Chapter 20: Coding approaches

Tess Tsindos

Learning outcomes

Upon completion of this chapter, you should be able to:

  • Describe coding.
  • Describe inductive and deductive qualitative coding.
  • Understand how to conduct qualitative coding.


What is coding?

Coding is a process whereby the researcher applies a label to describe the meaning of text and is the first step in qualitative analysis. It is important to be clear about the terminology used in coding and analysis. Erlingsson and Brysiewicz1 have provided the following succinct definitions, which can help the researcher understand the three different, commonly used terms:

  • Code – a label, a name (or tag) that describes the meaning of the text, usually in one or two words. It is about the meaning, not only the words in the transcript.
  • Category – active grouping or organisation of codes that are related to each other through content or context. A category answers questions about who, what, when or where. Category names should be factual and short. Codes can be grouped into subcategories and then moved into broader categories.
  • Theme – expresses the underlying meaning across categories. A theme is interpreted from the codes, categories and context or meaning of the transcript. It answers the questions why, how, in what way or by what means. Poetic and metaphoric language can be used; a theme is descriptive and uses verbs, adverbs and adjectives. A theme is more a statement than simply a word.

Deductive and inductive coding

There are two primary methods of coding: deductive and inductive.2 Deductive coding begins with a set of pre-established codes, which the researcher applies to the data set (e.g., interview transcripts). Deductive coding is more descriptive than interpretive, and is most often used for content analysis (see Chapter 21). For example, in a study about ‘falls’, a researcher asked participants for their perceptions and experiences of falls, whether they believed falls were a problem, and to what extent they are a problem. One participant stated:

Answer: I think they’re an unsolved problem. I think they’re something that we certainly need to manage better, and that’s not unique to [Hospital], it’s across the globe. And I don’t think that we’ve paid enough attention to it in the past. And I think we can learn from experiences, for instance, bacteraemia where there’s been a focus on prevention that’s worked so greatly. And then when we think about where we were 10/15 years ago with that, where we just more or less accepted that there’s going to be bacteraemia and people would be sick from them and die from them. And so I think falls is in that bracket now, where we need to put that concentrated effort into it. They should be preventable. We shouldn’t have them. We need to implement multiple, I think, projects or processes to prevent them. And I think you’re correct, there are many different types of patients, different environments, different type of buildings, different equipment available, that makes it fairly unique in each situation and in each ward even.

(Anonymous participant in unpublished falls study dataset collected by the authors)

The following are some of the deductive codes that could be generated from this answer. Remember that deductive codes come from the interview guide and prompts, and may also be also based on the literature. These are the codes the researcher is actively looking for in the data, rather than the codes that the researcher finds from the data (inductive).

  • Attitudes or perceptions
  • Unsolved problem
  • Learn from past lessons
  • Need to manage
  • Preventable
  • Ward layout
  • Patient characteristics
  • Access to resources.

Dacillo et al provided another good example of deductive coding. They examined ‘videoconferencing fatigue and its relationship with online student engagement (OSE) during the COVID-19 pandemic’.3(p1) The study examined existing concepts or ideas, informed by pre-existing theory – the surface meaning of their data. Dacillo’s first objective followed a deductive process because the themes had already been identified from the literature and the scale used (the five domains of videoconferencing).

Inductive coding works the opposite way to deductive coding. Codes are created based on what is revealed in the data set (e.g. interview transcripts). For example, inductive coding explores meanings and issues raised by the participant, such as fear of dying from falling, or the unique nature of each fall across the world. Inductive coding is more interpretive than deductive coding, and often leads to the development of new theories. Inductive coding can be used in thematic analysis and framework analysis (see Chapter 22), and in grounded theory (see Chapter 25) as open, axial and selective coding. For a good example of inductive coding, refer to Wang et al.4 Wang et al explored ‘the effects of initial contact with the clinical learning environment on first-year nursing students’ empathy levels and perceptions of professional identity’.4(p1) Diary recordings about their clinical learning experiences were collected and a thematic analysis was conducted on the content regarding professional identity in nursing students’ diaries. Based on the framework of grounded theory (see Chapter 25), the researchers adopted an inductive thematic analysis approach developed by Braun and Clarke.3 Wang discovered that first-year nursing students’ initial contact with the clinical learning environment helped them to enhance empathy and shape professional identity. Returning to Dacillo et al3, their second objective followed an inductive process because the themes were formed on the basis of emergent, lower-level concepts gleaned from the narratives.

