Primary, Secondary and Evidence-Synthesis Research

Overview

Health researchers can generate evidence in three broad ways. Primary research collects new data specifically for the current study. Secondary data research analyses data that already exist, often collected for clinical care, administration, surveillance or an earlier study. Evidence synthesis systematically identifies and integrates findings from multiple existing studies.

These categories describe the source and level of evidence generation, not research quality. A well-designed secondary analysis may answer an important question more efficiently than a new study. A poorly conducted primary study remains weak despite collecting original data. An evidence synthesis is only as trustworthy as its question, search, eligibility decisions, appraisal and synthesis.

Why this matters

Researchers sometimes default to primary data collection because it feels more original. This can duplicate existing work, delay answers and use participants or funds unnecessarily. Conversely, existing data may be convenient but poorly matched to the research question. A systematic review may appear easier than field research but requires extensive planning, searching, screening, appraisal and transparent synthesis.

The first decision is therefore not “which dataset is available?” It is “what evidence is needed to answer the question, and which source can provide it with acceptable validity, feasibility and ethics?”

Primary research

DESIGN PROFILE

Primary research

Core structure: Researchers prospectively collect new data, specimens, observations or accounts for the current research objectives.

Best suited for: Questions requiring variables, timing, populations, interventions or experiences not adequately represented in existing sources.

Important limitations: Greater time, cost and participant burden; recruitment challenges; possible loss to follow-up; and need to build and quality-assure all data-collection systems.

Primary research may be quantitative, qualitative or mixed methods. It includes surveys, cohort studies, trials, diagnostic studies, interviews, focus groups, observations and laboratory research.

Its major advantage is control over what is measured, when and how. Researchers can define eligibility, standardise instruments, establish time order and collect variables directly aligned with the question. However, this control is not absolute. Non-participation, measurement error, protocol deviations and missing data can still affect validity.

Primary research is justified when:

• the needed data do not exist;
• existing measures are invalid or insufficient;
• the population or setting is not represented;
• a new intervention must be assigned and evaluated;
• prospective timing is essential; or
• participant experience or context must be generated directly.

Secondary data research

DESIGN PROFILE

Secondary data analysis

Core structure: Researchers use an existing dataset to answer a question that may differ from the original purpose of data collection.

Best suited for: Efficient study of large populations, rare outcomes, long time periods, routine practice, health services, surveillance and questions where suitable variables already exist.

Important limitations: Limited control over variable definitions, missingness, coding, measurement timing, changes in practice, linkage and the population included in the dataset.

Sources include electronic health records, administrative claims, registries, surveillance systems, national surveys, biobanks, trial datasets and earlier research databases. Secondary analysis is not merely downloading data and running a model. Researchers must understand data provenance: why the data were created, who was included, how variables were recorded, how coding changed and what information is absent [1].

A variable collected for billing may not represent clinical severity. A diagnosis code may reflect rule-out evaluation rather than confirmed disease. A laboratory system may include repeat tests but not distinguish clinical episodes. A registry may exclude patients treated outside participating centres.

Evidence-synthesis research

DESIGN PROFILE

Evidence synthesis

Core structure: Researchers use transparent, reproducible methods to identify, appraise and synthesise findings from multiple existing studies.

Best suited for: Summarising what is known, estimating combined effects where appropriate, comparing bodies of evidence, identifying uncertainty and informing guidelines or future research.

Important limitations: Publication and selective-reporting bias, heterogeneity, poor primary-study quality, incomplete retrieval and inappropriate synthesis of dissimilar evidence.

Evidence synthesis includes systematic reviews of interventions, diagnostic accuracy, prognosis, prevalence, qualitative evidence and other questions. It may include meta-analysis, but meta-analysis is a statistical technique rather than a synonym for systematic review. A systematic review can be rigorous without pooling when studies are too heterogeneous.

A scoping review is usually used to map the extent, characteristics or concepts of a broad evidence field, clarify definitions or identify gaps. A systematic review is generally better when the objective is a focused answer based on explicit eligibility, appraisal and synthesis. The review type should follow the purpose rather than the desired label [2].

The Cochrane Handbook provides current detailed guidance for systematic reviews of healthcare interventions, including planning, searching, selection, data collection, risk of bias, analysis and interpretation [3]. PRISMA 2020 supports transparent reporting but is not a substitute for sound conduct [4].

Comparing the three sources

Choosing a source of research evidence

illustrative

Table 1. Primary, secondary and evidence-synthesis research

Feature	Primary research	Secondary data research	Evidence synthesis
Main unit	Newly recruited participants, specimens, settings or observations	Existing participants, records, events or datasets	Existing studies or reports
Control over measurement	Usually greatest, if planned well	Limited by original data purpose and systems	Limited by reporting and methods of included studies
Typical efficiency	Lower; new systems and recruitment required	Often higher once access and cleaning are secured	Search and review workload can still be substantial
Main bias concerns	Recruitment, measurement, adherence, follow-up	Data provenance, coding, missingness, selection, temporal changes	Publication bias, selective reporting, search limitations, study heterogeneity
Ethics and governance	Participant consent and study-specific review as applicable	Data access, lawful basis, privacy, consent scope and governance	Usually no new participant contact, but protocol integrity and responsible representation remain important
Typical contribution	New observations or intervention evidence	New answer from existing information	Integrated understanding of the existing research record

Selecting the appropriate route

DECISION GUIDE

Choose the evidence source

1. Has the question already been adequately answered by a current high-quality synthesis?
2. Do suitable existing datasets contain the required population, variables, timing and outcomes?
3. Would new primary data materially improve validity or relevance?
4. What participant burden, time, cost and governance requirements accompany each option?
5. Is a focused systematic review, broad scoping review or another synthesis appropriate to the objective?

