Music Therapy Increases Social and Emotional Well-Being in Persons With Dementia: A Randomized Clinical Crossover Trial Comparing Singing to Verbal Discussion

Abstract

The number of people living with Alzheimer’s disease and related dementias (ADRD) is growing proportional to our aging population. Although music-based interventions may offer meaningful support to these individuals, most music therapy research lacks well-matched comparison conditions and specific intervention focus, which limits evaluation of intervention effectiveness and possible mechanisms. Here, we report a randomized clinical crossover trial in which we examined the impact of a singing-based music therapy intervention on feelings, emotions, and social engagement in 32 care facility residents with ADRD (aged 65–97 years), relative to an analogous nonmusic condition (verbal discussion). Both conditions were informed by the Clinical Practice Model for Persons with Dementia and occurred in a small group format, three times per week for two weeks (six 25-minute sessions), with a two-week washout at crossover. We followed National Institutes of Health Behavior Change Consortium strategies to enhance methodological rigor. We predicted that music therapy would improve feelings, positive emotions, and social engagement, significantly more so than the comparison condition. We used a linear mixed model approach to analysis. In support of our hypotheses, the music therapy intervention yielded significant positive effects on feelings, emotions, and social engagement, particularly for those with moderate dementia. Our study contributes empirical support for the use of music therapy to improve psychosocial well-being in this population. Results also highlight the importance of considering patient characteristics in intervention design and offer practical implications for music selection and implementation within interventions for persons with ADRD.

“I’ve learned that people will forget what you said, people will forget what you did, but people will never forget how you made them feel” (attributed to Maya Angelou). These words carry added weight for individuals with Alzheimer’s disease and related dementias (ADRD). As they decline, persons with ADRD are more likely to forget what has just happened to them, but their emotions—and the effects of such responses on well-being—remain (Reschke-Hernández et al., 2020). This phenomenon has major implications for care and treatment implementation with this population.

As people with ADRD decline, significant deficits in memory, communication, and thinking undermine social initiative, autonomy, and well-being (Sheffler & Noggle, 2015). Although they retain the need for social interaction, worthwhile activity, and inclusion (Kitwood, 1997), such progressive cognitive and functional decline is increasingly associated with less activity engagement and social interaction (Cohen-Mansfield et al., 2009; Sheffler & Noggle, 2015), which can exacerbate behavioral symptoms (Kitwood, 1997; Sheffler & Noggle, 2015). With limited pharmacological treatments available for individuals living with ADRD, supportive activities that can facilitate engagement and well-being are important in dementia care (Kolanowski et al., 2018; Livingston et al., 2020).

A recent synthesis of systematic reviews indicated that numerous psychosocial interventions have emerged, including exercise, reminiscence, various forms of cognitive stimulation, and music (McDermott et al., 2019). While many have focused on cognitive outcomes (Livingston et al., 2020; McDermott et al., 2019), relatively few studies have included emotional or social measures of intervention effectiveness (Kolanowski et al., 2018; McDermott et al., 2019). Of those, most have focused on symptom management (e.g., anxiety, depression, agitation; Livingston et al., 2020; McDermott et al., 2019). Meaningful engagement holds the potential to trigger positive emotions and promote social and emotional well-being (Camp et al., 2015; Cohen-Mansfield et al., 2009; Reschke-Hernández, 2021; Reschke-Hernández et al., 2020). Thus, research that evaluates social and emotional benefits of person-centered psychosocial interventions is needed.

In the context of significant decline in cognitive function, persons with ADRD often retain the ability to respond to, engage, and take pleasure from music experiences (Baird & Samson, 2015). Emerging research suggests that music-based interventions may benefit this population (Soufineyestani et al., 2021; van der Steen et al., 2018), for instance, by reducing agitation (Ridder et al., 2013) and elevating mood (Baker et al., 2022; Cho, 2018; Särkämo et al., 2014; Werner et al., 2017). Music can evoke strong emotions (Juslin, 2013; Koelsch, 2010) and offers many flexible options for passive (e.g., background music, listening) and active engagement (e.g., playing instruments) and range of possible responses (e.g., sleeping, clapping, singing) among diverse individuals within various social contexts (Gfeller, 2008; Koelsch, 2010). With thoughtful design and implementation, music-based interventions can offer meaningful opportunities to exhibit self-efficacy and socially engage in an age-appropriate and inclusive way (Clair & Memmott, 2008). However, the extent to which specific music experiences benefit persons with dementia and at various stages of decline is unclear, and evidence to support music therapy for this population is limited (Soufineyestani et al., 2021; van der Steen et al., 2018).

To advance music therapy research and practice, Reschke-Hernández (2019, 2021) developed the Clinical Practice Model for Persons with Dementia through four integrative steps (complete description of the Model and its development provided in Reschke-Hernández, 2019, 2021). First, she examined music perception and cognition literature to consider why various musical elements (e.g., rhythm, tempo) might influence nonmusical behaviors, what purpose each element might serve, and how to arrange each element within various music-based interventions for this population (i.e., therapeutic function of music; Hanson-Abromeit, 2015). Next, she conducted an exhaustive literature review from the last 50 years and communicated with experienced healthcare professionals to collect music- and nonmusic-based intervention strategies for working with individuals with ADRD. Using a qualitative approach, she coded and categorized strategies, created a thematic structure, then received and implemented clinical expert input on that structure. Finally, components from steps 1 to 3 were arranged within a broader theoretical framework, which was again reviewed by the clinical experts.

