Effectiveness of a Disability Preventive Intervention for Minority and Immigrant Elders: The Positive Minds-Strong Bodies Randomized Clinical Trial

Abstract

Objective

To test the acceptability and effectiveness of a disability prevention intervention, Positive Minds-Strong Bodies (PMSB), offered by paraprofessionals to mostly immigrant elders in four languages.

Design

Randomized trial of 307 participants, equally randomized into intervention or enhanced usual care.

Setting

Community-based organizations in Massachusetts, New York, Florida, and Puerto Rico serving minority elders. Data collected at baseline, 2, 6, and 12 months, between May 2015 and March 2019.

Participants

English-, Spanish-, Mandarin-, or Cantonese-speaking adults, age 60+, not seeking disability prevention services but eligible per elevated mood symptoms and minor to moderate physical dysfunction.

Interventions

10 individual sessions of cognitive behavioral therapy (PM) concurrently offered with 36 group sessions of strengthening exercise training (SB) over 6 months compared to enhanced usual care.

Measurements

Acceptability defined as satisfaction and attendance to >50% of sessions. Effectiveness determined by changes in mood symptoms (HSCL-25 and GAD-7), functional performance (SPPB), self-reported disability (LLFDI), and disability days (WHODAS 2.0).

Results

77.6% of intervention participants attended over half of PM Sessions; 53.4% attended over half of SB sessions. Intent-to-treat analyses at 6 months showed significant intervention effects: improved functioning per SPPB and LLFDI, and lowered mood symptoms per HSCL-25. Intent-to-treat analyses at 12 months showed that effects remained significant for LLFDI and HSCL-25, and disability days (per WHODAS 2.0) significantly decreased 6-months after the intervention.

Conclusions

PMSB offered by paraprofessionals in community-based organizations demonstrates good acceptability and seems to improve functioning, with a compliance-benefit effect showing compliance as an important determinant of intervention response.

Trial Registration

ClinicalTrials.gov:

INTRODUCTION AND OBJECTIVE

Older adults (65+) in the U.S. are rapidly becoming more diverse, with one third of elders expected to belong to racial/ethnic minority groups by 2040 (1); more than one fifth with limited English proficiency (2); and 12% immigrants (3). As the number of individuals aged 65+ rises to 81 million by 2040 (1), nearly 20% will suffer from one or more mental health conditions (4), with mood disorders the most important risk for premature disability (5). Sociomedical models define disability according to functioning, which is influenced by mental and physical factors (6). Disability is the extent to which chronic and acute conditions affect functioning in body systems, physical and mental actions, and activities of daily life (6). Compared with non-depressed elders, depressed elders have a relative risk of 1.67 and 1.73 for incident disability in activities of daily living and mobility, respectively (7). Black (8), Asian (9) and Latino elders (10) are at particular high risk for disability as well as immigrant elders, who exhibit worse mental health than native-born elders (11). Yet, minority and immigrant elders, particularly Asian (12, 13) are less likely to obtain disability prevention services (10, 14) than their white counterparts. This is a missed opportunity, given that mood disorder treatment has been shown to reduce disability days by 40%−45% in those with severe to moderate depression (15).

Traditional approaches that address disability in clinical or primary care contexts have failed to reduce disparities in treatment access (14) and quality of care (16). Usual clinic-based services do not incorporate evidence-based mental health interventions, (17) lack patient compliance (17, 18), and lack resources for training and supervision (19). Yet, evidence-based approaches exist for improving coping (by treating mood symptoms; 20) and for functional restoration (by improving mobility; 21). The addition of a functional disability prevention component can reduce mental health symptoms (22) and build resistance or leg power, as power correlates better than strength with functional activities (23). But this combination of treating mood symptoms plus functionally oriented exercises for preventing disability progression has not been tested in the community. Alternative models of care for preventing disability (24) are vital to contend with minority and immigrant elders’ limited recognition of mood symptoms (25), stigma (26), limited English proficiency (27), and reduced access to professional care (28).

A combined cognitive-behavioral therapy-based (CBT) and exercise program is an ideal approach, given extensive literature on the effectiveness of CBT, such as Healthy IDEAS (29) and IMPACT (30), for better coping with mood symptoms; and exercise training for improving physical functioning in older, minority adults (31–33). Yet, these are not typically offered in diverse languages or for varied cultural groups. Implementation poses challenges, particularly workforce shortages (34) to deliver evidence-based care (4) in languages other than English (35). Community health workers (CHWs) have been used to deliver evidence-based treatments (36), tackle personnel shortages (36), increase diversity, and offset the lack of bilingual/bicultural clinicians (37). However, few studies have worked with CHWs to provide a psychosocial intervention tailored to minority and immigrant elders in the US (38). Expanding capabilities at elder-serving Community-based organizations (CBOs) through CHWs may be an option to reduce service disparities (39).

We evaluated the acceptability and effectiveness of a disability prevention intervention - Positive Minds-Strong Bodies (PMSB) - conducted by CHWs and exercise trainers for minority and immigrant elders in CBOs. We hypothesized that elders who received PMSB would improve in mood symptoms and decelerate deterioration in physical functioning compared to those receiving enhanced usual care.

METHODS

Trial Design

This randomized controlled clinical trial tested concurrent mental health (Positive Minds-PM) and physical functioning through exercise training (Strong Bodies-SB) intervention sessions (see Methods 1, Supplemental Digital Content [SDC]). It was approved by the institutional review boards for Massachusetts General Hospital/Partners HealthCare and New York University, with ceded reviews for partnering CBOs (see Table 5, SDC) conducting human subjects research. All participants provided informed consent.

