Developing consensus-based recommendations for the delivery of dementia services for the LGBTQIA+ community in the Republic of Ireland

Design

The traditional Delphi consensus process (as described in Hsu & Sandford, 2007), which involves multiple iterations with highly trained and specialised Delphi participants, is not well suited for a population of people with dementia. The consensus process used here was adapted to ensure accessibility to people with dementia, older people and care partners (Morbey et al., 2019). The research included older LGTBQIA+ people with and without dementia through the research process (with guidance from Swarbrick et al., 2019). There was substantial member involvement throughout (see section on public and patient involvement; PPI). A six-phase process took place in order to reach consensus on recommendations and prioritised needs.

Procedure

Six phase consensus process:

Phase 1 – Development. The PPI advisory group for the research was created as a first step. A questionnaire was then adapted, from Fredriksen-Goldsen & Kim (2017), following a thorough review of literature. PPI members decided on the inclusion of items, length of questionnaire, format of questionnaire, rating scales and accessibility issues (to include language used). The research team and PPI group worked in close collaboration to ensure final questionnaire distributed was appropriate for both people with dementia and LGBTQIA+ community.

Phase 2 - National survey of older LGBTQIA+ people and care partners. The questionnaire, adapted and piloted in Phase 1, was distributed. Postal and online completion options were used, Participants also had the option of completing the questionnaire over phone/video call.

Phase 3 - Interviews with key stakeholders - Older LGTBQIA+ adults and care partners were interviewed using an online platform (but provided with the option of a telephone interview if preferred). Those who indicated an interest in doing so, through the online questionnaire, were invited to an individual interview. The aim of the interviews was to gain more in-depth information on needs that were not captured in the survey and to further discuss future care needs.

Phase 4 - International review of best practice. This involved a review of literature in the area, as well as policies and frameworks developed in other countries. As part of this phase, we also interviewed international experts working in dementia care for LGBTQIA+ community. The aim was to develop a picture of what best practice in the area looked like internationally.

Phase 5 - (Virtual) consensus meeting. The aim of the consensus meeting was to agree a set of needs and recommendations. The consensus meeting involved the PPI advisory group, those who took part in the individual interviews, and representatives from voluntary and healthcare backgrounds working with people with dementia. A purposeful sampling strategy was used to ensure a diverse group with varied experiences and backgrounds. Because of the virtual nature of the meeting and because participants may have been experiencing cognitive impairment, the number of participants included in the meeting was kept low (maximum 10).

The meeting used a modified nominal group technique to ensure participation of all members and was guided by similar research in the area (Brett et al., 2017; Keegan et al., 2021; Schneider et al., 2016). The nominal group technique was used because it reduces the burden on participants and results can be obtained quickly and presented back to the group (McMillan et al., 2016). This was deemed appropriate considering the additional scoping work (phases 2–4) that preceded the consensus meeting.

Results of phases 2, 3 and 4 were presented to the group, along with the needs and recommendations that came from the research. Sli.do, an online polling tool, was used to facilitate the adding, voting, and ranking of items. Because Sli.do automatically calculates the ranking, it was possible to work through all the stages of the meeting in real time (see stages below):

As recommended and used elsewhere (e.g. Keegan et al., 2021), the modified nominal group technique used consisted of four stages- 1) silent generation when participants had the opportunity to think about any additional items they wanted to add; 2) round robin where participants added those items they felt were missing- this was done anonymously; 3) clarification – participants were provided with the opportunity if they chose to discuss any new items or seek clarification; 4) private voting and ranking of items. There were two rounds of voting on both the “Needs” and “Recommendations” once the additional items were added. The first round of voting was to rank the importance of each item. From this the “Top 10” list of items were identified and participants then had the chance to rank the items in order of importance (in Phase 6).

Phase 6 – Final Member checking- Following the consensus meeting, the final results were distributed for comment and agreement. This also involved the final ranking of items – the “Top 10” lists were ranked in order of importance.

Public and patient involvement

This research was led throughout by PPI members. The research funding application was developed in conjunction with The Alzheimer Society of Ireland and a member of their Dementia Research Advisory Team. On commencement of the research, the PPI Advisory Group was set up. This group was recruited from the target population (but not considered to be research participants) and advised on all aspects of the research process. The PPI group were involved in the adaptation and development of the questionnaire, advising on recruitment strategies, and working with the wider group to decide on the ranked needs and recommendations that were brought forward from the consensus process.

