A Cluster Randomized Trial of Tai Chi vs Health Education in Subsidized Housing: the Mi-Wish Study

Abstract

Background/Objectives:

Tai Chi (TC) may benefit older adults with a variety of diseases and disabilities. We tested the hypothesis that TC improves physical function in older adults living in low-income housing facilities.

Design:

Cluster randomized controlled trial.

Setting:

Subsidized housing facilities in Boston, Massachusetts and neighboring communities.

Participants:

Volunteers were recruited from 15 facilities. The 180 randomized participants were ≥ 60 years of age, able to understand English and participate in TC, expected to remain in the facility for 1 year, and able to walk independently.

Intervention:

Tai Chi classes were conducted in the housing facilities twice weekly for 1 year and compared to monthly health promotion educational classes and social calls.

Measurements:

The primary outcome was physical function measured by the Short Physical Performance Battery (SPPB). Secondary outcomes included other aspects of physical and cognitive function, and falls.

Results:

An interim analysis revealed less improvement over 12 months in SPPB scores among TC participants (+0.20 units, 95% CI −0.20 to +0.60, p = 0.69) vs. control participants (+0.51 units, 95% CI +0.15 to +0.87, p = 0.007), a difference of −0.31 units (95% CI −0.66 to 0.04, p = 0.082). This met the criterion for futility, and the Data Safety Monitoring Board recommended trial termination. There were no differences in 6- or 12-month changes favoring TC in any secondary outcomes or adverse events.

Conclusions:

In older adults with multiple chronic conditions living in subsidized housing facilities, 6 and 12 months of twice-weekly TC classes were not associated with improvements in functional health.

INTRODUCTION

With the growth of the older population, the number of frail, older adults living in low-income housing is rapidly expanding.¹ Health promotion activities provided in these facilities are variable and insufficient to reduce the large burden of disease, disability, and associated costs that afflict this large segment of the population.² Previous studies have shown multiple benefits of Tai Chi (TC) in older adults with a variety of disabilities and diseases;^3–10 however, they have rarely targeted those living in subsidized housing. The multisystem effects of TC exercise suggest that it may improve health and reduce health care utilization in older adults at risk of functional decline. We tested this possibility by conducting a cluster randomized controlled trial called the Mi-WISH Study (Mind-Body Wellness In Subsidized Housing) of twice-weekly TC classes (with interim video reinforcement) compared to monthly health promotion educational classes and social calls for up to one year in low-income senior housing facilities located in Boston, Massachusetts and neighboring communities. Our specific aims were to test the effect of TC vs. Health Education (HE) interventions on: 1) physical function as measured by the Short Physical Performance Battery (SPPB), and 2) health care utilization and cost. The results of the first aim are presented here.

METHODS

Study Subjects:

The study methods have been previously described.¹¹ We planned to enroll 320 subjects from 16 subsidized housing facilities, block randomized to TC or HE interventions, with 8 facilities in each arm. Eligible facilities included low-income state- and federally-funded housing developments for older residents. Subject inclusion criteria included living in a selected housing facility, age ≥ 60, able to understand instructions in English, able to participate safely in TC exercises at least twice a week, expected to remain in the facility for 1 year, and able to stand and walk independently.

Subjects were excluded if they were already participating in TC exercises or had any unstable or terminal illness, were unable to maintain posture sitting or standing, were unable to hear, see, or understand TC instructions and assessment questions. We screened for cognitive ability over the telephone and accepted those who could recall at least 2 of 3 words in 3 minutes, and 2 of 3 elements of the study protocol.¹¹

Interventions:

Core TC training components included: gentle dynamic stretching and strengthening, slow integrated weight-bearing movements, efficient posture, heightened body awareness and inner focus, mindful diaphragmatic breathing, and healing imagery.¹² Two classes were delivered at set times each week by TC instructors in a designated community room within each housing facility for up to 52 weeks.

TC exercises included 9 core movements following the traditional Cheng Man-Ch’ing’s Yang-style.¹³ These 9 movements were progressively added to warm-up and cool down exercises over the first 24 weeks of the intervention.

Eight experienced TC instructors administered the intervention under the direction of a principal investigator (PW), who provided training sessions prior to the start of the trial, systematically observed classes to assess fidelity of treatment, and provided ongoing feedback to instructors. Two instructors were assigned to each housing facility. Participants were given practice DVDs, a well-illustrated training manual, and instructions for a minimum of 20 minutes of daily home practice on 3 non-class days each week.

