Remotely supervised home-based online chair yoga intervention for older adults with dementia: Feasibility study

Abstract

Background:

and Purpose: Social isolation and caregiver burden call for an innovative way to deliver a chair yoga (CY) intervention to older adults with dementia who cannot travel to a community center. During a remotely supervised CY session, the yoga instructor can monitor each participant’s pose and correct poses to optimize efficacy of CY and reduce chances of injury. This study assessed the feasibility of a remotely supervised online CY intervention for older adults with dementia and explored the relationship between CY and clinical outcomes: pain interference, mobility, risk of falling, sleep disturbance, autonomic reactivity, and loneliness.

Methods:

Using a one-group pretest/posttest design, a home-based CY intervention was delivered remotely to 10 older adults with dementia twice weekly in 60-minute sessions for 8 weeks. Psychosocial and physiological (i.e., cardiac) data were collected remotely at baseline, mid-intervention, and post-intervention.

Results:

The results indicated that remotely supervised online CY is a feasible approach for managing physical and psychological symptoms in socially isolated older adults with dementia, based on retention (70%) and adherence (87.5%), with no injury or other adverse events. While there were no significant findings for pain interference, mobility, sleep, or social loneliness longitudinally, emotional loneliness showed a significant increase, F(1.838, 11.029) = 6.293, p = .016, η² = 0.512, from baseline to post-intervention. Although participants were socially connected to other participants via a videoconferencing platform, emotional loneliness increased during the pandemic period.

Conclusion:

A home-based remotely supervised online CY is a feasible approach for socially isolated older adults with dementia who are unable to travel to a facility.

1. Introduction

Dementia is characterized by progressive deterioration of brain function, resulting in cognitive impairment that affects learning, memory, and other cognitive functions [1]. Dementia also involves deterioration of physical function, such as mobility and gait performance; in particular, persons with vascular dementia decline more rapidly than those with Alzheimer’s disease [2]. Cognitive impairment in dementia is associated with behavioral and psychological symptoms, such as agitation, aggression, apathy, hallucination, depression, anxiety, and/or sleep disturbance [3]. Thus, dementia is a major cause of limitation in activities of daily living (ADL) in older adults and one of the main contributors to institutionalization [4].

Although increased prevalence and incidence rates have become a major public health issue [5], no cure for dementia has been identified [1]. To manage symptoms associated with dementia, such as memory and behavioral and psychological symptoms of dementia (BPSD), pharmacological management is often prescribed and used by older adults with dementia. However, drug adverse events [6,7] can contribute to admission to emergency departments, hospitalization, poor quality of life, mortality, and increased health care costs. Thus, safe, effective, and evidence-based nonpharmacological treatment is needed. Mind-body intervention, as a part of a nonpharmacological approach, can be used to reduce such symptoms [8].

Yoga, a type of mind-body intervention, consists of breathing techniques, physical poses, lifestyle education, and meditation [9]. Yoga has been shown to have beneficial effects on decreasing cognitive decline [10,11]. One theorized mechanism is the ability of yoga to support self-regulation and resilience through decreasing sympathetic and hypothalamic-pituitary-adrenal stress responses and supporting parasympathetic states [11-14]. Parasympathetic activity can be non-invasively measured by recording cardiac rhythms (via electrocardiogram or pulse) and then analyzing respiratory sinus arrhythmia (RSA; the primary component of high-frequency heart rate variability [HRV]) from the overall HRV rhythm [15]. Systematic reviews have demonstrated lower parasympathetic activity through measures of diminished HRV in people with dementia, including dementia with Lewy bodies and cognitive impairment [16,17]. Decreased parasympathetic HRV has been shown to be related to poor cognitive performance [18] and has been associated with many adverse health conditions, including chronic pain, neurodegenerative disorders, anxiety, and pain [19-22].

Increased physical activity may improve physical function in both older adults with dementia and their caregivers [23]; thus, those with dementia should keep physically active [24]. While research has demonstrated that cardiovascular exercise can increase parasympathetic HRV and cognitive fitness in persons with Alzheimer’s disease [25], research is needed to examine the feasibility of chair yoga (CY) in this population, in particular those who are physically frail.

CY, practiced sitting in a chair or standing and using a chair for support [26], has been shown to be a safe, nonpharmacological, noninvasive, and low-impact intervention for older adults with dementia [27]. CY combines flexibility, balance, strength, breathing, relaxation, and mindfulness training. CY includes stationary poses that use isometric contraction and relaxation of various muscle groups, as well as dynamic postures, breathing practices, imagery, and muscle relaxation practices [27].

