Thematic analysis is a popular method for systematically analysing qualitative data, such as interview and focus group transcripts.
It is one of a cluster of methods that focus on identifying patterns of meaning, or themes, across a data set.
It is relevant to many questions in perioperative medicine and a good starting point for those new to qualitative research.
Systematic approaches to thematically analysing data exist, with key components to demonstrate rigour, accountability, confirmability and reliability.
In one study, a useful six-step approach to analysing data is offered.Anaesthesia research commonly uses quantitative methods, such as surveys, RCTs or observational studies. Such methods are often concerned with answering what questions and how many questions. Qualitative research is more concerned with why questions that enable us to understand social complexities. ‘Qualitative studies in the anaesthetic setting’, write Shelton and colleagues, ‘have been used to define excellence in anaesthesia, explore the reasons behind drug errors, investigate the acquisition of expertise and examine incentives for hand hygiene in the operating theatre’. 1
General-purpose thematic analysis (termed thematic analysis hereafter) is a qualitative research method commonly used with interview and focus group data to understand people's experiences, ideas and perceptions about a given topic. Thematic analysis is a good starting point for those new to qualitative research and is relevant to many questions in the perioperative context. It can be used to understand the experiences of healthcare professionals and patients and their families. Box 1 gives examples of questions amenable to thematic analysis in anaesthesia research.
Thematic analysis involves a process of assigning data to a number of codes, grouping codes into themes and then identifying patterns and interconnections between these themes. 2 Thematic analysis allows for a nuanced understanding of what people say and do within their particular social contexts. Of note, thematic analysis can be used with interviews and focus groups and other sources of data, such as documents or images.
Thematic analysis is not the same as content analysis. Content analysis involves counting the frequency with which words or phrases appear in data. Content analysis is a method used to code and categorise textual information systematically to determine trends, frequency and patterns of words used. 3 Conversely, thematic analysis focuses on the relative importance of ideas and how ideas connect and govern practices. Thematic analysis does not rely on frequency counts to indicate the importance of coded data. Content analysis can be coupled with thematic analysis, where both themes and frequencies of particular statements or words are reported.
Thematic analysis is a research method, not a methodology. A methodology is a method with a philosophical underpinning. If researchers report only on what they did, this is the method. If, in addition, they report on the philosophy that governed what they did, this is methodology. Common methodologies in qualitative research include phenomenology, grounded theory, hermeneutics, narrative enquiry and ethnography. 4 Each of these methodologies has associated methods for data analysis. Thematic analysis can be combined with many different qualitative methodologies.
There are also different types of thematic analysis, such as inductive (including general purpose), applied, deductive or semantic thematic analysis. Inductive analysis involves approaching the data with an open mind, inductively looking for patterns and themes and interpreting these for meaning. 2 , 4 Of note, researchers can never have a truly open mind on their topic of interest, so the process will be influenced by their particular perspectives, which need to be declared. In applied and deductive thematic analysis, the researcher will have a pre-existing framework (which may be informed by theory or philosophy) against which they will attempt to categorise the data.4, 5, 6 For semantic thematic analysis, the data are coded on explicit content, and tend to be descriptive rather than interpretative. 6
In this review, we outline what thematic analysis entails and when to use it. We also list some markers to look for to appraise the quality of a published study.
Before embarking on qualitative research, as with quantitative research, it is important to seek ethical review of the proposed study. Ethical considerations include such issues as consent, data security and confidentiality, permission to use quotes, potential for identifying individuals or institutions, risk of psychological harm to participants with studies on sensitive issues (e.g. suicide or sexual harassment), power relationships between interviewer and interviewee or intrusion on other activities (such as teaching time or work commitments). 7
Qualitative research often involves asking people questions during interviews or focus groups. Merriam and Tisdell stated that, ‘The most common form of interview is the person-to-person encounter in which one person elicits information from the other’. 8 Information is elicited through careful and purposeful questioning and listening. 9 Research interviews in anaesthesia are generally purposeful conversations with a structure that allows the researcher to gather information about a participant's ideas, perceptions and experiences concerning a given topic.