To control for the effects of social interaction experienced during the group-based TC intervention, we utilized a health education (HE) intervention in which participants attended monthly educational group sessions within a common area of each housing facility. Sessions were led by research personnel and content experts and included material from education forms produced by the American Geriatrics Society. Sessions contained 30 minutes of lecture and 30 minutes of group discussion. Every participant in both arms received monthly social calls. The same research assistant called each participant to establish a social connection and acknowledge life events.

Randomization Procedures:

Housing facilities were paired by size and municipality. Within each pair, facilities were randomized 1:1 to TC or HE.¹¹

Blinding/concealment:

Participants were recruited and baseline assessments were completed prior to randomization. During follow-up, assessors were blinded to the participant’s treatment arm by concealing their site’s intervention assignment, making sure that assessment questions did not refer to a particular intervention, and scheduling assessments on days that the intervention was not taking place.

Study Variables:

The primary dependent variable was the SPPB total score,¹⁴ which showed significant improvement during our pilot study.¹⁸ Secondary outcomes included other aspects of physical function (i.e., the timed-up-and go (TUG) test, gait, standing balance, grip strength, self-reported physical activity, functional capacity), cognition (i.e., executive function, memory), person-centered measures (i.e., balance confidence, health-related quality-of-life, depression, exercise self-efficacy, satisfaction with the interventions, and expectancy for improvement), and falls. Falls were defined as any event in which the participant unintentionally came to rest on the ground or other lower level, not as a result of a major intrinsic event or an overwhelming external hazard.¹⁵ Falls, emergency visits, hospitalizations, and adverse events were recorded during monthly calls, at the beginning of every TC or HE session, and from participant’s reports on the MiWiSH hotline, throughout the intervention period. We also collected demographic variables and covariates, including: age, sex, race, primary language, income, education, comorbidity score, body mass index, and history of falls. Attendance at TC or HE classes was recorded, and participants randomized to TC recorded hours of home practice.

Data Analysis:

The effect of the TC intervention on SPPB was estimated from a shared-baseline linear mixed model with fixed effects for age, visit, age x visit interaction, and treatment x post-baseline visit interaction and random effects of site, participant within-site, site × time, and participant × time interactions. The shared baseline across treatment groups reflects homogeneity of the population with respect to treatment prior to random assignment and adjusts for effects of baseline level in a manner equivalent to ANCOVA.¹⁶ The random effects acknowledge the expected correlations among participants within a site and repeated longitudinal assessments of each participant. The model accommodates unequal enrollment across sites and loss to follow-up. All participants were included in the primary analysis according to their treatment assignment, following the intention-to-treat principle. Secondary continuous outcomes were analyzed using the same model. Counts of falls were analyzed in a similar mixed effect Poisson regression with offsets for log-time at risk. Nominal p-values and step-down Bonferroni-adjusted p-values were calculated for analyses of secondary outcomes.

Interim analysis:

One interim analysis for efficacy, futility, and sample size re-estimation was planned after participants enrolled in the first 8 of 16 facilities had completed their 12-month assessment. The efficacy threshold for stopping the trial was based on a two-sided Haybittle-Peto boundary at alpha=0.002 for the 12-month change in SPPB total score. A non-binding proposal to stop the study for futility was based on a beta spending rule, linear in information time with a final beta=0.15. Results of the efficacy and futility analysis were shared between an unblinded statistician and the Data Safety Monitoring Board (DSMB). Because the futility boundary was reached on the interim analysis, all available data were included in a single final analysis with treatment effects on SPPB. Due to early termination of the study, we examined efficacy outcomes primarily at 6 months, when the data available for analysis were most complete.

Protection of Human Subjects:

The protocol was approved by the Hebrew SeniorLife Institutional Review Board and all participants provided written informed consent. The study was registered at Clinicaltrials.gov ().

RESULTS

Between August 3, 2015 and October 20, 2017, recruitment activities were launched at 15 sites. Of nearly 600 individuals pre-screened, 354 provided written informed consent and were assessed for eligibility (Figure 1). Of those, 182 were determined eligible. One site with only two participants was abandoned due to poor recruitment prior to randomization. The other 14 sites, including 180 participants, were randomized. The first 8 sites were the basis for the interim analysis.

Figure 1.