Data collection tools

The original questionnaire, developed and administered by Fredriksen-Goldsen & Kim (2017), was adapted as appropriate to the Irish context. The survey was based on the National Health, Aging, and Sexuality/Gender Study (Fredriksen-Goldsen & Kim, 2017) and adapted with PPI group. Guides to survey design and implementation were also followed as described by Thayer-Hart and colleagues (2010) and Oppenheim (2000).

The questionnaire, which was hosted on QuestionPro, consisted of a number of different section- Demographics, Community and Service-Use, including health and dementia-care services; LGBTQIA+ identity, and Discrimination. A total of forty-six items were included in the questionnaire and both full and partially completed questionnaires were accepted.

The questionnaire went through several rounds of revisions after consulting with the PPI group. Before the final version of the questionnaire was sent for data collection, it was piloted with a small sample of people not included in the main study. The purpose of the pilot study was to test run the data collection method, identify potential issues, and examine its viability. A number of small edits and clarifications were made at this stage, but nothing that changed the overall content. A copy of the questionnaire can be found in the Extended data.

Online interviews and the consensus meeting were facilitated by using a secure Zoom account. The online interviews were audio-recorded. The consensus event was not recorded but notes were kept during the meeting (by MHO) and a record of decisions was kept.

Participants

Participants were eligible to participate if they:

People were ineligible to take part if they provided paid care for someone from the LGBTQIA+ community or were based outside of Ireland (exception for “international expert interviews”).

Participants were recruited through an email or social media post from gatekeepers at relevant organisations – e.g. The Alzheimer Society of Ireland, LGBT Ireland, TENI, Linc and other relevant local and national LGBTQIA+ organisations in Ireland. We used national LGBTQIA+ websites and magazines (Gay Community News), and radio and television interviews (TG4 and Radiό na Gaeltachta) to recruit participants. We also advertised through social media- Facebook and Twitter and paper versions of the questionnaire were available in a number of LGBTQIA+ community resource centres.

In order to consent to taking part in the survey, participants confirmed that they read the information sheet, and ticked boxes associated with the inclusion criteria to provide their consent. Participants did not need to reveal their name/signature in order to participate in the survey. Informed consent was obtained in writing for interview participants. The guidance on evaluation of capacity to consent from the British Psychological Society (2020) was followed. As such, obtaining consent was seen as a continuing process, not a one-off decision. It was unlikely that the person with dementia would lose capacity to consent during their involvement in the study, but their willingness to continue was checked regularly. Interview participants, with the exception of international experts, were invited to take part in the consensus meeting at recruitment stage and provided written consent to take part in the consensus meeting. Other consensus participants, representatives from voluntary and healthcare backgrounds, provided oral consent and written confirmation via email. The meeting was not recorded, and no personal information was collected during the consensus event- only a record of the scoring and ranking as a group was collected so oral consent was sufficient.

Sampling stopped after no more responses were recorded in the survey for a period of two weeks. Anyone who expressed an interest in the taking part in an interview was interviewed. Expert interview sampling was guided by principles of data adequacy (Levitt et al., 2017).

Analysis

Data from the questionnaires was analysed descriptively and via content analysis. The responses were exported into an Excel file and screened for errors and omissions to ensure data integrity. Descriptive statistics were calculated, which include totals (n), percentages, as well as the means and standard deviations. Open-ended text was analysed using content analysis.

Qualitative data from interviews was analysed using reflexive thematic analysis. This was an iterative, recursive process. All interviews were audio-recorded and transcribed. All transcriptions were deidentified during the transcription process and audio-recordings were deleted immediately after transcription. Data were analysed as described by Braun and Clarke’s (2021) six-phase data analysis process. This entailed three main tasks: familiarisation with data- the transcripts were read and re-read and initial codes were developed; coding and theme identification- ongoing formation and revision of themes; and the reviewing and refining of themes. This facilitated an inductive approach to identifying, analysing and reporting the themes identified within the data collected (Braun & Clarke, 2012). The first author was responsible for data analysis of both the survey and interview data, and verification of data integrity was conducted by the second author.

