Concept Mapping to Define Components for an Occupation-Based Intervention for Parkinson’s Disease and Anxiety

Study design

Consensus methods, such as nominal group technique and the Delphi methodology, are commonly used to generate and/or validate the components of a health intervention [25]. However, these approaches struggle to engage diverse groups, particularly rarely heard groups [26]. In contrast, Group Concept Mapping (GCM):

Therefore, GCM was used to identify the components needed for a new occupation-based complex intervention. It involved qualitative and quantitative methods in a structured, sequential process to organize the ideas of a group about a topic— people with PD living well with anxiety— and involved brainstorming, sorting, and rating (qualitative processes) combined with multivariate statistical analysis (quantitative) [27]. These ideas are then visually represented using two-dimensional concept maps [27]. This versatile method can be implemented in paper form or using a secure online platform called ‘Groupwisdomtrademark’ [28], enabling participants to contribute at their convenience [26].

Ethical approval was secured from the University of Plymouth Faculty of Health Staff Research Ethics and Integrity Committee (Reference 2473).

Participants and recruitment

Sample size in GCM studies varies depending on the question and data collection method [29]. The recruitment target of n = 40 was based on previous studies that have reported attrition of 10% between the GCM stages and reaching data saturation in stage 5 [29–31]. The sample was divided into three groups— people with PD, care partners, and occupational therapists (13–14 participants per group)— to avoid potential power differences between group participants. Occupational therapists were included based on feedback from patient and public involvement (PPI) stakeholders who contributed to the study design and felt that it was important to include therapists’ perspectives. Twelve weeks were allocated for recruitment.

Purposive sampling was used and was targeted to ensure a diverse range of respondents— recruitment materials targeted age (for all groups) and diagnosis duration (for people with PD). Inclusion criteria for our target sample were people with PD with a formal diagnosis of PD with self-reported PD-related anxiety, any care partner of a person with PD-related anxiety, and occupational therapists working with people with PD or with expertise in working with people with anxiety. Formal assessment of anxiety status was not used to avoid missing rich experiential data.

People with PD and care partners were recruited via the Parkinson’s UK Research Support Network and the Peninsula Parkinson’s Research Interest Group (PenPRIG). Occupational therapists were recruited via the Royal College of Occupational Therapists specialist sections for neurological practice (SSNP) and mental health (SSMH). A member of each organization acted as a gatekeeper and sent out email advertisements to their distribution lists. The gatekeeper protected the autonomy of prospective participants and allowed them to decide freely whether they wished to participate. The email invitations included a copy of the participant information sheet and the lead researcher’s (CL) contact details. Interested participants contacted the lead researcher, who answered any questions presented and then sent a weblink to ‘Groupwisdomtrademark’, which required participants to complete an informed consent form.

Participants were asked to engage in three out of five stages of the GCM: brainstorming, sorting, and rating (see below). Each stage was open for two to four weeks; an invitation and two email reminders were sent to participants at each stage.

Data collection

There were five specific stages: brainstorming ideas in response to the focus prompt, using open text (stage 1), idea synthesis by the research team (stage 2), participants sorting the brainstormed statements into groups that made sense to them (stage 3), participants rating each statement on its importance and feasibility (stage 4), and data analysis (stage 5) [27]. Participants could access ‘Groupwisdomtrademark’ at a convenient time and place during each stage and contribute as much data as they wished; no data limiters were used. Participants initially provided demographic information, including their role in the research (i.e., if they were a person with PD, care partner, or occupational therapist), gender, age, and how long they had been diagnosed with PD (if appropriate).

Stage 1: Brainstorming

Participants were asked to generate statements based on their thoughts and experiences in response to the focus prompt “one thing that would help a person with Parkinson’s to live well with anxiety might be . . .  ”. There was a two-week window for participants to complete the brainstorming stage.

Stage 2: Idea Synthesis

The list of the participants’ raw statements was exported to a spreadsheet, interrogated, and edited for grammar and readability without changing the wording. Statements expressing multiple ideas were split into single-idea statements; these were the only edits performed. All statements were assigned a keyword generated from the participants’ language to express their meaning. Statements with similar keywords were scrutinized together, and duplicates were removed. The result was a final set of refined, unique statements for use in the later stages [27].

