01 · The Question
Who actually generated the evidence in this study?
A paper may discuss “older adults,” “university students,” “patients with diabetes,” or even an entire national population. Yet the data might come from volunteers at one university, patients attending selected hospitals, members of a particular database, or respondents who met several eligibility criteria.
Those distinctions matter. Major reporting guidelines ask researchers to report eligibility criteria, recruitment sources or settings, and methods of participant selection precisely because readers need this information to interpret who was actually studied and to whom the results might apply.
03 · What You Need to Know
How to determine what population was actually studied
Start by separating the target population from the people observed
The target population is broadly the population about which researchers wish to make inferences. The people who actually enter a study are usually a narrower group.
Terminology varies somewhat across disciplines. In epidemiological and registry contexts, authors may distinguish the target population from a source or accessible population and then from the study population or sample. One registry methods reference, for example, describes the target population as the group to which findings are intended to apply, the accessible population as those available through participating sites, and the actual study population as those who can be identified, invited, and agree to participate.
Target population
The broader population about which the investigators ultimately want to make an inference.
Source or accessible population
The population from which potential participants can realistically be identified or recruited.
Study sample
The people or units who actually enter the study after the recruitment and eligibility processes.
The labels are less important than keeping the groups conceptually separate. Different methodological traditions use these terms somewhat differently, so check how the paper itself defines them rather than assuming universal terminology.
Read the methods, not just the introduction
The introduction tells you the scientific population of interest. The methods tell you where the evidence came from.
STROBE recommends that observational studies report the setting, locations, relevant dates, eligibility criteria, and sources and methods of participant selection. CONSORT 2025 similarly requires trial settings and locations as well as participant eligibility criteria, and its explanatory guidance emphasizes that these details help readers judge applicability and generalizability.
When reading, therefore, look for concrete details: country, institution, clinic, school, community, registry, database, recruitment dates, age restrictions, diagnostic criteria, inclusion criteria, exclusion criteria, and other conditions governing entry into the study.
Eligibility criteria can substantially narrow the population
A paper described as studying adults with depression might actually include only adults aged 18 to 60 with a particular diagnostic definition, receiving care at participating outpatient clinics, without specified comorbidities, and willing to enter a trial.
That narrower description matters. CONSORT's explanatory guidance notes that inclusion and exclusion criteria are needed to judge to whom trial results apply and are central to considerations of external validity.
Do not treat eligibility criteria as administrative details buried in the methods. They help define the population represented by the study.
Setting is part of the population context
“Patients with hypertension” is less informative than “adults receiving hypertension care at four urban tertiary hospitals.” Participants recruited from specialist clinics may differ from people treated in primary care or those not receiving healthcare at all.
CONSORT 2025 specifically emphasizes reporting settings and locations because healthcare organization, resources, baseline risk, and social or cultural environments can affect the applicability of trial findings. Similar reasoning extends beyond clinical research. Students recruited from an elite residential university, for example, need not represent students in community colleges, distance-learning institutions, or universities in other countries.
The recruitment pool is not necessarily the final sample
A study may begin with thousands of potentially eligible individuals and end with a much smaller group. Some may fail eligibility screening. Others may decline participation, never respond to an invitation, withdraw, or provide unusable data.
STROBE recommends reporting numbers at relevant stages, including those potentially eligible, assessed for eligibility, confirmed eligible, included, completing follow-up, and analyzed, along with reasons for non-participation where appropriate.
Therefore, “5,000 people were invited” does not mean that 5,000 people were studied.
Who enrolled and who was analyzed may also differ
Suppose 800 participants enroll, but the primary analysis includes 617 because some participants have missing outcome data. Which number defines the study?
Both numbers may matter, but they answer different questions. The enrolled sample describes who entered the study. The analytic sample identifies whose data directly contributed to a particular analysis. CONSORT 2025 asks trial reports to define who is included in each analysis and how missing data are handled, while STROBE asks observational studies to report participant numbers across study stages.
