Improving the depth of data quality or increasing confusion? Reflections on a data analysis involving members of a self‐help group for relatives of people living with dementia

Abstract Background Public involvement in research to improve data quality and to empower different stakeholders is good scientific practice, but rarely implemented across all research phases. Objective This article reports on an attempt to involve members of a self‐help group for relatives of people living with dementia as co‐researchers in the data analysis in a short‐term format. Methods One researcher identified statements about assistive technologies from 17 interviews with people living with dementia and informal caregivers. Two researchers and six co‐researchers independently assigned pre‐defined values to these statements. Subsequently, we compared the values of the researchers and co‐researchers. Results The members of the self‐help group identified four original values not considered by the researchers: consent, inclusion, participation and respect. Discussion The involvement of co‐researchers led to an improvement in the depth of data quality through the joint identification of values concerning assistive technology. Language barriers between researchers, co‐researchers and interview participants impeded the data analysis. Conclusion The challenges and benefits of a participatory data analysis shown here can provide a basis for recommendations for target group‐specific research involvement. Our recommendations relate to the recruitment of co‐researchers, requirements for conducting a participatory data analysis and the participation degree of people involved. Patient or Public Contribution The group of co‐researchers participating in the data analysis consisted of relatives of people living with dementia.


| BACKG ROU N D
Active public involvement in research including lay people as coresearchers 1 is an integral part of international scientific efforts. 2 Patient and public involvement aims to broaden perspectives of all parties involved by incorporating experience-based knowledge into theory-driven research. [3][4][5] For successful research involvement, the continuous participation of such lay co-researchers in all research phases is recommended. 2,[5][6][7] While public actors are increasingly engaged in the design phase of research, data collection and dissemination of scientific findings, 8 the public is rarely involved in data analysis. [8][9][10][11] However, participatory data analysis can increase the depth of data quality. 11 Public involvement in the data analysis phase of research seems to be particularly difficult compared to other phases 10 since prior methodological knowledge is crucial to analyse data. 12 An additional barrier in participatory data analysis is that the data often consist of large amounts of complex material 10 so that the analysis requires a lot of time 10,13 and concentration of the co-researchers in the analysis. 10 Not all groups that can make a valuable contribution to data analysis have the necessary time and resources to participate in the research phase.
The prerequisites for a participatory data analysis are challenging for any lay person, but are particularly challenging when inviting lay people with cognitive disabilities such as dementia patients and their caregivers with limited time resources to participate. In dementia research, expanding the understanding of researchers through patient and public involvement is essential to find care solutions tailored to the needs of people living with dementia as care recipients and their relatives as caregivers. Evidence from projects involving people living with dementia 12,14-16 and their caregivers 17 highlights the feasibility of participatory data analysis with these key stakeholders in dementia care. However, there is still a need to extend the existing evidence and resources in terms of structural requirements for patient and public involvement and appropriate recruitment strategies. 17 The focus of our work was on caregivers of people living with dementia who have a key role in care 18 and are confronted with specific caregiver burden. 19,20 Caregivers' specific perspective may broaden academic researchers' understanding of dementia. At the same time, caring relatives' constant obligations to the person living with dementia limit their availability for participation as co-researchers. In conclusion, there is a need for the involvement of caregivers while considering their specific resources. Here, we aimed to conduct a participatory data analysis with relatives of people living with dementia. We conducted a one-time, short-term data analysis session, accommodating the limited resources of the target group. We invited members of a self-help group for relatives of people living with dementia as our co-researchers in this study.
Our research question was whether the inclusion of informal caregivers of people living with dementia improves the depth of data quality by enriching the data analysis with additional results or rather complicates the analytical process, leading to confusion. The challenges and opportunities of participatory data analysis are highlighted to provide recommendations.