Stage 3: Sorting Activity

Each statement in the final set was assigned a number and listed in random order on the ‘Groupwisdom^trademark’ software. Each participant was asked to sort statements into groups that were meaningful to them by dragging and dropping them onto a blank workspace in the software. Participants were also asked to give each group a relevant name. For the data to be approved for further analysis, participants must have sorted≥75% of the statements into meaningful themes [27].

Stage 4: Rating Activity

Stage 4 ran simultaneously with stage 3 to reduce participant time burden. They were asked to rate each idea on a five-point ordinal scale for importance and feasibility to facilitate the identification of priority items. Importance ranged from 1 (‘not at all important’) to 5 (‘very important’), and feasibility from 1 (‘not realistic at all’) to 5 (‘very realistic’). Participants were asked to rate each statement with all the others to ensure that the full range of the scale was used. Similar rating questions have been used in other GCM studies [32]. Both activities were closed after three weeks.

Stage 5: Analysis

Qualitative and quantitative techniques were used to analyze the data. The ‘Groupwisdom^TM’ software combined the sorted data into a similarity matrix, showing each statement as a position on an X-Y axis indicating how participants sorted statements. The similarity matrix produced a two-dimensional point map using a multidimensional scaling process. The closer points represented statements that were more often grouped together because the participants thought they were conceptually similar [27]. The software generated a stress value that indicated the degree to which the distances between each point on the map were representative of the sorting data. Lower stress values suggest greater stability and goodness-of-fit. Stress values below 0.365 are desirable for GCM research studies [27, 31].

Hierarchical cluster analysis was applied to the data using Ward’s method to group statements with similar meanings into clusters representing how participants grouped their ideas [33]. This statistical technique was performed in the ‘Groupwisdom^TM’ software and involves drawing boundaries onto the point map to produce a cluster map. Clusters are successively merged, resulting in several viable cluster maps. A final cluster ‘solution’ is then selected through qualitative interpretative analysis [27]. In this research study, the first author (CL) consulted with the rest of the research team (KB, JM, IS) to evaluate maps of six to nine clusters to determine the final cluster solution. This process involved checking whether the statements within each cluster represented the overall theme and comparing the cluster maps to each other to check for similarities and discrepancies between cluster statements and titles.

Cluster labels were suggested by the software for the final cluster solution based on the verbatim statements made by the participants during the sorting activity. The research team refined the cluster labels for grammar and readability, but the wording was unchanged.

The rated data were then organized into a pattern match as a visual representation of the data. This allowed comparison of each cluster’s mean importance and feasibility [27] and helped highlight priority themes for future intervention development. Individual go-zone graphs were created for each cluster, indicating which statements within the cluster were the most important and feasible. Go-zones are bivariate graphs divided into four quadrants by the mean scores generated for importance and feasibility [27]. The upper right quadrant contained the statements that scored above the mean for importance and feasibility. Statements in the ‘go-zone’ were considered priority statements for intervention development (Supplementary Table 1). These visual representations of the data are intended to aid in the appraisal and subsequent use of the results in a future intervention development study.

Rigor and validity

The participatory approach of the GCM methodology promoted both the credibility and real-world applicability of the findings by ensuring that the results directly represented the participants’ experiences and ideas. A senior researcher (KB) peer-reviewed the cluster map to ensure that the headings represented the participants’ language.

Study population

The final sample consisted of 83 who participated in at least one activity and 64 who participated in more than one activity. The final sample was predominantly female. The recruitment target for people with PD was exceeded (n = 72). The care partner (n = 6) and occupational therapist (n = 5) groups were under-recruited. There was attrition between the study phases (Fig. 1) with 38% (n = 27) of people with PD, 33% (n = 2) of care partners, and 80% (n = 4) participating in the final phase.

Fig. 1

Line graph summarizing participation across each GCM activity.

Figure 1 shows how many participants from each sample group took part in each activity. During brainstorming, participants provided 198 ideas in response to the focus prompt, which were synthesized into a list of 119 unique statements (Supplementary Table 1). One participant’s sorting data was rejected because less than 75% of statements were sorted. Forty-four data sets from the sorting activity were approved for analysis. From the rating activity, one participant’s feasibility rating data set was rejected due to rating less than 75% of the statements, with 33 approved for analysis.

Because the participants used the ‘Groupwisdom^TM’ software at different times for the activities open for specified periods, some participants skipped, did not have the opportunity, or chose not to answer the demographic questions (n = 12). Participant characteristics of respondents to demographic questions are shown in Table 1.

