Review article: A primer for clinical researchers in the emergency department: Part III: How to write a scientific paper
Andrew Davidson, MBBS, MD, FANZCA, Director, Associate Professor; David McD Taylor, MD, MPH, DRCOG, FACEM, Emergency Physician, Professor; Franz E Babl, MD, MPH, FRACP, FAAP, FACEP, Paediatric Emergency Physician, Associate Professor.
A/Professor Franz E Babl, Emergency Department, Royal Children's Hospital, Parkville, Vic. 3055, Australia. Email: email@example.com
In this series we address key topics for clinicians who conduct research as part of their work in the ED. Analysis of research data does not represent the completion of a project as the findings need to be communicated to clinicians and other researchers in the field. In this section, we describe how to write up clinical research data for publication in a peer-reviewed journal. We also describe the editorial and peer-review process.
Publishing the results is a crucial part of a medical research project. In effect, a study never happened if the results are not published and researchers have an ethical obligation to publish their findings. For many, writing a paper seems like a daunting task, inevitably followed by a sometimes nerve-racking and, at times, frustrating review process. However, when their paper is finally published authors will inevitably be proud to see their study in print, adding to humanity's vault of knowledge.
In this paper we present an outline of the principles of what makes a good paper along with some tips and suggestions. We also outline the review process.
Throughout this series of reviews1 we have emphasised the importance of the research question. The research question is also important when writing the paper. How many times have you read a paper and been left baffled by the article, or given up half way through, asking yourself ‘I don't really know what this paper is actually about?’ This is a sign that the authors either do not have a good research question, or they have not clearly defined it.
Simply put, a paper states the research question, indicates why it is important, says how it was answered, gives the answer and then discusses the answer in terms of its validity and context. Before writing the paper, think carefully about what the research question is, and when writing keep the paper focused on the question. As discussed in previous reviews, a good question is well defined, novel, relevant, plausible and answerable.
General principles and rules
Aim to have the paper submitted as soon as possible after the study is finished. It is easy to lose momentum after a study is finished and the results have been presented at a conference, in a thesis or as an abstract. The longer it is delayed the harder it is to find the time to write. An organised researcher can write much of the paper before the study is finished. The Introduction and Methods sections can be written based on the study protocol and the format of the Results and Discussion sketched out well beforehand.
Authorship and acknowledgement
As authorship can be a contentious issue, it should be determined well in advance and all discussions should be transparent and documented. Ideally, authorship would be determined during project planning although there must be flexibility for those whose contribution wanes or those who are later brought into the team. There are standard guidelines for authorship2 and the issue is discussed in greater depth elsewhere.3 It is necessary that all authors read and approve the last draft of the paper before you submit it.
Remember to acknowledge all funding sources and those who contributed to the study but do not warrant authorship. Any conflicts of interest must be declared.
One paper or two?
Duplicate submission and duplicate publication (generally unacceptable unless editors are notified in special circumstances) and acceptable secondary publications of the same materials are clearly set out in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals by the International Committee of Medical Journal Editors.4 The more frequent issue for emergency researchers is that in some projects large amounts of data are collected and it might be that some secondary outcomes are important enough to warrant their own paper. There must be a balance between including large amounts of data about secondary outcomes in a single paper and the temptation to split the results among multiple papers. If the latter is excessive this is known as ‘slicing the salami’ publication. A publication combining as many components as possible will have a bigger scientific impact.5 In principle, separation of data between papers is acceptable, but this must be acknowledged in each paper. Some journals expect all such related papers to be submitted to the same journal.
Plagiarism is a major concern for journals. Take care to adequately reference all statements of fact. If possible, avoid copying phrases or sentences verbatim. If this is necessary, place them in parentheses along with a reference. Although self-plagiarism is not well defined, the reuse of large parts of a previously published text in a word-for-word copy is unacceptable.6 There are now several programs that can check for plagiarism very quickly and many journals use this software (e.g. iThenticate®, iParadigms, Oakland, CA, USA).
Writing good English
Journals differ in their publication format. Always follow the ‘Instructions to Authors’ very carefully. Not following the instructions implies (i) you have been rejected elsewhere and cannot be bothered changing the format, or (ii) you are lazy or cannot be bothered with detail. In either case the editors' confidence in your ability to conduct a study rigorously might be undermined.