CONSORT Flow Diagram – The Mi-WiSH Study

Participants were well matched across treatments (Table 1), with mean (± standard deviation) age of 75 ± 9 years, average cognition (MMSE), moderate functional impairment (SPPB), low mobility (PASE, 30% lower than the average in the PASE validation study)¹⁷, normal gait velocity, and an 8% average decline in velocity during the dual task of counting backwards. Participants reported 5.1 ± 2.7 health problems, with 48% reporting a fall and 15% reporting loss of more than 10 pounds over the previous year, percentages typical of a frail older adult sample and similar to those of the pilot trial.¹⁸

Table 1.

Baseline participant characteristics, mean ± SD (range) or number (percentage)

			Treatment Assignment
Variable	Level	Overall (n = 180)	Health Educ (n = 87)	Tai Chi (n = 93)	P Value
Age at enrollment		75.3±8.8 (59.6, 98.9)	74.6±8.6 (59.6, 95.0)	75.9±9.1 (59.8, 98.9)	0.36
Female Sex		120 (66.7%)	57 (65.5%)	63 (67.7%)	0.75
Hispanic/Latino		8 (4.4%)	3 (3.4%)	5 (5.4%)	0.72
Race	Asian	4 (2.2%)	2 (2.3%)	2 (2.2%)	0.013
	Black/African American	58 (32.2%)	23 (26.4%)	35 (37.6%)
	Caucasian/White	112 (62.2%)	62 (71.3%)	50 (53.8%)
	Other/Refused/Unknown	6 (3.3%)	0 (0.0%)	6 (6.5%)
Any falls in previous year		84 (47.5%)	39 (45.3%)	45 (49.5%)	0.65
Number of falls in the previous year		0.98±1.75 (0.0, 15.0)	0.79±1.06 (0.0, 4.0)	1.15±2.20 (0.0, 15.0)	0.17
Heart disease		25 (13.9%)	14 (16.1%)	11 (11.8%)	0.52
Atrial fibrillation		29 (16.5%)	12 (14.0%)	17 (18.9%)	0.42
Angina		21 (11.7%)	10 (11.6%)	11 (11.8%)	1.00
Hypertension		120 (67.0%)	58 (66.7%)	62 (67.4%)	1.00
Hypercholesterolemia		92 (53.2%)	46 (54.1%)	46 (52.3%)	0.88
Diabetes mellitus		59 (33.0%)	28 (32.6%)	31 (33.3%)	1.00
Ulcers or GERD		56 (31.5%)	29 (33.7%)	27 (29.3%)	0.63
Anemia		31 (17.3%)	18 (20.7%)	13 (14.1%)	0.32
Cancer (excl. skin cancer)		41 (22.8%)	25 (28.7%)	16 (17.2%)	0.076
Depression		60 (33.3%)	26 (29.9%)	34 (36.6%)	0.43
Osteoarthritis		87 (48.6%)	42 (48.8%)	45 (48.4%)	1.00
Spinal stenosis or disc disease		34 (19.2%)	16 (18.8%)	18 (19.6%)	1.00
Gout		23 (12.8%)	13 (14.9%)	10 (10.8%)	0.50
Asthma or COPD		43 (24.0%)	25 (29.1%)	18 (19.4%)	0.16
Eye disease		102 (57.6%)	50 (58.1%)	52 (57.1%)	1.00
Health problems (number†)		5.1±2.7 (0.0, 16.6)	5.2±2.7 (1.0, 16.6)	5.0±2.7 (0.0, 14.0)	0.68
BMI (kg/m2)		29.3±7.0 (15.1, 57.2)	29.1±6.1 (16.3, 50.3)	29.6±7.8 (15.1, 57.2)	0.64
20Q CES-D total score		11.9±8.2 (0.7, 42.9)	11.5±8.0 (0.7, 42.9)	12.3±8.4 (0.7, 33.2)	0.55
MMSE total score		25.7±3.6 (12.0, 30.0)	26.0±3.8 (12.0, 30.0)	25.5±3.3 (16.0, 30.0)	0.40
SF-12 Physical Component Score		40.7±10.8 (18.5, 58.9)	40.8±10.9 (18.5, 58.5)	40.6±10.7 (18.5, 58.9)	0.90
SF-12 Mental Component Score		52.1±9.8 (21.0, 71.3)	52.3±10.0 (21.0, 66.6)	51.8±9.8 (26.5, 71.3)	0.74
SPPB total score		8.14±2.82 (1.0, 12.0)	8.36±2.89 (1.0, 12.0)	7.92±2.76 (1.0, 12.0)	0.30
TMT B-A difference		118±75.3 (‒11, 274)	121±73.0 (0.0, 274)	116±77.7 (‒11, 258)	0.64
ABC score		71.7±23.1 (3.1, 100)	73.6±22.2 (8.8, 100)	69.9±23.9 (3.1, 100)	0.29
PASE total score		77.5±48.6 (0.0, 389)	83.3±46.3 (2.3, 198)	72.1±50.4 (0.0, 389)	0.13
TUG time (s)		15.7±7.7 (7.2, 45.0)	15.1±7.8 (7.2, 43.9)	16.2±7.6 (7.6, 45.0)	0.34
Gait NW velocity (m/s)		1.01±0.25 (0.26, 1.53)	1.04±0.26 (0.26, 1.53)	0.99±0.25 (0.37, 1.44)	0.18
Gait DT velocity (m/s)		0.93±0.26 (0.22, 1.48)	0.96±0.26 (0.22, 1.48)	0.90±0.26 (0.31, 1.39)	0.11