A structured interview is when the researcher has already decided on a set of questions to ask. 9 If the researcher will ask a set of questions, but has flexibility to follow up responses with further questions, this is called a semi-structured interview. Semi-structured interviews are commonly used in research involving thematic analysis. The researcher can also use other forms of questioning, such as single-question interview. Semi-structured interviews are commonly used in anaesthesia, such as the studies from our own research group.10, 11, 12
Interviews are usually recorded in audio form and then transcribed. For each interview or focus group, a single transcript is created. The transcripts become the written form of data and the collection of transcripts from the research participants becomes the data set.
The design of interview questions significantly shapes a participant's response. Interview questions should be designed using ‘sensitising concepts’ to encourage participants to share information that will increase a researcher's understanding of the participants' experiences, views, beliefs and behaviours. 13 ‘Sensitising concepts’ describe words in questions that bring the participants' attention to a concept of research interest. Examples of sensitising concepts include speaking up, teamwork and theoretical concepts (such as Kolb's experiential learning cycle or Foucauldian power theory in relation to trainee learning and operating theatre culture). 14 , 15 Specifically, the questions should be framed in such a way as to encourage participants to make sense of their own experience and in their own words. The researcher should try to minimise the influences of their own biases when they design questions. Using open-ended questions will increase the richness of data. Box 2 gives examples of question design.
Bias is an inclination or prejudice for or against someone or something, whereas positionality is a person's position in society or their stance towards someone or something. For example, Tanisha once had an inexperienced anaesthetist accidentally rupture one of her veins whilst they were siting an i.v. cannula in an emergency situation. Now, Tanisha has a bias against inexperienced anaesthetists. Tanisha's positionality—a medical anthropologist with no anaesthesia training, but working with many anaesthesia colleagues, including her director—may also inform that bias or the way that Tanisha interacts with anaesthetists. Reflexivity is a process whereby people/researchers proactively reflect on their biases and positionality. Biases shape positionality (i.e. the stance of the researcher in relation to the social, historical and political contexts of the study). In practical research terms, biases and positionality inform the way researchers design and undertake research, and the way they interpret data. It is important in qualitative research to both identify biases and positionality, and to take steps to minimise the impact of these on the research.
Some ways to minimise the influence of bias and positionality on findings include:
(i) Raise awareness amongst the research team of bias and positionality.
(ii) Design research/interview questions that minimise potential for these to distort which data are collected or how they are collected.
(iii) Researchers ask reflexive questions during data analysis, such as, ‘Is my bias about xxx informing my view of these data?’
(iv) Two or more researchers are involved in the analysis process.
(v) Data analysis member check (e.g. checking back with participants if the interpretation of their data is consistent with their experience and with what they said).
Before embarking on the study, researchers should consider their own experiences, knowledge and views; how this influences their own position in relation to the study question; and how this position could potentially introduce bias in how they collect and analyse the data. Taking time to reflect on the impact of the researchers' position is an important step towards being reflective and transparent throughout the research process. When writing up the study, researchers should include statements on bias and positionality. In quantitative research, we aim to eliminate bias. In qualitative research, we acknowledge that bias is inevitable (and sometimes even unconscious), and we take steps to make it explicit and to minimise its effect on study design and data interpretation.
Qualitative research typically uses systematic, non-probability sampling. Unlike quantitative research, the goal of sampling is not to randomly select a representative sample from a population. Instead, researchers identify and select individuals or groups relevant to the research question. Commonly used sampling techniques in anaesthesia qualitative research are homogeneous (group) sampling and maximum variation sampling. In the former, researchers may be concerned with the experiences of participants from a distinct group or who share a certain characteristic (e.g. female anaesthesia trainees), so they recruit selectively from within the group with this shared characteristic to gain a rich, in-depth understanding of their experiences. Conversely, the aim with maximum variation sampling is to recruit participants with diverse characteristics to obtain a broad understanding of the question being studied (e.g. members of different professional groups within operating theatre teams, who have diverse ages, gender and ethnicities).