Reviewers also get very annoyed by simple typographic errors. It is important to give the impression that you care about the paper and the research that you do. After many drafts and endless cutting and pasting, it is easy for simple errors to creep in. Always proof-read your last draft before you press the send button. A good idea is to give your draft to a colleague who is not familiar with the study. They will often pick up unidentified errors, and might identify sections that need clarification.
Although everyone has their own writing style, there are some general rules that should be followed. A paper will be better received if you use US or British/Australian spelling, as appropriate for the journal. In general, papers are written in the third person. However, occasional use of the first person might lighten the paper, but excessive use sounds unprofessional. The tense is usually past tense although the present tense might be appropriate in some circumstances. For example: ‘We conducted a prospective observational study’ and ‘Snake bites are rare in Australia’. Traditionally, the passive tense is used as it makes a paper sound neutral and impartial; however, excessive passive tense is dull and awkward to read.
Remember that many readers will not have English as their first language. Avoid long sentences. Use simple words, and avoid obscure words. With each sentence think of what words can be omitted without losing meaning. For example: ‘In previous studies nitrous oxide has been shown to increase the risk of vomiting’ versus ‘nitrous oxide increases the risk of vomiting’.
Avoid jargon and, in general, only use abbreviations that are recognised by the journal. Occasionally, a new abbreviation for a word or phrase might be useful. Do this only if the abbreviation is to be used extensively in the text. Always define the abbreviation at the first use of the word or phrase and remember that all abbreviations must be defined in the main text, even though they might have also been defined in the abstract. Generic names should be used for drugs unless there is a specific comparison of brands.
Structure of a paper
A paper describing an original study is usually divided into Abstract, Introduction, Methods, Results and Discussion sections. Case reports have a case description and discussion. A narrative review might or might not have an abstract, depending on the journal, followed by a structured review of the topic. In contrast, a systematic review identifies and analyses the relevant literature in a particular structured manner. Thus, they should have an introduction outlining which question will be answered by a review of the literature, a methods section describing the search strategy and how the papers were classified or outcomes synthesised and analysed, a results section describing the findings, and last a discussion section. There are guidelines on reporting research studies and these include checklists to ensure completeness. The best known guideline is the Consolidated Standards of Reporting Trials (CONSORT) statement for randomised controlled trials. Others are listed in Table 1. A library of reporting guidelines and corresponding checklists can be found via the Equator website.7–12
Table 1. Guidelines for reporting studies
|Randomised trials||Consolidated Standards of Reporting Trials||CONSORT||7|
|Meta-analyses of randomised trials||Quality of Reporting of Meta-analyses||QUORUM||8|
|Diagnostic studies||Standards for Reporting of Diagnostic Accuracy||STARD||9|
|Observational studies in epidemiology||Strengthening the Reporting of Observational Studies in Epidemiology||STROBE||10|
|Meta-analyses of observational studies||Meta-analysis of Observational Studies in Epidemiology||MOOSE||11|
The title should be concise, relevant to the research question, and accurately describe what the study is about. As a label, not a sentence, a title should not be a statement of fact. For example, ‘A randomised trial comparing fentanyl and morphine for analgesia after forearm fractures’ is preferable to ‘Morphine is better than fentanyl for analgesia for forearm fractures’. The authors should present their data and the readers should decide for themselves if they think morphine is indeed better. A descriptive term should be included in the title to provide an insight into the methodology used to answer the question. For example, ‘Ovarian torsion among women with abdominal pain: A retrospective cohort study’ or ‘Does nebulized lidocaine reduce the pain and distress of nasogastric tube insertion in young children? A randomized, double-blind, placebo-controlled trial’.
The abstract is very important and often is all that is ever read. Thus, it must succinctly describe the aim of the study and, in a sentence or two, why the question is relevant. The methods should be brief and the results section should focus on the primary outcome. The conclusion must be carefully written and be absolutely justified by the results. Do not speculate or make recommendations that are not justifiable.
The Introduction is all about the research question. Accordingly, it should begin with brief and succinct background information that sets the scene demonstrating that the question is relevant and novel. This in turn is followed by the aim of the study. Often, the research question is not explicitly stated as a question but is implicit in the aim of the study. For example, although a question might be ‘Is fentanyl better than morphine for analgesia for forearm fracture reduction?’, it is implicit in the stated aim of ‘To determine if fentanyl is superior to morphine for analgesia for forearm fracture’.