Abbreviations:

GERD - Gastroesophageal reflux disease; COPD - Chronic Obstructive Pulmonary Disease; CES-D - Center for Epidemiological Studies Depression Scale; MMSE - Mini-Mental State Examination; SF-12 – 12-Item Short Form Health Survey; SPPB - Short Physical Performance Battery; TMT - Trail Making Test; ABC - Activities-specific Balance Confidence Scale; PASE - Physical Activity Scale for the Elderly; TUG - Timed Up & Go Test; NW - Normal Walking Condition; DT - Dual Task Condition.

^†mean imputation of missing data for up to 20% of conditions queried. Others not listed in Table 1 were: pacemaker implanted, congestive heart failure, kidney disease, liver disease, stroke, Parkinson’s disease, multiple sclerosis, Alzheimer’s disease, epilepsy, syncope in past year, any health problem, and rheumatoid arthritis.

The interim analysis revealed less improvement over 12 months in SPPB among participants at sites randomized to TC (+0.20 units, 95% CI −0.20 to +0.60, p=0.69) vs. HE (+0.51 units, 95% CI +0.15 to +0.87, p=0.007), a difference of −0.31 units (95% CI −0.66 to 0.04, p=0.082). This was nearly 1 unit below the threshold of 0.56 units in favor of TC that was recommended as a non-binding futility criterion. After reviewing the interim analysis, the DSMB recommended that the trial be terminated. The National Institute of Aging, the study sponsor, accepted the DSMB’s recommendation but permitted the completion of 6-month follow-up visits at the 5^th and 6^th pairs of sites where participants were already enrolled and receiving study interventions. Given the availability of data on 6-month follow-up from 6 of the planned 8 pairs of sites as opposed to only 4 pairs with 12-month follow-up, the Steering Committee voted to shift the primary endpoint from 12-month change to 6-month change prior to breaking the blind for any secondary outcomes. While all 180 randomized participants were included in analyses, only 144 participants from 6 pairs of matched facilities received the TC or control intervention. In total, 135 participants enrolled at the active sites completed their 6-month follow-up. Of these, only 67 participants from the first 8 sites completed the 12-month assessment (Figure 1).

There were no statistically significant differences in 6-month change in any of the clinical, functional, cognitive, or affective outcome measures between participants at sites randomized to TC vs. HE (Table 2). In fact, participants receiving TC tended to experience less improvement in SPPB score compared to participants receiving HE. The upper 95% confidence bound for 6-month change in SPPB is less than one-quarter of the minimum clinically important difference of 2 units. In the subset followed for 12-months, those receiving TC experienced significantly greater increases in Trail Making B times (adjusted for Trails A times; +14 sec for TC vs. −15 sec for HE, p=0.017) and significantly greater decreases in verbal fluency for animal names (−0.58 names for TC vs. +1.08 names for HE, p=0.041).

Table 2.