Participants

Eligible participants were 60+ years of age; fluent in English, Spanish, Cantonese, or Mandarin; and scored 5 or more on either the Patient Health Questionnaire (PHQ-9; 40), the Generalized Anxiety Disorder 7-item Scale (GAD-7; 41), or the Geriatric Depression Scale (GDS-15; 42). They also scored between 3 and 11, representing minor to moderate disability, on the Short Physical Performance Battery (SPPB; 43). Participants were excluded if they disclosed substance use disorders; received mental health treatment within the prior three months or had an appointment within the next month; lacked capacity to consent (44); were homebound; had a neuro-musculoskeletal impairment; or their physician did not provide medical clearance for exercise or advised against it. Regular exercise was not an exclusion criteria and was examined across assessments (see Table 6, SDC). If potential participants scored 4 or 5 on the Paykel suicide questionnaire, they were referred to emergency services.

Participants were linked to CBOs and community clinics serving low-income minorities or immigrants in Massachusetts, New York, Florida, or Puerto Rico. Sessions were held at these locations, in participants’ homes, or, for PM, by phone. All PM sessions were one-on-one, whereas SB sessions often involved 2–3 participants.

Procedures

Potential participants (N=1,057) at partner CBOs were screened for mood symptoms and physical functioning and compensated with a $10 gift card. ‘ Participants doing regular exercise were not excluded, and information regarding current fitness service use was measured across assessments. Eligible participants (N=381; 36.0%) were invited to complete a 2-hour baseline assessment ($25 compensation). In total, 307 (80.6%) accepted and were randomized to intervention or control conditions (see Figure 1, CONSORT diagram). Participants took part in 2- and 6-month follow-up interviews ($25 compensation) and a final (post-intervention) 12-month follow-up ($50). Interviewing research assistants (RAs) were blinded to participant study condition. All interviews were audiotaped. To ensure standardized administration, we conducted quality control of each RA’s first two interviews and a random sample of 15% of interviews.

Figure 1.

Study flow chart (CONSORT Diagram)

Interventions

PM (adapted from the CERED intervention; 45), is a psychosocial intervention consisting of 10 one-hour individual sessions over a 6-month period, focused on psychoeducation, mindfulness, cognitive restructuring, noticing and overcoming unhelpful thoughts, and creating a self-care plan (see Methods 1, SDC). A five-stage approach was used to structure the cultural and linguistic adaptation process (46; see Methods 3, SDC). The PM sessions were administered by CHWs recruited in collaboration with Site Leaders and trained by the primary author and licensed supervisors (see Methods 2, SDC). Training involved two intensive days of workshops on core intervention elements and delivery, followed by four months of CHW role plays for all 10 sessions and 75% minimum fidelity checks.

SB, offered concurrently with PM, was adapted from the Increased Velocity Exercise Specific to Task (InVEST) (21) exercise training program for enhancing functioning and preventing physical disability in elders. Participants engaged in approximately three SB group sessions per week over 12–14 weeks for a total of 36 sessions. Sessions consisted of a series of 10 functional exercises with three sets of 10 repetitions each (47) administered by exercise trainers supervised by the second author.

The control condition offered three components of enhanced usual care: A call by research staff every two weeks to administer the PHQ-9, GAD-7, and the 5-item suicide questionnaire, thus mimicking administration in the intervention group; empathetic support if the participant expressed concern; and an NIH booklet about caring for one’s mental and physical health.

Material preparation, intervention protocols, and weekly supervision were conducted in English, Spanish, Cantonese, or Mandarin. Our culturally diverse clinical supervisors, community advisory group, and research team evaluated the cultural equivalence of our intervention and tracked changes in wording or presentation, to compare understanding across groups.

Supervision, Fidelity, and Compliance to PMSB

All PM sessions were audio recorded. Each CHW’s first two sessions and 15% of each CHW’s sessions were randomly reviewed for intervention fidelity using a checklist of session activities in weekly supervision with licensed clinical supervisors. The SB sessions were regularly videotaped, with 8.85% of sessions checked for implementation quality. CHWs and exercise trainers received biweekly supervision and feedback on fidelity and issues of clinical significance. PM session fidelity was 80.5% (ranging from 76.9% to 83.7%); whereas fidelity for SB sessions was 65.9% (ranging from 65.2% to 66.7%).

Outcome Measures

We translated and back translated the standard English versions of our primary outcome measures (see Table 1, SDC). When the back translation revealed ambiguities, we engaged a multinational panel of experts to resolve them. As our study addressed disability in a sociomedical model (6), we operationalized disability as the presence of elevated mood symptoms for depression (PHQ-9 scores; 40) and/or anxiety (GAD-7 and GDS-15 scores; 41, 42) together with mild to moderate physical disability as determined by physical performance tests (SPPB scores; 48). Primary mental health outcome measures included the Hopkins Symptom Checklist (HSCL-25) for overall mood (range 1 to 4; higher scores are worse; internal consistency (Cronbach’s alpha) for the full sample α=0.92; range of α=0.90–0.95 for the four languages) (49, 50), and the GAD-7 for anxiety (range 0 to 21; higher scores are worse; overall α=0.82; range of α=0.78–0.88) (51). Main physical function outcomes included the SPPB for measurement of functional performance (range 0 to 12; higher scores are better) (52); the Late Life Function and Disability Instrument (LLFDI) for self-reported disability (range 32 to 160; higher scores are better; overall α=0.95; range of α=0.91–0.97) (53), and the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) for self-reported level of functioning (range 12 to 60; higher scores are worse; overall α=0.84; range of α=0.81–0.86) (54, 55). Acceptability of the program was defined by satisfaction with treatment (measured with six-item scale; see Methods 4, SDC) and attendance to more than 50% of invited sessions (i.e., more than 5 PM and/or more than 18 SB sessions). This attendance threshold has been used in other studies with similar populations (56, 57), and was appropriate for elders not recognizing need for services and not actively seeking treatment. Immigrant populations might also be reluctant to participate due to distrust in institutions (58).

