Mixed Methods and Qualitative Evidence Synthesis

Introduction

Health-care decisions rarely depend on one kind of evidence. Quantitative research may estimate the effectiveness of an intervention, the frequency of an outcome, or the association between exposure and disease. Qualitative research explains how people experience illness, why services are used or avoided, how professionals interpret policy, and under what circumstances an intervention is acceptable and feasible. For health-care managers and researchers, the central challenge is not simply to collect both kinds of data, but to integrate them in a logically defensible manner.

Mixed-methods research combines quantitative and qualitative approaches within a single study or programme of inquiry. Qualitative evidence synthesis brings findings from multiple qualitative studies together to develop a more comprehensive interpretation of a phenomenon. Both approaches are particularly valuable in health-care management because service performance depends simultaneously on measurable outcomes, organisational processes, professional behaviour, patient priorities, social context and implementation conditions.

This chapter examines three major mixed-methods designs—convergent, explanatory sequential and exploratory sequential designs—and three influential approaches to qualitative evidence synthesis: thematic synthesis, meta-ethnography and realist-informed synthesis. It also introduces GRADE-CERQual, a framework for assessing confidence in findings from qualitative evidence syntheses. The chapter treats integration as an analytic task rather than as the simple juxtaposition of methods.

Learning objectives

By the end of this chapter, the reader should be able to:

1. Explain the philosophical and practical foundations of mixed-methods health research.
2. Distinguish convergent, explanatory sequential and exploratory sequential designs.
3. Plan integration at the design, methods, interpretation and reporting stages.
4. Conduct thematic synthesis, meta-ethnography and realist-informed synthesis appropriately.
5. Assess confidence in qualitative synthesis findings using GRADE-CERQual.
6. Judge the strengths, limitations, ethical issues and managerial usefulness of integrated evidence.

1. Foundations of mixed-methods research

Mixed-methods research is a methodology in which qualitative and quantitative approaches are intentionally combined to address a research problem. The defining feature is integration: the two forms of evidence must influence one another or contribute jointly to a conclusion. A report that places a survey and a set of interviews in separate chapters, without relating their findings, is better described as a multi-method study than a genuinely mixed-methods study.

The rationale for mixing methods commonly includes four purposes:

• Complementarity: one method elaborates, clarifies or enriches findings from another.
• Expansion: different methods examine different dimensions of a complex problem.
• Development: findings from one phase inform sampling, measurement or intervention design in another phase.
• Triangulation: partially independent forms of evidence are compared to examine convergence, divergence or inconsistency.

These purposes are not mutually exclusive. A study of missed outpatient appointments might quantify attendance patterns, interview patients about barriers, and use both sets of findings to redesign scheduling systems.

Philosophical foundations

Mixed-methods researchers draw on several philosophical positions. Pragmatism is frequently associated with mixed methods because it prioritises the research question and the usefulness of knowledge over allegiance to a single method. A pragmatic design can accommodate both numerical estimation and interpretive understanding when each is necessary to answer the question.

Other researchers use a dialectical or critical realist position. A dialectical approach treats conflicting findings as potentially productive rather than as evidence that one method has failed. Critical realism distinguishes observable events from underlying mechanisms and contextual conditions; this perspective is useful when researchers ask why an intervention works for some groups but not others.

Philosophical flexibility does not mean methodological inconsistency. Researchers must still specify what counts as evidence, how different claims will be compared, and how the final inference will be justified.

Core design decisions

Four decisions should be made explicitly:

1. Priority: Is qualitative or quantitative evidence dominant, or are both given broadly equal status?
2. Timing: Are data collected concurrently, or does one phase follow another?
3. Point of integration: Will integration occur during sampling, data collection, analysis, interpretation or reporting?
4. Purpose: Is the aim explanation, development, triangulation, intervention design or expansion?

A useful notation represents the order and emphasis of phases. In the notation, “QUAL” indicates qualitative work, “QUAN” quantitative work, “+” indicates concurrent activity, and an arrow indicates sequence. Thus, QUAL + QUAN describes a concurrent design, whereas QUAN → QUAL describes a quantitative phase followed by a qualitative phase.