Consensus on a topic was decided if a certain percentage of the votes fell within a prescribed range (Miller, 2006). This range was set at 70% of consensus participants agreeing that an item was important. Each of the needs and recommendations were calculated and ranked as they were scored by participants – this was done by adding the total score for each item and dividing it by the number of overall votes (McMillan et al., 2016). This ranked score was used to identify the “Top 10” needs and recommendations and presented to participants for discussion. The second author was involved in the score calculation and the first author reviewed the scores and informed the consensus participants.

Credibility and trustworthiness of the data was ensured through a number of triangulation strategies. This included having multiple observer/observations/data collection methods, as well as including two data analysts. The researchers immersed themselves in the data to ensure rich descriptions. Working with PPI Advisory group increased the validity of findings and the likelihood of collecting data that was useful to the group under study.

Survey responses

Participant demographics. Forty-nine responses were recorded in the survey with a 46.94% completion rate. Participants were aged between 50 and 75 years old. 2.78% of survey participants were LGBTQIA+ people living with dementia, 13.89% were a care partner of an LGBTQIA+ person with dementia and 83.33% were LGBTQIA+ adults over the age of 50.

57.14% self-identified as female, 39.29% self-identified as male and 3.57% selected epicene. 82.14% of participants were cisgender, 14.29% were transgender and 3.5% were unsure.

40.7% of survey participants identified as gay, and 40.7% identified as lesbian. 3.7% identified as bisexual and 14.8% identified as queer¹.

Community participation. The data presented in Figure 1 indicates that there is a strong will and a need for socialization within the older LGBTQIA+ community. However, as 21.43% stated “As I grow older, I feel increasingly excluded from the community” and 8.57% stated that they have no contact with the LGBTQIA+ community, there is an indication that despite the general desire to be involved with the LGBTQIA+ community, older LGBTQIA+ adults between 50 and 75 years of age become more isolated from the LGBTQIA+ community.

Figure 1. Participant community support.

Participants were asked “Prior to COVID-19, were you involved in any of the following activities?” and respondents answered Yes to a multitude of activities which are seen in Table 1.

However, when asked about attendance of non-LGBTQIA+ specific services prior to COVID-19, fewer participants were in attendance. This is illustrated in Table 2. Only two activities were attended by participants in the majority of cases, these were LGBTQI+ groups and Cultural/arts groups.

When presented with a list of services and asked whether they would like LGBTQIA+-specific versions of those services, a large majority stated that they would like LGBTQIA+ specific versions of those services to be introduced, as seen in Table 3.

Many of those who selected ‘other’ made suggestions for other LGBTQIA+-specific services, including an LGBTQIA+ medical support group or forum, an LGBTQIA+ retirement group, LGBTQIA+ specific support for people in the Traveller community, LGBTQIA+ care homes and residential living arrangements, and LGBTQIA+ specific ageing brain support.

Healthcare access. The majority of participants found it easy to access health information except in the instances where that health information was LGBTQIA+ specific. For example, the majority of participants declared that it was easy to find information on health issues that concern them, such as screening or regular health treatments, understand what their doctor said to them, judge the quality of health information from different sources, and to get the information they need from their doctor. However, when asked whether they found it easy to find health information from general sources that address the needs of LGBTQIA+ people, the majority of participants stated that this was difficult or very difficult, as illustrated by Figure 2 below.

Figure 2. How easy is it to obtain LGBTQIA+ healthcare information?

Participants were then asked how relevant their LGBTQIA+ identities were in a healthcare context. As illustrated in Figure 3, most participants believed that being LGBTQIA+ is relevant in a healthcare context.

Figure 3. How relevant is LGBT identity to healthcare?

The ethos of a care home also had the potential to contribute to anticipated disrespect. As one survey participant stated:

Like in prior research on this population, we found that many LGBTQIA+ older adults felt it necessary to conceal their gender/ sexual identity when accessing care/healthcare or stopped expressing their gender or sexual identity. As one survey participant stated:

Additionally, another survey participant stated,

Discrimination. Participants reported that they have experienced multiple forms of abuse throughout their lifetime and within the past 5 years, as seen in Table 4.

An optional question surrounding the types of discrimination faced by participants was also included in order to gain a deeper understanding into where participants faced the most of their lifetime discrimination. As illustrated by Table 5, older LGBTQIA+ people have experienced discrimination in occupational, healthcare, and civil contexts.

