• No results found

5.4 Data Analysis Procedures

5.4.1 Study I and Study II

A thematic analysis (Braun and Clarke, 2006) influenced by the hybrid-thematic analysis approach (Fereday and Muir-Cochrane, 2006) was performed in the analyses in Study I and Study II. Although there were some slight differences in how the analytical procedure was conducted for these studies, the overall procedure is based on the same analytical approach. The description in this section, therefore, refers to both studies. However, because different conceptual frameworks were used to analyse the interview data (due to dissimilar research questions), the two frameworks are therefore described separately.

Thematic analysis is a method that searches, identifies, analyses, and reports themes that emerge from the data considered important in descriptions of the phenomenon under investigation (Braun and Clarke, 2006; Daly et al., 1997). In this iterative approach to analysing data, the researchers immerse themselves in the data, develop initial codes, search for themes, and then review and define the themes. A hybrid-thematic analysis methods approach was applied since it combines an inductive (data-driven) and a deductive (theory-driven) approach to thematic analysis (Fereday and Muir-Cochrane, 2006). This approach was perceived

appropriate for the research questions in Study I and Study II because knowledge on social care managers’ actions and influences in implementation of EBP is scarce.

An immersion of the raw data was first performed after the completion of the interview transcripts to familiarize with the data. This was followed by an initial deductive abstraction analysis guided by pre-defined dimensions based on theory (through thematization (Kvale, 2008)) from previous research on managers’ implementation leadership actions, including what influences their leadership, which informed the interview guide. NVivo (Version 10) was used to conduct this analysis. In Study I, this related, for example, to line managers’ understanding of their role in implementing EBP and to the descriptions of how they practically work with implementation. In Study II, this related, for example, to the line managers’ perceived support in implementing EBP and the factors that influenced their implementation leadership. The pre-defined dimensions guided the initial coding of the data.3

Each interview was coded by two authors independently. The inter-rater reliability of each code was tested whereby two individually coded transcripts was compared between three of the authors using the same guide (as conducted in NVivo). Another inductive approach to analysing the transcripts was taken in which the relevant direct and indirect statements relating to the research questions and codes were, when relevant, generated. This resulted in an expansion of a previous code from the initial pre-defined dimensions. Themes that had emerged from these approaches were iteratively reviewed, discussed, and defined by the authors of Study I and Study II.

The analytical procedure also took an additional deductive approach in the use of conceptual models for each study that structured the themes that emerged from the data and gave them a practical perspective. Different theoretical frameworks were used for Study I and Study II since they aimed to answer different research questions. This analysis is described in Sections 5.4.1.1 and 5.4.1.2.

It is worth emphasising that this analytical procedure described above was not a clear linear process. Rather, the data were read and re-read carefully and were continuously discussed by the studies’ authors in a re-iterative process.

5.4.1.1 Use of conceptual model – Study I

The core implementation phases, as described by Fixsen et al. (2005), were used as a conceptual framework to deductively analyse which actions line managers take when leading implementation. This framework was chosen because it is based on an extensive review of the implementation literature (Fixsen et al., 2005) and has previously been used in practice and in research (National Board of Health and Welfare, 2013; Ogden et al., 2012; Sullivan et al.,

2008). In the use of this framework, certain managerial and leadership actions are sorted according to when they occur in the implementation process. There are four phases in the process.

The four phases are described as the “core phases” that should occur to ensure effective implementation. The phases are exploration (needs assessment, readiness for change),

installation (assessment of resources needed), initial implementation, and full implementation

(sustainability). Results from the deductive and inductive coding of the data that related to the line managers’ practical implementation work were re-read and summarised according to these four implementation phases. This was a reiterative process in which the authors agreed on the content of the final themes. The themes “Perceptions of evidence and EBP” and “Perceived role in implementing EBP” were generated from the initial deductive analysis and the inductive analysis of the data. However, they were not included in this additional deductive approach because these do not specifically concern managers’ actions in leading implementation (nevertheless, the two themes are included in Study I).

The data analyses for Study I were conducted separately for social services and for older people care as a result of the initial reading of the transcripts that showed notable dissimilarities between managers in these settings. For Study II, no examination of differences was required as the procedure was not pertinent to the research question.

5.4.1.2 Use of conceptual model – Study II

The CFIR was used as a conceptual framework for Study II to describe the data in relation to previous findings on influential contextual factors for implementation (Damschroder et al., 2009). The framework identifies the following five domains (that include 26 constructs and 13 sub-constructs) that are relevant in the study of effective implementation: Intervention

characteristics (8 constructs), Outer setting (4 constructs), Inner setting (5 constructs), Characteristics of individuals (5 constructs), and Process (4 constructs). This comprehensive

framework synthesises 19 existing theories and frameworks. Therefore, the framework was suitable for this research and because it focuses on what works across multiple levels of context. In line with Kaplan et al. (2010), three of the five CFIR domains were considered primarily as context: Outer setting, Inner setting, and Characteristics of individuals. After numerous discussions by the authors on which domains and constructs were most appropriate for Study II, Characteristics of individuals was excluded. The reason for this exclusion was the partial overlap between constructs in the Inner setting domain when considering the themes and codes generated from the raw data. Accordingly, it was decided to include the categories “Cosmopolitanism” and “External policy and incentives” in the Outer setting domain and to include all constructs in the Inner setting domain, including the sub-constructs “Learning climate” and “Leadership engagement”. In addition, the sub-construct “Leadership engagement” was divided into “Closest manager” and “Senior management” as the data revealed that these actors influenced the line managers differently as far as leading

implementation. The data were condensed and summarized according to these constructs. This was a reiterative process and the authors of Study II agreed on the final themes and their inherent codes. Figure 6 presents the final included domains, constructs and sub-constructs from the CFIR in Study II.

Figure 6. The CFIR domains, constructs and sub-constructs (numbered under each construct) included in the data analysis for Study II.