2. Principal mixed-methods designs

Convergent design

In a convergent design, qualitative and quantitative data are collected during a similar period, analysed separately using appropriate procedures, and then integrated. The design is suitable when both forms of evidence are needed to answer the same broad question and neither phase must logically precede the other.

For example, a hospital may measure staff burnout using a validated questionnaire while conducting focus groups about workload, leadership and emotional demands. Quantitative analysis might show high burnout in emergency departments; qualitative analysis might reveal that unpredictable rostering and inadequate recovery time are central explanations. Integration occurs when the researcher compares the findings and develops a joint interpretation.

Integration can produce:

• Convergence: both strands support a similar conclusion.
• Complementarity: each strand addresses a different aspect of the issue.
• Expansion: the strands reveal distinct dimensions.
• Discordance: the strands appear inconsistent and require investigation.

Discordance should not be concealed. A survey may show high satisfaction because respondents value clinical competence, while interviews may reveal frustration with communication. The apparent contradiction may reflect different meanings of “satisfaction,” social desirability, sampling differences or variation between subgroups.

Explanatory sequential design

An explanatory sequential design begins with quantitative data and follows with qualitative inquiry:

QUAN → QUAL → integrated explanation

The qualitative phase explains unexpected, statistically important or clinically meaningful quantitative findings. Sampling may be purposive: researchers might select participants with unusually high or low scores, people from contrasting sites, or cases that represent important subgroups.

Consider a telemedicine evaluation in which overall use is high but rural older adults have low uptake. Interviews may explain that unreliable connectivity, lack of confidence with devices and concerns about privacy limit use. The qualitative results do not replace the quantitative estimate; they explain its pattern and identify actionable mechanisms.

Common risks include treating the qualitative phase as a token explanation, selecting interviewees only because they confirm the statistical result, or allowing a small qualitative sample to make population-level claims. Researchers should preserve the interpretive depth of interviews and distinguish explanation from generalisation.

Exploratory sequential design

An exploratory sequential design starts with qualitative inquiry and uses its findings to develop a quantitative phase:

QUAL → QUAN → integrated testing or estimation

It is useful when a construct is poorly understood, an existing instrument is culturally inappropriate, or researchers need to identify locally meaningful outcomes. For example, interviews with informal caregivers may identify “constant readiness,” fragmented communication and financial uncertainty as dimensions of burden. These themes can inform questionnaire items, response scales and hypotheses for a subsequent survey.

The transition from qualitative themes to quantitative measures requires care. Researchers should document how concepts were operationalised, test content validity, pilot the instrument and examine reliability and construct validity. Quantification can make a concept visible to administrators, but it may also narrow or distort its meaning.

Other sequential and embedded designs

A study may embed a qualitative process evaluation within a randomised trial. The trial estimates intervention effects, while interviews and observations examine fidelity, acceptability and implementation. Alternatively, a study may use a multiphase design in which several linked qualitative and quantitative phases address programme development, testing and scale-up.

The design must match the causal and interpretive claims. A qualitative process evaluation cannot by itself demonstrate clinical effectiveness, and a statistically significant effect does not establish that an intervention was acceptable, equitable or sustainable.

3. Integration in practice

Integration is strongest when planned before fieldwork. A protocol should specify the research question, sampling relationship, data transformation procedures, analytic sequence and method for producing a final inference.

Connecting, building and merging

Three practical mechanisms are widely used:

• Connecting: results from one phase determine sampling or procedures in the next phase.
• Building: findings from one method inform an instrument, intervention or interview guide.
• Merging: qualitative and quantitative findings are brought together for comparison and interpretation.

A joint display is a useful merging device. It places findings side by side, often by case, subgroup, outcome or theme. For instance, a table may show clinic-level waiting time, patient-reported access and interview evidence about registration barriers. Joint displays make agreement and discrepancy visible and reduce the risk that one strand is ignored during interpretation.

Integration and inference

The final conclusion should indicate what is known from each strand and what is learned only through their combination. Researchers should avoid “methodological assimilation,” in which qualitative findings are reduced to counts, or “methodological dominance,” in which one strand is treated as the only legitimate evidence.

A transparent report should describe:

• the contribution of each method;
• the basis for resolving or retaining contradictions;
• whether integration occurred at design, analysis or interpretation;
• the limits of transferability and statistical generalisation; and
• how the integrated findings affect policy or practice.

