A multiple case study design using the realist approach was chosen because it allowed in-depth study of development and sustainment of DFIs within the real-life context of Dutch location-based DFCs (24). Multiple methods of data collection were performed, including semi-structured interviews, participating observations, available documents and focus groups.
Case selection
Selected cases were DFCs that were officially recognised as such according to Dutch Alzheimer Association’s criteria (13). Within the cases of Dutch DFCs, a DFI was our unit of analyses. A purposeful sampling method (25, 26) was used based on maximum variation of DFCs in terms of rurality (urban vs. rural sites), the geographic scatter in the Netherlands, duration of being a DFC and characteristics of DFIs within a DFC (8). This sequential sampling method was established together with the advisory panel of Mentality. A multiple recruitment strategy was adopted by identifying cases through the cooperation of the Dutch Alzheimer's association, the advisory panel of Mentality, and the network of the research team at national and/or regional level.
Twenty-five potential DFCs were identified and were invited by the research coordinator (JP) to take part. Nine DFCs were willing to participate, and were considered in terms of the maximum variation and the opportunity for identifying contextual factors, mechanisms and outcomes. Four DFCs were selected by the research team and advisory panel as best possible sites.
After the initial selection of DFCs, additional information was sent to key informants, such as policy officers. The key informants informed and consulted other stakeholders involved in the development and sustainment of DFIs, such as social and health professionals and/or volunteers, before making definite commitments. After this step, a letter of commitment was signed by the local policy officer for each case.
Table 1 provides an overview of included cases and their characteristics.
Table 1 Included Cases and Their Characteristics
DFC
Selection criteria
|
A.
|
B.
|
C.
|
D.
|
DFC since
|
2015
|
2019
|
2018
|
2015
|
Urban vs rural
|
Rural
|
Urban
|
Rural
|
Urban
|
National geographic dispersion
|
South-east
|
West
|
North
|
South
|
Dementia friendly initiative (DFI)
|
Intergenerational gardening and adapting the physical environment
|
Alzheimer Café and Odense housing
|
Dementia-friendly museum
|
Dementia-friendly choir and shared living room for elderly, including people with dementia
|
Recruitment of participants within the DFCs
Recruitment of participants within DFCs was organized around each DFI as unit of analysis. Participants were stakeholders who were actively involved in the development and sustainment of one or more DFIs. The aim was to include a variety of stakeholders who were involved at both the municipal and implementation level of one or more DFIs. Given time constraints, we set a target of 7–12 participants per DFC for project manageability.
Participants were recruited through a purposeful sampling and snowball sampling; key informants in each DFC identified participants who were active and engaged in the development and implementation of DFIs. These participants assisted in the identification of other eligible participants.
Data collection
Following the realist and case study methodology, multiple methods of data collection were used. These were semi-structured interviews, participating observations, available documents and focus groups. In all cases, data collection started with semi-structured interviews, interspersed with observations and documentation. Data collection was divided among four researchers. One pair (MT and JP) collected data in A and B. Another pair (ML and RJ) collected data in C and D. All data were collected between January 2019 and July 2019. See Table 2 for an overview of data collection.
Semi-structured interviews
Semi-structured interviews were conducted using a topic guide following the realist principles. They included open-ended and exploratory questions for developing program theories (27). The purpose was to gain insight into the nature of the DFI; the involvement of the interviewee in the DFI; and their experiences with and perspectives on how the DFI was developed, implemented and sustained. Special attention was given to contextual aspects and mechanisms leading to outcomes. All interviews were audio recorded with the permission of the participant(s). Each interview lasted approximately 40–60 minutes, and took place face to face, on location. In total, 29 professionals, 10 volunteers and 1 person with dementia were interviewed. The topic guide is available in Additional File1
Participating observations
The purpose of the participating observations was to gain insight into the implementation of DFI and to collect information regarding the context, such as the social and physical environment. Open notes were made during and after the observation sessions, and were used for probing contexts, mechanisms or outcomes during interviews and focus groups. Each observation lasted approximately 40–60 minutes and took place at the location of the DFI. In total, 11 participating observations of DFI were performed.
