The Health Research Process

Overview

The health research process is the organised sequence through which an important uncertainty is converted into evidence that can be appraised and used [1,2]. It begins before data collection and continues after publication. The process includes identifying a problem, reviewing existing knowledge, formulating a question, selecting a design, developing a protocol, obtaining relevant approvals, collecting and managing data, analysing findings, interpreting uncertainty, disseminating results and considering implementation.

These stages are often drawn as a straight line, but real research is iterative. A literature review may show that the original question is already answered. A pilot study may reveal that recruitment is unrealistic. Ethical review may require changes to consent or data handling. Early quality checks may identify a measurement problem that must be corrected. Iteration is appropriate when it is transparent and occurs before decisions are influenced by the final results.

Why this matters

Starting data collection without a complete process creates avoidable problems. Researchers may discover too late that their outcome was not measured, that important variables were defined inconsistently, that the available sample cannot answer the primary objective or that ethics approval does not cover a planned use of the data.

A structured process also clarifies responsibility. Investigators, statisticians, data managers, laboratory personnel, field workers, clinicians, community representatives and ethics committees contribute at different stages. Their involvement is most useful when it occurs before irreversible decisions are made.

The research cycle

1. Identify and prioritise the problem

A research problem is an important uncertainty, gap or practical difficulty that warrants systematic investigation. It should be grounded in clinical experience, surveillance, policy needs, patient priorities, theory or previous research. Priority depends on burden, potential benefit, equity, feasibility, ethical acceptability and whether new research is genuinely needed.

2. Review existing knowledge

The team examines what is already known, how previous studies were conducted, where evidence is uncertain and whether another primary study is justified. The literature review informs the question, design, measurements and analysis. It also helps prevent unnecessary duplication.

3. Formulate the question and objectives

The question specifies what the study seeks to know. Objectives translate the question into actions that can guide design and analysis. The primary objective should determine the central design and sample-size rationale. Secondary and exploratory objectives must not obscure the main purpose.

4. Develop a conceptual and methodological plan

The team defines the intended inference, target population, setting, study design, sampling strategy, variables or phenomena, operational definitions, measurement procedures and principal threats to validity. At this stage, alternatives should be considered rather than selecting a familiar design automatically.

5. Develop the protocol and supporting plans

The protocol is the prospective account of why and how the study will be conducted. It should specify the rationale, objectives, methods, population, time frame and expected outputs [3]. Supporting documents may include an analysis plan, data-management plan, standard operating procedures, case report forms, interview guides, consent materials, training manuals and monitoring plans.

PROTOCOL CHECKLIST

Decisions to complete before data collection

• primary research question and objective;
• study design, setting and study period;
• target, source and study populations;
• eligibility, sampling and recruitment procedures;
• variables, constructs, instruments and measurement time points;
• primary analysis or qualitative analytical approach;
• sample-size or information-power rationale;
• data-quality, confidentiality and security procedures;
• ethical, governance and registration requirements; and
• dissemination and authorship plans.

6. Assess feasibility and pilot procedures

Feasibility work asks whether the planned study can be delivered. Pilot testing may examine recruitment rates, comprehension of consent materials, timing, instrument performance, data-entry workflow, intervention delivery or interview procedures. A pilot is not a small definitive effectiveness study. Its conclusions should focus on feasibility questions.

7. Obtain approvals and register where appropriate

Health research may require ethics approval, institutional permission, data-owner approval, regulatory authorisation, trial registration or other governance review. Requirements depend on jurisdiction, study type, intervention and data source and should be verified locally.

8. Prepare the research team and systems

Team members require role-specific training and standardisation. Databases should be tested, access permissions assigned, variable names and coding documented, and monitoring procedures activated. The final approved documents must be version controlled so that all sites use the same materials.

9. Recruit and collect data

Recruitment and data collection follow the protocol. Screening, eligibility, consent, refusals, withdrawals, protocol deviations and losses to follow-up should be documented. Quality checks should occur during collection rather than only after the study closes.

10. Manage and analyse data

Data management includes secure capture, validation, coding, cleaning, linkage, derivation, backup, access control and archiving. Analysis follows the pre-specified question and data structure. Departures from the planned analysis should be justified and reported.

11. Interpret the findings

Interpretation asks what the findings mean, how uncertainty and bias affect them, whether alternative explanations remain, and to whom the conclusions apply. It should integrate methodological limitations rather than append a generic limitations paragraph after strong claims have already been made.

12. Disseminate and share responsibly

Results should be reported regardless of whether they are statistically significant, favourable or commercially attractive. Failure to address important questions efficiently or to report completed research contributes to avoidable research waste [4]. Dissemination may include journal articles, theses, reports, policy briefs, conference presentations, participant summaries, public communication, datasets or code, subject to consent and governance.

13. Support implementation and evaluate impact

Research evidence does not implement itself. Findings may require adaptation, stakeholder engagement, resource assessment, training and evaluation. Implementation can reveal new questions, beginning another research cycle.

Research documents and their functions

Core documents across the research process

illustrative

Table 1. Purposes of common research documents

Document	Main function	Typical timing
Research proposal	Justifies the importance, feasibility and resources for the planned study	Before funding or institutional approval
Study protocol	Specifies prospectively why and how the study will be conducted	Before approvals and data collection
Statistical analysis plan	Defines analysis populations, outcomes, estimands, models and planned sensitivity analyses	Before access to final comparative results
Data-management plan	Defines data flow, coding, quality, access, security, retention and sharing	Before database construction and collection
Standard operating procedure	Standardises a recurring operational task	Before staff training and study conduct
Study report or manuscript	Describes what was done, what was found and how findings should be interpreted	After analysis

Note: Terminology and required documents vary by institution and study type.