The Clinical Practice Model for Persons with Dementia holistically considers strengths and needs of a person with dementia. Biological, psychological, and social characteristics of an individual can exacerbate or mitigate signs and symptoms of dementia (Reschke-Hernández, 2019, 2021). Such personal aspects inform intervention design, which in turn influences an individual’s direct and indirect responses to treatment. Any aspect of the person, environment, or intervention could positively or negatively impact treatment outcomes. Reschke-Hernández proposed that psychosocial responses are the most immediate, direct, and salient outcomes of psychosocial interventions for persons with dementia. She further presents a schema regarding why particular intervention strategies are important and how to implement strategies within six interrelated thematic areas: Cognition (compensate for declines in declarative memory and processing speed), Attention (promote interest and motivation while minimizing distraction), Familiarity (enhance recognizability of stimuli), Audibility (facilitate auditory perception), Structure (offer redundant structure and predictability), and Autonomy (offer choice and opportunity for self-efficacy, matched to capabilities). This model informed the design and implementation of a singing-based music therapy intervention for nursing home residents with dementia in our study.

Past research has suggested that small-group music therapy that offers appropriate levels of challenge relative to degree of decline can result in purposeful activity (Hanson et al., 1996). Individual sessions may offer personalized attention and be more appropriate in certain circumstances (e.g., hospice, in-home); however, small groups are more cost-effective (i.e., appealing to stakeholders), which may improve access to music therapy. Music-making often occurs within a social context and can promote social engagement and cohesion (Trost et al., 2017)—necessary considerations for individuals who may retain limited social initiative and are often socially isolated (Sheffler & Noggle, 2015).

Music therapists design a wide variety of music experiences in practice. Prior research has suggested that an important consideration is the demand of the activity relative to the person’s degree of cognitive decline. In a study by Hanson et al. (1996), participants with ADRD were exposed to an equal number of intervention types (instrument play, singing, movement) at high and low demand levels in small group sessions. Each session included all three types of music experiences. The most active participation occurred when the degree of challenge aligned with participants’ cognitive decline (i.e., what has been termed the “Goldilocks principle”). This study offered important information regarding the design of music experiences relative to participant characteristics.

Nonetheless, two limitations in the study by Hanson et al., (1996) reflect a gap in the evidence base for music interventions with this population (Reschke-Hernández, 2019; Soufineyestani et al., 2021; van der Steen et al., 2018). First, lack of a control or well-matched comparison condition limits the ability to isolate intervention effects and consider underlying mechanisms. Second, many authors have examined a holistic session (i.e., numerous activities). This design introduces possible carryover effects and limits conclusions that can be drawn regarding each intervention’s effectiveness. Although a variety of reasons explain choices made in prior research designs (e.g., question under investigation, practicality), there are limitations to understanding music-based intervention effectiveness and best practice.

In the present study, we isolated a singing-based intervention. Singing is common cross-culturally across the lifespan (Mehr et al., 2019) and is a foundational tool in music therapy (Matney et al., 2018). We gain extensive implicit song knowledge through a lifetime of listening, which may render it less cognitively demanding than activities that rely on spoken language (Patel, 2014). Singing is associated with widespread activation in bilateral brain structures, including the primary auditory cortex, motor, and premotor areas (Vuust et al., 2022), and these areas are largely spared of AD-related pathology (Domoto-Reilly et al., 2019). Singing-based interventions have been examined in individual (Clair, 1996; Ridder et al., 2013) and group sessions (Baker et al., 2022; Hanson et al., 1996; Werner et al., 2017); however, these studies examined several activities within a session. Others have utilized recreational group singing (i.e., for enjoyment, with focus on music-making) as a comparison to music therapy (i.e., goal-directed use of music within a therapeutic relationship, for example, Baker et al., 2022; Särkämo et al., 2014; Werner et al., 2017), which may also be a beneficial social activity for this population (Baker et al., 2022; Soufineyestani et al., 2021). However, past studies that have compared recreational group singing and music therapy also compare different professionals (e.g., music educator or choral director vs. music therapist; Baker et al., 2022; Särkämo et al., 2014; Werner et al., 2017), number and variety of activities, dosage, and activity group sizes (which may impact degree of individualized attention, for example, large singing group vs. small music therapy group; Baker et al., 2022; Werner et al., 2017). Such differences make it difficult to interpret what about the two conditions led to observed differences. By directly comparing a singing-based intervention to a similar nonmusic condition, we can evaluate the effectiveness of this intervention at influencing social engagement and affective responses in persons with ADRD and consider underlying mechanisms of the music intervention.

The purpose of this randomized clinical crossover trial was to examine the impact of a music therapy intervention, grounded in the Clinical Practice Model, on social and emotional responses of persons with ADRD. Specifically, we compared small group live singing of participant-preferred songs with guitar accompaniment to an analogous nonmusic condition: verbal discussion. Both conditions used the same structure, strategies, and dosage, with active music versus verbal engagement being the key difference. Our dependent variables were feelings, emotions, and social engagement. Based on literature reviewed above (Gfeller, 2008; Koelsch, 2010; Trost et al., 2017), we hypothesized that social music-based interventions offer more means to interact and engage than social activities that rely on verbal conversation. Therefore, we predicted that the music therapy intervention would result in more constructive social engagement than the nonmusic verbal discussion condition. Also, based on literature (reviewed above) suggests that activities that include familiar and preferred music are more emotionally evocative than familiar activities that do not include music (Cho, 2018; Juslin, 2013; Koelsch, 2010; Reschke-Hernández et al., 2020), we predicted that the music therapy intervention would result in more improved feelings and positive emotions (pleasure) than the nonmusic comparison.