Documentation
Documents related to organisational policy papers, evaluations of and proposals for DFIs were studied. These documents, collected and delivered by the interview participants, aimed to contextualise and supplement data from interviews and observations. In total, 17 documents were collected.
Focus groups
In each case, interim findings from the interviews, observations and documents were presented as summary reports during a focus group meeting with the participants from interviews and observations, as well as other relevant participants who could not be interviewed individually. This was a basis for member checking, discussion to highlight unanswered questions and deepen the findings for theory development. Each focus group lasted approximately 60–90 minutes and took place at a location, chosen by the participants. In total, 17 professionals, 10 volunteers and 2 carers participated in the focus groups. Table 2 shows the overview of participants, specified per case and data collection method.
Table 2 Overview of Participants by Case and Data Collection Method
|
A.
|
B.
|
C.
|
D.
|
Total Interviews
|
7
|
9
|
12
|
12
|
Interviews professionals
|
5
|
5
|
9
|
10
|
Interviews volunteers
|
2
|
4
|
2
|
2
|
Interviews people with dementia or carers as co-developers
|
0
|
0
|
1
|
0
|
Total Participating observations
|
3
|
5
|
1
|
2
|
Activities involving participatory observation
|
Walking and working in the garden.
|
lecture and discussion in Alzheimer café, reading paper and playing board games in Odense housing
|
Presentation and discussion about Egyptian pharaohs
|
Participation during choir practice
|
Total Documents
|
4
|
5
|
5
|
3
|
Content of documentation
|
Proposals for government grants, policy papers
|
Evaluations, policy papers
|
Policy papers
|
Policy papers
|
Total Number of Participants in Focus Groups
|
4
|
13
|
7
|
4
|
Professionals
|
2
|
8
|
6
|
1
|
Volunteers
|
2
|
4
|
1
|
3
|
Carers as users of DFIs
|
0
|
1
|
0
|
1
|
Data extraction
Data extraction started per case after data from interviews, participating observations and documentation were collected. To ensure consistency and transparency, definitions of ‘contexts’, ‘mechanism resources’, ‘mechanism response’ and ‘outcomes’ were used (28–31). See Box 1 for definitions.
Box 1. Definitions for Data Extraction
Context pertains to the backdrop of an intervention (32). Context includes the pre-existing organisational structures; the cultural norms and history of the community; the nature and scope of pre-existing networks; and geographic location effects, such as previous experience with dementia-friendly initiatives (32).
Mechanisms are not interventions. They are the – often invisible – forces, powers, processes or interactions that lead to (or inhibit) change. They can be found in the choices, reasoning and decisions that people make as a result of the resources; the interactions between individuals or groups; and the powers and liabilities that things, people or institutions have as a result of their position in a group or society (33). Mechanisms are ‘triggered’ when (program) resources (e.g. information, money, expertise) interact with specific features of the context (individual, interpersonal, organizational, or institutional) (21).
Mechanisms resources refer to what is triggered among the context of participants/stakeholders (30, 32).
Mechanisms response refers to the responses of the participants, all that suggests a change in people’s minds and actions (30, 32).
Outcomes are the results of how people react to the mechanisms. Outcomes are either intended or unintended and can be proximal, intermediate and final (31, 34).
Outcomes can be labelled as:
• observed (the participant stated during interviews or observations that it had happened);
• anticipated (it had not happened yet but the participant expected it to); or
• implied (no explicit mention of the outcome was made but the data enabled the research team to infer, tentatively, that the participant had observed or anticipated it) (31).
Labelling levels of change in mechanisms:
• Individual changes include individuals’ skills and knowledge relating to dementia and DFIs, as well as the motivation, attitudes, commitment and values that affect individual behaviour (31).
• Interpersonal and network change refers to the relationships and networks between individuals and groups that influence development or sustainability of DFIs (31).
