3 Design and background
3.6 Methods
A multi-method approach was used in the project. An overview of the methodological approaches used in the project is provided below. As reviewed in Section 2.6, no previous studies in this area have used a multi method approach, most likely because of their small scale and limited sample sizes.
Specific details of how these methods were used in each study are described in the chapters which describe each study (Chapters 4 – 8).
3.6.1 Qualitative methods
Qualitative research encompasses a wide range of data collection and analysis methods (Willig, 2013). Because qualitative methods can be used to create research designs that do not pre-suppose answers or frameworks, they are useful in studies which aim to describe or explore a phenomenon which is not well understood (Willig, 2013). They also have the advantage of allowing the identity and perspective of the researcher to be incorporated (Braun & Clarke, 2006). This latter feature was important in this project because of the influence of the researcher’s lived experience on the design and execution of the project, as described below (Section 3.8).
As discussed above, the project aimed to comprehensively describe the young people who are involved in an extensive program of youth mental health research. In light of the limited existing work in the area, an approach to qualitative research which focused on description, rather than interpretive interplay with existing theories, was chosen. One approach which does this is Qualitative Description (QD). As outlined by Sandelowski (2000, 2010) and Neergaard, Olesen, Andersen and Sondergaard (2009), QD should be the method of choice when a description of a phenomena is the aim of the study. The approach involves low inference interpretation of participants’ experiences, which given the limited literature to date, made it an appropriate method to investigate the aims of the project.
QD can involve a range of data collection methods (Neergaard et al., 2009; Sandelowski, 2000, 2010), including those which were employed throughout the project: semi-structured focus groups (Study 1), semi-structured interviews (Study 2),
and an analysis of existing data (Study 3). The rationale for choosing each of these methods is described in the relevant chapters.
The reporting of results of a QD study involves staying close to the data and describing participants’ experiences in a language similar to the informants’ own language. In the YIMHR project this style of reporting allowed the young people’s voices and experiences to emerge with minimal additional interference from the researcher.
A criticism of the QD approach is that it can lack rigour. Because of this, it is important to use strategies to enhance the rigour of the method (often referred to as validity criteria) and to reflect on what benefits these strategies have achieved. The strategies employed in this project for this purpose are those proposed by Milne and Oberle (2005): authenticity, credibility, criticality and integrity. Authenticity involves ensuring that the informants are free to speak, that their voices are heard and that their perceptions are accurately represented. Credibility involves capturing and portraying a truly insider perspective. Criticality involves reflection on the critical appraisal applied to every research decision and integrity involves reflecting on researcher bias,
informants’ validations and member checking and researcher triangulation. Reflection on the use of these strategies in the project can be found in Chapter 9, Section 9.5 (Strengths and Limitations).
All qualitative data were managed using NVivo for Mac 10.1.13, and were analysed using the analytical method proscribed by Sandelowski (2000, 2010),
qualitative content analysis. Specifically, the form of analysis used was that referred to as ‘conventional content analysis’ by Hsieh and Shannon (2005), with specific
terminology borrowed from Braun and Clarke’s (2006, 2013) guides to successful qualitative research. As defined by Hsieh and Shannon (2005), the goal of qualitative content analysis is to provide knowledge and understanding of the phenomenon under study’.
Common to many forms of qualitative analysis, this is a constructivist approach which acknowledges the existence of a range of meanings that could have arisen from the data, and that pre-existing knowledge cannot be eliminated from the analysis (Braun & Clarke, 2006). For all of the qualitative studies in this project, the following approach was followed.
The data collected- either transcripts (Studies 1 and 2) or the pre-existing applications (Study 3) - were read repeatedly by the researcher (RR) to achieve a deep immersion in the data and the information. In all studies, this immersion phase occurred
over a number of months to ensure that this immersion was sufficiently deep. During these periods, data for other studies and research for the broader project was also being conducted. These co-occurrences resulted in the immersion phases being complemented by broader thinking about the topic, which informed the analysis.
The data were then read word-for-word and labels, called ‘codes,’ were assigned to the most basic segments or elements of the text (designated ‘text extracts’) that can be derived from the data. The number of codes applied to each text extract vary
depending on the possible multiple or complex meanings of the extract. As many codes as required to capture this meaning are applied to the section of text. Where the data being analysed was collected using a protocol (e.g. a focus group or interview), then codes are applied without direct reference to this.
Following the completion of the initial coding, the codes are reviewed. Where the codes are found to be very similar, for example ‘CALD’ and ‘culturally and linguistically diverse,’ they are merged into one code prior to step 2.
After the codes have been reviewed, and collated as necessary, themes and features are identified. Braun & Clarke describe a theme as: ‘patterned meaning across a dataset that captures something important about the data in relation to the research question, organised around a central organising concept’(page 337, Braun & Clarke, 2013).
A feature is similar to a theme but does not have a central organising concept. It simply clusters together many different ways that an idea is evident in the data. Features are used to inform the interpretation of the themes.
Themes are generated from the codes by reviewing the codes and the related text excerpts, and grouping these into related areas. This is an iterative process, which took place over a substantial period of time (between 2 weeks and 2 months). The time between stages allowed RR to reflect on each section of the text. Where no clear theme emerged, or where there was substantial apparent overlap, the broader research team was involved to assist with overcoming these challenges.
These ‘initial’ themes are reviewed using Braun & Clarke’s (2013) levels of thematic reviewing as a guide. Level one involves reviewing the text extract attached to each of the codes and assessing whether the codes which had been assigned to a theme formed a coherent pattern. Where they do not, the theme was reworked by moving some of the codes to other themes or renaming the theme. Level two, which is conducted after Level one, involves considering the validity of each of the themes in relation to each other.
3.6.2 Quantitative methods
Quantitative methods, specifically online cross-sectional surveys and
longitudinal studies have a range of different advantages to qualitative methods. These include the ability to use pre-existing measures; sampling of large populations,
including through online dissemination; the ability for participants to be less readily identifiable; and the ability to ascertain if participants change over time (deVaus, 2002). The importance of these attributes for the project are outlined below.
Surveys facilitate the use of standardised measures, enabling the comparison of participant groups within studies, and with other studies of comparable populations (deVaus, 2002). Where possible, concepts were measured using pre-existing scales which were selected on the basis of their psychometric properties and appropriate length (deVaus, 2002). The CRC developed a list of Standard Measures which the organisation expected all researchers to use throughout their projects (Young and Well Cooperative Research Centre, 2013). Where possible, the specific measures used were drawn from this list. Researcher-devised questions were only used where suitable scales were not available.
Large-scale survey methods, and online delivery in particular, facilitate de- identification. Ensuring participants were not readily identifiable was an important factor for the YIMHR project because of the potential pre-existing relationship between the researcher and young people who had been part of the CRC.
The administration of surveys online, thus allowing for sampling of larger populations than was possible using alternative methods, was important in Study 5. The target population for this study was all young people who were involved in the CRC (estimated to be 1000) and all young Australians aged 18 to 25. Attempting to access a cross-sectional sample of these populations using either paper-based surveys, or more in-depth qualitative methods, would not have been practical.
Finally, longitudinal surveys facilitate comparisons within the same population sampled at multiple time points (Haslam & McGarty, 2003). This feature was used in Study 4 to compare the young people who were accepted into the YBT with those who were not.