As with quantitative research, the purpose of sampling is to recruit sufficient numbers of participants to enable identification of patterns or richness in what they say or do to understand or explain the phenomenon of interest, and where collecting more data is unlikely to change this understanding.
In qualitative research, data collection and analysis often occur concurrently. This is because data collection is an iterative process both in recruitment and in questioning. The researchers may identify that more data are needed from a particular demographic group or on a particular theme to reach data saturation, so the next participants may be selected from a particular demographic, or be asked slightly different questions or probes to draw out that theme. Sample size is considered adequate when little or no new information emerges from interviews or focus groups; this is generally termed ‘data saturation’, although some qualitative researchers use the term ‘data sufficiency’. This could also be explained in terms of data reliability (i.e. the researcher is satisfied that collecting more data will not substantially change the results). Data saturation typically occurs with between 12 and 17 participants in a relatively homogeneous sampling, but larger numbers may be required, where the interviewees are from distinct groups or cultures. 16 , 17
For data sets that involve 10 or more transcripts or lengthy interviews (e.g. 90 min or more), researchers often use software to help them collate and manage the data. The most commonly used qualitative software packages are QSR NVivo, Atlas and Dedoose.18, 19, 20 Many researchers use Microsoft Excel instead, or for small data sets the analysis can be done by hand, with pen, paper and scissors (i.e. researchers cut up printed transcripts and reorder the information according to code and theme). 21 NVivo and Atlas are simply repositories, in which you can input the transcripts and, using your coding scheme, sort the text into codes. They facilitate the task of analysis, rather than doing the analysis for you. Some advantages over coding by hand are that text can be allocated to more than one code, and you can easily identify the source of the segment of text you have coded.
Qualitative data analysis is ‘the classification and interpretation of linguistic (or visual) material to make statements about implicit and explicit dimensions and structures of meaning-making in the material and what is represented in it’. 22
Several social scientists have described this analytical process in depth. 2 , 6 , 22, 23, 24, 25 For inductive studies, we recommend researchers follow Braun and Clarke's practical six-phase approach to thematic analysis. 26 The phases are (i) familiarising the researcher with the data, (ii) generating initial codes, (iii) searching for themes, (iv) reviewing themes, (v) defining and naming themes and (vi) producing the report. These six phases are described next.
In this step, the researchers read the transcripts to become familiar with them and take notes on potential recurring ideas or potential themes. They share and discuss their ideas and, in conjunction with any sensitising concepts, they start thinking about possible codes or themes.
The first step in Phase 2 is ‘assigning some sort of short-hand designation to various aspects of your data so that you can easily retrieve specific pieces of the data’. 2 The designation might be a word or a short phrase that summarises or captures the essence of a particular piece of text. Coding makes it easier to summarise and compare, which is important because qualitative research is primarily about synthesis and comparison of data. 2 , 25 As the researcher reads through the data, they assign codes. If they are coding a transcript, they might highlight some words, for example, and attach to them a single word that summarises their meaning.
Researchers undertaking thematic analysis should iteratively develop a ‘coding scheme’, which is essentially a list of the codes they create as they read the data, and definitions for each code. 25 , 26 Code definitions are important, as they help the researcher make decisions on whether to assign this code or another one to a segment of data. In Table 1 , we have provided an example of text data in Column 1. TJ analysed these data. To do so, she asked, ‘What are these data about? How does it answer the research question? What is the essence of this statement?’ She underlined keywords and created codes and definitions (Columns 2 and 3). Then, TJ searched the remaining data to see if any more data met each code definition, and if so, coded that (see Table 1 ). As demonstrated in Table 1 , data can be coded to multiple codes.