However, access and transportation to sites where CY classes are conducted is a barrier to participation, particularly for older adults in rural areas. Most older adults with dementia are cared for at home by a caregiver to maintain the living situation. Due to the COVID-19 pandemic, many older adults have not been able to participate in group-based in-person CY classes because of the requirement to remain at home for physical distancing.

Social isolation and caregiver burden have underlined the need for an innovative way to deliver the CY intervention to older adults with dementia and their caregivers. Online intervention could eliminate the fear of being infected and the need to travel to an exercise place [28]. Technological interventions for assisting older adults with dementia (e. g., application development, use of robotics, sensors, locator devices, reminders, virtual related technology) have created the potential for home-based CY interventions with real-time monitoring through a secure videoconferencing platform (e.g., Zoom, WebEx). Unlike online exercise programs that are recorded, in a remotely delivered intervention consisting of live CY sessions the yoga interventionist can correct poses to optimize benefits of CY and reduce chances of injury. Studies suggest that adults with dementia may be able to focus on the CY intervention better in a comfortable home environment rather than in a community center with associated distractions [28,29].

No research has been conducted on online CY for older adults with dementia via videoconferencing. Therefore, the purpose of this study was to assess the feasibility of conducting a remotely supervised, home-based, online CY intervention for older adults with dementia and to assess the relationship between online CY and clinical outcomes. The primary aim of the study was to assess the feasibility (retention, adherence, and safety) of conducting a remotely supervised, home-based, online CY intervention and completing outcome measures virtually for community-dwelling older adults with dementia, as measured by retention rate (percentage of enlisted participants completing the CY intervention and all three data collection times). The secondary aim was to examine the relationship between the intervention and chronic pain, physical function, or psychological symptoms. Finally, an exploratory aim was to evaluate the ease and ability of caregivers and participants to record cardiac data remotely for offline analyses of the effect of the intervention on parasympathetic regulation (i.e., RSA) and overall heart rate (HR).

1.1. Theoretical framework

The theoretical framework for this study is aligned with the unified theory of acceptance and use of technology (UTAUT). The theory has identified critical factors and contingencies related to the behavioral intention to use a technology [30,31]. UTAUT consists of four major constructs that influence behavioral intention to use a technology as applied to this study. The first construct is performance expectancy (the degree to which using an online technology will provide benefits to participants in performing certain activities). All participants in this study accessed the videoconferencing via Internet. The second construct is effort expectancy (the degree of ease associated with participants’ use of the CY intervention). It was relatively easy for the caregiver to access the intervention through videoconferencing at home. The third construct is social influence (the extent to which participants perceive that important others will agree that they should use a particular technology). The family caregivers agreed that, during the COVID-19 pandemic, older adults needed to keep physically active, since they were unable to attend in-person social activities. The fourth construct is facilitating conditions (participants’ perceptions of the resources and support available) [30,31]. In this study, participants were told that a research assistant (RA) would be available at each session and data collection point to monitor safety and provide technical support.

2. Methods

2.1. Design

A longitudinal study using single-arm repeated measures design in which all participants were assigned to the online CY group was utilized, following feasibility trial guidelines [32]. The twice-weekly CY intervention was designed and developed by the research team for use with older adults with dementia based on traditional Hatha yoga. The sessions incorporated aspects of regular yoga practiced while seated in a chair or standing and holding a chair for support. This method provided a sense of security and a safe environment. In a recent meta-analysis [33], a moderate dose (60–120 min per week) was found to be the optimal mind-body exercise, including yoga and tai chi.

2.2. Participants

Inclusion criteria were (a) 60 years or older; (b) living in the community, not institutionalized (c) diagnosed with dementia (e.g., Alzheimer’s disease, Lewy body, vascular, frontal lobe); by a neurologist or other health care provider; (d) Montreal Cognitive Assessment [MoCA]) < 26, considered to indicate moderate cognitive impairment [34]; (e) ability to ambulate at least 30 feet independently and safely with minimal assistance (e.g., cane or walker); (f) access to desktop or laptop computer and the Internet; (g) a supportive caregiver to accompany participant in each online class session and to provide information on participant’s symptoms. We excluded participants with concurrent medical conditions that could confound interpretation of outcome measures, pose a safety risk, or preclude successful completion of the protocol, such as (a) presence of psychiatric disorders (e.g., bipolar, schizophrenia), (b) alcohol or chemical dependency (c) in need of assistance by another person (e.g., holding the arm), to ambulate, or (d) wheelchair use.