Concept map

Multidimensional scaling of the data from all participant groups resulted in a point map with a stress value of 0.252. This score indicates the goodness of fit of the sorting data, i.e., the sample data are likely to represent the data that would be expected in the actual population [34]. After the hierarchical cluster analysis, several cluster maps were generated and reviewed by the research team and a patient and public involvement (PPI) representative. Consensus was reached on an eight-cluster solution with the following cluster names taken directly from the participants’ contributions: Access to information, Professional help, Peers and groups, Support from others, Coping, Self-help, Lifestyle changes, and Exercise. The cluster map is shown in Fig. 2. A point on the map represents each numbered statement (see Supplementary Table 1).

Fig. 2

Cluster map showing eight priority themes for intervention development. The map shows the distribution of each statement as a numbered point.

Pattern match

A pattern match compares the eight clusters’ mean rating scores for importance and feasibility (Fig. 3). The clusters are ranked along the vertical lines representing importance (left) and feasibility (right). The pattern match shows that three clusters were rated higher for importance than feasibility (Professional Help, Peers and groups, Support from others). The remaining clusters were rated higher for feasibility than importance (Self-help, Lifestyle changes, Access to information, Coping, Exercise). The overall correlation between the two rating patterns shows no to negligible relationship between importance and feasibility ratings across clusters (r = –0.07).

Fig. 3

Pattern match comparing the mean importance and feasibility rating scores at the cluster level (upper limit = 5). Higher mean scores are closer to the pattern match’s top (r = –0.07). Importance (n = 46); feasibility (n = 33).

Go-zone graphs and statements

Go-zone graphs that were generated for each cluster are presented in Fig. 4. Statements in the go-zones are identified as actionable priorities. Those with high importance and high feasibility were of particular interest, as they are considered priority themes for intervention development. Those with high importance but slightly lower feasibility are areas for future intervention development but are not immediate priorities.

Fig. 4

Go-zone plots for each of the eight clusters. Statements are shown as numbered points (see Supplementary Table 1). The upper right quadrant of each plot represents statements considered priorities for intervention development. The upper left quadrant represents areas for future development. Importance (n = 46); feasibility (n = 33).

The eight clusters represent a common understanding among participants of the key components for future intervention development.

This cluster was rated as the most important (Fig. 3). There are 18 statements in this cluster. A common theme within this cluster is increasing clinicians’ condition-specific knowledge, such as:

Participants felt it was important and feasible for clinicians to understand anxiety in PD, which is consistent with previous research on clinical expertise in care teams and observational data [35].

This cluster was ranked second for importance but last for feasibility (Fig. 3) and consists of 11 statements. Access to people and groups with experience of PD-related anxiety was a common theme:

While superficially like the previous cluster, the 11 statements in this cluster highlight the specific qualities that people with PD would like to see people, in their social context, exhibit to support them in their anxiety:

The participants identified this cluster as important and feasible (Fig. 3), which suggests that educating people in their social context about people with PD living with anxiety should be a core component of the intervention.

The 17 statements in this cluster relate to ways in which people with PD can help themselves to manage their anxiety so that they can live well with anxiety and PD. This cluster was rated fourth in importance and first in feasibility (Fig. 3), suggesting that the intervention should include a substantial self-help component in order to:

Relaxation techniques that are easy and quick to implement were identified in the go-zone. These findings suggest that relaxation techniques commonly used for anxiety (e.g., box breathing exercises) should be central to the new intervention [36].

The concept of coping as a strategy for reducing the unpleasant feelings associated with anxiety is embodied in this cluster. The 16 statements in this cluster focused on coping with the emotional impact of anxiety, for example:

Addressing these pressures may provide a viable method of coping with anxiety for people with PD. Several statements in this cluster (statements 45, 108, 111) highlight learning to accept living with a long-term incurable illness as an important and feasible component of the proposed intervention (Fig. 3).

Statements in this cluster relate to people with PD not only having access to information about their condition when they need it but also being educated about the condition that they are living with, for example:

Statements in this cluster suggest the provision of written information as important and feasible (Fig. 4).