| THE PROJEC T SAMI
The participatory data analysis was carried out within the ongoing study 'Sensor-based individualized activity management system for people with dementia in nursing facilities' (SAMi) with focus on assistive technology. The study aims at the development of a sensorbased assistance system to support people living with dementia and caregivers in inpatient care. For this purpose, sensor data of people living with dementia were collected with regard to their motion, location and the living context. Furthermore, relevant stakeholders within the dementia care sector were asked about their expectations for a support system and design suggestions. We identified people living with dementia and their caregivers as key stakeholders as people living with dementia will be the direct users of an assistive technology and caregivers often support the person to be cared for in the usage.
Assistive technology can positively impact the life of people living with dementia [21][22][23][24] and reduce caregiver burden, 23,25 but the willingness to use technological innovations is relatively low in older people. 24,26 According to the Technology Acceptance Model (TAM), the willingness to use technological innovations depends on the perceived usefulness and the perceived ease of use. 27 Early and continuous stakeholder engagement in research to gain knowledge about the needs of future user groups has the potential to increase the acceptance of new technologies by prospective users. 21,23 In order to develop recommendations for assistive technology, namely a need-based indoor navigation aid, semi-structured interviews were conducted with people living with dementia (n = 10) and informal caregivers (n = 10). Data were collected by one researcher (AK) from October 2019 to February 2020. The project and the results are described in detail in another article. We followed the Value Sensitive Design approach, which includes the identification of values of future users for the design of assistive technology. 28,29 A value is defined as something that has a meaning in the lives of individuals or groups. 28,29 In terms of assistive technology, values express certain conditions that technology must meet in order to be used. 28 The technology design approach should be strengthened by a participatory data analysis.

| ME THODS
The participatory data analysis was part of a multi-stage process consisting of data collection by one researcher (AK), a data analysis by two researchers (AK, SK) and a participatory data analysis with the co-researchers. The methodological approach used here is shown in Figure 1

| Data collection
The data collection was done by one researcher (AK). Interview participants were recruited from the memory clinic of the Rostock University Medical Center and the geriatric ward of a regional hospital. Seventeen interviews with people living with dementia (n = 10) and informal caregivers (n = 10) were conducted. Three interviews were dyadic interviews at the request of people living with dementia. The interviews took place in the rooms of the memory clinic, in the rooms of the hospital and in the private apartments of the interviewees. The semi-structured questionnaire for the interviews included questions on the current use of assistive technology, barriers to use them and expectations of user-friendly technologies.
The interviews were recorded on tape and written down pseudonymized as smooth verbatim transcripts.

| Data analysis
Firstly, one researcher (AK) identified the interview statements related to assistive technology in the transcripts and formed a subset of data only including these statements to reduce the large amount of data to a manageable amount similar to a previous study. 15  Secondly, the researcher who conducted the interviews (AK) and another researcher (SK) independently attributed values to the statements related to assistive technology. Data analysis was carried out by two researchers in order to obtain higher reliability. We decided to use an a priori defined list of values (Table 1) to increase the comparability of results between the researchers and co-researchers as well as to prioritize previously collected values to achieve inter-rater reliability. 11 The values were derived by a consensus process between interdisciplinary academic researchers and a literature review related to the core topics selected by the interdisciplinary research group 30 and can be found in Table 1 Thirdly, our co-researchers conducted the data analysis by assigning values to the statements. There was one member of the self-help group who was also an interview participant; otherwise, there was no overlap between the interview participant group and the co-researchers. Here, we followed the recommendation from a previous study 13 to involve co-researchers in data analysis who did not participate in the data collection. Access to the self-help group was provided by directly addressing the group leader as gatekeeper.
At the previous group meeting, the group leader had reported on the project and asked the participants for their agreement to participate.
The group meets once a month for two hours.
At the beginning of the meeting of the self-help group, the researcher (AK) acts as moderator of the analysis session and explained the purpose of the study and the method of participatory data analysis. There was no additional training for the participants to keep the effort for the co-researchers low and the duration of the participatory data analysis short. The participants received F I G U R E 1 Methods used for data analysis with first researcher (AK), second researcher (SK) and co-researchers, abbreviations: nnumber, created with Power-user

| Ethics
Written informed consent was obtained from all participants of the participatory data analysis. For the procedure, an ethical approval was obtained from the Ethics Committee of the Rostock University Medical Center (A 2020-0064) on 20 March 2020.