Method

The Institutional Review Board at the University of Iowa approved this study. We piloted all procedures with an independent participant sample (reported in Reschke-Hernández, 2019). This eight-week trial followed a protocol (NCT03643003) and involved two conditions: singing-based music therapy intervention and verbal discussion comparison. This trial consisted of the following activities (illustrated in Figure 1): (a) enrollment, stratification by facility and gender, and random assignment to order, (b) one week of standard care (SC), (c) two weeks (six 25-minute sessions) of either music therapy or verbal discussion, (d) a two-week SC “washout,” (e) crossover, and (f) one week of SC. We used a crossover design to efficiently compare the two study conditions, examine short-term ­responses (i.e., feelings, emotions, social engagement), and minimize confounds due to variability and change over time typical of persons with dementia (Lezak et al., 2012; Sheffler & Noggle, 2015). Given the brief intervention period, we elected to use a two-week washout to minimize the likelihood of any carryover effects. We conducted an a priori power analysis using data from related research (Cho, 2018; Millard & Smith, 1989; Ray & Mittelman, 2017) and the pwr.t.test function within the pwr package (Champely et al., 2018) for R Version 3.5.1 (R Core Team, 2018), α = .05 (two-tailed) and δ = .80. Based on this analysis, our target sample size was 30.

Figure 1.

Participant flow diagram for the within-subjects randomized crossover study.

Participants

Descriptive and neuropsychological data are displayed in Table I. We enrolled all those who met the inclusion criteria. We were interested in post hoc analyses regarding dementia severity and gender (refer to subsequent section), but we did not sample based on these characteristics given the scope and in situ nature of this study. Participants were at least 65 years old, resided at their facility for at least three months, and had a formal dementia diagnosis. We excluded candidates who did not speak or understand English fluently, had a history of severe mental illness, received hospice services, or were bedridden at enrollment. A legally authorized representative provided consent for each participant. Thirty-two persons with dementia (6 men and 26 women) ages 65–97 years old ($\mu =84.13$, mdn = 83.5, SD = 8.44) from three Iowa care facilities volunteered for this study. None of the participating facilities offered music therapy as part of SC. All participants were white and non-Latinx and on average held a high school level of education ($\mu =12.78$ years, range: 8–18 years). There was no participant mortality. Care staff completed the Katz Index of Independence in Activities of Daily Living to quantify autonomy (Katz et al., 1970). Music preferences (per family members, care staff, and participants when possible) were diverse but clustered around popular standards and big band (e.g., Doris Day, Frank Sinatra), country (e.g., Hank Williams, Johnny Cash), and singalong songs (e.g., You Are My Sunshine, Home on the Range, Iowa Corn Song). Most participants had little to no formal music training (two were professional musicians) and enjoyed social music engagement such as church choir and dancing. Participants received a personalized music CD and a Bluetooth-enabled boom box at the conclusion of the study.

Table I.

Participant Demographics, Cognitive Function, Independence, and Health

Measure	Men (n = 6)				Women (n = 26)
	Handedness: 1 left, 4 right, 1 mixed				1 left, 25 right, 0 mixed
	Range	Mean	Median	SD	Range	Mean	Median	SD
Age (years)	79–92	85.83	86.00	5.35	65–97	83.73	82.00	9.04
Education (years)	8–18	14.00	15.00	3.58	9–16	12.50	12.00	1.58
CDRa	2–3	2.17	2.00	0.41	1–3	2.31	2.00	0.74
MoCA beginningb	0–10	4.50	3.50	3.89	0–17	5.46	3.50	5.92
MoCA endc	0–10	4.50	4.50	3.45	0–17	4.75	3.00	5.26
Katzd	0–5	3.00	3.00	1.90	0–6	3.12	3.00	1.99

Note. ^aThe Clinical Dementia Rating discriminates between levels of dementia severity and ranges from 0 (least impaired) to 3 (greatly impaired; Morris, 1993). ^bThe Montreal Cognitive Assessment is a screening test of cognitive function and ranges from 0 (severe impairment) to 30 (unimpaired; Nasreddine et al., 2005). ^cTwo women were hospitalized at follow-up and unavailable to complete the MoCA at the end of the study. ^dKatz Index of Independence in Activities of Daily Living (Katz et al., 1970) ranges 0 (greatest dependence) to 6 (independent).

Care Facilities

All three participating facilities met Medicaid certification standards and offered licensed Chronic Confusion or Dementing Illness (CCDI) units (i.e., locked unit with staff who have at least minimum dementia care training). Facility 1 was a 92-bed for-profit in a small rural town (population 2,626). Eight residents from the 12-bed CCDI unit participated. Facility 2 was a 34-bed nonprofit in a small rural town (population 2,832). All 10 residents from the CCDI unit participated along with four residents from the main health center (14 total participants). Facility 3 was a 120-bed nonprofit in a suburban community (population 20,881) and 10 residents from the 25-bed CCDI unit participated.

Cognitive Status

Review of participants’ care center charts indicated that more than half of our sample had unspecified dementia (n = 18, 56%; probable Alzheimer’s disease, n = 9, 28%; vascular dementia, n = 4, 12%; dementia with Lewy bodies, n = 1, 3%; mixed dementia, n = 1, 3%). We used the Clinical Dementia Rating (CDR) scale to rate dementia severity within two weeks prior to the first study session. This measure consists of semi-structured interviews with the patient and an informant. Scores range from 0 (least impaired) to 3 (greatly impaired), and the CDR discriminates well between severity levels (Morris, 1993). Of the 32 participants, 4 scored at level 1 (mild), 15 at level 2 (moderate), and 13 at level 3 (severe dementia).

We administered the Montreal Cognitive Assessment (MoCA) as an estimate of change in cognitive status during the study (in conjunction with the CDR and again within 1 week after the final SC week; Nasreddine et al., 2005). On average, participants’ cognitive status remained stable from the beginning (N = 30, $\mu =5.03$, SE = 0.98) to study end ($\mu =4.70$, SE = 0.90; t(29) = .86, p = .40, d = −.06; n.b., two women were hospitalized at follow-up and excluded from this analysis). We observed a floor effect in 25% of this sample: eight participants obtained a 0 at both the beginning and end of the study.