At the end of the Introduction, the reader should have a clear idea of what the research question is, why it is important and what the investigators aimed to do in the study. The Introduction should be brief. A long introduction often indicates that there is no coherent question. Similarly, if it takes a thousand words to persuade the reader that the question is relevant, then chances are that it is not! The Introduction is not a literature review and only key references should be used.
This section describes what was actually done in order to answer the research question. It should be detailed enough so that the reader could repeat the study exactly. Ethics committee approval must be stated as well as details about who gave consent or if consent was waived. If the study is a trial of a device or drug, then study registration details should be included (e.g. the Australian and New Zealand Clinical Trial Registry [ANZCTR; http://www.anzctr.org]). Some journals will not accept a paper if the trial was not registered.
Overall, the Methods is a relatively easy section to write as it can be compartmentalised into relevant sections, each of which can be written separately. There is also a standard order in which these sections are written. Headings, order and content of each Methods section are outlined in detail in study type-specific reporting guidelines and their checklists.7–12 The Methods should describe the study design, setting (where it was performed) and period (when it was undertaken). It is often useful to briefly describe the characteristics of the institution, including its nature (e.g. tertiary referral), patient mix (paediatric, mixed, adult) and annual patient census.
The Methods should next describe the subjects, the inclusion and exclusion criteria and how they were selected. If there was randomisation of patients or blinding, these should be described in detail. Any study intervention should also be fully described as well as the outcome measures. This section should also outline all other data that were collected. Measurements and tools used to describe or collect data should be referenced if they have been previously described elsewhere, and definitions of data points should be clearly described.
The statistics section should outline the analysis plan and clearly indicate the primary outcome and secondary outcomes. By definition, there can only be one primary outcome. There should be details of descriptive statistics (mean, median or proportions, standard deviation for means and interquartile ranges for medians) and inferential statistics (e.g. chi square, anova, t-test, logistic regression). Degrees of uncertainty should be expressed with 95% confidence intervals wherever appropriate. There should be a justification of the sample size recruited. This usually involves an a priori sample size calculation to ensure that the study is adequately powered. The sample size calculation should describe the variables associated with the calculation and any assumptions made. Briefly describe any pilot studies that helped provide estimates of outcome frequencies or variability, or tested feasibility. Post-hoc power analyses are rarely appropriate.
Present the results in the order that data collection is described in the Methods section. It is customary to start by describing how many subjects were enrolled and how many were excluded from analysis and why. Then present the demographic data so the reader can appreciate the types of patients enrolled. The primary outcome must be presented before secondary outcomes. Wherever possible, most data should be presented in tables or figures. Never repeat information in text, tables and figures. Important outcomes can, however, be highlighted or summarised in the text. For example, it is permissible to state in the text that ‘Almost all patients in the fentanyl group . . .’ while reporting the actual proportion in a table.
Tables and figures should be clear. Be careful that numbers in the table add up where appropriate. Be consistent in reporting, for example the number of decimal places. It is very important not to comment on the data in the Results. Simply report the data and statistical results and only comment on what they actually mean in the Discussion.
Start the Discussion with the answer to the research question. For example, ‘This study clearly indicates that fentanyl is the superior analgesic for . . .’. Keep the Discussion focused on the research question and the primary outcome. After stating succinctly what the study found, relate the findings to previous similar studies. Again, this is not a literature review. The aim is to provide some synthesis of the topic given your results and within the context of earlier research. For example: ‘Our findings are consistent with those of Miller et al. but not those of Johnson et al.’ Do not introduce new results in the Discussion.
When comparing your study with previous studies, outline the strengths of your study but avoid hyperbole and unjustified assertions. Avoid claiming to be the first to do something as this is unnecessary, boastful and often wrong. As all studies have limitations, be honest and complete in their description. Almost invariably, reviewers will pick up on limitations not discussed. Any attempt to hide or play down limitations will almost certainly be counterproductive. Try and give the reviewers nothing but the good parts of your study to write about.
There should be some discussion of the mechanism that might explain the results, but this should not be unduly speculative. Interesting secondary outcomes should be mentioned, but avoid excessive or tangential discussion about outcomes that were not the focus of the research question.
For a research project to be translational, it must have some impact. Indicate the possible implications of your study but do not overstate these, for example, how could clinical practice be guided by the study or how does the study lead on to further research.