Outcome analyses, mixed model estimates ± SE (95% confidence interval)

	6-month Change		Tai Chi vs. Health Educ
Measure	Health Educ (n = 87)	Tai Chi (n = 93)	Difference (Tai Chi minus Health Education)	P Value
SPPB total score	0.18±0.19	0.05±0.18	‒0.123±0.248 (‒0.62, 0.37)	0.62
20Q CES-D total score	‒0.47±0.74	‒0.06±0.69	0.404±1.002 (‒1.60, 2.41)	0.69
SF-12 Physical Component Score	‒0.64±1.04	0.22±0.99	0.854±1.429 (‒2.00, 3.71)	0.55
SF-12 Mental Component Score	‒0.71±1.19	‒1.81±1.16	‒1.104±1.655 (‒4.41, 2.21)	0.51
TMT B-A difference	‒9.55±7.37	‒8.19±6.91	1.362±9.276 (‒17.2, 19.9)	0.88
ABC score	‒2.02±1.87	‒0.11±1.77	2.136±2.560 (‒2.98, 7.26)	0.41
PASE total score	‒4.14±5.37	‒3.42±5.17	0.720±7.291 (‒13.9, 15.3)	0.92
Gait NW velocity (m/s)	0.01±0.02	0.01±0.02	0.004±0.023 (‒0.04, 0.05)	0.85
Gait DT velocity (m/s)	‒0.00±0.01	0.02±0.01	0.022±0.020 (‒0.02, 0.06)	0.27

Baseline age was a strong predictor of physical function. While recognizing that physical function does not decline strictly linearly with increasing age, participants’ SPPB total scores were 0.77 units lower on average at baseline for each additional decade of age (95% CI 0.15 to 1.39, p = 0.015). Nevertheless, age had very little effect on the average magnitude of change in SPPB total score over 6 months (0.02 units greater decline per decade of age, 95% CI −0.33 to +0.29, p = 0.90). In addition, age had very little association with the estimated effect of TC treatment relative to HE on 6-month change in SPPB total score (0.19 units different between older and younger participants based on a median split, 95% CI −0.97 to +1.36, p = 0.75).

Seventy-two percent of TC participants attended 50% or more of scheduled sessions over the first 3 months of the intervention, while 84% of HE participants attended 50% or more of their classes. Those participating in TC at 3, 6, and 12 months attended 68%, 68%, and 64% of scheduled classes, while those participating in HE classes attended 76%, 78%, and 75%.

Musculoskeletal complaints and falls were the most commonly recorded treatment-emergent adverse events (Table 3), with numerically higher by not significantly higher rates reported by TC participants. The overall rate of falls was 0.46/patient-year among participants at sites randomized to TC vs. 0.26/patient-year among those randomized to HE, an incidence rate ratio of 1.76 (95% CI 0.58 to 5.31, p=0.25). The rate of falls was lower among TC participants who reported more hours of practice (−11%/100 hours of class time and practice, p=0.002), but was unassociated with reports of musculoskeletal adverse events (p=0.94).

Table 3.

Treatment-emergent adverse events by MedDRA system organ class and preferred term

		Health Educ			Tai Chi			Overall
MedDRA System Organ Class	Preferred Term	# Evts	# Sbjs	% Sbjs	# Evts	# Sbjs	% Sbjs	# Evts	# Sbjs	% Sbjs
Overall	Overall	35	22	33%	79	53	69%	114	75	52%
Injury, poisoning and procedural complications	Overall	16	11	16%	27	21	27%	43	32	22%
	Fall	16	11	16%	26	21	27%	42	32	22%
Musculoskeletal and connective tissue disorders	Overall	6	4	6%	33	29	38%	39	33	23%
	Arthralgia	0	0	0%	5	5	6%	5	5	3%
	Musculoskeletal discomfort	1	1	1%	21	18	23%	22	19	13%
Nervous system disorders	Overall	0	0	0%	6	6	8%	6	6	4%

Note:

# Evts = total number of reported events of a given classification, # Sbjs = total number of subjects reporting at least one instance of a given classification, % Sbjs = percentage of subjects who initiated treatment who experienced at least one instance of a given classification.

Only system organ classes or preferred terms reported by >5% of participants in a given treatment group are reported

In secondary analyses, we examined the characteristics of the subsets of participants who demonstrated improvements in functional or gait outcomes over 6 months, defined as changes that exceeded 1 unit for SPPB or 0.05 m/s for dual-task walking velocity. Participants experiencing improved SPPB scores over 6 months after exposure to TC vs. HE were more likely to be African American (p=0.007) and have lower verbal fluency (p=0.018) and lower PASE scores (p=0.020) at baseline. Participants experiencing improved dual-task walking velocity over 6 months after exposure to TC vs. HE were more likely to be slow to stand (p=0.012) and walk (p=0.014) at baseline.