4. Qualitative evidence synthesis

Qualitative evidence synthesis systematically brings together findings from qualitative studies to generate a new interpretation, explanation or theory. It differs from a conventional literature review because it uses explicit searching, eligibility criteria, appraisal, coding and synthesis procedures.

A synthesis may be aggregative, seeking to summarise recurring findings, or interpretive, seeking to develop concepts and explanations beyond the individual studies. The choice depends on the question. “What barriers are reported?” may suit thematic synthesis, whereas “How do patients negotiate trust in digital care?” may benefit from meta-ethnography.

Preparing a synthesis

The protocol should specify:

• the review question and phenomenon of interest;
• population, setting and context;
• eligible study designs and publication types;
• databases and supplementary search methods;
• procedures for screening and data extraction;
• approach to quality appraisal;
• synthesis method; and
• strategy for assessing confidence in findings.

Searches should be sufficiently sensitive to identify diverse terminology. Researchers may search bibliographic databases, citation networks, reference lists, grey literature and relevant organisational sources. In qualitative reviews, a narrow intervention-effect search can miss studies using different language for the same experience.

Study appraisal is not a mechanical exclusion exercise. Tools such as the Critical Appraisal Skills Programme checklist or the Joanna Briggs Institute tools can support structured judgement, but reviewers must consider whether methodological limitations affect the contribution of a study to the synthesis.

5. Thematic synthesis

Thematic synthesis is particularly useful when the aim is to identify patterns across qualitative studies while retaining an explicit connection to participants’ accounts. It generally involves three stages.

Line-by-line coding

Reviewers code the findings sections of included studies, including participant quotations and authors’ interpretations. Coding should remain close to the data initially. For example, statements such as “I did not know whom to call,” “the instructions changed each time,” and “my family translated the letter” might receive codes relating to uncertainty, inconsistent information and dependence on informal support.

Descriptive themes

Related codes are grouped into descriptive themes that remain relatively close to the content of the original studies. “Fragmented communication” might combine uncertainty, inconsistent instructions and unclear responsibility.

Analytical themes

Analytical themes go beyond the findings of individual studies to provide an explanatory interpretation. “Administrative ambiguity shifts coordination work to patients and families” is an analytical theme because it proposes how and why the problem operates.

Reviewers should maintain an audit trail showing coding decisions, theme development, disagreements and revisions. Software can assist with retrieval and organisation, but it does not perform interpretation. Multiple reviewers may code independently and discuss differences; consensus should not be confused with validity, and reflexivity remains essential.

6. Meta-ethnography

Meta-ethnography is an interpretive synthesis method developed to translate concepts across qualitative studies. Its distinctive feature is the relationship between concepts rather than simple aggregation of findings.

A classic process includes:

1. Getting started and clarifying the research interest.
2. Deciding what is relevant and selecting studies.
3. Reading studies closely.
4. Determining how studies relate to one another.
5. Translating studies into one another.
6. Synthesising translations.
7. Expressing the synthesis in a usable form.

A study’s concepts are often distinguished as:

• First-order constructs: participants’ own words and meanings.
• Second-order constructs: original study authors’ interpretations.
• Third-order constructs: the synthesis team’s interpretations across studies.

Translation may be reciprocal, where concepts are sufficiently similar to be compared, or refutational, where studies contradict one another. A line-of-argument synthesis then connects translated concepts into a broader account. For example, studies of medication adherence may describe forgetting, mistrust, cost and treatment fatigue. A meta-ethnography might develop the argument that adherence is negotiated within changing relationships among illness identity, treatment burden and confidence in the health system.

Meta-ethnography requires intensive reading and conceptual sensitivity. It is poorly suited to a very large number of thinly reported studies or to questions requiring only a frequency-based inventory of barriers. Translation should preserve context; a concept developed in specialist cancer care should not automatically be treated as identical to a concept in community mental-health care.

7. Realist-informed qualitative synthesis

Realist-informed synthesis asks how, why, for whom and under what circumstances an intervention or programme produces outcomes. It is grounded in a context–mechanism–outcome configuration:

Context (C) + Mechanism (M) → Outcome (O)

This is not a deterministic equation. It is a conceptual representation: mechanisms are processes triggered in particular contexts, and outcomes are contingent rather than guaranteed. A mechanism is not merely an intervention component; it may involve reasoning, response, trust, motivation, perceived legitimacy or changes in opportunity.

