The STROBE Checklist: A Practical Guide to Reporting Observational Studies in 2026

Most medical researchers know that randomized controlled trials need a CONSORT checklist. Far fewer have the same clarity about what they owe a journal when they submit a cohort study, a case-control analysis, or a cross-sectional survey. The answer is STROBE, and the honest assessment is that most authors file the checklist without understanding what the 22 items actually require. That matters because journals are increasingly enforcing compliance during desk review, and reviewers are more likely to request major revisions on incomplete observational reports precisely because observational evidence is harder to interpret when its context is missing.

The STROBE statement (Strengthening the Reporting of Observational Studies in Epidemiology) was published simultaneously in October and November 2007 across several leading journals, including PLoS Medicine, the Annals of Internal Medicine, BMJ, and Epidemiology. The group behind it was international, drawn from epidemiologists, statisticians, journal editors, and methodologists, and the checklist was designed to describe minimum reporting expectations rather than prescribe how studies should be designed or conducted. That distinction matters. STROBE does not tell you how to handle confounding. It tells you to report what you actually did.

In September 2025, JAMA Network Open published STROBE-Equity (2025;8(9):e2532512), a formal extension led by researchers at McMaster University and co-published in The BMJ, which adds 10 items specifically for studies reporting on health equity, disparities, and social determinants. If your cohort study includes data stratified by race, ethnicity, socioeconomic position, disability status, or geographic marginalization, STROBE-Equity now applies alongside the core checklist. The main STROBE statement alone is no longer enough for that work at journals that follow BMJ Group or JAMA Network editorial standards.

Which Study Designs STROBE Covers

STROBE applies to three observational study designs: cohort studies, case-control studies, and cross-sectional studies. The checklist has two layers. Eighteen items are shared across all three designs. Four items differ by design, because what you need to explain about participant selection, follow-up, and outcome assessment varies depending on how the study was structured. If you have used a nested case-control design or a case-cohort study, the case-control version of the checklist is the closer fit, and the methods section should say so explicitly.

What STROBE does not cover is equally worth knowing. Randomized trials use CONSORT. Diagnostic accuracy studies use STARD. Systematic reviews use PRISMA. Prediction model studies use TRIPOD. If your observational study sits at the intersection of these (a diagnostic cohort, a registry-based prediction study, an analysis embedded in a systematic review), EQUATOR Network guidance suggests specifying which primary checklist applies and noting any supplementary guidance. The Explanation and Elaboration document published alongside STROBE in PLoS Medicine gives worked examples for ambiguous cases and remains freely available online.

Pharmacoepidemiology studies using administrative data have their own extension, GRACE, developed through collaboration between the International Society for Pharmacoepidemiology and several major journals. Molecular epidemiology studies have STROBE-ME. Antimicrobial stewardship studies have STROBE-AMS. These extensions supplement rather than replace the core 22-item checklist, so the submission process typically involves both documents. When in doubt, the EQUATOR Network maintains a searchable database of over 500 reporting guidelines organized by study type, and checking it before submission takes less than five minutes.

How the 22-Item Checklist Works in Practice

The standard submission process for a journal that requires STROBE involves providing a completed checklist alongside the manuscript. Each of the 22 items carries a page number reference pointing to where in the manuscript the relevant information appears. This is not a box-ticking exercise. A page reference to a blank methods section or a one-line summary where a full account is expected signals to the editor that the manuscript is not ready.

The 22 items follow the structure of the manuscript: title and abstract (item 1), background and objectives (items 2 and 3), study design and methods (items 4 through 12), results (items 13 through 17), and discussion through funding disclosure (items 18 through 22). Each item group reflects where in the paper the information should actually appear. Authors who consolidate everything into the methods section and skip reporting in the results or discussion still fail the checklist, even if the underlying information exists somewhere in the manuscript.

The companion Explanation and Elaboration (E&E) document is as important as the checklist itself. It clarifies what each item actually means, gives examples of adequate and inadequate reporting from published papers, and explains the rationale behind requirements that can seem arbitrary when read in isolation. Authors who fill out the checklist without reading the E&E frequently misinterpret what items 7, 9, and 12 require, which are consistently the items generating the most revision requests at methodologically rigorous journals.

The Methods Section Items Journals Check Most Carefully

Research auditing STROBE compliance across published observational studies in clinical journals has consistently found the weakest reporting on a cluster of methods items: study size justification (item 10), sources of potential bias and how they were addressed (item 9), handling of quantitative variables when grouped or categorized (item 11), and statistical methods for confounding adjustment including subgroup and sensitivity analyses (item 12). These are not difficult concepts, but they are the ones authors most often describe vaguely or skip entirely.

Item 10 is one of the most commonly absent. If you did not conduct a formal power calculation before collecting data, as is common in administrative data studies or registry analyses, the item still requires an explanation. A statement that the study used all available records from the registry is acceptable, but it should also describe what precision the available sample size provides, or acknowledge that the study is exploratory and note its implications for interpreting confidence intervals. Saying nothing leaves the editor without a way to assess whether the study was adequately sized to detect the effects it reports.