DISCUSSION

This is the first cluster randomized trial of a 12-month TC intervention in older adults living in subsidized housing facilities – a population with multiple co-morbidities, high health care utilization, and great need for lifestyle interventions that can promote physical and mental health. On the basis of previous TC intervention studies for a variety of health conditions^5,19–21 and a pilot study by us in senior housing facilities,¹⁸ we hypothesized that this multimodal exercise, delivered by experienced trainers, would improve functional status (SPPB scores) and quality of life in this cohort of older adults. However, the results of this study did not support our hypothesis.

Many clinical trials evaluating the effects of TC training on balance, mobility, physical function, and fall prevention in older adults have demonstrated clinically-meaningful, and sometimes quite large benefits.^21–24 However, a small number of well-designed and adequately powered studies indicate no apparent benefit of TC.^25–28 Explaining lack of effectiveness in any one of these trials, including the present study, is difficult, because comparisons of interventions like TC are likely confounded by differences in study design, eligibility criteria, interventions (i.e., TC style and specific forms), control groups, duration and intensity of exposure, quality and fidelity of instruction, and intervention adherence rates.^29–31

Our previous pilot study targeted a similar population of frail older adults living in senior housing facilities and demonstrated significant improvements in SPPB (11%) and TUG (7%) scores among participants receiving TC training over just 12 weeks.¹⁸ The Mi-WiSH trial was similar to our pilot study in utilizing nearly identical TC training and educational interventions, including many of the same instructors. However, primary outcomes were assessed over a much shorter time interval in the pilot study (12 weeks vs. 6 to 12 months). Average age was somewhat greater in our pilot study (86.5 vs. 75.4 years), but it is unlikely that being younger explains our null findings in the Mi-WiSH trial. The positive effects of TC on balance and mobility have been observed in adults across a wide range of ages, including those younger and older than Mi-WiSH participants.²¹

Low class attendance may contribute to null findings, resulting in a low exposure or ‘dosage’ of TC.²⁵ Over the first six months of the Mi-WiSH study, average participant attendance in TC classes was 68% and dropped to 64% at 12 months. This contrasts with an average 83% attendance rate in our pilot study.¹⁸ As expected, studies with very low adherence rates (e.g., 24%)²⁶ have reported null findings, and those with higher adherence rates (e.g., 78%)³² have reported positive outcomes. However, numerous large-scale studies with attendance rates similar to or lower than ours have reported significant benefit from TC.^33–36

Multiple factors underlie poor health outcomes in people living within subsidized housing, including poverty, social stresses, environmental influences, psychological issues, and biological predispositions to negative health outcomes.^37–39 The observed ineffectiveness of TC in Mi-WiSH may have resulted from the unique sociodemographic characteristics of participants and a failure to differentially target critical environmental, social, and psychological determinants of health. Although higher levels of frailty may be associated with reduced responsiveness to TC,⁴⁰ positive outcomes have been reported in multiple studies that included frail adults.^41–45 Moreover, our exploratory analyses of Mi-WiSH data did not find frailty or sociodemographic factors to be associated with a worse response to TC.

The cluster-randomized design and early termination of our study may have resulted in a sample size that was too small to demonstrate statistically significant effects for our secondary outcome measures. If so, we would have expected to observe consistent trends toward benefit from TC in similar functional or cognitive variables, but this was not the case. In fact, the upper confidence bound on our estimate of functional improvement excludes a clinically significant benefit from TC, arguing against imprecision as a cause of our null finding.

Tai Chi is believed to be a low-impact and safe exercise, and a recent systematic review concluded that serious adverse events attributable to TC are rare.⁴⁶ Nevertheless, TC is expected to result in occasional minor musculoskeletal complaints which could deter adherence and reduce effectiveness. In Mi-WiSH, TC was associated with more musculoskeletal complaints compared to the more sedentary HE intervention. Less expected in Mi-WiSH was a higher (but statistically insignificant) rate of falls in the TC group. This is similar to findings in one large RCT,²⁵ but contrary to a larger and more recent trial.³² Fall rate was not associated with musculoskeletal complaints and was lower among TC participants who reported more hours of practice. The higher fall rate in the TC group could reflect closer tracking of adverse events at twice weekly vs. monthly sessions in the HE group.

In this cluster randomized pragmatic study of a diverse cohort of older adults with multiple chronic conditions living in subsidized housing facilities, 6 to 12 months of twice-weekly TC classes was not associated with improvements in specific measures of functional or cognitive health. Our findings do not exclude the possibility that TC may still confer emotional, social, and health benefits that were not measured in this study. Given its promise, popularity and value as a low-impact, mind-body, and socially rich activity, further research in subsidized housing populations is warranted.