Benefits of Personalised Music in Dementia - a Feasibility Study

This research will adopt a mixed methods design for the study involving both quantitative and qualitative data collected and analysed simultaneously, then compared or related and interpreted.

Up to 10 participating care homes will be selected by Quantum Care Home Group in line with the inclusion/exclusion criteria. A total of 80 participating residents living with dementia, 80 informal carers and 20 care home staff will be selected from the participating care homes.

This study will explore:

1. The most effective and efficient methods for creating personalised playlists, including developing automated approaches, using technology (e.g. Apps and music recommendation algorithms).
2. The most easy-to-implement delivery methods for delivering personalised music to residents in care homes.
3. The most effective ways to assess responses to personalised music from people with dementia, including trialling automated analysis of changes in facial expression in video footage.
4. Initial impressions of whether factors such as timing of listening affect the observed impacts of listening to personalised music or whether listening prior to difficult care events appears to be beneficial (e.g. dressing, washing, toileting, sundowning, etc.).

A personalised playlist of about 100 songs will be created using responses to the music preferences questionnaire provided by the resident and their informal carer/Consultee/attorney (or paid carer). These questionnaires will be provided by the research team to the potential participants, identified by care home staff, together with the information pack and consent forms.

A music listening session will be arranged where the resident and their informal carer/Consultee/attorney will be played 30 seconds excerpts of each of the tracks (in 15 minute blocks over a period of up to 1 hour). The resident will be asked to wear a monitor on their wrist, similar to a watch, in order to monitor movement and pulse rate. The session will be filmed to enable analysis of emotional responses to music using machine learning analysis of changes in facial expression, as well as directly observed movements of hands, feet and/ or head. Where necessary the music listening session can be split into multiple shorter sessions to accommodate resident or care home requirements. Responses to the 30 second excerpts will be used to generate a refined personalised playlist for each participant, anticipated to be between 10 and 20 tracks.

The refined personalised playlists will be played to residents in music listening sessions throughout the duration of the study. How long music is played for and how frequently will, for the feasibility study, be determined by the resident, care home staff and/or informal carers. Different delivery methods for delivering personalised music will be tested including use of wireless headphones, smart speakers or other wireless speakers and their effectiveness and ease of use assessed through feedback from residents and focus groups and interviews with informal carers and care home staff. Duration and frequency of music listening sessions will, where possible, be logged through the listening history in the music app used and/or through data recorded by care home staff or informal carers.

A short training session will be provided in each participating care home to train staff and informal carers on delivering personalised music listening and collecting observational data (e.g. on timing of listening and observed effects on mood and behaviour). Fidelity, acceptability and appropriateness of the interventions will be assessed by the research team through the observation of selected personalised music listening sessions.

Approaches to assessing responses to the music that will be used include analysis of facial expression changes and movement in video footage, heart rate changes and qualitative information provided in reports from a member of the research team, residents, care home staff and/or informal carers. Impacts on well-being of residents over time will be assessed using scores derived from inventories for behavioural and psychological symptom assessment as a proxy, e.g. The Neuropsychiatric Inventory Nursing Home Version (NPI-NH), including for depression/apathy and anxiety/agitation. The Health-Related Quality of Life Measure for people living with dementia (DEMQOL) and DEMQOL-Proxy (administered with the informal carer/Consultee) will also be completed. It is expected that these tools will be used before the intervention and between 1 and 3 further times for each resident participant at appropriate intervals during the period of the study.

For this study the use of qualitative and quantitative assessment tools is for the purpose of developing and testing the appropriateness of personalised music listening and assessment methodologies rather than to draw conclusions about efficacy or otherwise of the intervention itself (which would be the subject of further studies subject to subsequent ethical approval). The quantitative data generated, therefore will only be used to supplement qualitative assessments to help infer/indicate factors for further investigation in subsequent feasibility studies and a longer-term larger scale intervention study i.e. to inform the design of those studies. Therefore, in this feasibility study, analysis of the quantitative measures is not intended to produce statistically significant data (the statistics involved will be mostly descriptive and, in some cases, inferential). Together with subsequent feasibility studies, it is intended to inform the design of the larger scale intervention study in which the sample size will be calculated to produce statistically significant results.