Item 9, on bias, deserves particular attention. The E&E document distinguishes between reporting the sources of bias the team considered and describing any steps taken to limit those biases. These are different obligations. Many authors report that recall bias was a limitation without explaining whether the interview structure was designed to minimize it, whether self-report was validated against records, or whether a sensitivity analysis was run excluding subgroups most prone to it. That fuller account is what item 9 expects, not a list of risks with no account of what the research team did about them.

Item 7, which covers variables, requires you to define all outcomes, exposures, and confounders, including how they were measured. If a variable was derived from an ICD code list, name the codes or cite the published derivation algorithm. If a self-reported outcome was based on a validated instrument, name the instrument and the scoring method. If a potential confounding variable was included based on a directed acyclic graph (DAG), mention the DAG approach even briefly, since it signals a principled rather than ad hoc approach to covariate selection. Editors reviewing for internal consistency will check whether the variables in the results tables match the definitions in the methods. They usually do not match as precisely as authors assume.

Reporting Confounding: Where Most Papers Fall Short

The handling of confounding in the methods is the area where observational study reporting diverges most sharply between papers that satisfy reviewers and papers that generate lengthy revision requests. STROBE item 12 requires you to describe all statistical methods, including those used to control for confounding, but many authors treat this as a single sentence naming their regression model. That is not enough.

What editors and methods reviewers want to see is an account of how confounders were selected. Was variable selection based on prior literature, clinical knowledge, a predefined causal model, or a data-driven approach? If you used stepwise selection, that should be stated. If you used a propensity score method, describe the variables included in the propensity model, the matching or weighting approach used, and how covariate balance was assessed after matching or weighting. If you used multivariable regression, list the covariates and the rationale for their inclusion. The common failure is to report a Table 1 with fifteen baseline variables and then run a regression described only as "adjusted for baseline characteristics" without specifying which characteristics were included and why, or whether the selection was made before or after viewing the outcome data.

Missing data is a connected problem that item 12 also addresses. Complete-case analysis is acceptable but must be named as such rather than left unstated. If a substantial proportion of a key variable is missing, a note on whether the missing data appeared random or showed a pattern related to exposure or outcome status matters for interpreting the results. Journals at NEJM, The Lancet, and BMJ have increasingly expected either multiple imputation or clearly labeled sensitivity analyses around missing data assumptions for studies where the extent of missingness in a key variable is non-trivial. Whether or not the target journal explicitly requires this, stating what you did is always better than silence.

Residual confounding deserves explicit mention in the methods or limitations even when it cannot be quantified. If unmeasured variables are plausibly related to both your exposure and your outcome, saying so, and reasoning about the probable direction of the bias it would introduce, is more credible than omitting the issue. Reviewers at strong journals will raise it regardless. Anticipating that critique in the paper itself, with honest reasoning, is almost always the better strategy.

Results Section: What STROBE Requires Beyond Tables

The results section of most observational studies looks similar at first glance: a participant count, a Table 1 with baseline characteristics, and a Table 2 with the main adjusted estimates. STROBE formalizes what belongs in each of these and adds items that many authors omit because they feel redundant with information already in the tables.

Item 13 asks for a description of participant flow, including numbers at each stage of screening, eligibility, and inclusion. For a cohort study, this includes the follow-up experience: how many participants contributed data at the end of the study period, how many withdrew or were lost, and over what time frame. This information is often buried in methods or absent entirely. Journals that enforce STROBE during desk review frequently return papers because authors have not provided a flow diagram or equivalent table that accounts for participant numbers at every stage. For case-control studies, the number of controls per case and any refusals or non-participants among controls should appear here, not only in the methods.

Item 16 requires both unadjusted and adjusted estimates with their confidence intervals. If you are reporting only adjusted estimates because the unadjusted ones were confounded beyond interpretable range, a note explaining that decision belongs in the text. Reviewers reading a table with only adjusted odds ratios cannot assess whether the adjustment changed the direction of the effect or merely attenuated its magnitude. That comparison is often the most scientifically meaningful part of the results. Reporting both, and commenting briefly on the difference, strengthens rather than weakens the paper.

Item 17, on other analyses, encompasses subgroup analyses, sensitivity analyses, and analyses for effect modification. These belong in the results section with clear labels, not buried in a supplementary appendix that reviewers may not request access to. If you conducted a sensitivity analysis excluding participants with incomplete follow-up, or re-ran the primary analysis using a different exposure definition or a more conservative outcome threshold, those should appear in the paper with their results described in the text. A table without commentary does not satisfy this item.

The Discussion: Limitations as a Required Section

STROBE item 19 requires a discussion of limitations. That sounds obvious, but the item specifies something more than a list. The E&E document describes the expected content as a discussion of sources of potential bias or imprecision and a realistic consideration of the direction and magnitude of potential bias. An undirected list of limitations with no commentary on which are most likely to affect the conclusions does not satisfy this item.

Reviewers at leading epidemiology journals often focus their critique precisely here. They want to know whether residual confounding is plausible given what is known about the literature. They want to know whether information bias in the exposure measurement would have attenuated or inflated the effect estimate and in which direction. They want to understand whether the outcome definition captures the clinical entity of interest or a proxy that may differ systematically in subgroups. Stating these limitations with that specificity, even when the study cannot resolve them, is more credible than either omitting them or treating the limitation section as a brief disclaimer before the conclusion.