Participants also reported some day-to-day discrimination that they have experienced. For example, 39.13% participants reported that people do things to humiliate and devalue them a few times per year. 39.13% also reported that people suggest that they are inferior to others a few times per year and 39.13% report that they are treated with less curtesy and respect than others, a few times per year. However, the majority of participants never receive poorer service in shops or restaurants, are never made to feel less intelligent than others and have never had someone threaten to ‘out’ them to someone who they did not wish to disclose their identity to.

When asked why they experience discrimination several reasons were given by participants. One participant cited the lack of hate crime legislation in Ireland as causing a lack of protection form discrimination. Some participants mentioned that they were new to the community they were living in, others cited ageism, and many simply cited the fact that they were a gender or sexual minority.

Microaggressions were also experienced to some degree by participants. For example, 50% reported that people were dismissive of their “alternative” family structures and stable relationship and 54.55% of participants experienced negative stereotypes, a few times per year. However, between 13.64% and 31.82% did not experience the microaggressions listed in the survey.

Participants also displayed strong resilience in adverse situations, with 54.55% agreeing with the statement “I tend to bounce back quickly after hard times.”. 30.43% of participants also agreed with the statement “I usually come through difficult times with little trouble.”, with 21.74% strongly agreeing.

Identity management. All of participants in the survey have disclosed their sexuality or gender identity at least once, and the majority have disclosed this this more than once. 42.86% agreed that they are open about their sexuality whenever it comes up, and 38.10% strongly agreed. 40.91% agreed when they were assumed to be heterosexual/cisgender that they would correct them and 22.73% strongly agreed.

Participants emphasized the need to understand and respect the variety of networks and life experiences of LGBTQIA+ people. Many people noted that they are often assumed to be heterosexual or cisgender until told otherwise which can cause discomfort and puts the responsibility of disclosing gender and sexuality on the service user. As a survey participant stated:

61.9% strongly disagreed with the statement “I make things up to hide my sexual orientation or gender identity”, further suggesting that older LGBTQIA+ adults are open about their sexuality. A majority of participants also indicated that they display objects in their homes to suggest their sexual orientation or gender identity, which may be relevant for care providers entering the home.

42.86% strongly disagreed with the statement “I feel uncomfortable dealing with health professionals and official organizations where my LGBTQI+ identity is known”. However, different experiences were expressed with regards to the statement “I have to work harder for my concerns to be heard and acted upon by health professionals where my LGBTQI+ identity is known.”- see Figure 4.

Figure 4. I have to work harder for my concerns to be heard and acted upon by health professionals where my LGBTQI+ identity is known.

Because of anticipated fears experienced by LGBTQIA+ older adults with regards to entering a care facility, it was suggested by numerous participants that a service that is LGBTQIA+ positive should display visual signs of acceptance, such as badges, flags, symbols, leaflets or pictures of same sex couples, however this should not be done without the adequate training of staff members.

Multiple participants also suggested the creation of an LGBTQIA+ accepting environment. LGBTQIA+ dedicated dementia services that reflect the heterogenous needs of the LGBTQIA+ community need to be created. For instance, one participant stated that:

Some participants did not like the idea of differentiated services and would prefer to be in a mixed setting. As a survey participant stated:

Another survey participant similarly stated:

The suggestion of having an advocate assigned to people was also discussed in the interviews. The findings of the interviews will be discussed now.

Interview data

Inductive thematic analysis was performed on the six interviews conducted with international experts and LGBTQIA+ people over the age of 50. Two main themes were derived from these data:

Identity suppression and anticipated concerns. Many participants reported that they anticipated some forms of disrespect, such as homophobia, transphobia, humiliation, rudeness or isolation, if they were to enter a nursing home or become dependent on formal care. A transgender interview participant cited prior negative healthcare experiences as influencing his concern that he may be mistreated and humiliated in a care context. He stated:

An international expert stated that she would “rather die than be cared for in this place” when she visited a care home with a very large crucifix on the wall.

This suggests that even if a particular religious run service is LGBTQIA+ inclusive, the religious ethos alone may deter LGBTQIA+ service users from making use of it.