A proposal and protocol may overlap, but they serve different purposes. A proposal often persuades a funder or institution that the research should be supported. A protocol is the operational source of truth for study conduct. A manuscript must report what actually occurred, including deviations from the protocol.

The process is iterative but not arbitrary

Iteration should improve alignment. For example, a pilot may show that a planned six-minute walking test cannot be completed safely by the target population, requiring a revised outcome. This is a legitimate protocol-stage change.

By contrast, repeatedly changing outcomes or analyses after seeing which results are favourable creates a risk of selective reporting. The timing of decisions matters. Version control, protocol amendments and transparent labelling of exploratory work help distinguish planned refinement from outcome-driven flexibility.

DECISION GUIDE

When a proposed change arises

1. Does the change improve the ability to answer the original question?
2. Has the team seen outcome data that could influence the decision?
3. Does the change affect participant risk, consent, privacy or burden?
4. Is an ethics, governance, registry or funder amendment required?
5. How will the change and its timing be documented and reported?

Interface with statistical analysis

Statistical input should begin during question and protocol development. The analysis depends on the design, unit of observation, outcome type, time structure, clustering, repeated measurements and missing-data process. Sample-size planning also depends on the primary objective and estimand.

The analyst should not receive an undocumented spreadsheet only after data collection. Early collaboration helps define analysable variables, preserve identifiers needed for linkage or repeated observations, avoid ambiguous coding and ensure that the planned design can support the intended inference.

Worked example

WORKED EXAMPLE

Mapping the medication-error question through the process

Problem clarification: Reported medication errors increased after a new electronic reporting portal was introduced. The team first distinguishes actual error occurrence from reporting behaviour.

Evidence review: Previous hospital data and published studies suggest that errors cluster by ward and medication-administration workload. Training alone may not address system-level causes.

Question: Among adult inpatient wards, does a combined medication-safety intervention reduce the rate of preventable administration errors over six months compared with usual practice?

Protocol decisions: The team defines wards as the delivery unit, specifies the error definition, uses independent record review in addition to voluntary reports, documents patient-days as the exposure denominator, plans standardised reviewer training and considers contamination between wards.

Approvals and preparation: The team confirms the governance pathway for use of incident and clinical-record data, limits identifiable access, pilots the review form and tests inter-reviewer agreement.

Conduct and analysis: Recruitment and data collection follow a pre-specified period. The analysis accounts for ward-level delivery and differing patient exposure. Deviations and missing records are documented.

Interpretation: A reduction in independently verified error rates would be interpreted alongside intervention fidelity, reporting behaviour, ward differences and confidence intervals.

Dissemination and implementation: Results are reported to staff and governance committees, and future implementation is evaluated rather than assumed.

Suggested protocol wording

Common mistakes

Knowledge check

Why should the literature review occur before the final study design is selected?

Answer and reasoning

The literature may show that the question is already answered, identify important design weaknesses to avoid, provide validated measurements and clarify the remaining gap. It therefore shapes whether new research is needed and what design is appropriate.

A pilot study recruits 20 participants and shows a statistically significant improvement. Can the intervention be declared effective?

Answer and reasoning

Usually not. A pilot study is primarily intended to assess feasibility, procedures and uncertainty needed for a definitive study. Its small size and operational focus generally do not support a definitive effectiveness conclusion.

When is iteration in the research process problematic?

Answer and reasoning

Iteration becomes problematic when changes are driven by knowledge of the results but are presented as prospectively planned, or when changes affecting risk, consent or data use are implemented without required review. Transparent, documented refinement before outcome-driven decisions is different.

What is the main difference between a protocol and a manuscript?

Answer and reasoning

The protocol prospectively specifies what the researchers intend to do. The manuscript reports what was actually done and found, including justified departures from the protocol.

Key take-home messages

SUMMARY

Key take-home messages

• The health research process begins with a justified problem and continues through dissemination, implementation and new learning.
• The stages are iterative, but important changes must be prospective, documented and ethically governed.
• The primary question and objective should drive design, measurement, sample-size planning and analysis.
• Protocols, analysis plans, data-management plans and standard operating procedures have distinct functions.
• Early multidisciplinary input prevents problems that cannot be corrected after data collection.

References

1. Fathalla MF, Fathalla MMF. A practical guide for health researchers. Cairo: World Health Organization Regional Office for the Eastern Mediterranean; 2004. Available from: https://iris.who.int/handle/10665/119703
2. World Health Organization Regional Office for the Western Pacific. Health research methodology: a guide for training in research methods. 2nd ed. Manila: World Health Organization Regional Office for the Western Pacific; 2001. Available from: https://iris.who.int/handle/10665/206929
3. World Health Organization. Recommended format for a research protocol. Geneva: World Health Organization; [cited 2026 Aug 3]. Available from: https://www.who.int/groups/research-ethics-review-committee/recommended-format-for-a-research-protocol
4. Chalmers I, Glasziou P. Avoidable waste in the production and reporting of research evidence. Lancet. 2009;374(9683):86–89. https://doi.org/10.1016/S0140-6736(09)60329-9