Randomization

307 eligible participants were randomized to the intervention (N=153) or control condition (N=154) within site, in a ratio of 1:1 for each two-person block. Participant randomization was stratified by site using random numbers (see Figure 1).

Intent-to-treat analysis

To evaluate intervention effectiveness on our five primary outcomes (SPPB, LLFDI, WHODAS, GAD-7, HSCL-25), we conducted separate intent-to-treat analysis of post-baseline outcome scores over time using multilevel linear regression models, contrasting intervention and control groups at treatment completion (6-months) and six months post-treatment (12-months), while adjusting for baseline scores. To account for the slope of outcome trajectories over time and possible different slope post-intervention, models included an indicator for time of assessment and a linear time trend, and their two-way interactions with intervention. Time of assessment was coded in months (2-, 6- and 12-months) and centered at 6-months, yielding codes −4, 0 and 6, respectively. The linear time trend was coded 0 for the 2- and 6-month assessments and 6 for the 12-month assessment, allowing to test for different slopes during and after the intervention. By centering time of assessment as described above, the intervention indicator represents the effect at 6-months, while the two-way interactions can be used to estimate the effect at 12-months. We used multiple imputation to account for missing data/incomplete assessments (see Methods 5, SDC).

When assessing intervention effectiveness on physical function outcomes (SPPB, LLDFI, WHODAS), 35 intervention participants who did not engage in SB sessions were excluded from analysis, given that no facilities were available after hurricanes hit towns in two sites. Their control counterparts, per the two-person block randomization, were also excluded to rebalance the sample (see Methods 5, SDC). Our multilevel models estimated random intercepts at site- and participant-level, and used robust clustered standard errors in all estimations (59).

We conducted four sets of sensitivity analyses. First, to account for multiple outcome assessment, we estimated a multivariate multilevel regression (60) to test treatment effects obtained from a joint model instead of modeling the outcomes separately. This strategy accounts for the correlation among outcomes and provides more powerful tests of treatment effects compared to more traditional approaches (e.g., Bonferroni adjustments or combining the outcomes into a composite measure; 61). Second, we examined whether our intent-to-treat results were robust to more traditional models that included only an indicator for intervention, three indicators of time (2-, 6- and 12-months) and their two-way interactions. Third, we tested if the interventions had differing effects based on race/ethnicity, language, or site to ensure results were not influenced by cultural differences. Finally, we examined whether results changed when adjusting for additional concurrent service use (psychopharmacology and/or fitness) and whether differential effects occurred among participants with moderate-to-severe baseline physical functioning/mental health symptoms.

Compliance analysis

We analyzed whether intervention effects varied by number of sessions received (i.e., intervention compliance). Compliance was defined as zero (control group or intervention participants with no PM and no SB sessions), 1+ session (at least one PM or SB session without full compliance), and full compliance to treatment protocol (10 PM and 25–36 SB sessions). All compliance analyses excluded the 35 intervention participants (and their control counterparts) who did not receive SB sessions. Analyses were conducted in Stata 15 (62), using two-sided tests at α=0.05 for statistical significance.

RESULTS

Baseline characteristics

Table 1 presents baseline sociodemographic and clinical characteristics. We used i-tests and chi- square tests to assess baseline differences between intervention and control groups and found no significant differences, suggesting successful randomization. At baseline, around 57.5% of participants rated their physical health as poor or fair, 87.3% reported chronic conditions, and 69.5% identified as immigrants, with 66.1% speaking a primary language other than English. Two-thirds of participating elders (66.7%) had an SPPB score suggesting intermediate to minimal physical impairment, while almost half (49.8%) were 75 or older.

Table 1.

Baseline characteristics of the intention-to-treat population (N = 307)