This cluster ranked seventh in importance and second in feasibility (Fig. 3) and contains 21 statements. The importance of engaging in meaningful activities coupled with lifestyle adjustments (such as improving diet and increasing physical activity) to manage anxiety is woven throughout the cluster, for example:

The 10 statements in this cluster relate to physical exercise such as dancing (statement 6), yoga (statement 113), and tai chi (statement 8). However, exercise was rated the least important and sixth most feasible for an intervention to support people with PD with anxiety (Fig. 3).

Strengths and limitations

A strength of the GCM methodology is its structured and systematic procedure for data collection that places the voice of the participants, most of whom were people with PD with anxiety, at the center. The eight clusters portray what the participants feel is important to include in a new intervention to help people with PD live well with anxiety. In addition, the large number of statements suggests a range of thoughts among the participants, not just those of dominant group members. Online delivery using the ‘Groupwisdom^TM’ platform allowed participants to contribute to the study at their convenience, thus facilitating participation.

This study is not without limitations. The sample size exceeded our target (n = 40) and was heavily weighted towards people with PD (n = 72). Attrition occurred despite the research team’s efforts to maintain study participation, such as automated email and text message reminders when requested by participants. This likely resulted in additional attrition bias that skewed the results and compromised the study’s external validity.

Anxiety tends to occur earlier in PD and may be an early symptom of neurodegeneration [68]. This is reflected in the average disease duration of the participants (median 5 years). Evidence suggests that people living with young-onset PD (YOPD, diagnosed before 50 years) are more likely to experience anxiety than people with a late-onset of the disease [69]. It would be reasonable to expect this to be reflected in the characteristics of the participants. However, we can only be confident that 1% of participants with PD meet the YOPD criterion (Table 1). The next age group (45–64) accounted for 39% of the sample. As these age groups are based on the UK census, we cannot determine how many participants would be considered YOPD. Although our sample would still represent those aged 65 and above (60%), it would be helpful to use adjusted age groups in future research to distinguish between early and late-onset PD.

Small numbers of care partners and occupational therapists participated compared to people with PD. Thus, we cannot be confident that our findings represent all the groups included in this GCM study as they are likely weighted towards the views and experiences of people with PD. The under-recruitment in both care partner and occupational therapist groups may partly be attributed to the COVID-19 pandemic, which has had a noted impact on non-COVID-19 research due to increased service demands [70, 71]. Our sample was predominantly female, including for the group of people with PD. This finding was unexpected, particularly as PD tends to be more prevalent in men and men are over-represented in most research [72]. Resultantly, our findings may not fully represent the views of male people with PD, care partners, or occupational therapists. Future research will likely need to change its recruitment strategy and sampling frame to address this imbalance.

Moreover, GCM does not claim to be a definitive methodology, so studies using different methods may be needed. Although accounted for in the design of this study and the GCM methodology, the loss of participants between stages means that potential data was lost. Finally, we only collected basic demographic data to identify gender, age, and diagnosis duration (if applicable). As we lack data on areas such as cultural background and socioeconomic status, we cannot be confident any one group was not over-represented, limiting the external validity of this study to the UK population, including those from seldom-heard groups.

While digital delivery provided a convenient means of conducting this research, the impact on potential participant engagement in this study cannot be ignored, as it is likely that those who are less digitally literate will not have participated [73].

Future research directions

The findings highlight eight priority components, presented as clusters, for consideration in the future intervention. Individual ideas from within the clusters rated highly on importance and feasibility are suggested as starting points for future intervention development. Clusters, such as “Coping” and “Support from others”, highlight broader individual and social factors that support participation in meaningful roles and activities and need to be considered in intervention development. The ability to access timely support from peers and the development of individual coping strategies will be important considerations for the next stage of intervention development [17]. Future research, such as co-producing the intervention with stakeholders, is needed to integrate the components identified in this study into a new intervention to help people with PD to live well with anxiety. Future studies should seek to remove barriers to research participation for care partners and occupational therapists, as providing convenient data collection is unlikely to be sufficient.

Conclusion

This novel participatory study used an online GCM methodology to ask people with PD, care partners, and occupational therapists to highlight priority components that provide starting points for the future development of an occupation- and community-based intervention for people with PD-related anxiety. The eight clusters identified in the findings are exercise, lifestyle changes, self-help, coping, access to information, professional help, peers and groups, and support from others. Statements in each cluster rated highly on importance and feasibility will be prioritized in intervention development. We plan to use these findings to co-produce an occupation-based intervention for living well with anxiety in PD.