Item 21, on generalizability, is a related area where vague answers cause problems. The external validity of a case-control study conducted in a single hospital system differs from a national cohort study, and the setting and methods sections should already give readers the information they need to assess this. The discussion should apply that information explicitly rather than state generically that the findings may not generalize to all populations. Which populations? On what basis? That level of specificity is not always available, but when it is, it belongs in the paper.

The STROBE-Equity Extension: What Changed in September 2025

The STROBE-Equity extension was published simultaneously in JAMA Network Open (2025;8(9):e2532512) and The BMJ in September 2025. It was led by Omar Dewidar, Larissa Shamseer, GJ Melendez-Torres, and colleagues affiliated with McMaster University, following a Delphi consensus development process and established methods for guideline extension development endorsed by EQUATOR. The 10 new items address reporting gaps that became apparent as observational health research increasingly included equity dimensions that the 2007 STROBE items could not fully capture.

The extension focuses on five areas. The first is conceptual framing: how the study defined health equity, and whether that definition drew on a recognized theoretical framework or was left implicit. The second is variable operationalization: how equity-relevant variables such as race, ethnicity, income, education level, disability status, and geographic context were defined, measured, and validated in the study population. This matters because administrative data sources code these variables differently, and the same label can mean different things across databases.

The third area is intersectionality: if the study analyzed multiple equity dimensions simultaneously, how was joint exposure to several disadvantaged statuses handled analytically? The fourth is community involvement: whether the study was conducted with participation from members of the populations under study, and at what stage of the research. This is a transparency requirement, not a moral requirement, and studies that did not involve the community should say so clearly rather than omit the question. The fifth is implications for equity: what the results mean for reducing or widening the disparities the study measured.

As of June 2026, no major publisher has formally mandated STROBE-Equity as a required submission document alongside the main STROBE checklist. The extension was, however, co-published in The BMJ and JAMA Network Open, and journals aligned with those editorial standards have been applying equity-sensitive review criteria even before any formal mandate. Authors submitting equity-relevant observational work to journals in those families, to The Lancet, or to specialty journals in global health, social medicine, or health disparities should expect reviewers to apply STROBE-Equity criteria whether or not it appears in the submission checklist. Preparing for it proactively is easier than revising for it after peer review.

How Journals Actually Enforce STROBE at Submission

Enforcement varies considerably across journals, which is itself a problem for authors trying to calibrate how much effort to invest. Some journals ask for the STROBE checklist in the submission system but do not cross-check it against the manuscript before sending the paper to review. Others have desk editors or statistical reviewers who verify checklist items before external review even begins. A number of journals, including several in the European Journal of Epidemiology family, have conducted systematic reviews of STROBE adherence across their published papers and incorporated checklist assessment into their peer review process.

In practice, the safest assumption is that the checklist will be checked. Journals that include STROBE as a required submission file, rather than listing it as a recommendation, do verify it at desk review. A manuscript that lists page numbers but cannot deliver the required content at those pages will be returned before any reviewer sees it. That is frustrating but ultimately efficient: it costs the author less time than a major revision request requiring a rewrite of the methods and results sections three months after submission.

Journals that list STROBE as recommended rather than required still apply it. The difference is editorial discretion. A clearly well-conducted study with one or two minor STROBE gaps may proceed to review with a note to address the checklist at revision. A study with multiple vague methods sections, an undescribed bias analysis, and a single sentence on study size will often receive a desk rejection regardless of whether the journal calls the checklist mandatory. The checklist, for these editors, is shorthand for asking whether the paper is written to the standard that makes it reviewable at all.

A Pre-Submission STROBE Audit

The most practical way to use STROBE is to print the checklist specific to your study design, fill in page numbers, and then go back to each cited page and read what actually appears there. Do not fill in the checklist from memory. Many authors who have been through a paper a dozen times genuinely believe a particular item is addressed when it is not, because the information exists in their head but was never transferred to the manuscript.

If your paper reports any equity dimension, download the STROBE-Equity checklist from the McMaster University project site and work through the 10 additional items with the same care. Given that journals in global health and public health are now applying equity-sensitive review criteria even when they do not formally require STROBE-Equity at submission, preparing the manuscript to meet those items during initial drafting costs little effort. Addressing them as a revision, after a reviewer has raised them specifically, is far more labor-intensive and delays publication without improving the science.

The final step worth taking is a consistency audit across sections. Do the variables defined in the methods appear in the results tables? Do the statistical methods described include all models shown in the paper? Does the limitations discussion agree with the results discussion on the likely direction of bias? Inconsistency across sections of an observational paper is the single most common reason a reviewer requests major revisions on an otherwise sound study. STROBE is largely a tool for preventing that inconsistency. The checklist asks you to describe what you did, where you did it, and what it means. Answering those questions completely, in the right sections, is what distinguishes a paper that moves through review from one that circles back to the authors twice before acceptance.

Further Reading