	Total (N = 307)		Intervention (N = 153)		Control (N = 154)		Statistic(df), p
	N	%	N	%	N	%
Demographics
Age
60–64	21	6.8%	9	5.9%	12	7.8%	χ2(2) = 0.96, p = 0.62
65–74	133	43.3%	70	45.8%	63	40.9%
75+	153	49.8%	74	48.4%	79	51.3%
Gender
Male	59	19.2%	30	19.6%	29	18.8%	χ2(1) = 0.03, p = 0.86
Female	248	80.8%	123	80.4%	125	81.2%
Race
White/Caucas ian	31	10.2%	18	11.90%	13	8.6%	χ2(5) = 3.27, p = 0.66
Black/African/African American	24	7.9%	12	7.9%	12	7.9%
American Indian	1	0.3%		0.0%	1	0.7%
Asian or Pacific Islander	102	33.7%	51	33.8%	51	33.6%
Hispanic	136	44.9%	64	42.4%	72	47.4%
Other	9	3.0%	6	4.0%	3	2.0%
Education level
Less than High School	111	36.2%	49	32.0%	62	40.3%	χ2(1) = 2.25, p = 0.13
HS Diploma, GED, Vocational School, or More	196	63.8%	104	68.0%	92	59.7%
Place of Birth
Outside of U.S	210	69.5%	103	68.7%	107	70.4%	χ2(1) = 0.11, p = 0.74
U.S	92	30.5%	47	31.3%	45	29.6%
Marital Status
Married/Cohabitating	96	31.3%	56	36.6%	40	26.0%	χ2(3) = 5.98, p = 0.11
Divorced/Separated	85	27.7%	36	23.5%	49	31.8%
Windowed	98	31.9%	50	32.7%	48	31.2%
Never Married	28	9.1%	11	7.2%	17	11.0%
Smoking Status
No	198	64.5%	99	64.7%	99	64.3%	χ2(1) = 0.01, p = 0.94
Yes	109	35.5%	54	35.3%	55	35.7%
Self-rated physical health
Excellent	7	2.3%	3	2.0%	4	2.6%	χ2(4) = 1.01, p = 0.91
Very good	22	7.2%	12	7.9%	10	6.5%
Good	101	33.0%	53	34.9%	48	31.2%
Fair	143	46.7%	69	45.4%	74	48.1%
Poor	33	10.8%	15	9.9%	18	11.7%
Self-rated mental health
Excellent	11	3.6%	5	3.3%	6	3.9%	χ2(4) = 0.12, p = 1.00
Very good	35	11.4%	18	11.8%	17	11.0%
Good	115	37.5%	57	37.3%	58	37.7%
Fair	126	41.0%	63	41.2%	63	40.9%
Poor	20	6.5%	10	6.5%	10	6.5%
Primary Languages
English	62	20.2%	33	21.6%	29	18.8%	χ2(4) = 1.37, p = 0.85
Spanish	113	36.8%	55	35.9%	58	37.7%
Mandarin	46	15.0%	24	15.7%	22	14.3%
Cantonese	44	14.3%	23	15.0%	21	13.6%
Mixed	42	13.7%	18	11.8%	24	15.6%
Clinical Characteristics
Suicidal Riska	20	6.5%	11	7.2%	9	5.8%	χ2(1) = 0.23, p = 0.63
Suicidal Attemptb	1	0.3%	1	0.7%	0	0.0%	χ2(1) = 0.98, p = 0.32
Any Chronic Conditions	268	87.3%	129	84.3%	139	90.3%	χ2(1) = 2.45, p = 0.12
Short Physical Performance Battery (SPPB)c
Low score (3–6)	100	33.2%	52	34.4%	48	32.0%	χ2(2) = 4.15, p = 0.13
Intermediate score (7–9)	141	46.8%	63	41.7%	78	52.0%
High score (10–11)	60	19.90%	36	23.80%	24	16.0%
	Mean	SD	Mean	SD	Mean	SD
Short Physical Performance Battery (SPPB)c
Chair Stands Score	2.08	1.24	2.11	1.26	2.05	1.23	t(303) = 0.46, p = 0.64
Balance Ordinal Score	3.54	0.93	3.57	0.90	3.51	0.96	t(291) = 0.59, p = 0.55
Gait Ordinal Score	1.93	0.89	1.95	0.95	1.92	0.82	t(303) = 0.38, p = 0.70
SPPB Total Score	7.48	2.17	7.58	2.30	7.38	2.03	t(303) = 0.80, p = 0.42
Late Life Function and Disability instrument (LLFDI)
Function Component (LLF)	117.58	26.05	118.42	25.96	116.75	26.20	t(305) = 0.56, p = 0.57
Disability Component- Frequency (LLDA)	48.54	8.96	48.67	8.33	48.42	9.57	t(305) = 0.24, p = 0.81
Disability Component- Limitation (LLDB)	31.76	11.82	31.28	11.30	32.24	12.33	t(305) = −0.71, p = 0.48
WHODAS 2.0	22.19	7.48	21.97	7.09	22.40	7.86	t(305) = −0.5, p = 0.62
Generalized Anxiety (GAD-7)	5.99	4.59	6.18	4.63	5.79	4.56	t(302) = 0.72, p = 0.47
Depression (PHQ-9)	7.98	4.84	7.93	4.69	8.03	4.99	t(304) = −0.18, p = 0.85
Geriatric Depression (GDS)	5.51	3.29	5.46	3.26	5.56	3.32	t(302) = −0.26, p = 0.79
Hopkins Symptom Checklist (HSCL-25)	1.62	0.44	1.62	0.42	1.63	0.46	t(305) = −0.18, p = 0.86

Notes:

^aSuicidal risk includes participants who responded “yes” to either (1) feeling that life was not worth living, (2) wishing they were dead, and/or (3) having thoughts of taking their lives.

^bExclusion criteria included considering suicide/having a suicidal plan and/or suicide attempt during screening. One participant in the intervention group disclosed considering suicide/having a suicidal plan at baseline.

^cEligible participants scored between 3 and 11 in the SPPB during screening, but two participants scored < 3 & two participants scored > 11 at baseline.

Assessment retention

Of the 307 randomized participants, 262 completed at least two post-baseline assessments (85.3%) and 232 completed all three post-baseline assessments (75.6% retention rate), with no significant difference in number of completed assessments between intervention and control (χ²[3]=1.1, p=0.79).