When the distinction affects interpretation, determine whether all recruited participants were included in the analysis.
The sample size does not define the population by itself
Knowing that a study included 1,200 participants tells you how many observations were available, but almost nothing about who those participants were.
A large convenience sample from one narrowly defined source can still differ systematically from the target population. Conversely, a smaller probability sample may have a clearer relationship to a defined population. Sample size and population representativeness are different methodological issues.
Do not infer representativeness from demographic diversity alone
A sample can contain participants of different ages, sexes, socioeconomic backgrounds, or ethnic groups without being statistically representative of the population from which researchers wish to generalize.
Representativeness depends on how the sample relates to the target population and how participants were selected, not simply on whether the demographic table looks varied. NHLBI appraisal guidance similarly treats careful definition of the target population and representation of that population as relevant to judging how well a study addresses its research question.
Participant selection deserves its own appraisal
Once you know the population from which participants came, the next question is how they got into the study. Recruitment through probability sampling, consecutive clinic enrollment, advertisements, convenience sampling, volunteer panels, registries, and other mechanisms creates different pathways from a source population to an observed sample.
That is why identifying the population should be followed by examining how participants were actually selected. The two questions are closely related, but not interchangeable: one identifies the relevant groups, while the other examines the mechanism that produced the sample.
| Group |
Question to ask |
Example |
| Target population |
Who do the researchers ultimately want the findings to inform? |
Undergraduate students in the country |
| Source or accessible population |
From whom could participants realistically be recruited? |
Students enrolled at four participating universities |
| Eligible population |
Who met the study's inclusion and exclusion criteria? |
Full-time undergraduates aged 18 or older at those universities |
| Enrolled sample |
Who actually entered the study? |
1,040 consenting students |
| Analytic sample |
Whose data contributed to the analysis of interest? |
912 students with the required data |
The exact population may change across analyses
A paper can contain several analytic samples. One outcome may be available for nearly everyone, another only for a subgroup, and a longitudinal analysis only for participants retained at follow-up.
So asking “What population was studied?” may eventually require a more precise question: “Which participants generated this particular result?” That distinction becomes especially important when interpreting subgroup, longitudinal, complete-case, or secondary analyses.
06 · What This Means for You
Describe the population narrowly before considering generalization
When appraising a study, first write a literal description of the people or units represented in its data. Include the source, setting, major eligibility restrictions, and any important difference between enrollment and analysis.
Only after doing that should you ask whether the findings might extend to a broader population. This order matters because otherwise it is remarkably easy to start with the population named in the introduction and quietly treat the observed sample as though it represented that population by definition.
A simple population framework
If the paper names a broad population
Check the methods for the actual recruitment setting, sampling frame, and eligibility criteria before adopting that label.
If participants came from selected institutions or locations
Include those restrictions in your description of the observed population.
If only some eligible people participated
Distinguish the eligible or invited population from the participants who actually enrolled.
If the analytic sample is smaller than the enrolled sample
Identify whose data contributed to the specific result you are interpreting and investigate the reason for the difference.
This gives you a defensible answer to a basic appraisal question: who does this evidence directly describe? Whether it can reasonably inform decisions about people beyond that group is a subsequent question about applicability and external validity.
07 · A Quick Checklist
Before describing the study population, trace who actually contributed data
When identifying the study population, check:
Identify the broader population the researchers appear to want their findings to inform.
Find the institutions, communities, databases, clinics, schools, or other sources from which participants could be recruited.
Record the geographical location and study setting when they materially define the population.
Read every important inclusion and exclusion criterion rather than relying on the broad participant label.
Determine how many people were potentially eligible, invited, enrolled, retained, and analyzed when those numbers are reported.
Check whether nonresponse, withdrawal, missing data, or exclusions changed who ultimately contributed evidence.
For the result you are interpreting, identify the relevant analytic sample rather than assuming every enrolled participant contributed.
Keep description of the observed sample separate from claims about representativeness or generalizability.
