What Study Design Was Actually Used Rather Than Merely Named by the Authors?

Does the design label match what the researchers actually did?

The methods section calls the research “cross-sectional,” “prospective,” “experimental,” “case-control,” or a “cohort study.” Should you simply accept that label?

Use it as a starting point, not as the end of the appraisal. Study-design terminology is sometimes incomplete, ambiguous, or used differently across disciplines. The more reliable approach is to reconstruct the architecture of the study: how participants entered, whether researchers assigned an intervention, how groups were formed, when exposures and outcomes were measured, whether participants were followed through time, and how the analysis used those observations.

Identify the design from the study's defining features

Reconstruct the study before naming its design

Study design is an architecture, not merely a label

Terms such as cohort, case-control, cross-sectional, randomized trial, crossover trial, and interrupted time series summarize important structural features. They are useful precisely because those structures affect what comparisons can be made and what biases or alternative explanations deserve attention.

STROBE focuses on three major analytical observational designs: cohort, case-control, and cross-sectional studies. The initiative itself emphasizes reporting what was planned, what was done, what was found, and what the results mean. It also explicitly notes that its reporting recommendations are not prescriptions for how studies should be designed or conducted.

For critical appraisal, therefore, use the design name as shorthand only after you understand the procedures behind it.

First ask whether researchers assigned the exposure or intervention

This is often the most useful first division.

This distinction is more informative than asking whether the authors use the word “experimental.” An educational researcher who administers a questionnaire and compares naturally occurring groups has not created an intervention merely by calling the study experimental. Conversely, a deliberately assigned instructional condition has an interventional feature even if the authors use an unconventional design label.

If an intervention was assigned, ask how allocation occurred

Assignment alone does not establish randomization. Determine whether participants, clusters, classes, clinics, communities, or other units were allocated using a genuinely random mechanism.

If allocation was random, identify the unit randomized. Individual randomization and cluster randomization create different study structures. If allocation was determined by availability, participant choice, alternating assignment, date, institutional policy, investigator judgment, or another non-random process, the study should not be treated as an individually randomized trial merely because it contains intervention and comparison groups.

Random sampling and random allocation answer different design questions

A study can randomly sample participants and remain observational. Another can recruit a convenience sample and then randomly allocate those participants to interventions.

Random sampling concerns how participants were selected from a population or sampling frame. Random allocation concerns how enrolled units were assigned to study conditions. Confusing these two forms of randomization can lead to a fundamental misclassification of the study.

For observational studies, ask how participants entered relative to exposure and outcome

The temporal and sampling structure can help distinguish common observational designs.

These descriptions are deliberately structural rather than merely terminological. Real studies can contain hybrid features, nested designs, repeated cross-sectional samples, or secondary analyses of existing data that require more precise descriptions.

“Prospective” and “retrospective” do not fully identify the design

These words describe temporal aspects of how data or events relate to the research process, but they are not complete substitutes for a design description.

Calling something a “retrospective study” does not tell you whether participants were selected according to outcomes, whether an existing cohort was analyzed, whether records were reviewed cross-sectionally, or how exposure and outcome were temporally related.

Similarly, “prospective” does not automatically mean cohort study, randomized trial, or stronger evidence. Describe what was actually done.

A longitudinal study is not automatically a cohort study

Longitudinal simply indicates that observations extend across time. Randomized trials, cohort studies, panel studies, repeated-measures experiments, and other designs can all be longitudinal.

If authors use “longitudinal” as the primary design label, ask what generates the longitudinal structure. Were the same individuals followed? Were different samples drawn repeatedly? Was an intervention introduced? How were exposures and outcomes ordered?

A pretest-posttest structure does not by itself establish an experiment

Measuring participants before and after something happens creates a temporal comparison. It does not tell you whether the change was caused by the intervention.

A single-group pretest-posttest study lacks a concurrent comparison group. A non-randomized controlled pretest-posttest design adds a comparison group but may still differ systematically between groups. A randomized pretest-posttest trial adds random allocation.

These designs can look superficially similar in a table containing “pre” and “post” columns while supporting different inferences.

Identify exactly how comparison groups were created

Groups may arise through random allocation, researcher assignment, naturally occurring exposure, participant choice, institutional membership, outcome status, or analytical categorization.

This mechanism is often more informative than the label attached to the groups. Before interpreting an effect or association, determine what the actual comparison was and how its two sides came to exist.

Cross-sectional does not simply mean “a survey”

A survey is a method of data collection, not automatically a study design. Surveys can be used in cross-sectional studies, longitudinal panels, cohort follow-ups, experiments, and other designs.

Conversely, a cross-sectional study need not use a questionnaire. It might analyze clinical examinations, laboratory measurements, administrative records, imaging, or other data collected or defined within a cross-sectional framework.

Keep the method of measurement separate from the architecture of the study.

Secondary data do not automatically make a study retrospective

Researchers can conduct a new analysis using data originally collected for another purpose. The relevant design depends on how the underlying observations were generated and how the current analysis defines participants, exposures, outcomes, and time.

A secondary analysis of a randomized trial remains grounded in randomized assignment for comparisons that preserve that randomization. An analysis of an established prospective cohort does not become cross-sectional merely because the analyst receives the dataset years later.

This is why you should inspect what data were actually analyzed rather than classifying the research according to when the analyst opened the file.

Some studies genuinely combine designs

A mixed-methods project may contain a trial plus qualitative interviews. A cohort may contain a nested case-control analysis. A randomized trial may include an observational mediation study. A repeated cross-sectional survey may be conducted before and after a policy change.

In such cases, forcing the entire paper into one label can obscure rather than clarify the evidence. Identify the design relevant to the specific result you are appraising.

Study design labels do not determine quality by themselves

Recognizing the design tells you what methodological questions to ask next. It does not automatically tell you whether the study was well conducted.

A randomized trial can suffer from missing outcomes, deviations from intervention, measurement problems, selective reporting, or flawed analysis. A carefully conducted observational study can provide valuable evidence for questions that cannot or should not be randomized.

Design identification is therefore a starting point for appraisal, not a quality score.

When an “experimental study” is not a randomized experiment

Common mistakes when identifying study design

Describe the architecture first and attach the label second

When a design label seems uncertain, write a short procedural description before trying to name it. This often resolves the ambiguity and prevents the label from doing more intellectual work than it deserves.

Once you have reconstructed these features, compare your description with the authors' terminology. Agreement gives you greater confidence that the label is informative. Disagreement tells you to keep the concrete design features visible throughout your appraisal.

Before accepting the design label, reconstruct the study

Questions about identifying study design from the methods

The methods define the design more reliably than the label

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.