| RE SULTS
A total of six relatives of people living with dementia from one self-  Table 2.
Through the research involvement of the members of the selfhelp group, an improvement in depth of data quality by four values was achieved compared to the values identified by the researchers (see Table 3). The values only found by the members of the self-help group were consent, inclusion, participation and respect.
The two researchers hardly differed in the values that they most often attributed to the statements: relief/respite, safety and emergency help. The co-researchers frequently assigned the value of 'independence', which was identified by the researchers much less often. 'Quality of life' as the second most frequently identified value among the members of the self-help group was only found once by the first researcher and not at all by the second researcher.
'Emergency help' was also the third most frequently identified value by the members of the self-help group.

| D ISCUSS I ON
We conducted a participatory analysis of qualitative data with members of a self-help group for relatives of people living with dementia. Our research approach allowed us to invite six relatives of people living with dementia as a key stakeholder group in dementia care and involve the caregivers as co-researchers while accommodating their limited time resources. This participatory data analysis enabled an improvement in the depth of data quality by four values.  17 The values only found by our co-researchers were consent, inclusion, participation and respect.

TA B L E 3
Values identified by the researchers and the members of the self-help group for relatives of people living with dementia with the number (n) of times they have been identified in the interview statements it seems advisable to give co-researchers the opportunity to add values and discuss them afterwards.
The self-help group leader's encouragement of the discussion promoted an informal atmosphere. This underlines the previously established importance of a mediator between those involved in research. 34,35 We expected that caring relatives, as experts on the living en-

| Participation degree
As criticized in previous studies, 12,38 the decision-making power in our study remains with the academic researchers. Academic researchers selected the statements provided to the co-researchers, selected the values, moderated the analysis session and led the discussion. Future investigations should aim on strategies to include co-researchers in a greater scope with appropriate effort for the participants.
All in all, we could only achieve a participation degree of consultation in which the power in the research process remains with the researchers. 39

| Key learning points
• Participatory data analysis with informal caregivers of people living with dementia enables greater depth of the analysis and improves data quality.
• Participatory data analysis was kept short due to the limited time resources of the co-researchers. However, it became apparent that several sessions would have been necessary to prepare the co-researchers adequately for the analysis and to evaluate the results together afterwards.
• The level of participation must be based on the preferences and abilities of the co-researchers. In line with the approach of person-centredness, co-researchers should determine their level of decision-making power in projects.

| Limitations
We did not discuss the values prior to the joint data analysis with the co-researchers. Questions about the meaning of a value from coresearchers could be answered individually, but it remains unclear whether all co-researchers understood all values. Furthermore, coresearchers were only confronted with a small set of data. The data basis for the given recommendations is still limited and needs to be expanded by future research.
A member of the self-help group was also an interviewee. It should be noted that self-help groups do not serve the purpose of research and that members should act as co-researchers in a setting different to group meetings when long-term involvement in research is planned. A self-help group may be understood as an opportunity to access the target group but does not replace a research advisory board.

| CON CLUS ION
In summary, there has been an improvement in the depth of data quality through participatory data analysis. The challenges of implementing the participatory approach inform the design of public participation in the analysis research phase. We derived recommendations for the recruitment of co-researchers, time requirements for conducting a participatory data analysis, the need for a mediator and the necessity of the determination of the participation degree by the co-researchers.

ACK N OWLED G M ENTS
The authors thank all co-researchers. Special thank is due to Sabine Schwarz for her support with the recruitment of the co-researchers and Anna GM Temp for her support in English editing.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available on request due to privacy/ethical restrictions.