art&scienceresearch methods nursing standard: clinical · research · education

Use of constant comparative analysis in qualitative research
Hewitt-Taylor J (2001) Use of constant comparative analysis in qualitative research.
Nursing Standard. 15, 42, 39-42. Date of acceptance: March 19 2001.
Summary
This article describes the application of constant comparative analysis, which is one method that can be used to analyse qualitative data. The need for data analysis to be congruent with the overall research design is highlighted.

T

HE AIM OF qualitative research is to portray the reality of the area under investigation, and to enhance understanding of the situation and the meanings and values attributed to this by individuals; it does not involve the quantification of facts (Rose 1994). Qualitative methods emphasise the value of individual experiences and views, as encountered in real-life situations.
This type of investigation is often useful in nursing, as many issues concern the quality of the lived experience of individuals, which cannot be reduced to numerical values using statistical analysis. Sometimes a mixed methodology might be adopted, with elements of qualitative and quantitative enquiry being included in a study.
The nature of qualitative enquiry means that volumes of ‘rich’, ‘deep’ data are produced, often from a variety of sources. While not seeking to reduce data to statistical evidence, qualitative data nevertheless requires systematic analysis.
Given the volume of data produced, the practicalities of analysing, co-ordinating and ordering data into a form from which conclusions can be drawn and recommendations made, can appear overwhelming. In the qualitative research paradigm, a variety of data analysis procedures are commonly used (Polit and Hungler 1993). This article describes the use of constant comparative analysis, a method of analysing qualitative data where the information gathered is coded into emergent themes or codes. The data is constantly revisited after initial coding, until it is clear that no new themes are emerging. It can be used in a study with a single method of data collection, or in situations where multiple data collection methods have been used.
The study used to illustrate the process of data analysis explored the use of self-directed learning (SDL) in paediatric intensive care nurse education (Hewitt-Taylor 2000). It involved a six-

month case study of a paediatric intensive care course (ENB 415), using documentary analysis, repeated interviews with teachers and students, observation of course processes and selected lessons, and student learning diaries. The study also included a survey of seven centres offering the course: this involved interviews with teachers, focus groups with students and a survey of the managers of all paediatric intensive care units in
England, using postal questionnaires. A field diary was used to record additional data and personal reflections. Jaquelina Hewitt-Taylor RGN,
RSCN, BA(Hons), PhD, is Senior
Lecturer, Distance Learning
Centre, South Bank
University, London.
Email: hewittja@sbu.ac.uk

Coding data
All qualitative data analysis methods involve coding data into themes, then categories, to form conclusions (Jasper 1994); this study used constant comparative analysis (Benton 1991,
Morgan 1993). All notes from the analysis of the course documents, interview transcripts, observation notes, lesson transcripts, learning diary summaries, questionnaire transcripts and additional notes from the field diary, were coded.
The coding process was carried out by reading each of these documents and attributing a code to sentences, paragraphs or sections. These codes represented a theme or idea with which each part of the data was associated. For example, the code
‘nursing and self-directed learning’ was attributed to data that suggested there might be issues in self-directed learning pertaining specifically to nursing. Sections of transcripts were given no code, one code or more than one code.
The codes were written on hard copies of each document next to the related section. The codes and their definitions were recorded in a separate file. For example:
I Code – nursing and self-directed learning.
I Definition – any reference to, or indication that there might be issues relating to, selfdirected learning which is specific to nursing.
I Abbreviation – NSDL.
A separate file was used to ensure that the use of each code remained consistent and to establish a clear decision trail that could be used by auditors or future researchers. During data coding, notes were made about how decisions had been reached, how the coding process had been

Online archive
For related articles visit our online archive at: www.nursing-standard.co.uk and search using the key words below.
Key words

I Research methods
These key words are based on subject headings from the
British Nursing Index. This article has been subject to double-blind review.