How to code qualitative data: an example
| Research question To what extent do you think the surgical safety checklist (SSC) has changed teamwork culture in New Zealand operating theatres? | ||
|---|---|---|
| Data (The following quotes are excerpts from written responses to the above question that the authors CD and JW independently wrote, and TJ coded) | Potential code | Code definition |
| ‘In New Zealand, we have spent a lot of time trying to build whole of team engagement with the SSC through a change in the way it is delivered and by introducing local auditors who observe SSC delivery and score it against a behavioural marking scale. I think this has had a big effect on the way the SSC is delivered’. (JW) ‘We changed it so it was everyone's responsibility to lead different parts of the checklist, and got rid of the paper. This really helped’. (JW) ‘SSC has significantly improved teamwork culture by encouraging all disciplines of the operating theatre team to speak up and take ownership of safety in the operating theatre’. (CD) | Team responsibility | Participant describes processes or behaviour that demonstrates the SSC promotes teamwork or is managed by the team (rather than by one person). This includes behavioural change. |
| ‘It started out being a paper checklist that a nurse was tasked with signing off to certify that the SCC had been done. We changed it so it was everyone's responsibility to lead different parts of the checklist, and got rid of the paper. This really helped’. (JW) | Embedding the checklist | Participant describes processes that have made use of the SSC routine. |
| ‘I think that the SSC, along with our own approach to implementing it in New Zealand, and possibly other initiatives, such as NetworkZ and OWR, is changing the culture in New Zealand operating theatres. I think it's a combination of things happening at the same time that's influencing the culture in the operating theatres to be more team oriented, more inclusive and less hierarchical’. (JW) | Other influences on cultural change | Participant describes influences other than the SSC on teamwork. |
| ‘SSC has significantly improved teamwork culture by encouraging all disciplines of the operating theatre team to speak up and take ownership of safety in the operating theatre’. (CD) ‘In particular, nursing staff say that because of the introductions in the SSC and because they are asked if they have any concerns, they feel more part of the team’. (JW) ‘The overall management of the patient also feels more like teamwork as concerns from each discipline are discussed so that one aspect of a patient care is not siloed from another’. (CD) | Communication | Participant describes how communication (as an element of teamwork) is influenced by SSC |
In thematic analysis of interview data, we recommend that code definitions begin with something objective, such as ‘participant describes’. This keeps the researcher's focus on what participants said rather than what the researcher thought or said.
There is no set rule for how many codes to create. 25 However, in our experience, effective manageable coding schemes tend to have between 15 and 50 codes. The coding scheme is iterative. This means that the coding scheme is developed over time, with new codes being created as more data are coded. For example, after a close reading of the first transcript, the researcher might create, say, 10 codes that convey the key points. Then, the researcher reads and codes the next transcript and may, for instance, create additional four codes. As additional transcripts are read and coded, more codes may be created. Not all codes are relevant to all transcripts. The researcher will notice patterns as they code more transcripts. Some codes may be too broad and will need to be refined into two or three smaller codes (and vice versa). Once the coding scheme is deemed complete and all transcripts have been coded, the researcher should go back to the beginning and recode the first few transcripts to ensure coding rigour.
The second step in Phase 2, once the coding is complete, is to collate all the data relevant to each of these codes.
In this phase, the researchers look across the codes to identify connections between them, with the intention of collating the codes into possible themes. Once these possible themes have been identified, all the data relevant to each possible theme are pulled together under that theme.
After the initial collation of the data into themes, the researchers undertake a rigorous process of checking the integrity of these themes, through reading and re-reading their data. This process includes checking to see if the themes ‘fit’ in relation to the coded excerpts (i.e. Do all the data collected under that theme fit within that theme?). Next is checking if the themes fit in relation to the whole data set (i.e. Do the themes adequately reflect the data?) This step may result in the search for additional themes. As a final step in this phase, the researchers create a thematic ‘map’ of the analysis.
When viewed together, the themes should answer the research question and should summarise participant experiences, views or behaviours.
Once researchers have checked the themes and included any additional emerging themes they name the final set of themes identified. Each theme and any subthemes should be listed in turn.
The report should summarise the themes and illustrate them by choosing vivid or persuasive extracts from the data. For data arising from interviews, extracts will be quotes from participants. In some studies, researchers also report strong associations between themes, or divide a theme into sub-themes.