Dependent Measures

All measures were designed for and validated with persons with ADRD (for detailed psychometric properties, see Reschke-Hernández, 2019). Seven trained data takers (all music therapy students) collected all dependent data. We utilized REDCap to enhance data management.

During all sessions, data takers used the Observed Emotion Rating Scale (OERS; Lawton et al., 1996, 1999) in conjunction with the Menorah Park Engagement Scale (MPES; Camp et al., 2015) in 5-minute observation intervals. Per the OERS, they recorded behavioral indicators of pleasure, anger, anxiety/fear, and depression/sadness, each scored as 0 (not indicated), 1 (up to half of the interval), or 2 (more than half of the interval; .78≤κ≤.89; Lawton et al., 1999). The MPES yields an engagement score profile: constructive, passive, other, and nonengagement, each scored 0 (not indicated), 1 (up to half of the interval), or 2 (>half of the interval), and altruism and disengagement/agitation (scored respectively as frequency and yes/no; all categories mutually exclusive; Cronbach’s α = .90; Camp et al., 2015).

To measure feelings, data takers administered the Dementia Mood Picture Test (DMPT) before and after each session. This self-report measure was designed for use with persons across dementia severity levels and comprises yes/no questions and simple face drawings for six emotions: good mood, bad mood, happy, sad, angry, and worried. It yields a single score that ranges from 0 (most negative) to 12 (most positive; .95≤κ≤.99; Tappen & Barry, 1995). Data takers abandoned this questionnaire with any participant who indicated in any way that they did not understand or did not want to answer the questions, including signs of increased agitation. Of the 32 participants, 24 provided DMPT data.

Quality Assurance Strategies

The National Institutes of Health Behavior Change Consortium has recommended quality assurance strategies to achieve methodological rigor in research (Bellg et al., 2004). Following these guidelines, all interventionists completed systematic training prior to leading sessions. An independent masters-level music therapist with extensive teaching and clinical experience with this population reviewed all study training materials for quality, content, and congruity. Training details are reported in Reschke-Hernández (2019) and included (a) instructional modules and quizzes about ADRD, the Clinical Practice Model, and procedures for protocol implementation and (b) practice sessions to ensure treatment fidelity across interventionists. Interventionists documented protocol adherence using self-monitoring checklists before and after each session, and they checked in weekly with Author 1. Three credentialed music therapists with deep familiarity with the protocol attended randomly selected study sessions of each interventionist and completed a rating checklist to evaluate and document treatment fidelity. All participants and facility and study personnel were masked to study aims and hypotheses.

Study Conditions

All participants completed two conditions: a singing-based music therapy intervention and a verbal discussion comparison. Each facility was randomly assigned to a condition in order to avoid indirectly affecting residents in the comparison condition. We then randomly assigned participants to small groups at their facility. Although group members’ personalities and behavior could impact one another’s responses, this arrangement offers social activity, is cost-effective, and reflects real-world care provision. To avoid having all men (who enrolled in fewer number than women) in a single group, we used stratified random assignment to allocate men to groups.

Participants at two facilities (n = 8 and n = 10) received music therapy first, and participants at one facility (n = 14) received verbal discussion first. Each condition was presented in a small group format (3–5 participants per group; 7 groups total), three times per week, for two consecutive weeks. Each session lasted 25 minutes and occurred on consistent days and times in the afternoon. Three board-certified music therapists (“interventionists”) led all sessions, were white women under age 30, and had one to five post-internship years of experience working with persons with dementia. Interventionists were masked to study aims, hypotheses, and measures and aimed to engage participants maximally (to the extent they wanted to engage), regardless of study condition.

Interventionists maintained a consistent environment across sessions. Because many older adults have some degree of hearing loss, the group sat in a semicircle within reasonable proximity to the interventionist (Wilhelm, 2016). The interventionist led from a seated position with the flexibility to stand and assist participants as needed. Interventionists adapted in real-time to provide an appropriate level of support and challenge to each participant, based on the Clinical Practice Model. They used three topics in the same order in both the singing and verbal discussion conditions, each for two consecutive sessions: (a) travel and places, (b) nature and hobbies, and (c) love and friendship. They used age-appropriate visuals, devoid of musical cues, to enhance topic comprehension (e.g., photograph of a farm).

Interventionists used the same strategies across conditions following the Clinical Practice Model. Here, we provide a summary of accompanying strategies that informed the implementation of the study conditions (refer to Reschke-Hernández 2019, 2021 for a detailed Model description and protocol). Interventionists provided extra processing and response time to compensate for participants’ declines in declarative memory and processing speed (Cognition theme) and used concrete statements, a conversation decision tree, and pacing matched to participants’ behavioral responses. To promote interest and motivation while minimizing distraction (Attention theme), they used age-appropriate humor, surprise, and novelty, and they assessed nonverbal attentional cues before interacting with participants. They carefully selected familiar and culturally relevant topics, music, and images from participants’ formative years and utilized ample repetition (Familiarity theme). Interventionists facilitated auditory perception (e.g., proximity), use of residual hearing, and perception of visual cues (e.g., pictures, gestures) to aid auditory understanding (Audibility theme). They used a variety of organization strategies such as an opening and closing, transitions, themes, and redundant cues (Structure theme). Finally, interventionists provided abundant choice and opportunities for self-efficacy and focused on strengths (Autonomy theme).