2.3. Intervention

Each intervention session began with a 5-min socialization time. During that time, the digital gallery view was used so that each participant and caregiver could see other participants and feel socially connected. During the CY session, the yoga interventionist was spotlighted in the Zoom screen to allow participants to see only the interventionist. This spotlighting enabled participants to focus on the yoga sessions without being distracted by other participants on the screen.

Participants attended remotely supervised 60-min CY sessions twice a week for 8 weeks (16 sessions). The online CY consisted of five components: (a) check-in (10 min), (b) breathing techniques and intentional practice (10 min), (c) physical postures (25 min), (d) guided relaxation and visualization (10 min), and (e) wrap up (5 min). Each CY session was conducted synchronously via a video application (Zoom), accessible by computer. Participants interacted on Zoom with other participants or with the facilitator to maintain social bonds while maintaining physical distance. Researchers have reported that older adults with dementia and their caregivers are socially isolated [35]; the group setting provided opportunities for social interaction while encouraging continuation of yoga practice [36].

Each CY session was conducted by a certified yoga interventionist with yoga teaching experience. Caregivers joined the sessions, monitored the participant for safety, and assisted in following the poses correctly. Safety check was completed by asking each caregiver to check whether the chair was stable, no object was near that might hinder the participant, and the caregiver would be available during the intervention. For fidelity check, another yoga therapist observed 10% of the sessions and evaluated the yoga interventionist’s performance based on a fidelity checklist.

2.4. Measures

Data were collected at baseline, after 8 sessions (4 weeks), and after 16 sessions (completion of the intervention) via Zoom (Table 1). Baseline evaluation included data on demographics (e.g., age, gender, race/ethnicity, marital status), health information (e.g., type of dementia [Alzheimer’s disease, Lewy bodies with dementia], medical history, current medications), mobility, risk of falling, pain interference, sleep disturbance, psychological measures (e.g., loneliness), and awareness of autonomic reactivity. Cardiac rhythms were measured, via pulse, during a posture shift protocol at each of the three evaluations.

Table 1.

Demographic Characteristics of the Participants (N = 11).

Characteristic and category	n
Age, years M = 80.8 (SD = 7.458), range = 68–96
Gender
Male	8
Female	3
Race
Caucasian	10
African American	1
Marital status
Married
Divorced	2
Widowed	1
Level of education
Less than high school	1
Some high school	1
Some college	1
College graduate	4
Graduate degree	4
Level of independence
Live independently	2
Some assistance with complex activities	2
Some assistance with basic activities	5
Completely dependent	2
Relationship to the caregiver
Spouse/partner	7
Daughter	1
Significant other	1
Other	2
Professional Service
Neurologist	7
Neuropsychiatrist	1
Physical Therapist	2
Social Worker	3

2.4.1. Primary outcome

Retention was defined as at least 70% of the participants who started the CY intervention completing the intervention and all three data collection points (baseline, mid-intervention, post-intervention). Adherence was defined as participants adhering to at least 70% of the twice-weekly 45-min CY intervention sessions (11 of 16 sessions). Safety was defined as ≤ 10% of the participants reporting adverse events associated with participating in online CY.

2.4.2. Secondary outcomes

Each evaluation included questionnaires (PROMIS PI-SF [37], Sleep Disorders Inventory [SDI] [38], de Jong Gierveld Loneliness Scale [39], Body Perception Questionnaire-Short Form [BPQ-SF] [40], the Timed Up and Go (TUG) test [41], and cardiac data collection (iom2 biofeedback device) to evaluate clinical outcomes. Evaluations were conducted by the RA during live videoconferencing via Zoom, with the caregivers’ assistance.

Pain Interference.

The 8-item PROMIS PI-SF V. 1.0-8a was administered to assess self-reported pain interference in various aspects of the participant’s life within the previous 7 days, using a 5-point response scale ranging from not at all to very much [37]. Scores can range from 8 to 40, with higher scores indicating more interference [42]. Alpha reliability ranges from 0.96 to 0.99 and construct validity is adequate [43].

Mobility.

The TUG test was administered to measure mobility and risk of falls [41]. Participants were timed to stand up from the chair, walk 10 feet, turn, walk back, and sit down. A 2- or 3- minute break between trials allowed participants to recover. The average of three tests was calculated in seconds, timed with a stopwatch. The mean time of two trials was used for analyses. A TUG score of ≥12 seconds identified participants who were at higher risk of falling. Interrater intraclass correlation coefficient (ICC) and interrater ICC are high for the TUG (0.99), and construct validity for mobility has been correlated with gait speed (r = −0.61) and Barthel Index (r = −0.78) [41]. In the current study, the RA instructed the participant to complete the test via Zoom. The caregiver assisted the participant to perform the test safely according to instructions from the RA.