Treatment retention

Of 153 participants enrolled in PM, 105 (68.6%) completed all 10 sessions, 13 (8.5%) completed 4–9 sessions, 15 (9.8%) completed 1–3 sessions, 10 (6.5%) declined or dropped out, and 10 (6.5%) discontinued because of medical reasons, death, or referral for suicidal ideation. Relative to participants who completed 1–3, 4–9 or all 10 PM sessions, participants who declined/dropped out were less likely to be smokers (χ²[3]=7.9, p=0.05) and more likely to self-report their physical health as excellent/very good (χ²[8]=16.2, p=0.04). Of 153 individuals enrolled in SB, 37 (24.2%) completed all 36 sessions, 21 (13.7%) completed 25–35 sessions, 39 (25.5%) completed 1–24 sessions, 21 (13.7%) declined or dropped out, and 35 (22.9%) did not initiate treatment due to hurricane-related lack of facilities in Puerto Rico and Florida, failed/delayed medical clearance because of hurricane, or death. Relative to participants who completed 1–24, 25–35 or all 36 SB sessions, participants who declined/dropped out were more likely to be Latino (χ²[8]=23.7, p<0.01) and less likely to be Mandarin-speakers (χ²[3]=17.8, p=0.02). No additional baseline differences were observed.

Acceptability

Regarding acceptability, 77.6% of participants attended 6+ PM sessions, 53.4% of participants attended 19+ SB sessions (excluding the 35 who did not receive SB sessions), and 49.2% of participants attended both 6+ PM and 19+ SB sessions (excluding the 35). Participants’ satisfaction rating (see Methods 4, SDC) indicated that 79% of participants were very satisfied with their sessions. This percentage was higher among participants who attended more than 50% of both PM and SB sessions (88% versus 66%, t[84]=−3.5, p<0.001).

Intent-to-treat analysis

Table 2 presents the results of the intent-to-treat analyses for 6-month follow-up, showing the intervention was effective in improving functioning per the SPPB Total Score (t[1775.7]=2.1, p=0.03) and the LLFDI Function Component (t[663.5]=2.4, p=0.02), as well as in reducing psychosocial distress per the HSCL-25 (t[817.3]=−3.1, p<0.01). Effect sizes (Cohen’s d) ranged from 0.23 to 0.28, suggesting small effects in the intent-to-treat analysis (63). There were no significant effects on GAD-7 (t[921.1]=−0.94, p=0.35) or WHODAS 2.0 (t[2619.0=−1.44, p=0.15) scores.

Table 2.

Intent-to-treat Analysis: Effectiveness of the PMSB Intervention at 6-month follow-up^a

	SPPB Total Score			LLFDI Function Component			WHODAS 2.0
	b (SE)	t (df)f	p-value	b (SE)	t (df)f	p-value	b (SE)	t (df)f	p-value
Intervention effectb	0.60 (0.28)	2.12 (1775.69)	0.03	6.00 (2.46)	2.44 ( 663.45)	0.02	−1.24 (0.86)	−1.44 (2618.96)	0.15
Timec	0.16 (0.04)	3.65 (1068.91)	<0.001	−0.03 (0.30)	−0.10 (1331.09)	0.92	−0.21 (0.12)	−1.77 (4133.85)	0.08
Intervention*time	0.11 (0.09)	1.29 ( 728.17)	0.20	1.05 (0.63)	1.67 ( 406.17)	0.10	−0.31 (0.25)	−1.26 (2185.27)	0.21
(Time-t*)+d	−0.16 (0.06)	−2.75 (2860.11)	<0.01	−0.39 (0.42)	−0.92 (2228.32)	0.36	0.47 (0.17)	2.67 (6026.14)	<0.01
Intervention*(Time-t*)+	−0.13 (0.12)	−1.07 (1475.40)	0.28	−0.93 (0.90)	−1.04 ( 519.83)	0.30	0.10 (0.36)	0.28 (1097.39)	0.78
Baseline measure of the outcome	0.54 (0.05)	10.81 (1966.07)	<0.001	0.75 (0.04)	19.5 (1307.81)	<0.001	0.64 (0.05)	13.58 ( 367.58)	<0.001
Effect size (Cohen’s d) of Intervention Effecte	0.28 (0.13)	2.12 (1775.69)	0.03	0.23 (0.09)	2.44 ( 663.45)	0.02	0.17 (0.11)	1.44 (2618.96)	0.15
		GAD-7			HSCL-25
	b (SE)	t (df)f	p-value	b (SE)	t (df)f	p-value
Intervention effectb	−0.49 (0.52)	−0.94 ( 921.14)	0.35	−0.12 (0.04)	−3.09 ( 817.32)	<0.01
Timec	−0.15 (0.07)	−2.08 (1030.39)	0.04	−0.01 (0.01)	−2.40 ( 912.39)	0.02
Intervention*time	−0.03 (0.14)	−0.25 (1039.91)	0.81	−0.02 (0.01)	−1.46 (1776.80)	0.14
(Time-t*)+d	0.22 (0.10)	2.16 ( 709.24)	0.03	0.02 (0.01)	2.21 (1358.17)	0.03
Intervention*(Time-t*)+	−0.02 (0.21)	−0.12 ( 878.56)	0.91	0.02(0.02)	1.14 (1113.15)	0.26
Baseline measure of the outcome	0.40 (0.04)	9.09 (1324.87)	<0.001	0.61 (0.03)	18.01 ( 841.04)	<0.001
Effect size (Cohen’s d) of Intervention Effecte	0.11 (0.11)	0.94 ( 921.14)	0.35	0.27 (0.09)	3.09 ( 817.32)	<0.01

Notes:

^aAnalyses on physical (SPPB, LLFDI, WHODAS 2.0) and mental health (GAD-7, HSCL-25) outcomes use longitudinal data of 237 and 307 participants, respectively, with three follow-up assessments per participant. Each outcome variable was measured three times at 2-month, 6-month, and 12-month follow-up. The unit observation is a specific follow-up assessment. Missing data was handled using multiple imputation.

^bThe reference group is participants in the control arm.

^cTime is a continuous variable which equals to −4, 0, and 6 for the 2-month, 6-month, and 12-month follow-up, respectively.