Interestingly a transgender interview participant cited very positive surgical and person-centred care experiences in a German hospital run by a Lutheran charity and a desire to utilise other religious care facilities in the future. This indicates, that though research has shown that religious ethos can have a direct negative impact on LGBTQIA+ service users, that this is not always the case when the care provided is person-centred at its core. He shared that he had very positive surgical experiences in Germany, but very poor experiences in Ireland, which has caused him to worry about what will happen when he is older and unable to challenge mistreatment.

Hearing and reading about accounts of older LGBTQIA+ people in care and suppressing their identity in care due to fears of social exclusion, discrimination, abuse or even differential treatment, can also contribute to older LGBTQIA+ people anticipating their own identity suppression. As one interview participant stated:

High staff turnover was also considered a contributing factor to identity suppression and anticipated disrespect as it reduces consistency in care and can increase concerns over acceptance when trying to express one’s identity. An international expert participant discussed an instance in which she interviewed a lesbian woman in her 80’s who only first disclosed her sexuality to the manager in her nursing home, following her husband’s death:

Finally, the inability to conceal was reported to leave older LGBTQIA+ people with dementia vulnerable, particularly in potentially unfriendly environments, confirming findings from previous studies. This was described by an older participant from LGBTQIA+ community:

Additional note on transgender experiences and identity. Unlike with the experiences of sexual minorities, trans identity is often overfocused on in a healthcare context, which can both waste a service-user’s time when trying to focus on non-transgender related health issues and can feel uncomfortable and unnecessary.

With regards to transgender dementia care, an international expert noted that there are two opposing schools of thought about gender affirmation in dementia. One school of thought, which the participant was opposed to, was to rigidly affirm the gender as expressed by the transgender person, before their diagnosis with dementia. This was opposed by the participant as although she acknowledged that many transgender people have had to fight for the duration of their lives for gender recognition, and do not wish to lose this, in many reported cases, transgender people with dementia can experience gender dysphoria and different gender identities can become more salient at different times, which can be confusing and distressing. She believed that to address someone as their previously preferred gender identity, whilst they are presenting or feeling like another would be “well-meaning coercion, but coercion none the less”. Instead, this participant suggested the second school of thought, which is a more person-centred “take me as I am” approach, in which care providers address a person as the gender that they are most comfortable with in that moment.

Creating an LGBTQIA+ affirmative ethos and workforce. Creating an explicitly LGBTQIA+ affirmative workforce was seen as paramount in creating safer and more welcoming care environments for LGBTQIA+ people with dementia, and a number of participants suggested that in order to create this affirming workforce, a number of steps needed to be set in place, including specific recruitment techniques, training, the use of visual signs of acceptance and the creation of dedicated services.

One participant noted the visibility of same sex couples of all ages in a Canadian LGBTQIA+ health care center was very positive.

In order to ensure that all new members of staff are LGBTQIA+ positive, or at least are open to being trained in LGBTQIA+ affirmation in care, it was suggested that care services advertise an explicit pro-LGBTQIA+ ethos. As one interview participant stated:

Some participants suggested the need to diversify the workforce by hiring more LGBTQIA+ care workers, who would know how to socialise with other LGBTQIA+ people. Other participants, however, stated that the identity of the care provider was not so relevant as their dedication and level of training. Mandatory and comprehensive training was suggested by multiple participants, as people who were more biased towards LGBTQIA+ people would most likely skip the training if it were not mandatory. It was also suggested by participants that training should reflect the understanding of identities as opposed to just sexual behaviours, along with an understanding of the effects of attitudes and biases. For instance, an interview participant stated:

When the topic of dedicated services was discussed, participants had differing views on this. An interview participant stated that though he had never considered the idea of an LGBTQIA+ specific service, that he was very interested in the idea.

Some participants did not like the idea of differentiated services and would prefer to be in a mixed setting. An international expert also pointed out that in many cases different demographics within the LGBTQIA+ community often desire different things, for instance, gay men typically preferred the idea of LGBTQIA+ specific services, whereas lesbian women, particularly those who live separatist lives, preferred women-only services, which highlights the need for choice in care for LGBTQIA+ older adults, as well as services that recognise the different needs within the LGBTQIA+ community. In some instances, it may be necessary to take part in sensitivity work with service users to prevent transgressions against people with more marginalized identities.

