen.
What an editor does in the first five minutes
When I open a new submission, I am not evaluating your science yet. I am answering four questions, in roughly this order.
Is this in scope? I read the title and abstract and ask whether our readers would plausibly cite this. If I have to work to justify the fit, that is already a bad sign.
Is there a result? Not a topic, not an aim, a result. I want a number, a direction, a finding. If the abstract ends with "implications are discussed," I assume there is nothing there.
Is the method adequate for the claim? A quick read of the design, sample, and analysis. I am looking for a mismatch between what was done and what is being asserted.
Is it compliant and clean? Ethics statement, trial or review registration, reporting checklist, author contributions, conflicts, data statement, references in roughly the right format, English that a reviewer can read without effort.
Fail question one or two and you get a scope rejection. Fail three and you get a methods rejection. Fail four and you get an unsubmission request or, at many journals, a rejection with an invitation to resubmit properly — which costs you weeks.
The ten most common desk rejection reasons
Out of scope for the journal, or an obvious shotgun submission.
No clear research question, or a question the data cannot answer.
Abstract that describes the study but not the findings.
Incremental or purely confirmatory work with no stated contribution.
Missing ethics approval, consent statement, or trial registration.
No reporting checklist where one is mandated (CONSORT, PRISMA, STROBE, and so on).
Sample or analysis plainly insufficient for the conclusions drawn.
Causal language on observational data.
Formatting and length so far off that the paper looks recycled from another journal.
Language quality that would make peer review a burden.
Note how many of these are fixable in an afternoon.
Title and abstract: the only part guaranteed to be read
Your title should carry the finding or, at minimum, the design and population. Compare:
Before: A study on the role of nurse-led interventions in diabetes care
After: Nurse-led telephone follow-up reduces HbA1c at 12 months in type 2 diabetes: a cluster-randomized trial
The second version tells me population, intervention, outcome, timepoint, and design. I can route it to the right editor immediately.
For the abstract, the single most common failure is the empty final sentence. Write the structured abstract so each section does real work, and make the results section numeric.
Before: Results showed significant improvements in the intervention group. Conclusions: the intervention may be beneficial and further research is warranted.
After: Mean HbA1c fell by 0.62 percentage points more in the intervention arm (95% CI 0.41 to 0.83; p < 0.001), with no difference in severe hypoglycemia (3.1% vs 2.8%). Conclusions: nurse-led telephone follow-up produces a modest, sustained glycemic benefit without increased hypoglycemia risk.
Report effect sizes with confidence intervals, not p-values alone. Name the comparator. If your result is null, say so plainly in the conclusion — editors reject vagueness far more often than they reject negative findings.
Introduction: three paragraphs, one gap, one question
A long introduction is a reliable signal of an inexperienced author. The job is narrow: establish that the problem matters, establish what is already known, and state precisely what remains unknown and what you did about it.
A structure that works almost everywhere:
Paragraph one: the clinical, scientific, or policy problem and its scale, with the best available source rather than a convenient one.
Paragraph two: what the existing evidence shows and where it stops. Be specific about the limitation — population, outcome, duration, setting — not just "studies are limited."
Paragraph three: your objective, stated as an aim or hypothesis, in one or two sentences.
The most common mistake is a gap statement that is really an absence statement. "No study has examined X in Y" is not a justification; plenty of things have never been studied because they do not matter. Say why the gap is consequential: what decision is currently being made without this evidence?
Before: To our knowledge, no study has assessed this intervention in rural primary care.
After: Existing trials were conducted in urban academic centers with on-site specialist support. Whether the effect persists where follow-up relies on a single practice nurse is unknown, and this is the setting in which most national guidelines expect the intervention to be delivered.
Methods: where desk rejections are actually decided
This is the section I read most carefully before deciding, because it is the section that cannot be fixed in revision. If the design cannot support the claim, no amount of rewriting helps.
Start by finding your reporting guideline on the EQUATOR Network before you write a word of the methods, not after. Use PRISMA for systematic reviews, CONSORT for randomized trials, STROBE for observational studies, STARD for diagnostic accuracy, CARE for case reports, SRQR or COREQ for qualitative work. Write the methods against the checklist item by item, then submit the completed checklist with page numbers. An editor who sees a populated checklist immediately treats the submission as serious.
Registration is non-negotiable where it applies. Clinical trials must be prospectively registered — ClinicalTrials.gov or another WHO primary registry — and the ICMJE recommendations are explicit that journals should not publish unregistered trials. Systematic reviews should be registered in PROSPERO. Put the registration number in the abstract.
The methods elements editors check for by reflex
Design named in the first sentence, with setting and dates.
Eligibility criteria that someone else could apply.
Sample size justification, or an explicit statement that the sample was one of convenience.