Sleep Disorders.

The SDI was administered to collect data on the frequency, severity, and level of distress with respect to seven sleep symptoms in the previous 2-week period. The number of items was recorded and the SDI score was derived as the product of the average frequency ratings and average severity ratings (range 0–12), with higher scores indicating more distress [38].

Loneliness.

The 6-item version of the de Jong Gierveld Loneliness Scale [39] was used as a general loneliness measure of overall, emotional, and social loneliness. This scale includes two subscales: emotional loneliness (EL, 3 items), referring to a feeling of missing an intimate relationship, and social loneliness (SL, 3 items), indicating a wider social network. Total scores range from 0 (not lonely) to 11 (extremely lonely). Cronbach’s alpha was moderate to strong (0.85–0.92). In the 6-item scale, three statements (e.g., I experience a general sense of emptiness”) refer to EL and three items (e.g., “There are plenty of people I can rely on when I have problems”) refer to SL. SL refers to the absence of a social network, which is a wider circle of acquaintances and friends who can provide a sense of belonging (i.e., companionship and being a member of a community). EL is defined as absence of an attachment figure in one’s life to whom one can turn, such as an intimate or close relationship.

Autonomic Reactivity.

The BPQ-SF [40] was used to measure the subjective experience of the function and reactivity of target organs and structures that are innervated by the autonomic nervous system. The BPQ-SF includes 20 items with a 5-point response scale ranging from never (1) to always (5), with total scores ranging from 20 to 100 and higher scores indicating more reactivity. The BPQ-SF showed good reliability and converged with measures of stress and somatosensory amplification. For this study, the subscale of autonomic reactivity was used. Internal consistency for the subscale was 0.85–0.88 and the test-retest reliability score was r = 0.78 [44].

Cardiac Rhythms.

Cardiac rhythms were analyzed from pulse data recorded using the iom2 biofeedback device (UNYTE), a transmission earlobe photoplethysmograph (PPG). The device consists of an earlobe optical sensor, pulse processor, and USB cable, connected to the participant’s home computer via an USB interface. The device was delivered to each participant’s home prior to data collection. Evaluations were conducted during live videoconferencing, during which the researcher instructed the caregiver on the placement of the iom2 and recording of the participant’s cardiac data, and then guided the research protocol. The research protocol consisted of a posture shift from sit, stand, and then return to sit for 3 min in each posture (i.e., sit1, stand, sit2), with the aim of eliciting changes in cardiac rhythms, specifically changes in HR and RSA. The researcher monitored the process for accuracy and the participant for safety. Participants were instructed not to talk and to keep their eyes open during the data recording. Once data were collected, caregivers were given instructions for securely uploading the files to the researcher for analyses of cardiac rhythms.

Analyses of RSA consisted of the following steps. Peaks of the pulse wave were extracted using CardioPeak and Segmenter software (Brain-Body Center for Psychophysiology and Bioengineering, University of North Carolina, Chapel Hill). Peak detections were visually inspected and edited offline with CardioEdit software (Brain-Body Center, University of Illinois at Chicago). Editing consisted of integer arithmetic (i. e., dividing intervals when detections were missed and adding intervals when spuriously invalid detections occurred). RSA, defined as the heart period variance within the frequency band associated with spontaneous breathing, was calculated with CardioBatch Plus software (Brain-Body Center for Psychophysiology and Bioengineering, University of North Carolina, Chapel Hill), using the statistical procedures developed by Porges [45]. CardioBatch Plus quantifies the amplitude of RSA using age-specific parameters that are sensitive to the maturational shifts in the frequency of spontaneous breathing. CardioBatch Plus calculates RSA as follows: (a) timing sequential peak intervals to the nearest millisecond, (b) producing time-based data by resampling the sequential peak intervals into 500 msec intervals, and (c) detrending the time-based series with a 21-point cubic moving polynomial [46], stepped through the time-based data to create a smoothed template, which is then subtracted from the original time-based series to generate a detrended residual series; (d) bandpass filtering the detrended time series to extract the variance in the heart period pattern associated with spontaneous breathing in adults (0.12–0.40 Hz); and (e) transforming the variance estimates with a natural logarithm to normalize the distribution of RSA estimates [47]. These procedures are statistically equivalent to frequency domain methods (i.e., spectral analysis) for calculation of the amplitude of RSA when heart period data are stationary and appropriately detrended [15,48]. Several studies have reported no significant differences or high correlation between interbeat intervals when using pulse versus electrocardiogram (ECG) systems [49-52]. However, comparisons between variability measures may be different dependent on statistical methodology or participant characteristics, as some studies reported differences [52] while other studies reported similarities [49-53]. The present study uses the same statistical methodology as that used by Heilman et al. [49] and Davila et al. [53], which demonstrated similarity in results of analyses between pulse and ECG measures.

