The Use of Music Playlists for People with Dementia: A Critical Synthesis

Research question

Literature search

The original search was a systematic scan of the following databases: JSTOR, PsychInfo, PubMed, and ProQuest, using the search terms “music AND dementia” in the title and abstract. An initial 765 articles were collated from this search.

Selection of studies for inclusion

Purposive sampling [17] was used to select the articles from health and psychology disciplines. All identified references were screened according to the following inclusion criteria: 1) participants had to have some form of dementia or be caring for someone with dementia, 2) the intervention studied must involve the use of pre-recorded music alone or in combination with other musical activities, and 3) the research had to be published between 2006–2016.

Articles not meeting the above inclusion criteria were excluded. Studies relating solely to people with mild cognitive impairment were also excluded, as were opinion or general discussion papers or review papers and studies involving only active forms of musical engagement such as singing, playing musical instruments, or songwriting. While the focus of the current review was not music therapy, where music listening was used in a formal music therapy setting, these were also included because of the potential for such studies to further illuminate the value of music listening for people with dementia.

All types of evaluative studies were eligible for inclusion. While articles were appraised for scientific quality according to conventional standards such as the Joanna Briggs Institute’s critical appraisal tools (2016), methodological quality did not form a basis for exclusion since a critique of the reliability of studies in the field was viewed as a central part of the function of the critical analysis and formulation of a synthesizing argument. Abstracts were screened according to these criteria in an initial screening process, leaving a total of 37 articles for review.

Titles and abstracts to be considered were then uploaded to Covidence, an online platform for conducting systematic reviews (http://www.covidence.org). Since abstracts did not always provide enough information to assess the studies according to the inclusion criteria, full texts of articles were then subjected to a closer appraisal for relevance by two reviewers, after which a further 9 were excluded, leaving 28 studies for inclusion in the review (Fig. 1).

Fig.1

PRISMA flowchart of article selection.

Data extraction

Evidence summary templates were created in Covidence and used for data collection regarding the characteristics of studies, participants, interventions and final outcomes. The data extracted also included details about:

Evaluation of studies

It was expected that the studies identified would be quite heterogeneous, and that a lack of comparable methodologies and outcome measures would mean the reviewed studies would not be amenable to a meta-analysis. Thus, a narrative synthesis model was used in exploring the data [15, 18]. This approach uses words and text to explore relationships in the data and to develop a theory based on a synthesis of the findings in the reviewed studies [19]. The studies were critically appraised by two authors independently of each other on the basis of methodology, sample size, the appropriateness of the musical intervention tested, and the appropriateness of the conclusions drawn to the results reported and other criteria drawn from the Health Care Practice Research and Development Unit (HCPRDU) evaluation tool for quantitative studies [20]. This tool was developed to assist in the critical appraisal of research studies, and is particularly useful for reviews that include both qualitative and quantitative studies. The two authors who conducted this appraisal subsequently discussed their relative assessments and reached agreement about the relative weight of the studies discussed. Studies were then grouped and sub-grouped as described below for comparative purposes, according to study design, intervention type, the outcomes reported and other study features. The lead author prepared a narrative analysis based on these findings and discussions, giving greater emphasis to the more robust studies considered, which was then checked independently by all other authors.

General characteristics of the reviewed studies

The 28 studies considered in this review were published in journals representing 8 different disciplines and fields of specialty (see Fig. 2), but were primarily found in health-related journals. A summary of the characteristics contained in this study can be found in Table 1. The studies originated from 13 different countries, representing Europe, Asia, North America, and Australasia. A large proportion of studies did not report the specific type of dementia that their participants had been diagnosed with (n = 13, 46%). A number reported focusing on participants with Alzheimer’s disease (n = 7, 25%), while the balance of studies reported that their participants were made up of those with Alzheimer’s disease and ‘other’ forms of dementia (n = 8, 29%). None of the studies distinguished between people with different forms of dementia within their results. Sample sources included residential aged care facilities (nursing homes) (n = 18, 64%), home care situations (n = 4, 14%), day activity centers for people with dementia (n = 2, 7%), hospitals (n = 2, 7%), and an assisted living facility (n = 1, 4%). One study did not report the source of theparticipants.

Fig.2

Fields of origin of reviewed studies.

A diverse range of study designs were used in the reviewed studies including randomized controlled trials (RCTs) with parallel treatment groups (n = 7, 25%), an RCT with a cross-over design within the treatment group and a parallel control group (n = 1, 4%), controlled experiments with cluster randomization (n = 2, 7%), controlled experiments with parallel treatment groups but no randomization (n = 2, 7%), crossover designs with randomization of the order of conditions (n = 2, 7%), crossover designs with a standard order for conditions (n = 8, 29%), single group experiments with a pre-post design (n = 3, 11%), single group experiments with continuous observation (n = 2, 7%), and a single case crossover design (n = 1, 4%). Thus only 10 studies in total (36%) had both a control group or condition, and some kind of randomization included in the design (Table 2). Of the seven studies that used RCTs with parallel treatment groups, two compared similar interventions or used comparable outcome measures. Since the aim of our review was both to look at the effect of various interventions on particular psychological and behavioral symptoms, as well as to look at the relative influence of other variables, multiple studies that compared similar interventions under similar conditions would have been required to conduct a meta-analysis. Thus, it was decided that insufficient comparable studies were available for the purposes of a meta-analysis in this instance.