A research programme may use all three routes. An evidence synthesis can identify a gap. Secondary data can estimate local burden. Primary qualitative work can explain implementation barriers. These studies are complementary when their roles are planned rather than combined without purpose.

Validity in secondary data research

Researchers should examine:

• provenance: who created the data, why and under which procedures;
• population coverage: which persons, institutions or events are included or excluded;
• variable meaning: coding systems, algorithms and operational definitions;
• time: changes in diagnostics, policy, coding and care pathways;
• linkage: match quality, duplicate records and linkage error;
• missingness: whether absence means not measured, not recorded, not applicable or truly absent;
• data cleaning: implausible values and derivations; and
• fitness for purpose: whether the dataset can support the intended inference.

Reporting guidance such as RECORD addresses studies using routinely collected health data and highlights the need to describe code lists, data cleaning, linkage and data-access limitations [5].

Validity in evidence synthesis

A review can be biased by an incomplete search, arbitrary exclusions, duplicate data, inconsistent extraction, inappropriate risk-of-bias assessment or selective emphasis on favourable results. Reviewers should pre-specify eligibility, search multiple relevant sources, document screening, extract data consistently, appraise study limitations and choose a synthesis method compatible with the evidence.

The unit of analysis is the study or study result, but studies may share participants or datasets. Multiple papers from one cohort should not automatically be treated as independent evidence.

Ethical and governance considerations

Evidence synthesis usually does not involve new participant contact, but ethical responsibilities remain. Reviewers should represent studies accurately, manage conflicts of interest, avoid unnecessary duplication and report findings even when the evidence is uncertain.

Interface with statistical analysis

Primary and secondary quantitative studies use statistical methods determined by design and data structure. Secondary datasets often contain clustering, repeated encounters, time-varying coding and informative missingness that require explicit handling.

Evidence synthesis may use meta-analysis when study estimates are sufficiently compatible, but pooling is not mandatory. The decision depends on the question, estimand, study design, population, intervention or exposure, outcome and heterogeneity. Detailed synthesis methods belong in the Biostatistics and specialised review lectures.

Worked example

WORKED EXAMPLE

Building an asthma research programme

Evidence synthesis: The team first conducts a focused systematic review of interventions intended to reduce repeat adolescent asthma admissions. The review identifies moderate evidence for discharge education but limited evidence in rural settings.

Secondary data study: Hospital and pharmacy records are linked to estimate local readmission patterns, medication dispensing and geographic variation. The team validates key coding algorithms and acknowledges that dispensing does not prove medicine use.

Primary qualitative study: Adolescents, caregivers and clinicians are purposively sampled for interviews about access, beliefs, school routines, transport and discharge communication. The findings identify barriers not available in routine records.

Integrated programme decision: The combined evidence supports adapting an existing intervention rather than creating an entirely new one. A later primary trial evaluates the adapted programme.

Each component answers a distinct question. The programme is stronger because it does not ask one data source to provide all forms of evidence.

Suggested Methods wording

Appraising the proposed source

APPRAISAL CHECKLIST

Questions for critical appraisal

• Does the selected evidence source contain the population, timing and measurements required by the question?
• For secondary data, are provenance, coding, linkage and missingness described?
• For primary research, is new participant burden justified by information not already available?
• For evidence synthesis, are the search, eligibility, appraisal and synthesis reproducible?
• Are conclusions limited by the weaknesses of the underlying data or studies?

Common mistakes

Knowledge check

A national registry contains 15 years of data but lacks the primary exposure needed for the research question. Is its large size sufficient justification for use?

Answer and reasoning

No. Size cannot compensate for absence or invalid measurement of a central variable. The dataset must be fit for the intended question, population, timing and inference.

What is the main distinction between a systematic review and a meta-analysis?

Answer and reasoning

A systematic review is a complete research process for identifying, selecting, appraising and synthesising studies. Meta-analysis is an optional statistical method for combining compatible numerical estimates within a review.

When may a scoping review be more appropriate than a focused systematic review?

Answer and reasoning

A scoping review may be appropriate when the purpose is to map a broad field, clarify concepts, describe types of evidence or identify gaps rather than answer a narrowly focused effectiveness or association question.

Why can secondary analysis require ethics or governance review even without new participant contact?

Answer and reasoning

The analysis may involve personal or sensitive information, uses beyond the original purpose, linkage, re-identification risk, access controls and retention. These issues require review according to applicable consent, institutional and legal requirements.

Key take-home messages

SUMMARY

Key take-home messages

• Primary, secondary and evidence-synthesis research differ by the source and level of evidence generation, not by an automatic quality ranking.
• Choose the source that best fits the question, population, timing, variables and intended inference.
• Existing data require careful assessment of provenance, coding, missingness, linkage and governance.
• A systematic review uses explicit reproducible methods; meta-analysis is optional.
• Multiple evidence sources can form a coherent research programme when each answers a distinct question.

References

1. Wickham RJ. Secondary analysis research. J Adv Pract Oncol. 2019;10(4):395–400. https://doi.org/10.6004/jadpro.2019.10.4.7
2. Munn Z, Peters MDJ, Stern C, Tufanaru C, McArthur A, Aromataris E. Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Med Res Methodol. 2018;18:143. https://doi.org/10.1186/s12874-018-0611-x
3. Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al., editors. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.5. Cochrane; 2024. Available from: https://www.cochrane.org/handbook
4. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. https://doi.org/10.1136/bmj.n71
5. 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. https://doi.org/10.1371/journal.pmed.1001885