• Organisational change refers to the systems, policies and procedures, practices, culture and norms within an organisation that affect the access and resources needed to develop and sustain DFIs (31).
• Institutional change relates to the wider environment within which individuals, networks and organisations operate. This includes the political system, civil society and the media, political and economic factors, and broader social factors (culture, norms, collective beliefs) that influence the development and sustainability of DFIs (31).
|
Following the realist approach, the researchers (MT and AOB) constructed context-mechanism-outcome configurations (CMOc) from each interview transcript in order to examine what worked for whom, under what circumstances, and why and how (22). Interviews were thus analysed using CMOc, which were drafted and placed (by AOB/ MT) into a data extraction form in Excel, which is available in Additional File 2. Next, data from field notes and documents was extracted and supplemented to the CMOc. Corresponding CMOc were triangulated, noting the data source and corresponding fieldnotes or documentation. If new or rival data were found, they were extracted in new CMOc.
Labelling CMOC on outcomes and levels of change
In the last steps of data extraction, outcomes of each CMOc were labelled according to whether they were observed, anticipated or implied by the data source, following the reasoning that any insight into relevant contextual factors or mechanisms must relate to an outcome (31). Each CMOc was also labelled on levels of change as defined by Punton (31) to deepen the data-extraction on mechanisms, since change occurs at different social strata, consistent with the development of DFI (31, 32, 35). Thus, researchers (AOB/MT) labelled each CMOC according to the level on which change took place – the individual, interpersonal, organisational or institutional (31).
Data synthesis
Data synthesis used an inductive, sequential and iterative approach (24, 36). Sythesis of information within each case was followed by a cross-case synthesis and finally overall synthesis. All authors participated in each step.
Within-case synthesis
Synthesis started by identifying concrete outcomes (‘what were changes so that DFIs were developed and sustained?’) and then working backwards (‘how and why did these changes occur?’ and then ‘under what circumstances will these mechanisms lead to these changes?’) (context) (31). As such, similar outcomes, such as an improved perception of the importance of a DFI or taking initiative, were clustered per case. Second, commonalities of mechanisms and contexts were also clustered, such as mechanisms referring to feeling important or contextual aspects such as stakeholders from various organizations. Third, based on the clusters, patterns in outcomes, mechanisms and contexts were outlined. An example of such an outline was contextual factors, such as stakeholders having a personal affinity with people with dementia and carers, leading to mechanisms such as feelings of being important or connected, which then themselves led to outcomes such as taking initiative or purposefulness in collaboration. The levels on which change occurred remained distinct. These outlines were compared with corresponding configurations and quotes from data sources, to check for consistency and explanatory power (37).
Cross-case synthesis
Cross case synthesis started by following the same three steps as within-case synthesis, namely clustering of the outcomes, followed by clustering of commonalities of mechanisms and contexts, and finally developing outlines. This resulted in ten outlines, including intermediate outcomes in which mechanisms remained distinct on levels of change. The ten outlines are available in Additional File 3. These outlines, including the accompanying levels of change, were discussed with all authors.
Configuring middle-range program theories
Next, the ten outlines were configurated into initial middle-range program theories (MRPTs) by clustering outcomes and selecting the most prevalent mechanisms, including their contextual factors. This was carried out by three researchers independently (MT, WK and LD) and overlap and differences were discussed until consensus was reached. After this step, the initial MRPTs were presented to all authors and the advisory panel for their feedback. They discussed the MRPTs using their field expertise and confirmed those MRPTs’ usefulness and applicability. Suggestions for wording led to the final MRPTs.
Formulating realist program theories at a midrange level, such as middle-range program theories (MRPT), enabled both the specification of contexts, resources responses leading to outcomes and the conceptualization and explanation of those outcomes (38). Such MRPTs constituted more granular hypotheses about specific causal links and processes to communicate the findings as concretely as possible in practice compared to the more abstract middle-range theories (21).