For example, a community health-worker programme may improve vaccination uptake when workers are trusted and services are conveniently available. The intervention supplies information and social support, but trust and perceived accessibility may be the mechanisms through which families act. Where workers are viewed as outsiders or clinics are distant, the same programme may have little effect.

Realist-informed qualitative synthesis generally proceeds through iterative searching, theory development, extraction of context–mechanism–outcome evidence, and refinement of programme theories. It can include qualitative studies, process evaluations, policy documents and selected quantitative evidence. The aim is not to calculate one pooled effect, but to explain patterned outcomes and refine transferable theories.

Its major strength is relevance to complex health programmes. Its challenge is that causal language can become speculative if mechanisms are inferred without adequate empirical support. Reviewers should distinguish observed mechanisms, plausible mechanisms and conjectural mechanisms, and show how each claim is supported.

8. Assessing confidence with GRADE-CERQual

GRADE-CERQual—Confidence in the Evidence from Reviews of Qualitative research—assesses how much confidence decision-makers should place in an individual finding from a qualitative evidence synthesis. It does not assign one overall score to an entire review. Confidence is judged finding by finding.

The four components are:

1. Methodological limitations: concerns about the design or conduct of the contributing studies.
2. Coherence: how well the evidence supports the review finding and whether there are unexplained contradictions.
3. Adequacy of data: the richness and quantity of data supporting the finding.
4. Relevance: how applicable the evidence is to the review question and context.

Confidence is commonly reported as high, moderate, low or very low. A review begins with high confidence in a finding and downgrades it when there are important concerns. This is a judgement framework, not a numerical score, and downgrading should be justified transparently.

A finding may have high coherence but low relevance to a particular country. Conversely, a highly relevant study may provide only thin data. Reviewers should avoid treating the number of studies as a substitute for richness, and should not automatically downgrade a finding because studies used different qualitative traditions.

GRADE-CERQual supports decision-making, but it does not determine policy by itself. Equity, feasibility, costs, legal duties, patient preferences and stakeholder priorities must also be considered.

9. Worked applied case: improving diabetes follow-up

The following is an illustrative case, not a report of an actual study.

Background and problem

A regional health system introduced nurse-led telephone follow-up for adults with type 2 diabetes. Administrative data showed a modest reduction in missed appointments but no consistent improvement in glycated haemoglobin. Managers needed to understand the mixed results before expanding the service.

Design

The evaluation used an explanatory sequential mixed-methods design. First, appointment records and clinical indicators were analysed across 18 clinics. Clinics were then purposively selected because they showed high, moderate or low engagement. Interviews were conducted with patients, nurses and practice managers.

Quantitative findings

Engagement was highest among patients already attending regular appointments. Patients with unstable housing, limited phone access and language barriers were more likely to miss telephone contacts. Clinical improvement varied substantially between clinics.

Qualitative findings

Interviews suggested that the telephone service worked when nurses had access to records, could schedule follow-up flexibly and were known to patients. It was less effective when calls came from unrecognised numbers, interpreters were unavailable, or the interaction focused narrowly on medication rather than food security, transport and emotional distress.

Integration and solution

A joint display linked clinic engagement patterns with qualitative explanations. The programme was redesigned to include scheduled calls, interpreter access, text reminders, escalation pathways and a brief social-needs screen. Managers also stopped treating non-contact as patient non-compliance and introduced alternative face-to-face and community options.

Outcome and learning points

The integrated evidence indicated that the service model was not uniformly ineffective; its outcomes depended on organisational readiness and patient circumstances. The case illustrates that quantitative data can identify variation, while qualitative evidence can explain mechanisms and guide adaptation. It also demonstrates why an average outcome may conceal inequitable access.

10. Quality, ethics and governance

Mixed-methods and synthesis projects raise several ethical and governance issues. Participants may disclose experiences of discrimination, unsafe care, domestic abuse or inability to obtain essential treatment. Interview protocols should include consent procedures, confidentiality protections, distress pathways and limits to confidentiality where legally required.