Nearly everything mentioned in an interview can be coded, including but not limited to:

  • setting and context
  • definition of situation
  • perspectives
  • ways of thinking about people and objects
  • process
  • activities
  • actions
  • events
  • conditions
  • consequences
  • strategies
  • relationship and social structure
  • meanings.

How to code

Before commencing coding, the researcher must accept that they are the data analysis tool. Because qualitative data analysis is subjective and inductive, the way the researcher sees the world will influence how data will be collected and interpreted. To ensure accurate coding, the researcher needs to acknowledge their own influences, such as education, gender identity, social class, professional role, preconceptions, ethnicity, and many more.

It is advisable to create a coding guide, which can be developed prior to data analysis, based on a literature review. The coding approach can also evolve through the process of coding. Codes can be factual (content analysis) or interpretive (thematic analysis or grounded theory). The coding guide will have names for codes, categories and themes. These names can come from the researcher, the participants and/or the literature. Most often, the researcher comes up with terms, concepts and categories that reflect what they ‘see’ in the data.

Coding can vary depending on the approach to data analysis. Coding for content analysis is most often deductive and categorical. Coding for thematic analysis5 and grounded theory are most often inductive, deductive, matrix and charting, open, axial and selective coding (see chapters 22 and 25). Thematic and grounded theory coding entails examining the text for codes that are discovered in the data. Table 20.1 provides some examples of various types of coding.

Table 20.1. Coding examples from health and social care research

Title The language of TV commercials’ slogans: a semantic analysis6 Measuring the patient experience of mental health care: a systematic and critical review of patient-reported experience measures7 Contexts and mechanisms that promote access to healthcare for populations experiencing homelessness: a realist review 8 Transforming while transferring: an exploratory study of how transferability of skills is key in the transformation of higher education9
CC Licence CC BY 4.0 CC BY-NC 3.0 CC BY-NC 4.0 CC BY 4.0
First author and year Noor, 2015 Ferdinand, 2020 Siersbaek, 2020 Bazana, 2018
Aim To analyse linguistic tools that copywriters of TV commercials use in relation to product specialism; to highlight the semantic property of TV commercial slogans; to pinpoint strategies employed by copywriters to influence viewers
To provide an overview of the psychometric properties and the content of available PREMs(p.2156) To identify and understand the health system contexts and mechanisms that allow for homeless populations to access appropriate health care when needed(p.16) To highlight issues in terms of institutional culture and race, as well as establish and highlight the problematic issue of skills transferability within higher education between academics(p.13-14)
Study design Descriptive qualitative Inductive qualitative Realist review Explanatory qualitative
Data collection Watching TV channels A comprehensive review of the published peer-reviewed literature was conducted using the MEDLINE bibliographic database A review of reviews and grey literature. Interviews
Coding analysis approach Semantic: explicit content of the data (factual codes), obvious meanings, also known as descriptive codes Latent: concepts and assumptions underpinning the data, underlying concepts, also known as conceptual codes Realist: reality is evident in the data; what objectively exists in the data (inductive and deductive coding) Constructionist: how reality is created in the data
Results The study unveils the underlying mechanism of encoding and decoding meanings of the discussed slogans of TV commercials. Provides a description and a critical analysis of the available PREMs for mental health care that can help professionals choose PREMs that best suit their needs. Access to health care for populations experiencing homelessness depends on adequately resourcing and training providers to meet the needs of patients in a welcoming and attentive setting without stigma and judgement. Services should be closely linked, and staff and providers should be empowered to take responsibility for providing flexible, responsive and opportunistic care in flexible settings. The contexts in which this is possible arise in a respectful, empathetic culture, which is created when managers and leaders value and champion it. The diversity of cultures does not allow for a one-size-fits-all approach to skills transferability and collaboration, and the diversification and allowed growth of new generation academics will only be realised by the institution once the institutional culture embedded in the policies and procedures of the university for a dynamic response to teaching.

Creswell10 developed a figure on page 244 of his book to demonstrate the coding process. While Creswell demonstrates these steps as a linear process, in fact they do go back and forth to refine codes, categories and themes. He commences the process by reading the text data and then dividing the text into as many segments of information as possible. Then codes are assigned to the segments, the codes are reduced, and finally codes are collapsed into themes. He presents a linear process that commences with many pages of texts and ends with everything reduced to five to seven themes. Refer to his book for the figure.