Where control groups or conditions existed, comparisons were between music listening and other audio conditions such as reading (e.g., [21]), between active musical engagement or music therapy, and music listening (e.g., [22]), or between music and standard care [23]. A small number of studies (n = 2, 7%) did not appear to have control conditions that were well matched to the music listening group, with the control engaging in more passive or more solitary activities than the treatment group, making it difficult to attribute where effects of the interventionlay.

Overall sample sizes tended to be relatively small in many studies, although they ranged from 1–120 with a mean of 36.6. Where control groups were used, group sizes ranged from 5–40 participants with a mean of 22.6. A total of 13 studies (46%) had sample or group sizes of less than 20, and only 3 studies (11%) reported group sizes of 40 or more. An attempt was made to look at effect sizes of results in order to determine how sample sizes influenced the results. However, only 6 studies reported effect sizes or sufficient information to allow effect sizes to be calculated for within-subjects analyses (21%) (Table 2). Nevertheless, a number of studies that reported non-significant results did have relatively small sample sizes, suggesting that more definitive findings could have been obtained with large samples. Attrition rates also tended to be high with over half the studies reporting drop-out rates of more than20% (57%).

The music interventions

The majority of studies included in the review related to musical interventions that were not led by a music therapist (n = 17, 61%). Six studies (21%) included sessions led by a music therapist, while 5 considered interventions involving music sessions facilitated by a caregiver after training by a music therapist. The length of single sessions of musical engagement across all studies ranged from 5 min to 4 h, m = 53.8 min. Several studies involved single listening sessions only (n = 3, 11%), while the majority involved daily (n = 4, 14%), twice daily (n = 1, 4%), weekly (n = 7, 25%), or semi-weekly (n = 11, 39%) sessions, over a period ranging from 3 weeks to 6 months. However, as demonstrated in Table 1, several studies (n = 15, 54%) did not adequately report the details of the intervention procedures in order for these details to be ascertained in eachcase.

Most studies involved individual music sessions (n = 15, 54%), 4 of which used headphones, while 6 played music over loudspeakers (5 did not report whether speakers or headphones were used). One study directly compared the use of headphones and speakers [24], with both conditions being found to have similar effects. However, this was a single case study. No other study attempted to assess the relative effects of using headphones versus use ofloudspeakers. Nine studies were group sessions in which only loudspeakers were used, and 4 studies involved caregiver and participant listening together.

A majority of studies selected ‘favorite music’ or music based on information from family members and caregivers (n = 14, 50%) (e.g., [25]). Of these studies, 9 did not use any specific protocol or give specific details about how these preferences or recommendations were sought, stating only, for example, that they used music that was biographically relevant (e.g., [26]). However, 5 studies used the Assessment of Personal Music Preference (APMP) established by Gerdner [27].

The remaining studies used music selected by the researcher (n = 11, 39%). Of these, one study used nature music (which included piano and sounds from nature such as birds, running water, and whale song) [28], one used music composed by the researcher [29], one used music from the era most often recalled by participants [22], two used classical or baroque music [30, 31], one compared Dean Martin and religious music [32], and five used music that was considered ‘age appropriate’ [33–37].

It was unclear in at least three of the studies that used researcher-selected music whether there was an attempt to select music designed to target specific symptoms. However, 6 attempted to match the music to the symptoms of interest, such as using calming music to reduce agitation (see for e.g., [29, 35]). Studies that used ‘favorite music’ did not attempt to match the music to any particular desired mood state, with the exception of Sakamoto and colleagues [22] who confined music selections to joyful music. Two studies used a combined approach, attempting to select music on the basis of both individual preferences and the specific mood effects desired[38, 39].

Relative effectiveness of different interventions

As demonstrated in Table 3, a number of studies (n = 11, 39%) reported positive outcomes from individual receptive music interventions that used either an established protocol for music selections or that was based on family recommendations, although one study did report positive results from classical music [30]. In non-therapist led sessions, several studies reported reduced agitation after the use of individualized playlists (n = 5, 18%) [40, 41]. Hicks-Moore and Robinson [42] for example found that listening to favorite music was effective in reducing agitation both alone and when combined with hand massage. Other studies reported reductions in anxiety [43], depression [5], and pain [44], and increases in quality of life [5] and performance on several cognitive domains [45]. Increases in caregiver satisfaction, relaxation, comfort, and happiness were also reported [46]. Where directly compared, there was little evidence that music therapy served advantages over non-therapist led listening [5, 47].