Three minutes of pulse data during each posture were edited and analyzed. RSA and HR were quantified during each sequential 15-second epoch and the averages in each condition were used in the data analyses. HR was included in analyses as a measure of overall cardiac activity, while RSA was included as a measure of high-frequency variability.

2.5. Recruitment

Twenty-one older adults with dementia were contacted and screened for eligibility; 20 met inclusion criteria, and one was excluded due to a serious comorbid condition (Parkinson’s disease). Of those 20, three declined to participate due to scheduling conflicts, four due to lack of interest, one due to too much data collection, and two for caregiver’s personal issues. Of the remaining 11 participants who met inclusion criteria and completed informed consent and assent forms, one dropped from the study before the start of the intervention. Ten participants began the online CY intervention, and three withdrew (two no longer interested, one with caregiver’s schedule conflict; see CONSORT flow diagram, Fig. 1). Of the 10 participants who began the intervention, 8 met criteria to furnish cardiac data from home (two participants had only tablet PCs and were not able to install the software for collecting cardiac data remotely). Of the 10 participants who started the CY intervention, 7 completed the intervention and post-intervention data collection, except cardiac data (5 participants completed cardiac data) after 3 participants dropped the intervention. Of the 3 participants, 2 were no longer interested in CY because 60-min sessions were too long to concentrate for those with advanced stages of dementia.

Fig. 1.

Flow of participants through the study.

Of the 8 participants who were able to furnish cardiac data from home, 5 met criteria to furnish cardiac data after 3 participants had dropped the intervention. Of the 5 participants who completed cardiac data, 4 furnished cardiac data for all postures, for each assessment session. Equipment error affected only 1 participant at one time point. These findings indicate the feasibility of remotely collecting cardiac data from home, with the equipment assistance of a caregiver and the guidance of a researcher through the protocol via Zoom.

2.6. Study Procedure

Prior to initiating recruitment, the study design was approved by the researchers’ affiliated universities; (a) Florida Atlantic University Institutional Review Board, USA, and (b) University of British Columbia Research Ethics Board, Canada. The RAs completed a training program in screening and data collection, using the specific measures to be used in the study, and completed a return demonstration of correct procedures. They were monitored by the researchers for data collection and screening.

A telephone meeting was held with potential participants and their caregivers to explain the purpose of the study and to show how to use Zoom and the video application. Then a virtual screening session was conducted via Zoom. When the person met eligibility, the RA reviewed the consent form and assent form with the participant and primary caregiver. Those who agreed to participate in the study completed and electronically signed an online consent form (participant) or assent form (caregiver). Prior to the online CY sessions, baseline data collection was completed by the RA with the caregiver’s assistance via live videoconferencing. The mid-intervention and post-intervention data collections were conducted in the same way as at baseline.

The RA sent a link to each participant’s caregiver via email to guide them in using the video application and data were collected online. After baseline data collection, the 8-week CY intervention was conducted. The participants and their caregivers were remotely supervised by the RA to ensure proper use of technique. An RA was available at each session to provide technical support. After the intervention was completed, the synchronous online CY sessions were offered to participants by the yoga interventionist so they could continue practice. The principal investigator and co-investigator monitored the process for accuracy and the participants for safety.

2.7. Statistical analyses

An intent-to-treat approach to data analysis was employed, using SPSS™ v. 28.0 for Windows (IBM Corp, Summers, NY) to perform all analyses. Preliminary analysis included descriptive statistics (mean, SD, range, frequencies) to identify sample characteristics. Retention was calculated by dividing the number of completing participants by the number of participants who completed baseline data collection (retained/enrolled X 100). Second-stage analysis included single-group repeated-measures analysis to examine the difference in outcomes (pain interferences, physical function [mobility], risk of falls, loneliness, and sleep disturbance at pretest (baseline), mid-intervention (after 8 sessions), and post-intervention (after 16 sessions). Exploratory analysis included a single-group repeated measures analysis of variance (ANOVA) to examine changes in HR and RSA in three postures (sit, stand, sit) across three data collection points (baseline, mid-intervention, post-intervention).