Linking survey, clinical and interview data creates re-identification risks. Data minimisation, role-based access, secure storage and transparent information governance are essential. Researchers should consider whether participants consented to linkage and whether secondary use is ethically justified.

Qualitative evidence synthesis usually uses published or publicly available material, but ethical responsibility remains. Reviewers should represent participants fairly, avoid decontextualising quotations, acknowledge colonial or extractive research practices, and consider whose voices are absent. Language translation requires attention to meaning rather than word-for-word substitution.

Researchers should declare conflicts of interest, preregister or publish protocols where appropriate, report deviations, and make analytic decisions auditable. Patient and public involvement can improve question selection, interpretation and dissemination, although involvement should not be used symbolically or as a substitute for rigorous analysis.

11. Technology, AI and contemporary developments

Digital tools increasingly support searching, deduplication, transcription, qualitative coding, translation and evidence mapping. Artificial intelligence may assist with screening or suggest candidate codes, but it should not be treated as an autonomous interpreter. Models can reproduce sampling biases, erase uncertainty, misread culturally specific language and generate plausible but unsupported summaries.

Best practice requires human verification, documentation of tools and prompts, protection of confidential data, and explicit reporting of where automation was used. AI-generated coding should be checked against source material and should never replace reflexive engagement with context, power and meaning.

Contemporary mixed-methods work is also expanding through implementation science, equity-oriented evaluation, participatory research and integration of routinely collected data with lived-experience evidence. These developments increase the potential value of synthesis while making transparency, data governance and reflexivity more important.

12. Best-practice checklist

A high-quality project should:

• begin with a question that genuinely requires more than one form of evidence;
• justify the selected mixed-methods or synthesis design;
• define priority, timing and integration procedures;
• preserve the integrity of qualitative and quantitative analysis;
• investigate discordant findings rather than suppressing them;
• use purposive and theoretically informed sampling where appropriate;
• maintain an audit trail and reflexive record;
• distinguish description, interpretation and causal explanation;
• assess synthesis findings with GRADE-CERQual where appropriate;
• report context, equity implications and limitations;
• protect privacy during data linkage and secondary analysis; and
• translate findings into feasible, evidence-informed action.

Chapter Summary

Mixed-methods research integrates qualitative and quantitative evidence to answer complex health-care questions. Convergent designs collect both strands concurrently; explanatory sequential designs use qualitative inquiry to explain quantitative patterns; exploratory sequential designs use qualitative findings to develop measures, hypotheses or interventions. Integration may occur through connecting, building and merging, and discordance can generate important insight.

Qualitative evidence synthesis systematically develops interpretations across studies. Thematic synthesis progresses from line-by-line coding to descriptive and analytical themes. Meta-ethnography translates concepts across studies and develops higher-order interpretations. Realist-informed synthesis explains how outcomes arise through context–mechanism–outcome configurations. GRADE-CERQual assesses confidence in individual synthesis findings using methodological limitations, coherence, adequacy and relevance. Ethical practice requires attention to consent, confidentiality, representation, equity, transparency and responsible use of digital and AI tools.

Key Terms

Mixed-methods research: Intentional integration of qualitative and quantitative approaches in one inquiry.

Convergent design: A design in which qualitative and quantitative data are collected concurrently and integrated.

Explanatory sequential design: Quantitative inquiry followed by qualitative work to explain findings.

Exploratory sequential design: Qualitative inquiry followed by quantitative measurement or testing.

Integration: The process of combining methods, findings or inferences into a coherent conclusion.

Triangulation: Comparison of different evidence sources to assess convergence or divergence.

Joint display: A visual or tabular device that combines qualitative and quantitative findings.

Qualitative evidence synthesis: Systematic interpretation of findings from multiple qualitative studies.

Thematic synthesis: Coding-based synthesis that develops descriptive and analytical themes.

Meta-ethnography: Interpretive synthesis based on translation of concepts across studies.

First-order construct: A participant’s own account or meaning.

Second-order construct: An original study author’s interpretation.

Third-order construct: The synthesis team’s interpretation across studies.

Realist-informed synthesis: Synthesis focused on how outcomes arise in particular contexts through mechanisms.

Context–mechanism–outcome configuration: A conceptual explanation linking circumstances, generative processes and results.

