STROBE for observational studies: a practical reporting guide

Stetleaf Academy

Reviewed by Alex Juusela, MD, MPH, FACOG, Maternal-Fetal Medicine Physician · Last reviewed 2026-11-05

Observational studies describe exposures and outcomes without randomly assigning the exposure under study. Their reports need to explain how participants entered the study, how variables were measured and how the analysis addressed alternative explanations. STROBE provides a reporting framework for cohort, case-control and cross-sectional studies. This guide approaches it from the author's side: what information should a reader be able to find in your manuscript?

Why it matters

The STROBE statement contains 22 reporting items for these three observational designs [1]. Its explanation and elaboration discusses the reasoning behind the guidance and gives examples [2]. The purpose is clearer reporting, not a numerical appraisal of study quality. A completed checklist does not show that bias has been removed or that an association is causal.

Observational reports can become difficult to interpret when their descriptions omit the source population, the timing of measurements or the reasons for adjustment. Clear writing cannot repair those design limitations. It can expose them so that readers understand the evidence being offered. Start with the study records and analysis documentation, not with a model paragraph borrowed from a different project.

Key ideas

Name the actual design

A cohort study follows people from an exposure or other starting point toward outcomes. A case-control study selects participants according to outcome status and compares exposures. A cross-sectional study examines exposure and outcome at a defined time or period without the same follow-up structure. Name the design and describe the timing rather than relying only on the broad label retrospective.

Choose the STROBE material that matches the design. Eligibility, sampling, matching and follow-up raise different reporting questions in different designs. If routinely collected health data are used, consider the RECORD extension alongside STROBE [3]. Do not assume that using a large database removes the need to describe participant selection.

Describe the route into the analysis

State the setting, dates, eligibility criteria and source population. Explain how participants were identified and selected. For case-control studies, describe the source of controls and why it fits the question. For cohort studies, explain follow-up. If matching was used, describe its criteria and how it was addressed in analysis [1,2].

Account for numbers at successive stages and explain exclusions. A table or flow account should allow a reader to distinguish people never eligible from people excluded because information was unavailable. Keep this account consistent with the sample sizes in the results tables.

Define and measure the variables

Describe exposures, outcomes and other variables in operational terms. Give the source of measurements, relevant time windows and any differences between groups. If a category is created from a continuous measurement, explain its boundaries. A phrase such as high exposure is not a reproducible definition unless the report tells readers how that category was assigned.

For routinely collected records, document how codes or algorithms identify the study population and variables [3]. Distinguish an absent entry from an observed negative finding. The reporting task is to make the measurement process visible, including its weaknesses, rather than treating every field in a dataset as equally trustworthy.

Make adjustment understandable

Explain which potential confounders were considered, why they were selected and how they entered the analysis. Report unadjusted and, where applicable, adjusted estimates with their precision, identifying the variables used for adjustment [1,2]. A reader should not have to reverse-engineer the adjustment set from a table footnote that says only fully adjusted.

Adjustment is not a promise that all confounding is gone. Avoid saying that a change after adjustment proves which factor caused the original association. Measured variables may be incomplete, and unmeasured factors may remain. The explanation of limitations should reflect those possibilities rather than convert an adjusted association into a causal conclusion.

Explain missing data and sensitivity analyses

Show how much information is missing for important variables and describe how missing data were handled. A complete-case analysis and an analysis using imputation make different demands on the report. Describe the actual method and its assumptions at a level that allows readers to understand the approach. Label additional analyses clearly rather than implying that every analysis was planned from the outset.

Discuss what the report cannot establish

Discuss likely sources of bias, imprecision and limits to applicability [2]. Specify the concern: selection into the dataset, exposure measurement, outcome ascertainment or residual confounding. General statements that every study has limitations do little work. Interpret the results in relation to the question, design and uncertainty, including findings that do not support the initial expectation.

Worked example

Illustrative example (invented). Not a real study.

An invented cohort examines whether travel distance is associated with attendance at a follow-up clinic. Nobody assigns where participants live. The team uses records from one clinic and follows attendance after the initial appointment.

State the design. The methods identify a cohort, its starting event and the follow-up window. They explain how travel distance was measured.
Account for selection. The dataset contains 500 eligible people. Distance is unavailable for 40, leaving 460 in a complete-case analysis. Of these, 180 are in the longer-distance category and 280 in the shorter-distance category.
Define categories. The report states the distance threshold and why it was chosen instead of using unexplained near and far labels.
Describe adjustment. The analysis plan identifies potential confounders and the reason for including each. The results distinguish the crude association from the adjusted estimate and give uncertainty for both.
Discuss the gap. The authors explain that the 40 excluded people may differ from those analyzed, and that unmeasured transport access could affect interpretation.

These invented counts demonstrate selection reporting only. They supply no observed association, adjusted result or evidence about clinic attendance.

Common pitfalls

Using retrospective as the whole design description. Explain selection, timing and follow-up.
Reporting only the final sample. Show how eligible people reached each analysis.
Leaving adjustment unexplained. Name variables and the rationale for their inclusion.
Treating missing records as negative outcomes. Describe what absence of information means.
Calling adjustment a cure for confounding. Discuss remaining uncertainty.
Treating checklist completion as methodological approval. Keep reporting and methodological appraisal distinct.

Quick self-check, in our own words

I named the design and described its timing.
I explained the source population and participant selection.
My variable definitions include measurement sources and time windows.
I accounted for exclusions and missing information.
I named the adjustment variables and described the analysis.
I kept conclusions proportionate to the design and uncertainty.
I considered RECORD if the data were routinely collected.

References

von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med. 2007;4(10):e296. doi:10.1371/journal.pmed.0040296. PMID: 17941714.
Vandenbroucke JP, von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. PLoS Med. 2007;4(10):e297. doi:10.1371/journal.pmed.0040297. PMID: 17941715.
Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med. 2015;12(10):e1001885. doi:10.1371/journal.pmed.1001885. PMID: 26440803.