3. Results

3.1. Sample characteristics

Mean age of the 11 participants who were consented to participate in the study was 80.73 years (SD = 7.458, range 68–96). A majority of the participants were male (n = 8), non-Hispanic White (n = 10), married (n = 8), and college graduates (n = 8). Mean MoCA score was 15.4 (SD = 6.407). Regarding the type of dementia, four participants had been diagnosed with Alzheimer’s disease, one with Lewy body dementia, and six with other dementias. About half (n = 5) required some assistance with basic activities. Of the primary caregivers, seven were spouses or partners. To manage dementia symptoms, 8 participants had been taking medication(s) for a mean of 27.4 months (SD = 3.0). Only four participants had used nonpharmacological treatment to manage symptoms (three used exercise and one attended music therapy). Three reported a history of falling due to cognitive impairment.

3.2. Primary aim Feasibility of the CY intervention

Program fidelity was measured in terms of (a) retention, (b) adherence, and (b) safety. Seven of the 10 participants who started the CY intervention completed the intervention and final data collection, for a retention rate of 70%. Participants completed a mean of 14 of the total 16 sessions (87.5% adherence rate), including 3 who attended all 16 sessions. No injury or other adverse events occurred during the intervention. The caregiver assisted the participant to complete the data collection under the RA’s supervision via Zoom meeting.

3.3. Secondary aim: CY intervention

The CY intervention was examined in terms of clinical outcomes (pain interference, mobility, risk of falls, cardiac rhythms, loneliness, and sleep disturbance) at baseline, mid-intervention, and post-intervention. Results of a repeated-measures ANOVA indicated that EL changed significantly over time from baseline through post-intervention, F(1.838, 11.029) = 6.293, p = .016, η² = 0.512, while SL did not change significantly over time, F(1.955, 11.732) = 1.80, p = .208, η² = 0.231. Bonferroni post hoc tests indicated that EL increased significantly from baseline (M = 0.430, SD = 0.535) to post-intervention (M = 1.42, SD = 0.975; see Fig. 2). However, EL did not change significantly from baseline to mid-intervention (M = 1.28, SD = 1.253) or from mid-intervention to post-intervention, although a significant change was observed from baseline to post-intervention.

Fig. 2.

The level of emotional loneliness at three data collection points.

A repeated-measures ANOVA across three data collection points showed no significant differences in outcomes for pain interference, mobility, risk of falls, or sleep disturbance: pain interference, F(2, 5) = 0.237, p = .224, η² = 0.221 (Table 2).

Table 2.

Differences in outcome variables at three data collection points.

Outcomes	Mean (SD)	F	Mean Square	Effect size: Partial Eta- square (η2)	p value
Chronic Pain		.237	21.625	.221	.224
Baseline	10.6 (4.541)
Mid-Intervention	9.3 (2.360)
Post-Intervention	12.1 (4.018)
Timed Up & Go		.516	125.9	.079	.505
Baseline	27.1 (27.3)
Mid-Intervention	21.0 (6.4)
Post-Intervention	23.5 (17.1)
Emotional Loneliness
Baseline	0.43 (0.535)	6.293	2.228	.512	.016
Mid-Intervention	1.28 (1.253)
Post-Intervention	1.42 (0.975)
Social Loneliness		1.800	1.023	.231	.208
Baseline	0.29 (0.488)
Mid-Intervention	1.00 (1.290)
Post-Intervention	0.85 (0.899)
Sleep Frequency		1.022	24.468	.146	.373
Baseline	7.29 (5.648)
Mid-Intervention	4.85 (5.336)
Post-Intervention	7.85 (6.414)
Sleep Disturbance		.857	3.387	.125	.418
Baseline	3.0 (2.309)
Mid-Intervention	2.0 (1.825)
Post-Intervention	3.0 (2.516)

Stability and balance-cardiac regulation indicators were positive. Risk of falling, defined at ≥ 12 s on the TUG test, decreased over the intervention period. Nine participants (81.8%) were identified at baseline and mid-intervention (4 weeks) as having a risk of falling; that number decreased to 7 at post-intervention (8 weeks). TUG test results and fall risk trend suggest that the intervention may have had a positive outcome on mobility and balance.