^dTime-t* is a continuous variable equal to 6 for the 12-month follow-up and 0 otherwise. t* denotes the time when treatment ends which equals to 6.

^eEffect size (Cohen’s d) of intervention effect was calculated using the outcome standard deviation for the full sample at baseline.

^fDegrees of freedom for statistical inference of individual coefficients calculated using Barnard and Rubin (1999; 69), which are a function of the number of imputations and the increase in variance of estimates due to missing data.

Results from our multivariate multilevel model, which accounted for multiple outcome assessment and the correlation among outcomes, indicated that the significance of our results was not influenced by modeling the five outcomes separately. After modeling the outcomes jointly, we conducted a joint global test to evaluate the null hypothesis that all treatment effects were zero. The joint test found a significant effect of the treatment through the five outcomes at the α=0.05 level (χ²[5]=14.3, p=0.01), and the estimated effect sizes were of the same magnitude (see Table 2, SDC). Our sensitivity analyses also suggested that PMSB effectiveness was not an artifact of cultural differences in protocols (see Tables 3–7, SDC). After adjusting for language, estimated effects were virtually the same, with no significant interaction effects between intervention and language groups or when adjusting for patient’s race/ethnicity or clinical site as well as for current pharmacotherapy and/or fitness service use and severity of baseline physical functioning/mental health symptoms.

Intent-to-treat analyses at 12-month follow-up (Table 3) showed that the intervention continued having significant effects 6-months post-intervention. The effect on the SPPB attenuated and became nonsignificant (t[3565.9]=1.7, p=0.08), but the effect size on LLFDI was slightly bigger although not significantly different from the 6-month effect size (t[850.1]=−0.28, p=0.78). The intervention effect on the WHODAS 2.0 became significant (t[1339.2]=−2.8, p<0.01), and there was still an improvement in mood symptoms per HSCL-25 (t[1465.2]=−2.8, p<0.01). The intervention continued having no effect on the GAD-7 score (t[1828.7]=−1.7, p=0.10).

Table 3.

Intent-to-treat Analysis: Effectiveness of the PMSB Intervention at 12-month follow-up (six-months post-intervention)^a

	SPPB Total Score			LLFDI Function Component			WHODAS 2.0
	b (SE)	t (df)d	p-value	b (SE)	t (df)d	p-value	b (SE)	t (df)d	p-value
Intervention effectb	0.48 (0.28)	1.73 ( 3565.9)	0.08	6.68 (2.38)	2.81 (1576.13)	<0.01	−2.47 (0.88)	−2.81 (1339.21)	<0.01
Baseline measure of the outcome	0.54 (0.05)	10.81 (1966.07)	<0.001	0.75 (0.04)	19.5 (1307.81)	<0.001	0.64 (0.05)	13.58 ( 367.58)	<0.001
Effect size (Cohen’s d) of Intervention Effectc	0.22 (0.12)	1.73 ( 3565.9)	0.08	0.26 (0.09)	2.81 (1576.13)	<0.01	0.33 (0.12)	2.81 (1339.21)	<0.01
		GAD-7			HSCL-25
	b (SE)	t (df)d	p-value	b (SE)	t (df)d	p-value
Intervention effectb	−0.84 (0.51)	−1.66 (1828.65)	0.10	−0.11 (0.04)	−2.79 (1465.16)	<0.01
Baseline measure of the outcome	0.40 (0.04)	9.09 (1324.87)	<0.001	0.61 (0.03)	18.01 ( 841.04)	<0.001
Effect size (Cohen’s d) of Intervention Effectc	0.18 (0.11)	1.66 (1828.65)	0.10	0.24 (0.09)	2.79 (1465.16)	<0.01

Notes:

^aAnalyses on physical (SPPB, LLFDI, WHODAS 2.0) and mental health (GAD-7, HSCL-25) outcomes use longitudinal data of 237 and 307 participants, respectively, with three follow-up assessments per participant. Each outcome variable was measured three times at 2-month, 6-month, and 12-month follow-up. The unit observation is a specific follow-up assessment. Missing data was handled using multiple imputation.

^bThe reference group is participants in the control arm. Coefficient was calculated using the time-by-intervention interactions from Table 2. Because Time is equal to 6 for the 12-month follow-up, the intervention effect at 12-months is calculated as: Intervention + (6*Intervention*Time) + (6*Intervention*(Time-t*)+)

^cEffect size (Cohen’s d) of intervention effect was calculated using the outcome standard deviation for the full sample at baseline.

^dDegrees of freedom for statistical inference of individual coefficients calculated using Barnard and Rubin (1999; 69), which are a function of the number of imputations and the increase in variance of estimates due to missing data.

Our results were robust to more traditional intent-to-treat analysis that included only indicators for time of assessment and their interactions with treatment. Except for SPPB, the intervention effects at 6- and 12-months were very similar in magnitude for LLFDI, WHODAS 2.0 and HSCL-25, with overall time-by-treatment significant interactions (F[3,2178.8]=3.0, p=0.03; F[3,3678.4]=3.0, p=0.03; and F[3,3321.5]=2.9, p=0.04, respectively; see Table 8, SDC). However, for all outcomes where the intervention had a significant effect using our original model, the estimated coefficient of the time trend (labeled Time-t* in Table 2) was statistically significant (p<0.05), suggesting that the inclusion of such time trend was adequate.