The singing-based intervention consisted of live, participant-preferred songs that reflected the session topic with acoustic guitar accompaniment. To focus specifically on the impact of singing, interventionists did not use any other types of music activities or accompanying instruments. They implemented approximately 85% interactive singing (21 minutes) and 15% (4 minutes) conversation, commentary regarding the song selections and theme, and intermittent verbal prompts to engage. In the comparison condition, interventionists prepared conversation starters and trivia to lead discussion for the entire session based around the same themes and strategies, but without any music.

Statistical Analysis

We used a separate linear mixed model for each dependent variable (feelings, emotions, social engagement). We used R 3.5.1 (R Core Team, 2018) to perform all statistical analyses with packages lme4 (Bates et al., 2015), lmerTest (Kuznetsova et al., 2017), and emmeans (Lenth et al., 2018). Each model included a random intercept for participant to account for repeated observations. As stated in our hypotheses, our primary interest was the direct comparison of study conditions. Post hoc, we considered fixed effects for order and additional factors that could inform clinical practice: session number (which could impact when an intervention becomes impactful), dementia severity (as in Hanson et al., 1996), and gender (evidence exists on sex and gender differences in dementia diagnosis, presentation, and treatment; Ferretti et al., 2020; Subramaniapillai et al., 2021), all coded as categorical factors. We used Akaike’s information criterion to select final variables included in each model and evaluated residual variance to ensure statistical assumptions were met. To facilitate interpretation, we converted all scales to a percent of maximum possible score prior to analysis [(score − minimum)/(maximum − minimum)] × 100. There was no indication of assumption violations for any model. We used Satterthwaite’s method to estimate denominator degrees of freedom and Tukey adjustment for multiple comparisons. Contrasts were specified from each model. Our level for statistical significance was .05.

Results

Feelings of Emotion

As previously mentioned, data takers administered the DMPT (Tappen & Barry, 1995) with 24 of the 32 participants before and after each session. We calculated change in feelings from pre- to post-session as our dependent variable. The best-fitting model for feelings included session number, order, condition, gender, and interaction between condition and gender. Order was nearly statistically significant whereby those who received music therapy second reported slightly more feeling elevation in response to all sessions ($\widehat{\mu }=8.00$, 95% CI = 3.44, 12.55) than those who had music therapy first ($\widehat{\mu }=2.77$, 95% CI = −1.30, 6.84; t = −2.03, p = .0563, β = −5.22). In support of our hypothesis, there was a significant main effect of condition (t = 2.09, p = .0381). As illustrated in Figure 2, participants’ feeling elevation was more pronounced after music therapy ($\widehat{\mu }=8.48$, 95% CI = 4.24, 12.72) than after verbal discussion sessions ($\widehat{\mu }=2.29$, 95% CI = −2.26, 6.85; β = 6.18).

Figure 2.

Change in feelings of persons with dementia in response to music therapy and verbal comparison conditions. Note. This figure illustrates feelings that participants (n = 24) self-reported in response to music therapy and verbal conditions, as measured by the Dementia Mood Picture Test (Tappen & Barry, 1995). A positive score indicates feeling elevation from pre- to post-session. Errors bars indicate standard error of the mean.

In an exploratory analysis, we found a significant interaction between condition and gender. Due to the small number of men enrolled in the study, these results should be interpreted cautiously. Women reported improved feelings in response to both conditions (t = −1.04, p = .3009; for verbal, $\widehat{\mu }=8.11$, 95% CI = 4.34, 11.88; for music, $\widehat{\mu }=5.43$, 95% CI = 1.50, 9.36; β = −2.68). By contrast, men reported diminished feelings in response to the verbal condition ($\widehat{\mu }=-3.53$, 95% CI = −11.80, 4.75) and an elevated response to the music therapy condition ($\widehat{\mu }=11.52$, 95% CI = 4.01, 19.03, t = 2.82, p = .0053, β = 15.05; for verbal condition, men versus women, t = −2.57, p = .0128, β = −11.64).

Observed Emotion

Data takers used the OERS (Lawton et al., 1996, 1999) to record behavioral indicators of pleasure, anger, anxiety/fear, and depression/sadness in 5-minute observation intervals during sessions. We used participants’ mean emotion scores for each session to account for varying observation interval numbers across sessions (participants had to attend at least 3 of 5 observation intervals for data from a given session to be included in analysis). The best-fitting model for emotions included session number, condition, and dementia severity. Data takers introduced a small amount of measurement error when observing emotions (intraclass correlation = .75).

Signs of pleasure included smiling, laughing, and affectionate behavior. In support of our hypothesis, there was a significant main effect for condition (t = 5.49, p < .0001). As illustrated in Figure 3, participants demonstrated more pleasure during music therapy ($\widehat{\mu }=42.07$, 95% CI = 34.61, 49.52) than during verbal discussion sessions ($\widehat{\mu }=28.13$, 95% CI = 20.68, 35.57, β = 13.94). In post hoc analysis, we found a significant main effect for dementia severity. Persons with moderate dementia ($\widehat{\mu }=44.37$, 95% CI = 35.42, 53.31) demonstrated significantly more pleasure than persons with severe dementia ($\widehat{\mu }=22.87$, 95% CI = 13.57, 32.17; t = 3.39, p = .0058, β = 21.50). Data takers observed negative emotions at low levels in both conditions, without sufficient data to conduct statistical analysis.

Figure 3.

Pleasure observed in persons with dementia during music therapy and verbal comparison conditions. Note. These figures illustrate (a) the main effect of condition on pleasure, demonstrated by all participants (N = 32) during music therapy and verbal conditions and (b) the main effect of dementia severity on pleasure, demonstrated by persons with mild (n = 4), moderate (n = 15), and severe dementia (n = 13) collapsed across conditions, as measured by the Observed Emotion Rating Scale (Lawton et al., 1996, 1999). The analysis regarding persons with mild dementia should be interpreted cautiously given the small sample size. Errors bars indicate standard error of the mean.