GRADE-CERQual: Framework for assessing confidence in individual qualitative synthesis findings.

Coherence: The degree to which evidence supports a synthesis finding.

Adequacy of data: The richness and quantity of evidence supporting a finding.

Reflexivity: Critical examination of how researchers influence and are influenced by the research.

Review Questions

Short Answer Questions

1. What distinguishes a mixed-methods study from a multi-method study?
2. Describe the main purpose of an explanatory sequential design.
3. What are the three stages of thematic synthesis?
4. What is meant by a third-order construct in meta-ethnography?
5. Name the four components of GRADE-CERQual.

Long Answer Questions

1. Compare convergent, explanatory sequential and exploratory sequential designs, including suitable health-care examples.
2. Explain how integration can occur through connecting, building and merging.
3. Describe the process and intellectual contribution of meta-ethnography.
4. Explain how realist-informed synthesis uses context–mechanism–outcome reasoning.
5. Discuss the ethical and governance challenges of linking qualitative, quantitative and routinely collected health data.

Analytical / Application Questions

1. A survey finds high satisfaction with a digital clinic, but interviews reveal substantial frustration with communication. Develop two possible explanations for the discordance and propose an integration strategy.
2. Design an exploratory sequential study to develop a measure of caregiver burden for a culturally diverse population. Explain how you would protect the meaning of qualitative concepts during instrument development.
3. A qualitative synthesis finds that community health workers improve service uptake only where they are trusted and services are accessible. Use a context–mechanism–outcome framework to propose an implementation strategy and identify what evidence would increase confidence in the finding.

Multiple Choice Questions with Answers

Question 1

Which feature most clearly distinguishes mixed-methods research from a multi-method study?

A. Use of more than one researcher

B. Integration of qualitative and quantitative evidence

C. Collection of data in more than one location

D. Use of a published research protocol

Correct Answer: B. Integration of qualitative and quantitative evidence

Question 2

In an explanatory sequential design, which sequence is correct?

A. QUAL → QUAN

B. QUAN → QUAL

C. QUAL + QUAL

D. QUAN + QUAN

Correct Answer: B. QUAN → QUAL

Question 3

What is the principal purpose of an exploratory sequential design?

A. To use qualitative findings to develop quantitative measures or hypotheses

B. To eliminate the need for qualitative analysis

C. To pool clinical trial effect sizes

D. To collect both datasets without integration

Correct Answer: A. To use qualitative findings to develop quantitative measures or hypotheses

Question 4

Which stage of thematic synthesis develops interpretations that go beyond the content of individual studies?

A. Line-by-line coding

B. Data extraction

C. Analytical theme development

D. Reference management

Correct Answer: C. Analytical theme development

Question 5

What is distinctive about meta-ethnography?

A. It calculates a pooled treatment effect

B. It translates concepts across qualitative studies

C. It excludes study authors’ interpretations

D. It uses only administrative datasets

Correct Answer: B. It translates concepts across qualitative studies

Question 6

In realist-informed synthesis, a mechanism is best understood as:

A. Only the physical component of an intervention

B. A guaranteed cause of an outcome

C. A process through which people respond in a particular context

D. A statistical significance threshold

Correct Answer: C. A process through which people respond in a particular context

Question 7

Which item is not one of the four GRADE-CERQual components?

A. Coherence

B. Adequacy of data

C. Relevance

D. Statistical power

Correct Answer: D. Statistical power

Question 8

A joint display is primarily used to:

A. Replace all qualitative interpretation

B. Combine and compare qualitative and quantitative findings

C. Calculate inter-rater reliability automatically

D. Recruit participants

Correct Answer: B. Combine and compare qualitative and quantitative findings

Question 9

In a convergent design, qualitative and quantitative data are usually:

A. Collected concurrently and integrated during interpretation

B. Collected only after a clinical trial

C. Collected exclusively from the same instrument

D. Kept permanently unrelated

Correct Answer: A. Collected concurrently and integrated during interpretation

Question 10

A synthesis finding is supported by relevant studies but by very thin data. Which GRADE-CERQual component is most directly concerning?

A. Adequacy of data

B. Randomisation

C. Statistical power

D. Blinding

Correct Answer: A. Adequacy of data