It is advisable to create a coding guide that can be developed prior to analysis, based on a literature review. Codes can be factual (content analysis) or interpretive (thematic analysis/grounded theory).

Coding can also be thought of as a simple, 3-step process involving deconstruction and reconstruction of text.

Step 1: Deconstruction, or fragmentation2

Deconstruction begins by assigning a code word or phrase that accurately describes the meaning of the text segment. Then line-by-line coding is conducted, first in theoretical research (grounded theory, framework) or segmentation (applied thematic analysis). More general coding involving larger segments of text is adequate for practical research. Deconstruction breaks down data and reconceptualises it, and makes comparisons between events, actions and interactions. Then conceptual labels are applied, and these are grouped into categories. Initial relationships between categories are developed.

Clustering is undertaken as the next action. After coding text, a list of all code words is developed, whereby similar codes are brought together (clustered) and redundant codes are reviewed. The objective of clustering is to reduce the long list of codes to a smaller, more manageable number (perhaps 25–30 codes).

The researcher should be looking for key phrases. Are there phrases that make some sense but are not necessarily describable? People often circle through the same ideas in their answers: are there topics that occur and recur? Are there local or common terms that are used in unfamiliar ways (e.g. ‘women’s troubles’, ‘bad blood’, or ‘difficulties’)? Are metaphors or idioms used (e.g. ‘rock-solid marriage’ or ‘cooking with gas’)? If yes, what do they represent?

Step 2: Reconstruction

Reconstruction involves specifying more rigorous codes than in the first step. Data is combined (put back together) to demonstrate new ways of making connections between a category and subcategories. Table 20.2 demonstrates the code-to-category reconstruction.

Table 20.2. Interconnecting the data

Deconstruction: code Reconstruction: category
Attitudes or perceptions

• Ward layout

• Patient characteristics

• Access to resources
Barriers or challenges to falls prevention

Step 3: Grouping

Grouping brings categories into a core category that is built from a core code. This is a point of theoretical integration, used in selective or theoretical coding with grounded theory. The story is built from categories.

After you have created codes and categorised them, you have come to the end of the coding process. After this process, you move on to creating and naming themes, and then describe the approach to the analysis (see chapters 21–25).


Coding is a process of assigning labels to text. Coding can be inductive or deductive. Each coding approach has its own techniques but should be chosen on the basis of the researchers’ approach to analysing their data.


  1. Erlingsson C, Brysiewicz P. A hands-on guide to doing content analysis. African Journal of Emergency Medicine. 2017;7(3):93-9. doi:10.1016/j.afjem.2017.08.001
  2. Saldana, JM. The Coding Manual for Qualitative Researchers. 3rd ed. SAGE; 2015.
  3. Dacillo MJF, Dizon JKM, Ong EJT, Pingol AML, Cleofas JV. Videoconferencing fatigue and online student engagement among Filipino senior high school students: a mixed methods study. Front Educ. 2022;7:973049. doi:10.3389/feduc.2022.973049
  4. Wang Q, Cao X, Du, T. First-year nursing students’ initial contact with the clinical learning environment: impacts on their empathy levels and perceptions of professional identity. BMC Nurs. 2022;21:234. doi:10.1186/s12912-022-01016-8
  5. Braun V, Clarke V. Thematic Analysis: A Practical Guide. SAGE; 2002.
  6. Noor M, Raza-e-Mustafa, Muhabat F, Kazemian B. The language of TV commercials’ slogans: a semantic analysis. Communication and Linguistics Studies. 2015;1(1):7-12. doi:10.11648/j.cls.20150101.12
  7. Fernandes S, Fond G, Zendjidjian XY et al. Measuring the patient experience of mental health care: a systematic and critical review of patient-reported experience measures. Patient Prefer Adherence. 2020;14:2147-2161. doi:10.2147/PPA.S255264
  8. Siersbaek R, Ford JA, Burke S et al. Contexts and mechanisms that promote access to healthcare for populations experiencing homelessness: a realist review. BMJ Open. 2021;11:e043091. doi:10.1136/bmjopen-2020-043091
  9. Bazana S, McLaren L, Kabungaidze T. Transforming while transferring: an exploratory study of how transferability of skills is key in the transformation of higher education. Transformation in Higher Education. 2018;3(a35). doi:10.4102/the.v3i0.35
  10. Creswell JW. Educational Research. Planning, Conducting, and Evaluating Quantitative and Qualitative Research. Pearson Education; 2002.