Social Engagement

As previously mentioned, data takers observed engagement using the MPES. As with observed emotions, we used participants’ mean engagement scores for each session to account for varying observation interval numbers across sessions. Constructive engagement (optimal for the type of singing and verbal discussion in this study) consists of active and observable social interactions when participants clearly engage and display interest in the activity, and their behavior is reasonably appropriate to the task (Camp et al., 2015). The best-fitting model for social engagement included an interaction term, with condition and dementia severity as the static predictors. Data takers introduced a minimal amount of measurement error for social engagement observations (intraclass correlation = .90).

In support of our hypothesis, there was a significant main effect for condition (t = 2.51, p = .0126). On average, participants demonstrated more constructive engagement during music therapy ($\widehat{\mu }=72.86$, 95% CI = 63.67, 82.06) than during verbal sessions ($\widehat{\mu }=65.99$, 95% CI = 56.88, 75.10, β = 6.87). In post hoc analysis, we found a significant interaction between condition and dementia severity (F = 5.00, p = .0074). As illustrated in Figure 4, persons with moderate dementia demonstrated more constructive engagement in music therapy ($\widehat{\mu }=88.84$, 95% CI = 77.32, 100.37) than in verbal discussion ($\widehat{\mu }=72.12$, 95% CI = 60.45, 83.79, t = 4.82, p <.0001, β = 16.73). During music therapy, persons with mild ($\widehat{\mu }=82.86$, 95% CI = 60.78, 104.95) or moderate dementia demonstrated significantly more constructive engagement than persons with severe dementia ($\widehat{\mu }=46.89$, 95% CI = 34.86, 58.92; for mild vs. severe, t = 2.91, p = .0170, β = 35.98; for moderate vs. severe, t = 5.09, p <.0001, β = 41.96). During the verbal condition, persons with mild ($\widehat{\mu }=81.14$, 95% CI = 59.51, 102.76) or moderate dementia demonstrated significantly more constructive engagement than persons with severe dementia ($\widehat{\mu }=44.72$, 95% CI = 32.61, 56.84; for mild vs. severe, t = 2.99, p = .0142, β = 36.41; for moderate vs. severe, t = 3.29, p = .0065, β = 27.39). Contrasts including persons with mild dementia should be interpreted cautiously given the small sample.

Figure 4.

Social engagement observed in persons with dementia during music therapy and verbal comparison conditions. Note. These figures illustrate the percentage of constructive (a) and passive (b) engagement demonstrated by participants with mild (n = 4), moderate (n = 15), and severe dementia (n = 13) during music therapy and verbal conditions, as measured by the Menorah Park Engagement Scale (Camp et al., 2015; types of engagement mutually exclusive). Analyses regarding persons with mild dementia should be interpreted cautiously given the small sample size. Errors bars indicate standard error of the mean.

Passive Engagement

Per the MPES, we also measured and included post hoc analysis of passive engagement. Passive engagement occurs when participants are reasonably assumed to be alert, oriented toward the group, and appear to be observing the activity (levels of engagement mutually exclusive; Camp et al., 2015). Although constructive engagement is ideal for the types of activities examined here, degree of engagement is a choice, passive engagement is preferable to nonengagement or disruptive behavior, and it is a reasonable and appropriate response to the types of activities studied here.

Session order was nearly statistically significant whereby those who received music therapy second displayed somewhat more passive engagement during all sessions ($\widehat{\mu }=43.92$, 95% CI = 35.22, 52.61) than those who had music therapy first ($\widehat{\mu }=32.84$, 95% CI = 24.38, 41.29; t = −2.02, p = .0532, β = −11.08). Although there was no a main effect of condition (t = .20, p = .8396, β = −.81), we found a significant interaction between condition and dementia severity (F = 6.26, p = .0022; refer to Figure 4). During music therapy, persons with severe dementia ($\widehat{\mu }=50.77$, 95% CI = 41.06, 60.48) demonstrated significantly more passive engagement than persons with moderate dementia ($\widehat{\mu }=23.07$, 95% CI = 13.71, 32.42; t = −4.11, p = .0004, β = −27.70). Persons with moderate dementia ($\widehat{\mu }=40.08$, 95% CI = 30.40, 49.77) demonstrated more passive engagement in verbal sessions than in music therapy (t = −3.37, p = .0109, β = −17.01). Data takers observed all other forms of engagement (other engagement, nonengagement, disengagement/agitation, and altruism) at low levels in both conditions, without sufficient data to conduct statistical analysis.

Discussion

The purpose of this multi-site randomized clinical crossover trial was to examine affective and social responses of persons with ADRD engaged in small-group music therapy sessions. We isolated a single type of music therapy experience (singing) and compared it to a verbal discussion activity without music. Both conditions were grounded in the Clinical Practice Model for Persons with Dementia (Reschke-Hernández, 2019, 2021). Negative emotions and behaviors that are challenging for caregivers (e.g., agitation) occurred at low levels in both conditions. In support of our hypotheses, the singing-based intervention yielded significant positive effects on feelings, emotions, and constructive social engagement, particularly for those with moderate dementia. These effects were significantly greater in response to the singing-based intervention than the nonmusic comparison activity that used the same structure and intervention strategies within six interrelated thematic areas (i.e., Cognition, Attention, Familiarity, Audibility, Structure, Autonomy). In the following sections, we discuss these responses to the two study conditions.