Compliance analysis

Table 4 shows that intervention effects at 6-month follow-up on physical function and symptom reduction varied by intervention compliance; the greatest effect was observed among participants who fully adhered to the protocol and completed target treatment sessions. Six-months after baseline, relative to those with zero sessions, participants who completed all 10 PM sessions and sufficient SB sessions (25–36 sessions) experienced significantly larger improvements in all physical outcomes and all but one mental health outcome. They had better physical performance per SPPB Total Score (t[7.9]=4.0, p<0.01), late-life functioning per LLFDI Function Component (t[8.0]=3.9, p<0.01), and significant reduction in disability per WHODAS 2.0 (t[7.9]=−3.3, p=0.01). Participants with full compliance also had significant reductions in mood symptoms per HSCL-25 (t[8.0]=−2.7, p=0.03). Full compliance still did not improve GAD-7 scores at 6-months (t[8.2]=−2.0, p=0.07). Intervention compliance effects on physical function improvement and mood symptom reduction were also significant at 12-month follow-up (Table 5). Twelve-months post-baseline, participants with full treatment compliance continued demonstrating significant improvement in SPPB Total Score (t[8.1]=2.9, p<0.05) and HSCL-25 (t[8.1]=−2.8, p<0.05). These participants also demonstrated significant LLFDI and WHODAS 2.0 effects for the 12-months outcomes, that were not significantly different from the 6-month effect size (t[1279.8]=−0.96, p=0.34, and t[1916.3]=1.15, p=0.25, respectively) The effect of full compliance at 12-months was the only time when an improvement in GAD-7 was observed (t[7.0]=−3.9, p<0.01).

Table 4.

Dosage Analysis of the PMSB Intervention at 6-month follow-up^a

	SPPB Total Score			LLFDI Funtion Component			WHODAS 2.0
	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value
0 PM & 0 SB (Reference)	--	--	--	--	--	--	--	--	--
1+ Session without full compliance	0.54 (0.38)	1.43 (7.81)	0.19	2.75 (2.25)	1.22 (5.39)	0.27	−0.33 (1.05)	−0.31 (7.74)	0.76
10 PM & 25–36 SB sessions (Full Compliance)	1.01 (0.25)	4.04 (7.94)	<0.01	10.28 (2.63)	3.92 (8.03)	<0.01	−3.11 (0.95)	−3.28 (7.93)	0.01
Baseline Measure of the Outcome	0.50 (0.06)	8.94 (6.97)	<0.001	0.73 (0.04)	17.38 (6.46)	<0.001	0.61 (0.06)	9.70 (7.12)	<0.001
Effect size (Cohen’s d) of Full Complianceb	0.47 (0.11)	4.04 (7.94)	<0.01	0.39 (0.10)	3.92 (8.03)	<0.01	0.42 (0.13)	3.28 (7.93)	0.01
		GAD-7			HSCL-25
	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value
0 PM & 0 SB (Reference)	--	--	--	--	--	--
1+ Session without full compliance	−0.48 (0.79)	−0.60 (7.80)	0.56	−0.12 (0.05)	−2.25 (7.02)	0.06
10 PM & 25–36 SB sessions (Full Compliance)	−1.71 (0.84)	−2.04 (8.23)	0.07	−0.17 (0.07)	−2.66 (8.00)	0.03
Baseline Measure of the Outcome	0.37 (0.07)	5.34 (6.52)	<0.01	0.50 (0.05)	9.54 (7.42)	<0.001
Effect size (Cohen’s d) of Full Complianceb	0.37 (0.18)	2.04 (8.23)	0.07	0.38 (0.14)	2.66 (8.00)	0.03

Notes:

^aAnalyses include 237 participants at 6-month follow-up. Missing data was handled using multiple imputation.

^bEffect size (Cohen’s d) of Full Compliance was calculated using the outcome standard deviation for the full sample at baseline.

^cDegrees of freedom for statistical inference of individual coefficients calculated using Barnard and Rubin (1999; 69), which are a function of the number of imputations and the increase in variance of estimates due to missing data.

Table 5.

Dosage Analysis of the PMSB Intervention at 12-month follow-up (six-months post-intervention)^a

	SPPB Total Score			LLFDI Function Component			WHODAS 2.0
	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value
0 PM & 0 SB (Reference)	--	--	--	--	--	--	--	--	--
1+ Session without full compliance	0.23 (0.28)	0.84 (7.60)	0.43	2.58 (3.23)	0.80 (7.44)	0.45	−1.29 (1.37)	−0.94 (7.69)	0.38
10 PM & 25–36 SB sessions (Full Compliance)	0.87 (0.31)	2.85 (8.10)	0.02	11.71 (1.93)	6.08 (8.17)	<0.001	−4.05 (0.72)	−5.61 (8.17)	<0.001
Baseline Measure of the Outcome	0.54 (0.07)	7.90 (7.13)	<0.001	0.76 (0.03)	22.67 (7.04)	<0.001	0.64 (0.06)	10.40 (6.78)	<0.001
Effect size (Cohen’s d) of Full Complianceb	0.40 (0.14)	2.85 (8.10)	0.02	0.45 (0.07)	6.08 (8.17)	<0.001	0.54 (0.10)	5.61 (8.17)	<0.001
		GAD-7			HSCL-25
	b (SE)	t (df)c	p-value	b (SE)	t (df)c	p-value
0 PM & 0 SB (Reference)	--	--	--	--	--	--
1+ Session without full compliance	−1.03 (0.85)	−1.21 (7.91)	0.26	−0.10 (0.06)	−1.73 (7.58)	0.12
10 PM & 25–36 SB sessions (Full Compliance)	−1.56 (0.40)	−3.89 (7.04)	<0.01	−0.15 (0.05)	−2.80 (8.07)	0.02
Baseline Measure of the Outcome	0.36 (0.08)	4.61 (7.13)	<0.01	0.49 (0.06)	8.38 (6.61)	<0.001
Effect size (Cohen’s d) of Full Complianceb	0.34 (0.09)	3.89 (7.04)	<0.01	0.34 (0.12)	2.80 (8.07)	0.02

Notes:

^aAnalyses include 237 participants at 6-month follow-up. Missing data was handled using multiple imputation.