However, results were not universally positive. Garland and colleagues [41], for example, found that while overall results of listening to favorite music in their study were positive, there was a widely divergent response from participants reporting that “a dramatic improvement in agitation for some was offset by neutral or negative outcomes for others” (p. 520).

Support for the value of group music listening interventions was less robust, with only two studies providing clear positive results for group receptive programs [29, 34]. One study found positive effects on emotional state for a music therapy group that mixed active and receptive, but no significant differences with the control group which participated in cooking activities [35].

Despite some positive results, some negative outcomes were also reported in relation to group receptive interventions. Nair and colleagues [31], for example, reported that after exposure to Baroque music, behavioral disturbances in residents of an aged care facility actually increased overall during weeks in which the music was played. The authors concluded that this was because the music used was not appealing to participants and that the use of individualized music selections would be more effective. Interestingly, Chang and colleagues [28] similarly found that behavioral symptoms were worse in weeks during which nature music was played to residents of an aged care facility. They argued that this may have been due to the fact that there was a time lag in the effect. However, most studies indicated that response to music tended to be immediate. Thus, the two papers taken together seem to suggest the relative ineffectiveness of researcher-selected music played in group settings in improving behavioral disturbances inpeople with dementia.

Seven studies directly compared active and receptive interventions, and tended to demonstrate that both types of musical engagement yielded some positive results. Positive results for both active and receptive were reported in 6 of the seven studies, although active engagement such as singing was reported as having stronger effects in 3 studies. Sakamoto and colleagues [22], for example, compared active musical engagement with a receptive intervention and a control group, each of which were administered by a group of trained facilitators including but not limited to music therapists. They found that while both active and receptive interventions caused increased parasympathetic nervous system activation, active engagement resulted in better overall improvements to mood. Holmes and colleagues [34] found that while participants demonstrated engagement with both the active and receptive music programs, engagement was greater for the active intervention.

Raglio and colleagues [47] also compared music listening with active music therapy, but found no significant differences between conditions, although trends in the data suggested that active music therapy had a greater effect on behavioral and psychological symptoms. However, while the brief report of this study states that no caregiver was present during the music listening, it did not clearly identify whether the music therapist also led the music selection for the listening condition. Therefore, while this study offers some support for the greater impact of active musical engagement on symptoms of dementia, it remains inconclusive about the necessity for a trained music therapist to be involved in facilitating music listening programs. In addition, the fact that the conditions compared active engagement with a therapist present, with receptive listening without anyone present makes it difficult to determine the degree to which the outcomes were influenced by the nature of the musical engagement involved, or by the personal contact with a therapist.

In a follow-up study [5], the same authors then compared music listening with active music therapy and a control group receiving standard care. It was found that all groups showed improvements in behavioral symptoms, depression, and quality of life, with no significant differences between groups. The authors attributed these results to the types of outcome measures used. However, this study also was not able to demonstrate that active music therapy serves any advantages over music listening.

Several studies did not clearly distinguish between active and receptive musical engagement in the results (e.g., [48, 49]). For example, Clement and colleagues [50] found that short-term benefits on emotional state were experienced by both the music and the cooking groups, while long-term benefit was experienced by the music group only. However, both the cooking and the music interventions involved active and receptive aspects, making it difficult to disentangle the relative effects.

A number of studies gave the reader only a vague idea of the level of involvement of a music therapist. For example in the study by Guetin and colleagues [38], a music therapist utilized software that produced individualized playlists designed to shape moods according to an inverted U-curve—with music initially intended to calm the listener and then to gradually increase arousal. The control group took part in a rest and reading session “under the same conditions and at the same intervals” (p. 38), but the study does not provide further details as to the reading condition, i.e., whether or not this was a live reading, a recorded reading, how the reading material was selected, or whether it too was designed to follow a particular mood curve. In addition, the authors noted that in their study the music sessions involved the development of a patient-therapist relationship, but did not describe whether the reading group was similarly facilitated by a therapist. The study reported significant improvements in anxiety and depression in the music group in comparison to the control group. However, while this was generally a robust study, the scarcity of detail in relation to the control group makes it difficult for the reader to judge whether or not the differing effects observed can be attributed to the music itself or to other situational variables.

An interesting sub-group of 5 studies concerned interventions where caregivers were trained by a music therapist to use music during their caregiving routine. For example, Särkämö and colleagues[26, 45] trained caregivers to use both singing and music listening, finding that both were able to improve mood, remote episodic memory and also had an effect on cognitive domains. The music listening group improved more than the control group in behavioral disturbances, although no significant group effects were observed on any scales at a 6-month follow up. However, 2 studies in which caregivers were trained to implement a music intervention were unable to report significant results, possibly due to methodological limitations.