Mechanisms That May Explain Responses to Study Conditions

Music and Social Engagement

Participants demonstrated significantly more constructive social engagement during the singing intervention than during the verbal comparison condition. Implicit musical memory for long-known musical styles or specific selections appears to be spared in persons with ADRD until severe stages of decline (Baird et al., 2015; Jacobsen et al., 2015). Music offers multiple means for engaging and responding (e.g., listening, singing, playing, moving, creating) through its various elements (e.g., rhythm, melody, lyrics), and it can be intrinsically rewarding (Gfeller, 2008). In essence, the music therapy intervention highlighted a relative strength (music processing and engagement) and promoted less effortful engagement (i.e., nonpropositional singing acquired via repetition; Patel, 2014), and interventionists strategically selected and adapted music to make it easier to participate.

By contrast, social conversation relies heavily on semantic and propositional spoken language—that is, volitional, conscious effort to form words into meaningful ideas (Patel, 2014). To follow along and socially engage in verbal sessions, participants had to comprehend what others said, pay attention and remember the discussion as it ensued, and formulate and coherently verbalize responses. Although interventionists strategically selected topics, used prompts, and adapted conversation to make it easier to participate, each task that is inherent to verbal discussion (comprehension, propositional speech, attention, working and semantic memory) highlights a deficit for persons at all dementia severity levels (Lezak et al., 2012), albeit to differing degrees.

Support and Challenge

A particularly important and clinically meaningful finding of this study (illustrated in Figure 4) is that persons with moderate dementia demonstrated high levels of constructive engagement during music therapy. The Ecological Theory of Aging suggests that older adults have a “zone of maximum performance potential,” whereby environmental demand (i.e., stress) aligns with a person’s capabilities (e.g., physical mobility, cognitive abilities) and varies over time. Too much or too little demand leads to negative affective, perceptive, and cognitive experiences, whereas optimal person-environment alignment (i.e., ecology) yields positive experiences (Lawton & Nahemow, 1973). The Progressively Lowered Stress Threshold model (Hall & Buckwalter, 1987) builds upon Ecological Theory and suggests that persons with dementia become less able to cope with internal stressors (e.g., pain; needs for affection and attention) and external demands of the care environment as they decline, which can lead to agitation and care resistance. Supportive care and age-appropriate activities that align with participants’ capabilities and needs can reduce challenging behaviors and facilitate positive experiences.

Our post hoc analysis indicated that persons with moderate dementia demonstrated more constructive engagement in music therapy than in verbal discussion. This finding suggests that participants with moderate dementia may have experienced better alignment of the singing-based intervention to their capabilities than those with either more or less severe dementia. The adaptations to the singing-based intervention may not have offered sufficient challenge for persons with mild dementia and, as in prior similar research (Hanson et al., 1996), too much demand to actively engage those with severe dementia. Additional research with a larger, more balanced sample of individuals across the decline trajectory is needed. However, our findings, in conjunction with those of Hanson and colleagues (1996), indicate that the type and demand of an activity, relative to decline, is likely an important consideration for music-based interventions for persons with ADRD.

Repetition

Interventionists followed strategies within the Familiarity theme of the Clinical Practice Model (Reschke-Hernández, 2019, 2021) to promote and enhance the recognizability of topics and songs. Within this theme, application of ample repetition may have been a key differentiating feature between singing and verbal conditions that influenced constructive social engagement, emotions, and feelings. Music therapists selected repertoire that contained a lot of repetition (e.g., repetitive lyrics, melody, rhythm patterns) and that enabled them to add repetition. Sometimes interventionists only sang the chorus, and often they repeated a song multiple times. Although repetition was a strategy in both conditions, repetition is inherent in music across structural elements, is associated with neural reward systems (Juslin, 2013; Koelsch, 2010), and tends to enhance our enjoyment of music. By contrast, adapting spoken language in such a repetitious fashion could actually become uninteresting, tedious, and socially odd (Margulis, 2014). Music therapists indicated the value of repetition in session notes, stating “She would sing on the third repetition,” “She would sing along when there was more repetition,” and “He usually joined in [with] repetition.” Repetition offers predictability, organization, and a sense of familiarity (Margulis, 2014; Patel, 2014), and it may have facilitated both engagement and pleasurable responses. Effects of repetition have largely been studied in healthy young adults. Future research could examine responses among persons with dementia using music that contains varying degrees of repetition to explore the role of this possible mechanism.

Music and Emotions

Participants self-reported that they felt better and demonstrated more pleasure in response to music therapy compared to verbal discussion. In this study, interventionists largely used music from participants’ formative years within participant-preferred genres (Familiarity theme, Clinical Practice Model, Reschke-Hernández, 2019, 2021). Familiar and meaningful music can trigger involuntary and effortless recall of autobiographical memories (Belfi et al, 2016; El Haj et al., 2012), which seems to play an important role in emotional responses to music (Juslin, 2013; Koelsch, 2010) including in persons with dementia (El Haj et al., 2012). While interventionists used familiar images and topics in the verbal condition, these prompts may have required more effortful engagement and explicit memory recall and subsequently less intense positive emotions (Belfi et al., 2016). Music therapists mimicked stylistic qualities such as syncopation and pulse accentuation to further enhance familiarity and promote interest and motivation (Attention theme, Clinical Practice Model, Reschke-Hernández, 2019, 2021). Such characteristics are implicated in the urge to move to music and are associated with enhanced arousal and pleasurable affective states (Juslin, 2013; Trost et al., 2017). If judicious choice and implementation of the music itself are key ingredients in music-based interventions with this population, then important considerations in music therapist training and continuing education include culturally sensitive repertoire selection (including awareness of bias and stereotyping) and cultivating the musical skills that are necessary to both emulate familiar musical styles and adapt to participant needs in the moment.

