Writing for PCCM: The 3,000-Word Structured Clinical Research Report

Authors of clinical research articles want their work noticed and read. Unfortunately, it seems that many subscribers have little time for in-depth reading of articles published in Pediatric Critical Care Medicine (PCCM). In a survey carried out by our publishers in 2019, in a self-selected sample of 1-in-7 out of about 2,000 subscribers, readers’ habits included: reading little or just “skimming” articles, 13%; reading about a quarter of the journal, 42%; or, reading about half the journal, 24%. Clearly, there is a problem for both subscribers and authors. In response, we plan to make articles shorter and more readable using a structured approach to content. Clinical Research Reports will be limited at 2,500 to 3,000 words (including abstract) and the logical progression, paragraph-by-paragraph, through an article will be structured for the readers’ benefit. This new approach does not mean an end to the Brief Report (i.e., the 1,500-word article including an abstract, 25 references, and two figures, and/or tables), but it does mean that the 4,000-word Clinical Investigation article will become a rarity after June 2021, needing Editor-in-Chief agreement before submission. The following outline on how and what to write in PCCM’s new structured Clinical Research Report is not original to our Journal. Others have used a similar approach (1), and I was taught it over 25 years ago as an essential skill for reviewing and handling articles submitted to the Archives of Disease in Childhood and British Medical Journal stable of journals (see reference 2 for an updated transcript of the course). TITLE The current convention for Titles is to be descriptive with just enough information. Titles used to be declarative or defining, at least that was until we realized readers often decided they knew enough by reading the title, and so moved onto the next article in a journal! Have a look at recent trends in titles of articles published in other journals. Be correct with your choice of words when describing your study. “Effect,” “Impact,” and “Incidence” are not determined in retrospective studies, instead use “Association of…” and “Prevalence” as the appropriate words. Avoid titles that convey a question. Invariably, it tells reviewers (and readers if it gets that far) that the authors still do not know an answer to the research question they set out to study. If the writers do not know, then why would a reader spend valuable time reading? Now, a common device in titles is to present the subject area: separated by a colon from the object in the research report, or methodology used. Make readers read the Abstract to find out more. ABSTRACT The hardest part of the article to write is the Abstract–therefore, as you prepare your submission to PCCM, leave it until the end. Then, construct a 300-word section using the following headings: Objective, Design, Setting, Patients, Interventions, Measurements and Main Results, and Conclusions. It is important to tell readers when the study was done, how many patients were in the study, and how many outcome events occurred. Do not obscure real numbers by only presenting percentages. MAIN SECTIONS OF THE REPORT In the main sections of the article, the remaining 2,200 to 2,700 words should be presented in 19 to 21 paragraphs, each around 110 to 140 words in length. Try and write simply. Use short sentences, certainly less than 30 words and preferably half that length. The article’s paragraphs are divided between four sections, Introduction–Methods–Results–Discussion, using a sequence of 2–5–5–7 paragraphs for shorter articles, or a sequence of 2–7–7–5 paragraphs for methodologically complex articles. INTRODUCTION (2 PARAGRAPHS) Readers of PCCM are experts in their own right and therefore do not need a general discourse about a topic. Write a short, focused Introduction with 10 to 15 references. Paragraph 1 provides the setting and context, with enough information to understand the reasoning in the next paragraph.Research In Context Use three bulleted points, each no more than 40 words, to provide more information. For example, I could state that most readers do not read all content in PCCM and we need a solution. Give a short outline of any additional evidence before the writing of your current report, that you were not able to say in paragraph 1 of the article. End with what you think is the bottom line. For example, this outline of the 2,500- to 3,000-word structured Clinical Research Report will focus on the 2-5-5-7 paragraph sequence for writing the Introduction, Methods, Results, and Discussion. Paragraph 2 should tell readers why and what the authors did in the study, and the ideas being tested. Hence, present the main or primary aim or hypothesis, followed by your secondary objectives. If you think that readers need to know more in order to understand your work, then use the Report in Context Box (which will not be counted as one of your 5 illustrations or as part of the total word count) to provide any extra detail that is necessary to understand the article. METHODS (5 PARAGRAPHS) Authors can use 5 paragraphs to describe what has happened to a single patient in their study, and how they have handled measurements and data about a patient. Paragraph 1 gives details of research ethics and/or Institutional Research Board (IRB) approval. Provide the local IRB number and date of approval, and whether consent for study inclusion was waived, or who gave consent. When reporting a clinical trial, provide the trial registration number from a registry that is a data provider to the World Health Organization International Clinical Trials Registry platform. For example, in the United States, use the 8-digit National Clinical Trial (NCT) Number assigned to your study when you registered with ClinicalTrials.gov. Also, provide the date when the protocol was first registered, and check that you are in full compliance with the International Committee of Medical Journal Editors (ICMJE) recommendations for clinical trial registration and reporting (3). Paragraph 2 clearly describes the who, what, where, and when. Provide the inclusion and exclusion criteria for patient selection. Check the details for reporting your research on the EQUATOR (Enhancing the QUality and Transparency of health Research) network website (4). Table 1 lists the most common types of studies submitted to PCCM, with the associated EQUATOR guideline. This list is not exhaustive and there are many other specialist guidelines available. For example, there is now a position statement on reporting