^bEffect size (Cohen’s d) of Full Compliance was calculated using the outcome standard deviation for the full sample at baseline.

^cDegrees of freedom for statistical inference of individual coefficients calculated using Barnard and Rubin (1999; 69), which are a function of the number of imputations and the increase in variance of estimates due to missing data.

DISCUSSION AND CONCLUSION

Our study is one of few (29) addressing disability prevention for ethnic/racial minority and immigrant elders in four languages. This trial shows that PMSB is a promising preventive intervention that moves culturally-competent services into the community to meet client needs. As such, it may be a strategy to address the great unmet need for disability services in elder populations (64). Uniquely, we combined physical and mental health interventions, administered by paraprofessionals, showing gains in elders’ physical and mental health functioning. Our effects on mood symptoms in a heterogeneous sample of minority elders was comparable to those observed with Latino adults (45) using similar mental or physical health interventions (65). Small intent-to-treat effects are also consistent with a secondary prevention intervention. Further, the physical function changes we observed were similar to those reported in the InVEST trial (31, 32). Our findings showed significant gains in directly measured (SPPB) as well as patient reported (LLFDI) outcomes, suggesting that decline of physical function with age can be decelerated, at least partially, independent of participant’s ethnicity and even in the presence of mental health conditions.

Although results are promising, we note the following limitations. Our sample of minority participants focuses on Latino and Asian ethnic/racial minorities and immigrants. We cannot be sure the findings generalize to other minority groups or have a sufficient sample for cross-group comparisons. We did not use equipment capable of measuring strength and velocity in a precise way, but relied on RA-administered assessments. Although outcome measures went through a rigorous translation process and demonstrated adequate internal consistency within language group, our sample was insufficient to carry out more formal measurement invariance analyses. Fortunately, random assignment makes it likely that significant effects are not due to problematic items, as problems would be balanced in treatment and control conditions.

The findings do suggest adequate acceptability for persons not seeking disability prevention services. More than half of participants attended >50% of the PM and SB sessions; and nearly half received the combined intervention at a moderate rate of compliance. PM acceptance could be explained by flexibility in session times offered, and bilingual/bicultural staff. However, SB compliance appeared low overall compared to a previous study of the same intervention in adults (21, 47) and other studies with elders (66). This may highlight minority elders’ competing health demands, since close to half rated their health as fair or poor and more than three-fourths had at least one chronic physical condition. Participants noted other scheduled activities, such as medical appointments; not feeling well enough; or transportation issues to the CBO, given reliance on public transportation. To improve compliance, we plan to offer SB by video, especially during winter months, to build a routine and desire for exercise starting at home. Starting prevention interventions early seems paramount to ensure decreased consequences of illness, such as frailty and late life disability (67).

Importantly, intervention participants reported feeling that the culturally tailored intervention responded to their needs. This aligns with literature showing that racial/ethnic and language concordance are important factors in racial/ethnic minority health experiences (68). Our work similarly addresses the Institute of Medicine’s report on the need to build a disability prevention workforce (4), with evidence for the use of paraprofessionals. Throughout the field of disability prevention, PMSB can be a valuable tool to address workforce shortages, disparities in service access, and disparities in service quality for minority elders.

Supplementary Material

1

Methods 1. Descriptive methods – PMSB intervention protocol

Methods 2. Descriptive methods – Training of Community Health Workers (CHWs)

Methods 3. Descriptive methods – Program tailoring to language groups

Methods 4. Analytical methods – Psychometric analysis of satisfaction measure.

Methods 5. Analytical methods – Data preparation.

Methods 6. Analytical methods – Data analysis.

Table 1. List of assessments/variables

Table 2. Intent-to-treat analysis of PMSB on main outcomes: results from multivariate model.

Table 3. Intent-to-treat analysis of PMSB on main outcomes by language evaluated at 6-month follow-up.

Table 4. Intent-to-treat analysis of PMSB on main outcomes by race/ethnicity evaluated at 6-month follow-up.

Table 5. Intent-to-treat analysis of PMSB on main outcomes by site evaluated at 6-month follow-up.

Table 6. Intent-to-treat analysis of PMSB on main outcomes by baseline mental health/fitness service use evaluated at 6-month follow-up.

Table 7. Intent-to-treat analysis of PMSB on main outcomes by baseline severity of physical functioning/mental health symptoms evaluated at 6-month follow-up.

Table 8. Intent-to-treat analysis of PMSB on main outcomes: results from alternative model.

Table 9. Correlation of PMSB main outcome measures at baseline and 12-month follow-up.

Table 10. Significant baseline differences between participants with missing/incomplete assessments and participants with complete assessments

Highlights.

• This multi-site clinical trial tests whether Positive Minds-Strong Bodies (PMSB), a disability prevention intervention offered in four languages, is acceptable to minority and immigrant elders and feasible to be offered by paraprofessionals.

• The 6-month intent-to-treat analyses showed significant intervention effects in improving late-life functioning and lowering mood symptoms while the 12-month analyses showed sustained significant effects on self-reported disability, disability days, and mood symptoms.

• The PMSB intervention shows promise in preventing disability in a diverse sample of minority and immigrant elders.