Documente Academic
Documente Profesional
Documente Cultură
Qualitative approaches
An overview of three
different approaches to the
interpretation of qualitative
data. Part 2: practical
illustrations
This paper continues the discussion of the essential features and methods
inherent within three approaches to the interpretation of qualitative data
presented in Part 1 (Priest et al 2002). Leslie Woods, Helena Priest and
Paula Roberts explore the use of grounded theory, qualitative content
analysis and narrative analysis through the use of an illustrative example
44
Qualitative approaches
Box 1: Transcript extract
Q: Can you tell me about how you have helped a patient to stop smoking.
Nurse: One was a lady who has smoked all her life from the age of
Zyban, and it was a case of this is my only chance, if I dont stop smoking
now then I wont ever She was almost in tears. She said you have to take
appointment. I gave her a leaflet to fill in outlining 10 reasons why she wanted
to stop smoking and a diary of a day giving how many times she smoked.
10
When I saw her the first time I wanted to make sure she could commit to
11
attending the seven-week programme and then it was really over to her to
12
talk about motivation for why she wanted to give up smoking. At each
13
14
15
16
potential risks of smoking. I try to apply it to her by looking at the risks that
17
are relevant to a woman of her age and the relevant length of time that she has
18
smoked
19
20
Nurse: I think the general level of awareness is quite good and that instils
21
a lot of guilt in patients because they then feel guilty that they are knowingly
22
doing something that is harming them. You have to take that on board when
23
you are talking to them because that is not very helpful. She then sees me
24
at the end of the first day that she has quit as a motivator, which is useful,
25
26
45
reflect the main concepts that are conveyed. Looking over the first 15
lines or so, there are a number of codes that can be applied to the text.
For example, in lines 3-6, issues are raised to which the codes
motivation, timing and urgency (as well as others) could be generated.
In fact, the code motivation can be applied to other pieces of this
interview extract, in fact all 26 lines!
In this example, it is apparent that the same data extract can be coded
under a number of different conceptual headings. This is quite normal in
grounded theory and is merely a reflection of the way we communicate
in everyday situations. That is to say that our interactions are rarely
limited to discussing discrete ideas, divorced of context; we tend to
communicate in a complex and often haphazard way with the result that
a single speech can address multiple issues.
After reading this extract several times it becomes clear that motivation
is an important concept that appears to have a number of features and
characteristics. Consequently, motivation is likely to become a category
as opposed to a single code and, as such, it is important to identify the
various properties or codes related to this central concept. In other words,
as the analyst, once you have identified motivation as a key issue, you
might start to ask the questions such as: what factors motivate people
to stop smoking? What factors sustain motivation? Looking at the text
in response to these and other questions, a number of factors start to
become apparent within this single transcript excerpt. For example
timing [lines 4-7], commitment [lines 8-12], positive reinforcement [lines
6-8, 12-14], health gain [lines 12-15], health risks/consequences [lines
15-18], guilt [lines 20-22] and so on. It is not important that you may
have used different terms to name these properties, but it is important
that all codes are clearly defined and accurately reflect the meaning
within the data. In that way, as each new instance of a defined concept
is identified, either later in the same interview or in other data, it can be
added to the coding system.
Axial coding
Axial coding is used to start to make tentative connections and
relationships between the first-level open codes and categories initially
NURSE RESEARCHER VOLUME 10 NUMBER 1
46
Qualitative approaches
generated. This is done following the coding paradigm (Strauss and
Corbin 1998) outlined in Part 1 in this issue of Nurse Researcher (Priest
et al 2002). So, in our example the central phenomenon of the model is
motivation. What led to this phenomenon was a desire to stop smoking;
that is, the causal condition. The properties of a desire to stop smoking
include codes we have generated, for example, health gain, although this
requires further confirmation at this stage and there are likely to be other
causal conditions. The codes of timing and urgency are intervening
conditions that relate to the central concept and the action/interactional
strategies. The action/interactional strategies related to the central concept
include the commitment to the education programme, outlining reasons
for stopping smoking and the prescribed treatment for the patient, as well
as the positive reinforcement of the nurse.
It should be remembered that open and axial coding are iterative
processes that occur concurrently with data collection. Movement
between both phases is often not overt, and it is possible for the analyst
to move back and forth in one coding session without necessarily being
aware of this. The use of the grounded theory approach allows the
connection of codes and categories in the data to be established and
theoretical propositions to be developed. For example, you might arrive
at a proposition that motivation to stop smoking is primarily triggered by
an acknowledgement of health risks and consequences within the
individual. As an analyst and researcher, you would go back to your data
to see if there was evidence to support or refute your proposition.
Moreover, this might become the focus of future data collection when
you might wish to interview patients at a smoking cessation clinic.
As stated at the outset, there are other approaches to the ways in
which data can be analysed in qualitative research, and the next to
be discussed is content analysis.
