Using Music to Develop a Multisensory Communicative Environment for People With Late-Stage Dementia

Amy Clare, DClinPsych; Paul M. Camic, PhD; Sebastian J. Crutch, PhD; Julian West, BA; Emma Harding, Bsc; Emilie Brotherhood, Msc


Gerontologist. 2020;60(6):1115-1125. 

In This Article

Research Design and Methods

This study employed a qualitative, longitudinal design influenced by classic grounded theory (CGT) (Glaser & Strauss, 1967) and video analysis as established by Griffiths (2013). According to Griffiths and Smith (2016), the use of visual CGT provides a way of eliciting meaning from people who express themselves primarily using nonverbal means as it allows for careful observation of subtle or nuanced communication. Classic grounded theory was used over other qualitative methods as it supports the development of a theoretical understanding grounded in social experiences (Urquhart, 2013).


Participants included eight adults with advanced dementia who lived in the same residential care home (Table 1). Diagnosis was made by National Health Service physicians prior to admittance to the care home. Level of dementia impairment was determined by the clinical dementia rating scale (Morris, 1997). This scale measures impairment from 0 (none) to 3 (severe) across six areas (memory, orientation, judgment/problem solving, community affairs, home and hobbies, and personal care). Additional participants were staff members from the care home (three to five per session) and professional musicians (three each session) trained in the Music for Life intervention.


Ethical Considerations and Consent. An ethics panel at Canterbury Christ Church University approved the study (ID: 18–075). Participants (given pseudonyms) were in the advanced stages of dementia and lacked the capacity to give consent; their legal guardians provided this on their behalves in line with Mental Capacity Act recommendations (MCA; Department of Health, 2005). No family guardian declined to give consent. The British Psychological Society's Code of Ethics (2009) was followed throughout the research.


The intervention involved 8-hr-long live music sessions adhering to a similar format each week: an opening piece based on a musical theme involving harp, flute, and oboe, followed by the "welcome song," which was same piece each week led by one of the musicians singing "welcome" to each resident and naming them individually. Improvised group and one-on-one music then followed and continued throughout the session. An ending piece based on the same musical material as the opening piece, completed each session. Musicians and participants played a range of instruments throughout the intervention (Table 2).

Data Collection. Video data consisted of one recorded music listening session, used as a "negative case" (Allen, 2017), further described below, followed by 8-hr-long weekly sessions (the intervention). All sessions were recorded using the Fly 360-degree video camera (360fly, Canonsburg, PA, United States) and analyzed by NVivo 11 software (version; QSR International, 2016), which allowed slowed-down video analysis (e.g., 0.25 s per frame) to be captured. The Fly 360-degree camera allowed 360-degree footage of the room and a continuous view of all participants, at any point in time, in order to better capture and understand interpersonal interactions.

Data Analysis. As in Griffiths and Smith's (2016) study, video was the primary data-gathering tool. Video data consisted of all verbal interactions, nonverbal interactions, observable behaviors and the sequence of interactions. Viewing data from a 360-degree perspective facilitated close observations of simultaneous activity across the group at any given moment as well as easily accessible forward and rewind abilities, thus aiding detailed description throughout.

To begin the analysis, four residents were chosen at random and were then observed for the whole of each session across five of the eight live music sessions (sessions one, three, five, and seven) and the recorded music session. The initial data from these observations helped to guide further theoretical sampling (Urquhart, 2013). A transcription was made of moment-by-moment participant observations: what they were doing, what was happening around them (including the music), as well as what preceded and followed the moment. The transcript was developed by watching each session, slowing down the recording and typing the moment-by-moment observations into a word document. This resulted in a narrative transcript for each of the sessions observed for each participant. Once a narrative transcript was made for each of the initial four participants, the next step involved coding using NVivo-11software. This initial coding involved constant comparison; once the first code was identified, it was compared to the next piece of data in order to see whether they were similar, and could be incorporated together, or whether a new code needed to be named. The codes generated from this initial coding were followed by a process of selective coding where the codes were grouped together into core categories, depending on whether they shared similar properties. Throughout coding, theoretical memos were written to help identify emerging relationships and conceptual categories. The process of coding involved an ongoing shaping and reshaping of the core categories, which contributed to theoretical development as additional data were coded in an iterative cycle. This process resulted in the emergence of a core category—a category to which all the other concepts and categories related. The aim of the analysis was to reach theoretical sufficiency whereby a sufficient depth of understanding is obtained to allow for a plausible theory to develop (Dey, 1999). Once the core category was identified, it was decided that theoretical sufficiency had been reached.

To achieve this, following the initial data analysis additional observation was then completed with the other four residents to test the initial concepts being identified. These latter four residents were not observed for the whole of each of the five sessions. Instead, using theoretical sampling, samples were taken from each session at time periods when the previously observed participants had all shown a noteworthy change (either an increase or decrease, as agreed by AC and PMC) in communicative actions. The coding process was repeated for the additional data, allowing for definition and clarification of categories and further conceptual understanding. Several variations of the emerging theory were developed and adapted as constant comparison took place prior to their consolidation. Once this stage of the analysis was completed, using theoretical sampling, a negative case was chosen, identified where only recorded music was played. The same musicians, staff, and residents also attended this session. The negative case was sampled looking for both similarities and differences across the 10 communicative actions seen in Figure 1. According to Allen (2017), negative case analysis can be used to strengthen qualitative rigor and allow for nuanced analysis and theory development.

Figure 1.

A multisensory communicative environment through music for people with late- stage dementia.

Quality Assurance. Several methods of quality assurance were used in order to monitor and reflect on the researchers' preconceptions and motivations and the potential impact of this on the analysis. A reflective research diary was completed throughout. The first and second authors examined and discussed all coded data and reviewed corresponding sections of video while all authors discussed and agreed components of the emerging theory. Throughout, Chiovitti and Piran's (2003) guidelines for grounded theory quality assurance were followed.