july 4/vol15/no42/2001 nursing standard 39

art&scienceresearch methods nursing standard: clinical · research · education

conducted, and any specific queries raised.
Data analysis was inductive, as the study sought to promote understanding of individual perceptions, not prove a preconceived theory.
Codes were, therefore, generated from the data, rather than predetermined. Although literature-based codes can provide a useful tool, they can impede the development of new ideas
(Strauss and Corbin 1990). Having coded the first transcript, each subsequent reading of this and other transcripts was carried out with this in mind. New codes were added as necessary.
After coding the hard copy of each document, the copy was then highlighted, cut and pasted.
Each coded section was put onto a new ‘window’ and stored on a computer file with the title code.
The name of the participant whose interview the code pertained to, and the line numbers from the transcript, were included on each coded section. It could then be traced to the original to provide further contextual details that might become necessary as data analysis proceeded.
These were then electronically stored in a file with the name of the code. This was an efficient way to transfer coded segments into a storage area, with minimum retyping or rewriting. One hard copy of each coded transcript was retained, in addition to the electronic copy. After final coding was complete, code files were printed and stored in files labelled with each code name.
The quality of data analysis depends on repeated, systematic searching of the data
(Hammersley 1981). In an attempt to achieve this, repeated coding was performed to review interpretations, in the light of new data gathered and as new codes were generated, until no new insights were being gleaned (Riley 1990).
Established coded sections were compared with other similarly coded segments to ensure consistency of application, as well as adherence to the definition of the code (Strauss and Corbin
1990). Where events or conversations had been recorded in more than one of the methods used
(for example, in observation and interviews), both transcripts were reviewed together after initial coding. On some occasions, events from interviews, observations, or learning diary entries in the field diary had also been recorded.
Then diary entries were reviewed to check if there was any evidence of extraneous circumstances influencing the researcher’s interpretation of events, or impinging on the event being recorded, to review any other interpretations that were perceived at the time.
Data collection and analysis are interwoven in qualitative research. Some authors suggest that they should proceed together, and concept development should be examined in subsequent encounters with the study participants (Belgrave and Smith 1995, MacKenzie 1994). Data should
40 nursing standard july 4/vol15/no42/2001

also be analysed as promptly as possible after collection so that qualitative elements of the encounter recorded in the data can be recalled as accurately as possible (Carey 1995). This, combined with the recording of additional notes in a research journal or diary, ensures that qualitative data analysis is as rich and detailed as possible. The process of data analysis was, therefore, commenced before completing the fieldwork, with preliminary analysis performed in a week of the events recorded.
Once coding was completed, the codes that had common elements were merged to form categories (Strauss and Corbin 1990). The coded sections of data were placed in categories in the data collection methods used. This was performed electronically; files were created for each category, containing copies of the codes that had been merged to form the category. The definitions of the categories and the codes placed in these were recorded in the same way as codes. Some codes were placed in more than one category. The categorised data were then printed and stored manually in files with the name of each category.
The categories derived from each data collection method were then clustered around each research question they contributed to answering.
A list was complied of categories that related to each research question, and some categories were used to address more than one question.
Once all the research questions had been allotted input from the categories, the information pertaining to each question was examined and reviewed to compile a report. The findings were finally checked against the diary entries to identify whether the researcher’s views recorded before or during the study had unduly influenced interpretation of the data gathered.
Boxes 1 and 2 illustrate a selection of data from one data collection method and how it was broken down into codes, categories and answers to each research question. Each data collection method was analysed, coded, and categorised in this manner.
Validation by respondents
The interpretation of the responses and emergent findings were discussed with the case study participants. Nolan and Behi (1995) suggest that in qualitative research, the findings should be presented to participants and their views explored. Others suggest that this should also be applied to qualitative data analysis (Silverman
1993, Wellington 1996). The negotiation of outcomes in discussing the emergent findings with participants was considered congruent with the equality of power and mutual respect of SDL, making the method of analysis in one respect

art&scienceresearch methods nursing standard: clinical · research · education

congruent with the area studied. However, this procedure does not fully validate the findings, and might only mean that the interpretation given is acceptable to respondents (Silverman
1993). Before discussing the interpretation of data with participants, it was necessary to decide how disagreement over interpretations would be handled. Although achieving a balance of power is important, respondents should not be given too much power in relation to defining the research interpretations (Riley 1990). The researcher felt that respondents might disagree with the conclusions reached, but that they should be able to recognise the descriptions in the study as accurate portrayals of events (Fetterman 1989).
Previous interview data, findings from observations, and learning diary entries were explored at subsequent interviews, and there were no significant differences of opinion concerning the findings. However, some concerns were expressed over the extent to which one of the students agreed with the emergent findings. On talking to the student about how the issues had been, she said that she agreed with the interpretations, but the researcher was not entirely convinced that she did. She might have agreed, but it might be that she did not want to disagree with the researcher, which would reflect her general disinclination to disagree with teaching staff.
This type of problem is relevant to any research where power issues might be perceived to exist between individuals involved in the study, and might affect their inclination to disagree with interpretations, for example, in a nurse-client, or a teacher-student relationship. This type of problem is recognised in the literature (Silverman
1993), and indicates that member checks are not necessarily adequate to achieve complete trustworthiness in qualitative research. However, they do contribute to the trustworthiness of the research.
During focus group interviews with students and interviews with teachers in the survey, points were summarised as the interviews progressed to promote accuracy and clarify the emergent interpretations of the participants’ views (Carey
1995).
Methods of analysis
Data analysis forms part of the research methods used in an enquiry, and should, therefore, be consistent with the philosophical underpinning of the study. While placing a method of analysis into the ‘qualitative’ or ‘quantitative’ paradigm is relatively straightforward, ensuring that the data analysis method is congruent with the more subtle elements of the study is sometimes less clear.
Qualitative analysis concerns words not numbers, and exclusive counting of the frequency with which codes occurred, or opinions