3.4. Exploratory aim: Remote collection of cardiac data for analyses of RSA and HR

Eight participants originally met eligibility criteria to provide cardiac data from home. Two participants were excluded from cardiac data collection due to having a pacemaker or not having a necessary webcam. All participants provided baseline data in all three postures, consisting of a posture shift from sit, stand, and then return to sit for 3 min in each pose (sit1, stand, sit2). Six of those participants completed mid-intervention data in all three postures and 4 provided post-intervention data in all three postures. Data from one participant was incomplete during the mid-intervention assessment due to equipment error. One participant withdrew from the study after completing the baseline assessment, and 2 participants withdrew after completing the mid-intervention assessment.

Only 4 participants provided complete data. An exploratory analysis was conducted to evaluate whether the physiological response patterns changed significantly by session and by posture. Repeated measures ANOVA (Session [pre/mid/post] x Posture [sit1, stand, sit2]) were calculated for RSA and HR. Descriptive statistics are shown in Table 3. There was a significant condition effect for HR, F(2, 12) = 17.95, p < .003, η² = 0.86, as there was an increase in HR from sit1 to stand, F(1, 3) = 16.25, p < .03, and a decrease in HR from stand to sit2, F(1, 3) = 26.01, p < .02. However, the condition effect for HR was no longer significant when age was added as a covariate. There were no other significant condition effects or interactions for HR or RSA (with or without including age as a covariate).

Table 3.

Descriptive statistics for respiratory sinus arrhythmia (RSA) and heart rate (HR) across three postures at three data collection points (N = 4).

Measure	RSA (Mean, SD)	HR (Mean, SD)
Baseline
Sit1	5.72 (1.90)	71.83 (16.52)
Stand	5.06 (1.21)	80.36 (20.37)
Sit2	5.53 (1.61)	72.81 (17.23)
Mid-Intervention
Sit1	6.25 (1.48)	65.92 (13.65)
Stand	6.21 (1.49)	71.09 (14.88)
Sit2	6.51 (1.68)	64.86 (14.06)
Post-Intervention
Sit1	6.21 (2.75)	67.20 (16.49)
Stand	5.86 (2.40)	75.40 (19.68)
Sit2	6.19 (2.66)	68.18 (18.06)

4. Discussion

The study findings showed that it is feasible to conduct a home-based remotely supervised online CY intervention with older adults with dementia. This finding is important, as older adults with dementia and their caregivers may be challenged in attempts to attend CY programs at community facilities. To the best of our knowledge, this is the first study to focus on evaluating a remotely supervised online CY intervention targeted at older adults with dementia and measuring clinical outcomes virtually under the remote guidance.

This intervention was well accepted by 70% of the dyads, who completed a mean of 14 of the 16 sessions, including 3 who attended all 16 sessions. In a study by Li et al. [54] of an online falls prevention program for older adults with mild cognitive impairment, 75% retention was achieved, similar to the retention rate in this study. The intervention showed strong program fidelity and acceptability by participants and their caregivers. Safety was checked by adverse events associated with CY, such as pain, muscle ache/cramping, fatigue, or any combination of these conditions that occurred during the sessions. No injury or other adverse events occurred during the intervention, indicating that online CY with caregiver support is safe. Although 70% retention was reached, 3 participants dropped the intervention; 2 reported that they were no longer interested in CY because the 60-min sessions were too long for older adults with advanced levels of dementia and their attention span and concentration decreased [55].

The results indicated that this remotely supervised online CY intervention was feasible for this population. Telehealth-based CY intervention was found to be convenient to both participants and their caregivers because it was easily accessible from home and did not require transportation or getting dressed, which reduced caregiver burden and stress [28,56]. Unlike online intervention studies that used a self-report questionnaire [57], this study used objective measures such as the iom2 biofeedback device and TUG test with the caregiver’s assistance to perform the test safely and according to instructions from the RA. This indicated that objective measures could be used by the caregiver under remote supervision by the RA.

While there were no significant findings for pain interference, mobility, sleep, or loneliness longitudinally for participants in the online CY intervention, EL increased significantly from baseline to post-intervention, even though the participants were socially connected to other participants via a videoconferencing platform. A twice-weekly CY session for 8 weeks may not be sufficient for older adults with dementia to decrease isolation. Participants were vulnerable to social isolation and loneliness due to cognitive impairment, communication barriers, and physical deterioration associated with dementia, which could increase the likelihood of mortality from isolation [58-60]. In addition, the pandemic situation made them more socially isolated, which could lead to increasing loneliness because of functional dependence on family members or support by community services [61]. The family caregivers (e.g., spouse or adult child) may have had more burden and stress from caregiving for the person with dementia during the pandemic.