The study will trial the use of facial expression change through automated analysis of video footage. The initial pilot showed a strong correlation between the scores, indicating that, after further calibration, absolute thresholds can be derived to classify a track as either high or low response. Deep-learning facial expression change analysis will be developed to determine the feasibility or otherwise of using automated analysis of facial expression change in video footage. This will involve:

• Acquiring videos at 30 frames per second while music is played, extracting 1 image per second and generating 3 features for each.
• Performing face detection using Microsoft Azure Cognitive Services Application Programming Interface (API) and extracting an emotion score from the facial expression in 8 categories: anger, contempt, disgust, fear, happiness, neutral, sadness, and surprise. The research team will then form a secondary score based on a linear combination of these expressions and assign it to each frame.
• The research team will then aggregate the values of all frames acquired while a track was played to form a total response score by assigning different weights to the frames and deriving first-order statistics from the resulting distribution. The researchers can then compare the scores to subjective observation.

The views of care home staff and informal carers/Consultees/attorneys about the intervention will be sought throughout the study, via the means of focus group discussions and interviews. These focus group discussions and interviews will be conducted with care home staff and families in order to explore their views of how personalised playlists have an impact on residents including their impressions of the effect of variables such as timing of listening or the impact on difficult care situations. Direct feedback will also be sought from participating residents themselves, where feasible.

Informed consent for participation will be obtained, including for participation by informal carers/Consultees and care home staff, as well as residents. If anyone participating in the study, in any way indicates that they do not want to take part or in any way indicates that they are unhappy or distressed, then their participation will be paused (e.g. a different track played or a different question asked) and if the distress persists, the session will be stopped.

Informal carers/Consultees/attorneys will provide qualitative assessments of any benefits or disbenefits to them arising from the impact of personalised music listening on the behavioural and psychological symptoms of their relative/friend (e.g. improving relationship with carers and family). These assessments will be provided through participation in structured focus groups. Informal carers/Consultees/attorneys will participate in up to 3 focus groups, conducted by a member of the research team, experienced in qualitative research, once the refined personalised playlist is completed, between 2 and 4 months later and at the conclusion of their participation in the study.

Care home staff will provide qualitative assessments of any benefits or disbenefits to them arising from the impact of personalised music listening on the behavioural and psychological symptoms of the resident, for example:

• improving relationship with carers and family;
• improved quality or ease of care provision such as easing of distress caused by difficult care events.; and
• ease of incorporation of personalised music therapy into the usual working practices of the care homes.

These assessments will be provided through participation in structured focus groups. Care home staff will participate in up to 3 focus groups, conducted by a member of the research team, experienced in qualitative research, once the refined personalised playlist is completed, 2-4 months afterwards and at the conclusion of their participation in the study. If the care home staff are unable to participate in a focus group, a semi-structured 1:1 interview will be conducted instead.

As individuals with more severe dementia may lack the capacity to communicate their music preference or emotional responses to listening to music, this study will both:

• explore using the outcome measures as an objective way of gauging their preferences and responses by comparing responses measured through biomarkers and subjective assessments given by individuals living with less severe stages of dementia and their carers, family and/or friends; and
• work directly with people with advanced dementia who are not able to communicate their music preferences or emotional responses to listening to music, instead gathering such information from their carers, family and/or friends and measuring responses through biomarkers and subjective assessments given by these individuals.

The eventual larger scale study will explore issues of cost effectiveness, in addition to clinical effectiveness.

Pilot sub-study: Benefits of personalised music in dementia - a pilot study to investigate mechanisms of action

12 of the 80 participants (who have capacity to provide informed consent) will be further invited to take part in a pilot sub-study to investigate mechanisms of action.The capacity of residents to give consent will be assessed and documented by care home staff. The participants will be provided with an information sheet about the sub-study, consent forms and an MRI Safety Screen Questionnaire. This pilot study will help develop the mechanistic rationale for a larger scale intervention study to explore the effects of personalised music on behavioural and psychological symptoms in people with dementia.

In January 2021 we received approval to add a sub-study to this project, to use a Music for my Mind developed Web Application (the App), to invite relatives, friends or carers of a person living with dementia to use the App to create and deliver a personalised music playlist to their loved one and seek their feedback on the use of the technology and how it can be improved, and on their perceptions of how listening to the playlist has impacted on the well-being of the person living with dementia and, by extension, their own well-being.

This App has been developed as a lateral response to the Covid-19 epidemic (since the main study cannot currently be progressed, as it requires access to care homes). Our overall goal is to benefit people living with dementia and this sub-study aims to explore new ways of achieving that.

The participants in this sub-study will be the relatives, friends and/or carers of people living with dementia (whether the person living with dementia lives in a care home, their own home or with a relative). The planned sample size is 1000 participants. The participants will be recruited entirely through digital means through our team and supporter networks; through local radio and local branches of partner charities; through social media.