The Conclusion should be short, related specifically to the research question and valid, that is, supported by the results. Do not end a paper saying more research needs to be done; this makes your paper sound as if it has no value in itself. Also, only conclude what your data allow you to and not your data in combination with the findings of others.
Ensure key statements of fact are referenced. It is not necessary to reference large numbers of papers that all make the same point. However, those that are cited should be entirely relevant, as recent as possible and of the best-quality studies undertaken. As a rule, cite only the original research articles, not review papers. If it is expedient to cite a review paper, be honest and state this. For example, you could state ‘In a review by Miller et al. . . .’. Make sure that references are done in the format required by the journal. Not doing so annoys both the editors and reviewers.
Choosing a journal
When choosing a journal, consider the journal's impact factor, the readership and the journal's reputation for efficiency and fair review. Like it or not, the impact factor is still important when it comes to securing grants and promotion. A journal is more likely to accept your paper if it will appeal to its readership. For example, studies on Australian snake bite management are more likely to be published in Australian journals. Journals differ in their efficiency. Consider your own previous experience and that of your colleagues. Last, do not be afraid to aim for a good journal and do not underestimate the value of your paper. Never submit a paper to two journals simultaneously.
Keeping to the word limit can be a challenge. Although the journals usually have some tolerance of papers that exceed the suggested word count, it is not uncommon to be asked to trim the paper, even before peer review. A particularly common criticism is that the discussion is too long. To see where a paper can be trimmed, check: are results duplicated in tables and text, or results and discussion? Is there discussion of previous literature that is not strictly necessary? Is there excessive discussion about secondary outcomes or excessive speculation about mechanisms and the implications of the study? Remember to keep the paper focused on the research question and the primary outcome.
After submission, your paper will be checked by the journal office for format and completeness. If there is some potential for publication, it will be sent for peer review. Alternatively, the editor might reject the paper immediately. The peer-review process is not perfect and it is best to approach it with a positive and patient attitude. Do not be disheartened by pages of comments and questions. These are often of considerable use in improving the paper.
When the reviewers' comments are received, respond to all their questions and comments. Editors look favourably on a reply letter with the reviewers' comments listed and, below each comment, a description of how the paper has been revised in response. You can disagree with a comment, but you have to provide a clear reason why you disagree. Some comments are just suggestions and might not need to be adopted. Do not ignore any questions and pay particular attention to any comments or questions posed by the editor as they are the ones who make the final decision. Thank the reviewers for their constructive comments and be polite and respectful in your reply. Your paper will inevitably be better after review and some reviewers spend many hours trying to help you produce a better paper. Good editors will filter out unhelpful comments from reviewers, but if some do slip through do not respond with sarcasm or rudeness.
If you are rejected, then do not be despondent. It might be that the journal does not think your paper fits their journal or that there is simply no page space available. If you have been rejected and the reviewers and editor have clearly misunderstood your paper, then you may appeal to the editor.
If your study question is clear, original and relevant, and your methodology is valid, then your paper will eventually be published somewhere. If you take care to write the paper well, then it will be published sooner and in a better journal.
We acknowledge grant support from the Murdoch Children's Research Institute, Melbourne, Australia and the Victorian Government's Operational Infrastructure Support Program.
DMT is a Section Editor (original research) for Emergency Medicine Australasia.
Anesthesia & Analgesia Guide for Authors. Anesth. Analg. 2009; 109: 217–31.
Coffin C. The scientific journal article: approaching the first draft. In: Barnard S, ed. Health Care Communications Group: Writing, Speaking, & Communication Skills for Health Professionals. New Haven: Yale University Press, 2001; 338.
Day RA, Gastel B. How to Write and Publish a Scientific Paper, 6th edn. Cambridge: Cambridge University Press, 2006.
Hall GM. How to Write a Paper. London: BMJ Publishing Group, 2003.
Holmes DR, Hodgson PK, Nishimura RA, Simari RD. Manuscript preparation and publication. Circulation 2009; 120: 906–13.
Hulley S, Cummings S, Browner W, Grady D, Newman T. Designing Clinical Research, 3rd edn. Philadelphia: Wolters Kluwer/Lippincott Williams & Wilkins, 2007.
Myles PS, Gin T. Statistical Methods in Anaesthesia and Intensive Care. Edinburgh: Butterworth Heinemann, 2004.