While a significant change in HR was found across three data collection points, this finding was no longer significant when age was included as a covariate, indicating the importance of considering age-related changes in cardiac activity across the lifespan [62,63]. As lack of statistical power may have contributed to nonsignificant findings, including more participants in future studies is important. This would allow for further evaluation of the effectiveness of CY on cardiac measures, including parasympathetic functioning via RSA. A larger sample size would also allow for assessment of vagal efficiency, a measure of the parasympathetic vagal control of HR [64-67] that can be calculated by evaluating change in HR and RSA across postures [67,68]. The ability of current participants to shift postures while providing cardiac data indicates that analyses of vagal efficiency would be feasible.

4.1. Limitations and implications

These findings should be interpreted in light of the study’s limitations. First, this pilot study was limited by a small sample size, consisting mainly of non-Hispanic Whites. Recruitment for the study during the pandemic was a challenge, requiring a variety of methods, such as social service networks, emails, and telephone calls. The study design was a single-arm repeated-measures design. Without a control group, the findings should be viewed with caution. Second, the findings may not be generalizable to older adults and caregivers who do not have access to a computer and the Internet. Third, although the online intervention is convenient for older adults and caregivers, technology issues may reduce motivation to participate in CY. Online CY may have some variability, and it may not confer the same efficacy as in-person professionally administered yoga sessions. Older adults with minimal computer skills may feel uncomfortable in accessing the online CY program, which requires the online class link, signing on to sessions, positioning the computer to view the instructor, and recording data [54]. However, online intervention can help to manage dementia symptoms at home before a crisis situation occurs, which can lessen caregiver burden and stress and improve quality of life for older adults with dementia [56]. The considerable time and cost associated with traveling to in-person yoga sessions over several weeks could be burdensome to patients. In particular, older adults with dementia tend to have reduced access to transportation, which limits participation. Home-based online CY has the potential to reach participants who otherwise would not have access to these interventions.

The study results can inform future research and practice in implementation of online CY or other exercise program for promoting health and wellness in older adults with dementia living at home. For future studies, a randomized control trial design with a control group and a larger sample with effect size would provide a higher level of empirical evidence to support use of the online CY program. Future clinical trials should be performed to examine the effects of online CY intervention for older adults with dementia on health outcomes and to assess the longterm sustainability of online CY intervention (e.g., 6 months post-intervention). Moreover, pairing CY with other nonpharmacological interventions may reduce clinical symptoms and increase efficacy.

It is important for caregivers to have proper training in accessing the online intervention and in collecting cardiac data and providing technical assistance as needed. Some caregivers with minimal computer skills may have had lack of confidence in accessing an online CY program that requires a class link, signing on to sessions, positioning the computer to view the instructor, and recording data at the beginning of CY intervention. Results of the study demonstrated the feasibility of training caregivers to collect cardiac data and psychosocial data in the home under the remote guidance of a trained RA. However, caregivers face some physical and cognitive challenges. Caregivers who are also older adults are likely to have age-related memory loss and have trouble with remembering passwords [69]. Deteriorating fine motor control could make it difficult to use a mouse or touch pad [69].

To minimize technology challenges in older adults with dementia and their caregivers in using a device measuring cardiac rhythms, technical support and resources should be provided during data collection and CY intervention. This telehealth approach can be used as a monitoring and coaching strategy for implementation of home-based CY and measurement of physiological data. In this study, the RA provided a manual with step-by-step procedures and pictures and had technical sessions with each caregiver and participant to assist in downloading the application and connecting earlobe optical sensor, pulse processor, and USB cable to the participant’s home computer via an USB interface. With this personalized technical support, caregivers in this study had no difficulty in reporting cardiac data. For future research, the process should be user friendly. A back-up plan with useful troubleshooting options should be developed to minimize technological challenges and establish an alternative communication plan, such as a video call in the smart phone that can be used in the event of challenges. Provision of computer hardware (with large screen), Wi-Fi connection, and manual addressing technological issues is recommended [69]. Time should be allotted for participants and caregivers to interact before each CY session to build rapport and relationships.

5. Conclusion

The present study suggests that remotely supervised online CY is a feasible approach for socially isolated older adults with dementia. This technology-based intervention could allow socially isolated older adults who are living at home, especially those in underserved communities where people are becoming more digitally connected, to receive remotely supervised CY. Remotely collected cardiac data and psychosocial data can provide a more complete assessment of the effects of an intervention. Online CY classes provide one means of reducing health disparities by providing access to interventions for persons who are unable to travel to a clinic or facility.