prediction models in critical care journals (5). Refer to whichever one of the resources used and provide a checklist demonstrating adherence to the guideline in a Supplemental Digital Content (SDC) file. TABLE 1. - EQUATOR Network Resources Available for Commonly Reported Clinical Research Reports Type of Study or Report EQUATOR reporting guideline Randomized controlled trial CONSORT: Consolidated standards of reporting trials Interview or focus group COREQ: Consolidated criteria for reporting qualitative research Qualitative research SPQR: Standards for reporting qualitative research: a synthesis of recommendations Using observational health datasets RECORD: Reporting of studies conducted using observational routinely collected health data Observational studies STROBE: Strengthening the reporting of observational studies in epidemiology Quality improvement studies SQUIRE: Standards for quality improvement reporting excellence Diagnostic test accuracy STARD: Essential items for reporting studies of diagnostic accuracy Prognostic studies TRIPOD: Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis Clinical practice guidelines AGREE: A tool to improve reporting of clinical practice guidelines Systematic reviews and meta-analyses PRISMA: Preferred reporting items for systematic reviews and meta-analyses Paragraph 3 describes the measurements used for the primary and secondary outcomes, which are consistent with the primary and secondary aims of the study (i.e., match them to what was hypothesized in paragraph 2 of the Introduction). How were data fields defined and are you sure of the validity of these values in the analyses? For example, take a commonly used measurement, length-of-stay in days, about which the definition is often omitted in a report. Was it calculated from discharge date minus admission date to give days? Alternatively, was it an exact determination in hours and minutes from date and time of discharge minus date and time of admission, divided by 24 to give days? The definition used will impact whether the value is a continuous variable, or a number on an integer scale with a measurement error up to 1 whole day (in which reporting to 1 decimal place has no meaning). If the secondary outcome uses ventilator-free-days (up to day 30 from admission) as a continuous variable with normal distribution, then this derivation will only be valid when the second of the two definitions for length-of-stay has been used. The final two paragraphs of the Methods section outline the statistical plan. Be prepared to consult a statistician because the Journal is likely to do so if your report is headed toward acceptance. If reporting a clinical trial, provide the date when the study’s statistical plan was first registered and whether the plan in the report differs from the original registration. A clinical trial will not be published in an ICMJE-compliant journal if the statistical plan was registered after study completion, or there is insufficient reasoning about why the statistical plan was changed from the original registration. The strictness of this approach is because of the risk-of-bias, the possibility that readers will apply the findings to their practice and, ultimately, patient safety (3). Paragraph 4 is also used for details of power and sample size, description of tests, model-building and assessments, and their underlying assumptions. For example, a common topic for PCCM articles is to compare outcomes between PICUs. Check the sample size of annual admissions needed for comparing each PICU by, say, an annual probability of death when the overall rate is < 5% (see reference 6 for a valuable read, the 95% and 99% confidence intervals in Figure 3 in the paper will surprise you). Finally, how was the problem of missing data dealt with? Paragraph 5 and the SDC are used for an outline of any uncommon statistical approaches. We will publish more PCCM Notes, Methods, and Statistics articles focused on writing and reading reports using data science (e.g., big data, machine learning and artificial intelligence for causal inference or prediction modeling), databases or registries, quality improvement methodologies, and qualitative research. Until then, authors will need to provide enough guidance for readers to navigate their report. This paragraph is also the place to mention adherence to standards for statistical reporting. For example, when assessing and evaluating the prognostic ability of different, new, or updated risk-adjustment models in PICU populations, PCCM now expects–when appropriate–adherence to approaches like the Intensive Care National Audit and Research Center standard (7). That is, presentation of the suitability of the model using all three of the following measures: Discrimination, as assessed by the area under the received operating characteristic curve (8); Calibration as assessed by the Hosmer-Lemeshow C*-statistic (9) and Cox’s calibration regression (10); and, Overall fit of the model as assessed by Brier’s score (11). RESULTS (5 PARAGRAPHS) The 5 paragraphs and illustrations (up to 5 Figures and/or Tables in the entire article) of the Results section should tell two sides of the report’s narrative, without duplication of content. Can a reader be tempted or drawn-in to reading the full article by looking at the Title, Abstract, and the sequence of 5 Illustrations? Keep tables simple with no more than 8 columns width and 40 rows length. You do not have to present everything in the main body of the article, use SDC files if necessary. Figures should have no more than two subcomponents. Because most studies in PCCM use a retrospective design, in which the aims or hypotheses came after the events being researched, I will use this study approach as an example for writing the Results section. Language must be consistent with the study design. Avoid any inference about causation throughout a retrospective report and write only about ‘association’. Such writing is not easy. Many statistical tests use mathematically accurate terms such as “controlled”, “predictors”, “effects”, and “risk”, and the writing should correct potential confusion about what such information means derived from a retrospective study (see section on Discussion paragraph 6 for more details). Typically, paragraph 1 (and perhaps the first table) describes the study sample with demographic summary details. Reporting numbers and descriptive statistics are important and certain general principles must be followed (look closely at references 12 and 13). Use an appropriate degree of precision that is consistent with your measurements. Report the total sample size