Content analysis
The interview may have been conducted as part of an exploratory study,
and therefore the main analytic categories would be known. If this was
NURSE RESEARCHER VOLUME 10 NUMBER 1
47
the case, key concepts in the interview questions would form the master
codes, i.e. help to stop smoking and awareness of risks. Thereafter,
second-level coding would follow on from these master codes. If content
analysis was to be undertaken on text without master codes being known,
the first step would involve reading the text several times to get a feel for
the master codes, from which further sub-codes would emerge. Let us
presume this is the case. If we were utilising a computerised software
package the main analytic categories may be labelled: age [from lines 2
& 3], quantity [line 3], medication [line 4], emotion [line 5], motivation
[lines 6, 12], motivator [lines 24, 26], nursing intervention [lines 8, 9, 13,
25], commitment [line 10] and risks [lines 15,16,19].
If we had text from several respondents, first-level coding would ensue
across all data sets to ascertain the master codes. Thereafter, second- and
third-level coding would ensue to ascertain core themes within each main
category. For example, under the master code nursing intervention, secondlevel coding would include text under the headings of: appointment [line
8], leaflet [line 8], diary [line 9] and carbon monoxide reading [line
13]. If several pieces of text, either from within one text or several texts
pertained to the same concept they would be copied and pasted under the
appropriate sub-code. Then, within each section of text, thematic
interpretation of meaning is undertaken. If several respondents use similar
words, manifest content analysis is possible; otherwise interpretation of
meaning is facilitated through latent content analysis. To illustrate derived
themes, respondents actual phrases may be utilised, in a reverse process
of deriving evidence to support theoretically derived concepts (see Woods
and Roberts 2000). For example, for the theoretically derived concept
motivator [lines 24, 26], evidence may readily be derived from the actual
text by reference to line numbering to illustrate the concept, for example,
I do another carbon monoxide reading which is substantially less than
her previous reading thats a really good motivator. This method of
qualitative content analysis can readily be crosschecked for inter-coder
reliability, and is thus a very robust form of coding.
The third and final approach to analysis of the transcript extract is
described below.
NURSE RESEARCHER VOLUME 10 NUMBER 1
48
Qualitative approaches
Box 2: Segment of transcript
Defining feature
(Gee 1986,1991)
Orientation
Thirty-five a day.
Complicating action
She said: You have to take me now or its not going to work.
I just thought,
Evaluation
Resolution
Coda
Narrative analysis
Following the principles described by McLeod and Balalmoutsou (2000),
a segment of the transcript [lines 2 8] has been selected for microanalysis. This segment can be identified as a story in that it has a beginning,
middle and an end, as well as a logical temporal sequence (Riessman
1993); furthermore, it is possible to identify the defining features of
orientation, complicating action, evaluation, resolution and coda, as outlined
by Labov and Waletzky (1967). The transcript has been transformed into
stanza form (Gee 1986, 1991), with Labov and Waletzkys (1967) key
elements identified. These features are indicated in the right-hand column
of Box 2, alongside the appropriate section of transcript.
NURSE RESEARCHER VOLUME 10 NUMBER 1
49
Reduced transcript
Sub-plots
Patients plight
50
Qualitative approaches
selection of such a short illustrative example could result in interpretations
being made that are out of context with the larger narrative.
As with other forms of qualitative research, data selected for narrative
analysis are open to multiple interpretations. As Riessman (1993) has
noted narratives are interpretive and, in turn, require interpretation: They
do not speak for themselves. This may lead to doubt about the
credibility of analysis (Ayres and Poirier 1996). Furthermore, rules of
story structure and grammar are culturally determined, and may not be
shared between researcher and participant (McLeod 1997). Thus, despite
following a set procedure, individual researchers may arrive at different
conclusions regarding the structure, shape and meaning of stories.
However, as Ayers and Poirier (1996) have argued, provided that the
analysis has been conducted responsibly, and remains faithful to the data
and the methods, then the conclusions drawn can be accepted as a valid
interpretation. Thus, narrative analysis may be considered as a viable
qualitative approach in nursing and health care research.
Conclusion
This paper has provided practical exemplars to illustrate the fundamental
analytical principles inherent within three qualitative research approaches
which can be applied to the practical analysis of interview or other textual
data. While necessarily brief, it is hoped that this practical overview,
which should be read in conjunction with Paper 1 (Priest et al 2000)
introducing the background philosophies and analytical principles of the
three different approaches, will assist researchers in making informed
decisions in their approach to data handling and interpretation, and inspire
them to seek out more detailed texts.
Leslie Woods PhD, BSc(Hons), Cert Ed, RGN, DN Cert, Lecturer,
University of Liverpool, England
Helena Priest PhD, MSc, BA, RMN, Dip N, Dip N Ed, Lecturer, Keele
University, Staffordshire, England
Paula Roberts PhD, MA Ed Mgmt, RMN, RGN, Cert Ed, Cert HEd,
ILTM, Senior Lecturer, Keele University, Staffordshire, England, and
Editor, Nurse Researcher
NURSE RESEARCHER VOLUME 10 NUMBER 1
51
References
Ayers L, Poirier S (1996) Virtual text and
the growth of meaning in qualitative
analysis. Research in Nursing and Health.
19, 163-169.
Emden C (1998) Conducting a narrative
analysis. Collegian. 5, 3, 34-37.
Gee JP (1986) Units in the production of
narrative discourse. Discourse Processes.
9, 391-422.