Tight word limits on many academic journals can make it difficult to include multiple quotes in the text. 27 One way around a word limit is to provide quotes in a table or a supplementary file, although quotes within the text tend to make for more interesting and compelling reading.
Ideally, each researcher in the team should be involved in the data analysis. Contrasting researcher viewpoints on the same study subject enhance data quality and validity, and minimise research bias. Independent analysis is time and resource intensive. In clinical research, close independent analysis by each member of the research team may be impractical, and one or two members may undertake the analysis while the rest of the research team read sections of data (e.g. reading two or three transcripts rather than closely analysing the whole data set), thus contributing to Phase 1 and Phase 2 of Braun and Clarke's method. 2
The research team should regularly meet to discuss the analytical process, as described earlier, to workshop and reach agreement on the coding and emergent themes (Phase 4 and Phase 5). The research team members compare their perspectives on the data, analyse divergences and coincidences and reach agreement on codes and emerging themes. Contrasting researcher viewpoints on the same study subject enhance data quality and validity, and minimise research bias.
There are a number of indicators of quality when reading and appraising studies.28, 29, 30, 31 In essence, the authors should clearly state their method of analysis (e.g. thematic analysis) and should reference the literature relevant to their qualitative method, for example Braun and Clarke. 2 This is to indicate that they are following established steps in thematic analysis. The authors should include in the methods a description of the research team, their biases and experience and the efforts made to ensure analytical rigour. Verbatim quotes should be included in the findings to provide evidence to support the themes.
A number of guides have been published to assist readers, researchers and reviewers to evaluate the quality of a qualitative study. 30 , 31 The Joanna Briggs Institute guide to critical appraisal of qualitative studies is a good start. 30 This guide includes a set of 10 criteria, which can be used to rate the study. The criteria are summarised in Box 3. Within these criteria lie rigorous methodological approaches to how data are collected, analysed and interpreted.
Another approach to quality appraisal comes from Lincoln and Guba, who have published widely on the topic of judging qualitative quality. 28 They look for quality in terms of credibility, transferability, dependability, confirmability and authenticity. There are many qualitative checklists readily accessible online, such as the Standards for Reporting Qualitative Research checklist or the Consolidated Criteria for Reporting Qualitative Research checklist, which researchers can include in their work to demonstrate quality in these areas.
As with quantitative research, qualitative research has requirements for rigour and trustworthiness. Thematic analysis is an accessible qualitative method that can offer researchers insight into the shared experiences, views and behaviours of research participants.
The authors declare that they have no conflicts of interest.
The associated MCQs (to support CME/CPD activity) will be accessible at www.bjaed.org/cme/home by subscribers to BJA Education.
Tanisha Jowsey PhD BA (Hons) MA PhD is a senior lecturer in the Centre for Medical and Health Sciences Education, School of Medicine, University of Auckland. She has a background in medical anthropology and has expertise as a qualitative researcher.
Carolyn Deng MPH FANZCA is a specialist anaesthetist at Auckland City Hospital. She has a Master of Public Health degree. She is embarking on qualitative research in perioperative medicine and hopes to use it as a tool to complement quantitative research findings in the future.
Jennifer Weller MD MClinEd FANZCA FRCA is head of the Centre for Medical and Health Sciences Education at the University of Auckland. Professor Weller is a specialist anaesthetist at Auckland City Hospital and often uses qualitative methods in her research in clinical education, teamwork and patients' safety.
