Chapter 4: Data Collection and Analyses
4.2 Analysis
A large variety of data were collected throughout this study, analysing this data demanded a number of techniques. Following the practices of many qualitative, exploratory methodologies (Silverman 2013, Lofland et al. 2007, Miles and Huberman 1994, Glaser and Strauss 1967), analyses began early in this study and were carried out concurrently to data collection. A variety of analysis techniques contributed to: the creation of the first interview schedule while formalising the observation field-notes, through producing the paired interview schedules, past the decision to cease data collection, until the findings chapters of this thesis were finalised. All analyses were shared with my research supervisors in order to maximise rigour. A discussion around the selection and use of each of these analyses follows.
4.2.1 Qualitative Analysis
In order to examine the qualitative data thoroughly, different types of analyses were used. As the combination of data types – rather than observations alone in the interactions – allowed for
more substantial claims (Teddlie and Tashakkori 2009, Robson 2002), the initial focus of the qualitative analyses was on the cases. Each case was analysed independently to identify variables and processes (which I refer to collectively as ‘concepts’) involved in psychosocial nursing (within-case analysis (Paterson 2010, Miles and Huberman 1984)). The cases were then analysed in relation to each other, to look for similarities and differences (between-case analysis (Burns 2010, Miles and Huberman 1984)). The interactions were included in the final ‘stage’ of qualitative analyses when the concepts arising from the cases underwent comparative analyses. These are explained below in order of occurrence.
All of the qualitative data were entered into an NVivo electronic qualitative analysis software project. This allowed coding of each case to identify emerging concepts (appendix 8a), comparison of the concepts between cases in the form of memos (appendix 8b), and diagrammatic representation of the concepts that emerged in the study (appendix 8c). The use of electronic packages aids managing large quantities of data but the responsibility for analyses remains with the researcher (Silverman 2011, Gibbs 2002, Hammersley and Atkinson 1995).
4.2.1.1 Within-Case Analysis
A ‘descriptive analysis approach’ (Miles and Huberman 1994) was taken. Each case was analysed during the data collection process, in order to develop interview schedules and identify issues for further exploration. As I typed up fieldnotes, I made reflective comments on sections of data that concerned psychosocial needs or nurses’ responses to these and made notes in the interview schedules. Each individual interview was created around these notes. As the study progressed, concepts that emerged recurrently were also added to an interview schedule template, so that each subsequent interview explored these concepts as well as anything new that arose from the observations.
Each individual case was searched for evidence relating to the research questions. Transcribing the interviews myself began this process (Silverman 2011, Lofland et al. 2006). As I transcribed interviews I made reflections on what had been said. As each interview write-up was completed I re-read the interview and identified further concepts. Once cases had been completed, all the data for each case were re-read in a search for additional concepts and to link concepts within the case.
Each emerging piece of relevant evidence was given a code identifying it as a significant concept (Glaser and Strauss 1967, Boyle 1994, Miles and Huberman 1994). For example, when nurses related their ability to deal with patients’ psychosocial needs to whether they knew this patient, I labelled this ‘familiarity’. Segments of data could be assigned more than one code. When I found further evidence in a case referring to a concept it was assigned the same code. Using NVivo I could then create documents collating each piece of evidence, under the appropriate code, to build up a picture of that aspect of the nurse’s response to each patient’s observed psychosocial need within each case. This comparison of words, or phrases, in a case to another part of the same case checks data and allows clarification (Glaser and Strauss 1967, Corbin and Strauss 2008, Creswell 2014).
4.2.1.2 Between-Case Analysis
The NVivo software enables printing of all segments of data relating to each code in one document and the creation of diagrams illustrating the relationships between codes. Doing this allowed me to explore occurrences of the same concepts in different cases to consider similarities and difference in the realisation of concepts between cases (Silverman 2011, Morse 1994). For example, when a nursing behaviour, such as how nurses responded to patient’s expression of a psychosocial need, was identified in one case, I could check all other cases for similar behaviours. When similar behaviour occurred, I could consider the factors involved and
When a regular pattern of concepts emerged, I was more alert to observing for future occurrence of these concepts. These frequently occurring concepts then became the focus for further data collection and analyses. The similarities and difference were compared to identify possible associations between concepts.
When patterns of concepts were less frequent, further exploration of them in this study was discounted. This process of funnelling data into categories is time consuming (Glaser and Strauss 1967, Silverman 2011) but allows for rigorous in-depth exploration of the key concepts in order to answer the research questions (Hammersley and Atkinson 1995). Funnelling also enables identification of dichotomous variables for comparative analyses (Glaser and Strauss 1967).
4.2.1.3 Comparative Analysis
More focussed analyses of all of the qualitative data, including interactions and encounters, occurred following the comparative approach advocated by Glaser and Strauss (1967) and Ragin (1987, 1994). This approach to analysis is particularly suitable for a small number of cases. The aim of this analysis is to determine whether there are specific factors which lead to specific outcomes – for example, what influences each nurse to behave in the way they do. This style of analysis is a process of recognising which factors are involved in which outcomes; and ruling out factors that have a different outcome under the same circumstances. This is done by forming pairs of variables (factors) that may affect each other within each case. All pairs of variables are then compared to the same pairs of variables in all of the other cases; if one pair of variables has an opposite outcome in a different case then these variables have been shown not to be related. These analyses identified associations for investigation in future studies, and indicated similarities and differences within this study’s data.
4.2.1.4 Summary of Qualitative Analyses
The qualitative data in the study’s 24 cases were analysed by identifying and coding the emerging concepts associated with nurses’ responses to patients’ psychosocial needs. A process of constant comparison and consideration of similar and different cases, with all of the qualitative data, enabled a narrowing down of concepts. Narrowing concepts allowed exploration of the key issues with potential associations to the nurses’ provision of psychosocial support.
Many of the variables identified in the qualitative analyses could be categorised, for example, the different ways nurses responded to patients’ psychosocial needs. Once a variable can be categorised, it can be counted and, therefore, undergo numerical analyses. These categorical variables, and many of the demographic variables, were analysed using one or more of the quantitative techniques described below.
4.2.2 Management of quantitative variables
Four main SPSS databases were created to manage the quantitative variables containing 1. A row per encounter,
2. A row per nurse, 3. A row per patients,
4. A row for one, randomly selected, encounter per nurse-patient interaction (this file was created to exclude impact of any individuals’ characteristics or pairs’ ‘relationship style’. For example, including data for all five encounters between one nurse and patient, when there are only two encounters for another nurse-patient pair would skew the results in favour of the first paring).
Quantitative demographic and organisational data and, where possible, concepts translated from the qualitative analyses, were added to the appropriate databases as variables. Exploratory analyses were undertaken to indicate whether there might be associations between possible dependent variables – nurse response to psychosocial need or nurse response style – and a number of independent variables, for example: the type of need, nursing experience, and patient care aim.
The lack of probability sampling, uncertain statistical representativeness of the samples and small sample sizes raised the question as to the appropriateness and value of inferential statistics. However, this does not discount the value of exploring quantitative data to support the qualitative findings in a hypothesis-generating as opposed to hypothesis-testing context. Simple descriptive analyses were therefore used to summarise relevant variables and cross- tabulations carried out to explore possible associations where appropriate.
4.2.3 Summary of Analysis
This study employed a variety of methods of qualitative data analyses to explore how patients expressed psychosocial needs and how nurses’ immediately responded to them. Constant comparative descriptive analyses of the qualitative data allowed identification and analyses of the key concepts associated with psychosocial needs and their support.