expressed, would diminish the essence of qualitative data since its most important element is generating precise and in-depth meanings
(Morgan 1993). It has been suggested that counting how often codes occur is helpful in clarifying whether reality is in accordance with the overall impressions gained by the researcher
(Morgan 1993, Polit and Hungler 1993,
Silverman 1993). However, this view is disputed, as it is possible that numbers alone will become the focus, with the loss of subtle nuances of meaning and individual views, which are the strength of qualitative research (Morse 1995). In group processes, not all participants will respond to all questions or be present in all situations, making counting the frequency of codes meaningless (Saint Germain et al 1993).
In qualitative research, there is debate concerning the most appropriate way to establish the quality of research. Given this debate, the philosophical assumptions that underpin the research design and the subsequent evaluation criteria should be determined at the outset of the study. The researcher had set out to demonstrate that the study findings were trustworthy in terms of dependability and confirmability, credibility and transferability (Lincoln and Guba 1985). In deciding on the method of data analysis to be used, it was, therefore, important to ensure that the data analysis methods would embrace these criteria. One criterion used for establishing the confirmability of research is the establishment of an audit trail (Lincoln and Guba 1985). This is intended to allow other individuals to understand and evaluate how the data was coded and categorised, why data was placed into these codes and categories, and how these were clustered to answer the research questions. In establishing credibility, analysis procedures included triangulation of data by comparing interpretations and recordings, and data collection methods. Member checks added to the credibility criteria, albeit with an acknowledgment that these have their own inherent problems. On coding and categorising the data, it is important not to lose contextual and descriptive elements of the data, which add to the transferability of the research. This was augmented by ensuring that data, which were placed into categories, could easily be traced to the original transcript, to review any additional contextual data.
Conclusion
In qualitative data analysis, the main focus is not on quantification of facts, but rather on identifying the meanings and values attributed by individuals in real-life situations, with idiosyncratic and personal views forming an important

REFERENCES
Belgrave L, Smith K (1995) Negotiated validity in collaborative ethnography.
Qualitative Inquiry. 1, 1, 69-86.
Benton D (1991) Grounded theory. In
Cormack DFS (Ed) The Research
Process in Nursing. Second edition.
London, Blackwell.
Carey M (1995) Comment: concerns in the analysis of focus group data.
Qualitative Health Research. 5, 4,
487-495.
Fetterman D (1989) Ethnography Step By
Step. London, Sage.
Hammersley M (1981) Making sense of
Qualitative Data. Milton Keynes, Open
University Press.
Hewitt-Taylor J (2000) Self-directed
Learning: Its Use and Importance in
Nurse Education. Unpublished PhD thesis, Sheffield University.
Jasper M (1994) Issues of phenomenology for researchers of nursing. Journal of Advanced Nursing.
1, 9, 309-314.
Lincoln Y, Guba E (1985) Naturalistic
Enquiry. London, Sage.
MacKenzie A (1994) Evaluating ethnography: considerations for analysis. Journal of Advanced Nursing.
19, 774-781.
Morgan D (1993) Qualitative content analysis: a guide to paths not taken.
Qualitative Health Research. 3, 1,
112-121.
Morse J (1995) The significance of saturation. Qualitative Health
Research. 5, 4, 516-523.
Nolan M, Behi R (1995) Alternative approaches to establishing reliability and validity. British Journal of Nursing.
4, 10, 587-590.
Polit D, Hungler B (1993) Nursing
Research: Methods, Appraisal and
Utilisation. Third edition. Philadelphia,
Lippincott.
Riley J (1990) Getting the Most Out of
Your Data. Bristol, Technical and
Educational Services.
Rose K (1994) Unstructured and semistructured interviewing. Nurse
Researcher. 1, 3, 23-33.