Matrix codes: 1A01, 2A01, 3A01
1. Shelton C., Smith A., Mort M. Opening up the black box: an introduction to qualitative research methods in anaesthesia. Anaesthesia. 2014; 69 :270–280. [PubMed] [Google Scholar]
2. Braun V., Clarke V. SAGE; Thousand Oaks, California: 2013. Successful qualitative research: a practical guide for beginners. [Google Scholar]
3. Bengtsson M. How to plan and perform a qualitative study using content analysis. NursingPlus Open. 2016; 2 :8–14. [Google Scholar]
4. Savin-Baden M., Major C.H. Routledge; Oxford: 2013. Qualitative research: the essential guide to theory and practice. [Google Scholar]
5. Guest G., MacQueen K.M., Namey E.E. SAGE; London: 2011. Applied thematic analysis. [Google Scholar]
6. Terry G., Hayfield N., Clarke V., Braun V.J.T. In: Willig C., Stainton Roger W., editors. Sage; London: 2017. Thematic analysis; pp. 17–37. (The SAGE handbook of qualitative research in psychology). [Google Scholar]
7. Wiles R. A & C Black; London: 2012. What are qualitative research ethics? [Google Scholar]8. Merriam S.B., Tisdell E.J. 4th Edn. vol. 108. John Wiley & Sons; 2015. p. 136. (Qualitative research: a guide to design and implementation). [Google Scholar]
9. Dexter L.A. North Western University Press; Evanston, IL: 1970. Elite and specialized interviewing. [Google Scholar]
10. Long J., Jowsey T., Garden A., Henderson K., Weller J. The flip side of speaking up: a new model to facilitate positive responses to speaking up in the operating theatre. Br J Anaesth. 2020; 125 :1099–1106. [PubMed] [Google Scholar]
11. Long J., Jowsey T., Henderson K.M., Merry A., Weller J. Sustaining multidisciplinary team training in New Zealand hospitals: a qualitative study of a national simulation-based initiative. N Z Med J. 2020; 133 :10–21. [PubMed] [Google Scholar]
12. Castanelli D.J., Weller J.M., Molloy E., Bearman M. Shadow systems in assessment: how supervisors make progress decisions in practice. Adv Health Sci Educ Theory Pract. 2020; 25 :131–147. [PubMed] [Google Scholar]
13. Taylor S.J., Bogdan R., DeVault M. 4th Edn. John Wiley & Sons; Hoboken, New Jersey: 2015. Introduction to qualitative research methods: a guidebook and resource. [Google Scholar]
14. Kolb D.A. FT Press; London: 2014. Experiential learning: experience as the source of learning and development. [Google Scholar]
15. Foucault M. The subject and power. Crit Inq. 1982; 8 :777–795. [Google Scholar]16. Francis J.J., Johnston M., Robertson C., et al. What is an adequate sample size? Operationalising data saturation for theory-based interview studies. Psychol Health. 2010; 25 :1229–1245. [PubMed] [Google Scholar]
17. Guest G., Bunce A., Johnson L. How many interviews are enough? An experiment with data saturation and variability. Field Methods. 2006; 18 :59–82. [Google Scholar]
19. Scientific Software Development GmbH . 2021. ATLAS.Ti.https://atlasti.com/ Available from: [Google Scholar]
20. Dadoose . 2021. Dadoose.https://www.dedoose.com/ Available from: [Google Scholar]22. Flick U. In: The SAGE handbook of qualitative data analysis. Flick U., editor. SAGE; London: 2014. Mapping the field; p. 170. [Google Scholar]
23. Bazeley P. Analysing qualitative data: more than ‘identifying themes’ Malays J Qual Res. 2009; 2 :6–22. [Google Scholar]
24. Morse J.M., Field P.A. 2nd Edn. SAGE; Los Angeles: 1995. Qualitative research methods for Health professionals. [Google Scholar]
25. Saldaña J. SAGE; London: 2012. The coding manual for qualitative researchers. [Google Scholar]26. Braun V., Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006; 3 :77–101. [Google Scholar]
27. Greenhalgh T., Annandale E., Ashcroft R., et al. An open letter to the BMJ editors on qualitative research. BMJ. 2016; 352 :i563. [PubMed] [Google Scholar]
28. Lincoln Y.S., Guba E.G. But is it rigorous? Trustworthiness and authenticity in naturalistic evaluation. New Dir Program Eval. 1986; 1986 :73–84. [Google Scholar]
29. Hannes K., Lockwood C., Pearson A. A comparative analysis of three online appraisal instruments’ ability to assess validity in qualitative research. Qual Health Res. 2010; 20 :1736–1743. [PubMed] [Google Scholar]