and group sizes. When using percentages, also report the numerator and denominator. Percentages are unnecessary when the denominator ≤ 30. Decimal places in percentages have little meaning with a denominator ≤ 100. Paragraph 2 gives sufficient information and analyses to understand the results of the primary outcome, which will be presented in paragraph 3. Often, the table shows multiple comparisons in clinical and demographic information about two outcome groups. In the past, reports have also used p values (‘p’ written lower case italic, or upper-case non-italic, no hyphen between ‘P’ or ‘p’ and value is needed) with these comparisons, and the question for readers is what is the author trying to say with all these comparisons? Corrections for multiple comparisons are unnecessary if readers are expected to do extra work and account for the multiple comparisons–but the author must make clear that readers need to do this informal assessment. If all the null hypotheses are true, 5% of the uncorrected p values will be < 0.05. This approach means that all comparisons should be reported, with no highlighting (or bold text) for the apparently ‘significant’ results. Alternatively, help the reader by providing a correction for the multiple comparisons. In this instance, the significance level is redefined as the chance of obtaining one or more significant results if all the null hypotheses in the ‘family’ of related tests are actually true (14). The simplest approach for a familywise error rate is the Bonferroni correction; divide 0.05 (5%) by, let’s say 25 comparisons reported in a table, and define a particular comparison as significant when the p value is < 0.002. Here, the 5% significance level applies to the entire family of comparisons rather than to each of the 25 individual comparisons. Some journals now take the view that because the p value is open to misinterpretation, it is better to replace them in any preliminary results with just estimates of effects or association and corresponding 95% confidence intervals (15). This approach certainly makes sense if the primary and secondary outcomes have been defined in the Methods (see Methods section, paragraph 3). Paragraph 3 (and likely your next illustrations) lays out the primary outcome. In retrospective studies this analysis often involves a multivariable logistic regression, with mathematical predictor or explanatory variables and some outcome such as mortality. A future article in the series of PCCM Notes, Methods, and Statistics will focus on regression models and confounding. However, there are three points to stress here. First, the number of outcome events in the study population will limit the number of variables that can be used in the model, or else there is risk of model over-fitting. This number is between 15 and 25 outcome events per variable studied (16), but fewer events per variable may be acceptable when propensity score matching is used (17). Second, report the regression coefficient (beta weight) for your primary explanatory variable and the associated confidence interval and p value. Third, be careful about declaring no association, or absence of any effect, which will usually require a process of exclusion. In this instance, it is more accurate to state failure to find an association between that variable and greater/lesser odds of the outcome of interest, which may be because of the study’s sample size, imprecise or invalid measurements, missing data, or reverse confounding (18) (see section on Discussion, paragraphs 1, 2, 5, and Paragraph 4 on the analyses for the secondary outcomes or Results require description of estimates and 95% confidence but no p Paragraph 5, the end of the Results section, can be used for any like the secondary outcome, statistical and reporting should be with estimates and 95% confidence but no p PARAGRAPHS) is about using a or structured Discussion (see reference 19 for an to the and reference for a a general do not content from the Introduction or give new or results. Here, the focus is on the significance of the study findings in to the the of the study and data, and any to clinical Paragraph 1 the two or three main points in the article (e.g., primary outcome, secondary outcome, and any 2, and 4 in each of the main points and them in the of the In an article of 2,500 to 3,000 words (including the there is no for ideas or that were actually studied in the report. Paragraph 5 is the place to any in the Be open about of the study and also whether any than the you from the the Use three bulleted points, each no more than 40 words, to provide more information. Give a short statement about what your study to the Avoid and do not the evidence that you have End with what you think is the clinical bottom what do you to do with this Paragraph 6 is the place in the Discussion authors meaning in the results and give some about how they are of the For example, regression will need to be terms Alternatively, what the results mean for Use a second information which can be the will not be counted as one of the 5 or toward the total word Give readers any extra and what you is the clinical bottom line. Paragraph is the final paragraph of the report. It could be used for a or to outline what to be in this particular of study, or what work is Avoid writing research is Try and be about a When using the of the structured article because your are you will need to two paragraphs from the Discussion if you want to them to the Methods and Results, or and use shorter For example, paragraph 6 on meaning can be the Results section and the paragraph of the Discussion. Paragraph of the Discussion can also be if the first paragraph in the Discussion is You have the to up to 40 references. Use 10 to 15 in the Introduction, some in the Methods, and that should leave up to for the Discussion. Try and be in your and make sure that you use PCCM’s of providing a to the 2,500- to 3,000-word Clinical Research Report is that it should help new writers prepare a for authors will know what to and of this content will be This of a whole article is not to be for word or to the between research authors and their Use the content of this article as a because the reviewers and will also to from readers will know what content should be where, and they can more this information and then about in-depth If you want to the short of an article with the then, the then, the paragraph of the Introduction and then, the first with two or three main and paragraphs of the and then through the illustrations and/or If the authors have now your full then now take your time and to the and the whole read from