Primary outcome defined and distinguished from secondary outcomes, with the time point.
Statistical analysis plan, including how missing data were handled and whether analyses were pre-specified or exploratory.
Ethics committee name and approval reference, consent procedure, or a documented waiver.
Data availability statement that says something concrete.
Missing data handling is the item most often skipped and most often raised by reviewers. One honest sentence — "Twelve participants (4.1%) had missing primary outcome data; these were handled by multiple imputation with 20 imputations, and a complete-case sensitivity analysis is reported in Supplement 3" — prevents a round of review.
Results: report, do not interpret
Lead with participant flow, then baseline characteristics, then the primary outcome, then secondary outcomes, then sensitivity analyses. Resist the urge to explain. Every interpretive sentence in the results is a sentence a reviewer will flag.
Two habits that improve credibility immediately. First, never report a p-value without the estimate it belongs to. Second, do not describe non-significant differences as "trends toward significance" — editors read that as a tell.
Tables should be self-contained. If I cannot understand Table 2 without the text, it needs a better title, defined abbreviations, and clear units. Figures should have axis labels a reader can interpret at a glance.
Discussion: state your contribution in the first sentence
Open with what you found, not with a restatement of the background. The first sentence of the discussion is a free opportunity to tell the editor why the paper is worth sending out, and most authors waste it.
Before: Diabetes remains a significant global health burden affecting millions worldwide.
After: In this cluster-randomized trial in 18 rural practices, nurse-led telephone follow-up produced a 0.62 percentage point greater reduction in HbA1c at 12 months than usual care, an effect comparable to that seen in urban specialist settings.
Then compare with prior work honestly, including studies that disagree with you. Then limitations — real ones, stated without hedging, with their likely direction of bias. "Our outcome assessors were not blinded, which may have inflated the observed effect" is far stronger than "some limitations should be acknowledged." Finally, a conclusion proportionate to the evidence. Observational data support association language, not causal language, and the mismatch is one of the fastest routes to rejection.
The administrative layer that quietly sinks submissions
Authorship should meet the four ICMJE criteria, and contributions are best declared using the CRediT taxonomy. Guest and gift authorship are editorial misconduct, and disputes after acceptance are handled under COPE guidance, which is slow and unpleasant for everyone. Settle the author list before submission, in writing.
Declare funding and conflicts even when you think there is nothing to declare — "The authors declare no competing interests" is a required sentence, not an optional one. Register for an ORCID iD; many journals now require one for the corresponding author.
Choose the journal before you finish the paper, not after. Read three recent articles of your type in that journal and match their structure, length, and reference style. If you are considering an unfamiliar title, run it through Think. Check. Submit. and check whether it is indexed in DOAJ if it is open access. A paper published in a predatory venue is, for most purposes, a paper that does not exist.
A pre-submission checklist worth ten minutes
Title states the design and the main finding or population.
Abstract reports numeric results with confidence intervals and a conclusion that follows from them.
Introduction ends with a clearly stated objective or hypothesis.
The correct EQUATOR checklist is completed with page numbers and uploaded.
Registration number present in the abstract where applicable.
Ethics approval, consent, and data availability statements are all present and specific.
Authorship meets ICMJE criteria; CRediT roles assigned; all authors have approved the submitted version.
Conclusions use language proportionate to the design.
Formatting, word count, and reference style match the target journal's instructions.
One colleague outside your subfield has read it and can state your main finding in a sentence.
That last item is the best predictor I know. If a smart reader outside your area cannot tell you what you found, an editor skimming at 8 a.m. will not either.
Keeping the thread from search to submission
Much of what gets papers desk-rejected comes from fragmentation: the literature was gathered in one place, the notes lived in another, the draft in a third, and the reporting checklist was reconstructed from memory at the end. Context leaks at every handoff, and what leaks first is precision — the exact inclusion criteria, the reason a study was excluded, the source behind a background claim.
That is the problem PACR was built around. It is an AI-native research workspace where searching the literature, reading and summarizing it, drafting, collaborating with coauthors, and preparing for publication all happen in one connected environment, so the reasoning behind a screening decision or a cited claim stays attached to the manuscript instead of evaporating between tools. A typical use might be running a scoping search, keeping the excluded-study rationale alongside the records, and drafting the methods with that trail still visible — which is exactly what you need when a reviewer asks how you got from 2,400 records to 31.
You can explore how literature search and screening work in the research side of the platform, or set up a workspace and build your next manuscript with the reporting requirements in view from day one.
Surviving desk review is not about writing beautifully. It is about making it effortless for a busy editor to see, in five minutes, what you asked, what you did, what you found, and that you followed the rules. Do that, and the paper gets sent out — which is all you are asking for at this stage.