There is a need for non-nuclear family structures and friend networks that are more relied upon by LGBTQIA+ people, to be respected and understood by professional dementia care providers, as in some instances, people in the LGBTQIA+ community experience alienation from their family of origin, decreasing their overall support system. This can be exacerbated if the LGBTQIA+ person has also migrated, particularly if leaving behind a country where LGBTQIA+ identities are more severely discriminated against. LGBTQIA+ people are also less likely to have children, decreasing their likeliness of having an intergenerational cohort. This is not always the case, as many LGBTQIA+ people do have children, adopt, or have younger relatives or friends, indicating that it is also important to avoid assumptions and ask questions about a person’s available network. It was suggested that an independent advocate should be triggered upon a dementia diagnosis, who could act in an older person’s best interests in cases where an individual’s social network was smaller, or in cases where unaccepting families of origin, or other potentially exploitative people are acting against the best interest of the service user.

Consensus process

Following the analysis of the interview and survey data, a consensus meeting was held with ten key stakeholders, who consisted of LGBTQIA+ people with dementia, LGBTQIA+ older adults, former caregivers of LGBTQIA+ older adults, and people who have worked with LGBTQIA+ older adults and/or people with dementia. The findings were presented in an oral presentation and the core needs and recommendations that came from the research were presented and clarified. There were ten core needs and sixteen recommendations derived from the data. Following the discussion, participants were asked to add any new needs or recommendations that they thought would be appropriate. These additional needs and recommendations were added to the full list. The complete unranked list of needs and the complete unranked list of recommendations can be seen in Table 6.

Following the addition of the new needs and recommendations by the stakeholders, the voting took place. After this vote participants were presented with the ten recommendations and needs that achieved the highest scores. Participants were then asked to rank the top ten recommendations- which recommendations should be prioritised and in which order. The final top ten need and recommendations, along with the associated rank score are presented below.

Top 10 needs identified

Top 10 recommendations

Limitations

This research had several limitations that require discussion. Firstly, the number of participants included in the research was small. This was anticipated by the research team and a number of steps were taken to ensure a consensus-based process- e.g the research contained multiple phases; the research was led by a representative PPI advisory group; the consensus meeting used purposive sampling to ensure representation across groups; and a number of recruitment avenues were used.

The questionnaire itself was lengthy and required a level of concentration that may have unintentionally excluded those with more severe dementia. Although the research included incomplete questionnaires (46.94% completion rate) and allowed for the questionnaire to be completed by or with a caregiver, there are likely people who were unable to take part because of the this. The challenges posed by the COVID-19 pandemic limited the possibility of in-person data collection. Although the research team placed paper versions of the questionnaires in LGBTQIA+ community centres, many older people were sheltering at home and not attending these locations.

The questionnaire did not capture the views of caregivers (only 13.89% of the total sample) and it is suggested that further research looks at this cohort separately, as we know that this group often has fears about the future that are coloured by their own experiences of caregiving (Price, 2011). We included caregivers in interview and consensus stages, but this is limited to a small number of caregivers.

As can be seen from the findings, all of the participants in the research have disclosed their gender or sexual identity to at least one other person. This indicates that there may be older people who have not disclosed to anyone at this point that have been excluded from the data.

Finally, the research team acknowledge that the term “older”, set for this research as 50+, will vary in terms of ethnicity and life expectancy due to health disparities. For instance, in 2016 only 3% of people from the Travelling community in Ireland were aged 65 or older and ageing in this community has been redefined as being aged 40+ as their life expectancy is 17% lower than the non-Travelling Irish community (Gibney et al., 2018). It may have been more prudent to set a lower age limit for people in the Travelling community to better capture their ageing experiences. Future researchers may also consider allowing participants to decide if they identify as “older” rather than having a cut-off for the research.

Underlying data

The data that support the findings of this study, including scores from the two rounds of voting, the questionnaire answers, and interview transcripts, are available on request from the corresponding author [S.M.H] with reasonable request. The interview transcripts of a participant who did not give consent for their transcripts to be shared will not be available. The dataset is not publicly available due to their containing information that could compromise the privacy of research participants. Due to the smaller sample of interview participants, and the specific and unique nature of some of the described life events of the participants that were paramount to analysis, de-identification is not sufficient to prevent possible recognition of the individuals.

Extended data

Open Science Framework: Dementia service needs and recommendations for LGBTQIA+ community. https://doi.org/10.17605/OSF.IO/P3UJE (Hynes, 2022).

This project contains the following extended data

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).