While the results of this study indicate overall positive emotion evocation, music therapists noted that several participants expressed longing for family members in response to various music selections. Often the participant did not recall that these individuals had died and inquired where they were and why they were not present. Music therapists validated but did not dwell on such responses and transitioned to a different song to prevent sustained experience of negative emotions (as in Reschke-Hernández et al., 2020). Such real-time assessment of responses to music, flexible adaptation of the intervention, and verbal processing may highlight important qualities of live, person-centered music therapy interventions and music therapist training. These findings also indicate a possible side effect of music interventions.

Limitations

Limitations regarding dependent variables, study personnel, and participants influence the interpretation and generalizability of the results. Although we utilized a reliable and valid measure of feelings, self-report assumes that an individual has access to and honestly reports on their internal, private state. It is possible that some participants did not comprehend questions on the DMPT or lacked capacity to reflect and report on their feelings. Despite a high degree of interrater reliability on observational measures, some data takers could have been more observant, meticulous, and skilled than others or might have favored a particular condition and inadvertently influenced observational data.

Although we minimized bias through masking, training, and quality assurance strategies, bias was likely present to some extent among data takers and interventionists. To minimize confounds among group leader training and clinical experience, music therapists led both conditions. While a strength of this study is that the conditions (including group leaders) were well-matched and grounded in the Clinical Practice Model, the Model has not yet been widely used, and interventionists might have unconsciously performed better when leading a particular condition. Level of training and experience could influence research replication or generalizability of results to practice (e.g., a novice might not attain such positive outcomes; one with better skills might procure better outcomes). Practice philosophy and approach influence intervention implementation and could lead to different outcomes than reported here. Although an inherent challenge of behavioral research, interventionist characteristics are nonetheless important considerations for application of these findings to clinical practice.

Individual characteristics and care structure (e.g., memory care unit at a facility, in-home care, nonspecialist unit) may influence responses to the music therapy intervention or comparison activity reported here. Our exploratory analysis indicated a sizeable effect of gender on feelings, albeit in a very small sample of men. Emerging research indicates sex and gender may influence signs and symptoms of dementia and response to treatment to varying degrees (Ferretti et al., 2020; Soufineyestani et al., 2021; Subramaniapillai et al., 2021). In post hoc analysis, we also found a large effect of decline on constructive engagement, which aligns with past research in music therapy (Hanson et al., 1996) and the Progressively Lowered Stress Threshold model (Hall & Buckwalter, 1987). Given the short duration of the intervention and washout period in this study, a longer intervention period could have a greater and longer lasting impact. Future research that utilizes a randomized controlled trial design or mixed methods approach could enhance our understanding of responses to music-based interventions and inform best practices.

This study sample consisted of white, English-speaking older adults from Iowa and limits the generalizability of our results. Importantly, lack of marginalized individuals in research is a significant gap in aging research (Brewster et al., 2019) and the music perception and cognition literature base that informs music therapy practice (Baker et al., 2020). We should strive to build relationships, trust, and reciprocity with organizations in partnership with clinicians to help address this research gap, improve culturally sensitive practice, and better serve persons with dementia and their families.

Significance and Conclusion

This study evaluated social and emotional benefits of a person-centered singing-based music therapy intervention among persons with ADRD. Results indicate that an intervention that is grounded in the Clinical Practice Model for Persons with Dementia (Reschke-Hernández, 2021) can yield clinically meaningful social and emotional benefits for this population. Data takers observed few instances of negative emotions, nonengagement, and agitation in response to both conditions, which carries positive implications for care staff and participant well-being. Further testing and refinement of the Model and research to identify, define, and reliably measure long-term responses without placing added burden on care staff is warranted.

We followed quality assurance strategies to strengthen methodological rigor. Rather than a holistic music therapy session, we examined a single, protocol-based intervention, and we included a nonmusic comparison. These choices enhance the reliability and validity of our findings and facilitated our ability to consider possible mechanisms that underlie this intervention. Furthermore, we used a linear mixed model rather than an ANOVA approach to analysis, which enabled us to account for individual variability—a particularly important feature for longitudinal research with persons with ADRD. Future music therapy research would benefit from such design and analysis considerations.

In conjunction with past research, an important implication for clinical practice regards consideration of the capabilities of the person relative to the challenge and support offered by an activity—a “one-size-fits-all” approach to music-based interventions is not appropriate. The intervention tested in this study, while more controlled than what music therapists typically do in practice, reflects a real-world form of treatment. It was grounded in facilitation principles and strategies, which are as important to understand as the intervention itself. While additional research is needed to evaluate the utility of the Clinical Practice Model for Persons with Dementia, this research contributes empirical support for the use of music therapy to enhance psychosocial well-being of individuals whose lives are significantly impacted by dementia.

The authors thank all the participants, their families, and participating care facilities and staff, without whom this research would not have been possible; Dr. Mary Adamek, Dr. Jeremy Manternach, Miranda Peyton, Elizabeth Ouverson, Bailey Bodeker, Valerie Wilks, Ann Fienup, Nicole Belluomini, Danae Molenkamp, Brenna Oates, and Laura Elliott Buckner for their assistance; and Joey Walker, Meghan Ross, Rachel Abbe, and Bethany Wheeler for their clinical expertise. This study was completed in partial fulfillment of the degree of Doctor of Philosophy at the University of Iowa. Full dissertation is available through The University of Iowa at https://doi.org/10.17077/etd.59oh-y06y.

Funding: This research was partially supported by grants from the University of Iowa Office of Outreach and Engagement (to ARH), the National Institute of Mental Health (P50 MH094258 to DT), the Kiwanis Neuroscience Research Foundation (to DT), and the National Center for Advancing Translational Sciences of the National Institutes of Health (UL1TR002537 to the University of Iowa). This research is also supported by Dr. Richard and Mrs. Ellen Caplan, the University of Iowa Music Therapy Area, and contributors to a University of Iowa GOLDRush crowdfunding campaign (to ARH).

Conflict of interest: None declared.