july 4/vol15/no42/2001 nursing standard 41

art&scienceresearch methods nursing standard: clinical · research · education

Box 1. A selection of data from interviews with nurse teachers
Category 1:
Assessment
Codes:
Assessment
Entrance criteria
Formative assessment
Peer assessment
Selection
Self-assessment
Summative assessment
Validation /Verification
Category 2:
Content
Codes:
Concerns over covering content Content
Curriculum
Obligation to cover content Timetables
Category 3:
Elements of self-directed learning
Codes:
Active participation
Building on experience
Flexibility
Ground rules
Individual learning styles
Individual needs
Learning contracts
Learning from experience
Negotiation
Pace
Personal learning
Practitioner’s experience

Previous experience
Relevance
Student-centred learning
Student choice
Student control
Student led
Category 4:
Employment
Codes:
Employment issues and self-directed learning
Manager’s expectations
Category 5:
Factors affecting self-directed learning Codes:
Assumptions regarding student needs
Attendance
Culture
Differing teacher perceptions Factors affecting self-directed learning
Learning resources
Obligation to use self-directed learning
Resources
Teacher’s ability to be self-directed Teacher expectations
Time factors affecting self-directed learning
Category 6:
Group issues

Codes:
Competitiveness
Group dynamics
Group needs
Group size
Peer learning
Peer support

Category 14:
Preparation for self-directed learning Category 15:
Self-directed learning
Category 16:
Student ability to be self-directed

Category 7:
Learning defined
Codes:
Nature of knowledge
Skill acquisition
What is learning

Category 17:
Student evaluation
Category 18:
Student views

Category 8:
Not self-directed learning
Codes:
Behaviourism
Prescriptive

Category 19:
Subjects
Category 20:
Teaching methods

Category 9:
Nursing issues and self-directed learning Category 21:
Teacher’s role

Category 10:
Objectives

Category 22:
Teacher/student relationship

Category 11:
Organisational Issues

Category 23:
Teacher’s views of self-directed learning Category 12:
Practice

Category 24:
What is self-directed learning?

Category 13:
Preferred teaching and learning styles Category 25:
Why students attend

Note: Codes are only shown for Categories 1-8, but they were allocated in a similar manner for Categories 9-25.

Box 2. Category clustering
Research question 1: What do nurse teachers on the ENB 415 course understand by the term self-directed learning?
Categories: Learning defined, not self-directed learning, preparation for self-directed learning, self-directed learning or developing self-directed learning, what is self-directed learning?
Research question 2: What are the views of nurse teachers on the ENB 415 course concerning the use of self-directed learning?
Categories: Assessment, content, elements of self-directed learning, factors affecting self-directed learning, group issues, nursing issues and selfdirected learning, objectives, organisational issues, practice, preferred teaching and learning styles, preparation for self-directed learning, student ability to be self-directed, subjects, teacher’s beliefs about self-directed learning, teaching methods, teacher’s role, teacher/student relationship, why students attend?
Research question 3: What do students on the ENB 415 course understand by the term self-directed learning?
Categories: Preparation for self-directed learning
Research question 4: What are the views of students on the ENB 415 course concerning the use of self-directed learning?
Categories: Preferred teaching and learning styles, preparation for self-directed learning, student evaluation, student views, teacher/student relationship, why students attend?
Research question 5: What are the views of purchasers of the ENB 415 course concerning teaching and learning strategies used on the ENB 415 course? Categories used: Employment, nursing issues and self-directed learning, practice

Saint Germain M et al (1993) Survey and focus groups in health research with older hispanic women. Qualitative
Health Research. 3, 3, 341-367.
Silverman D (1993) Interpreting
Qualitative Data. London, Sage.
Strauss A, Corbin J (1990) Basics of
Qualitative Research: Grounded
Theory Procedures and Techniques.
London, Sage.
Wellington J (1996) Methods and Issues in Educational Research. Sheffield,
Sheffield University Division of
Education.

part of the overall picture. However, there is a need for thorough and systematic analysis of the information generated by these processes. It is important to consider how data will be analysed at the design stage of any research, to ensure that the analysis procedures proposed are in keeping with the overall philosophy, and fall within the evaluative criteria of the study.
Successful analysis and presentation of qualitative data requires a systematic and ordered approach so that complex data that emerge

42 nursing standard july 4/vol15/no42/2001

from a variety of sources can be collated and presented in a manageable form. Constant comparative analysis is one method that can be used to identify broad themes and patterns, or categories that emerge from qualitative research studies. In these categories, the precise nature of each individual’s view can be captured and recalled and data can be presented in a logical sequence in relation to the research questions addressed in the study