Combined Psychoeducational and Exercise Training Intervention to Improve Mental Health and Physical Functioning in Late-Life: A Randomized Controlled Trial

Abstract

Objective:

To assess the effectiveness of an enhanced psychoeducational and exercise training intervention (Positive Minds Strong Bodies-Enhanced [PMSB-E]) on depression and anxiety symptoms and physical functioning.

Design:

Randomized trial of 427 participants, assigned to intervention or usual care.

Setting:

51 community-based organizations and 17 clinics in Massachusetts, New York, Puerto Rico, and Maryland, US. Data collected from November 2020 to October 2024, at baseline and months 3, 6, and 12.

Participants:

Adults aged 60+ or older with mild to severe depression or anxiety symptoms, minor to moderate mobility limitations, and fluency in English, Spanish, Mandarin, or Cantonese.

Interventions:

10 psychoeducational and 36 exercise sessions over 6 months compared to usual care (one-on-one phase; 1:1 ratio). Half of the intervention participants were offered an additional 6-month group maintenance phase (1:1 ratio). Control participants received up to 4 usual care calls.

Measurements:

Primary outcome was change from baseline to month 12 in depression and anxiety symptoms (HSCL-25) and physical functioning (SPPB and LLFDI). Secondary outcome was change in month 6.

Results:

Intervention participants showed improved SPPB scores in month 12 (standardized difference, 0.28; 95% CI, 0.07–0.49) and month 6 (0.34; 95% CI, 0.14–0.55). There were no effects on LLFDI or HSCL-25 scores, nor interaction effects by race and ethnicity. Completing the recommended number of sessions had a larger effect on SPPB scores.

Conclusions:

PMSB-E replicated previous findings of improving objective physical functioning but not self-reported physical functioning or mental health symptoms. We believe this is due to virtual administration during the pandemic.

Introduction and Objectives

Between 2020 and 2040, the older (65+) racial and ethnic minoritized population of the U.S. is projected to increase by 105%, compared to 26% of non-Latino Whites of the same age group.¹ Yet, according to the 2016 Health and Retirement Study, non-Latino Black and Latino older adults show greater signs of accelerated biological aging than their non-Latino White counterparts.² Researchers have suggested that these disparities in biological aging can be traced to disproportionate exposure to specific stressors (e.g., systemic poverty, racism).³ Such exposure requires constant adaptation from minoritized groups, which, in turn, wears down biological systems.³

Issues such as discrimination and traumatic experiences related to migration can also affect the physical and mental health of immigrant and refugee older adults in the U.S.⁴ U.S.-born minoritized, immigrant, and refugee older adults can have higher functional limitations, impairment, and disability than U.S.-born non-Latino Whites in mid and late life,^5,6 straining the capacity of disability prevention services. Minoritized, immigrant, and refugee older adults are also at greater risk for severity, persistence, and recurrence of psychiatric disorders,⁷ with less access to disability prevention than non-Latino Whites⁷ and poorer identification by providers.⁸ These patients confront barriers to accessing clinical services, such as stigma,⁷ self-reliance,⁹ limited English proficiency,⁷ workforce shortages,⁷ undocumented status,¹⁰ and low mental health literacy.⁷

Physical disability is a risk factor for depression¹¹ and vice versa.¹² Mobility limitations (e.g., balance and walking speed) predict later disability,¹³ which can be addressed through treatment of mood symptoms and exercises emphasizing functional abilities relevant to activities of daily living. A previous intervention, Positive Minds-Strong Bodies (PMSB), showed promise for reducing disparities. PMSB was a culturally adapted disability prevention intervention,¹⁴ administered in English, Spanish, Mandarin, and Cantonese, by community health workers (CHWs): frontline public health workers identifying with the community being served.¹⁵ It demonstrated high levels of acceptability and effectiveness for physical functioning and reduction of mood symptoms among minoritized and immigrant older adults.¹⁴ The present study tested the effectiveness and implementation of an enhanced version, PMSB-E, with group sessions from months 7–12 and exercise videos to scale the intervention¹⁴ in low-resource community settings. We tested this enhanced version for three reasons. First, to examine whether it was similarly effective across diverse racial and ethnic groups. Second, to respond to participants who requested conducting the intervention as a maintenance activity after the intervention period, to enhance its effect. Third, to accept an intervention as evidence-based, at least two positive trials are required.¹⁶

Methods

The Mass General Brigham institutional review board approved the study, with cede reviews from sites participating in research activities and collaboration agreements for recruitment-only sites. Participants provided verbal informed consent. This study followed CONSORT guidelines for RCTs.

Procedure

We recruited participants through outreach at clinics and CBOs (senior centers, elder housing facilities, and cultural organizations) with racially and ethnically diverse, mostly economically disadvantaged clients. At CBOs, study staff conducted virtual and in-person presentations to assess interest. We obtained clinician referrals at clinical sites and called patients from a registry, following an outreach letter. Study staff administered a two-part screener, including self-reported measures and physical functioning in-person using the Short Physical Performance Battery (SPPB).¹⁷ Between November 20, 2020, and April 15, 2024, we assessed 1287 individuals for eligibility.

Participants

Inclusion criteria were age ≥60; fluency in English, Spanish, Mandarin, or Cantonese; mild to severe depressive or anxiety symptoms using the Patient Health Questionnaire-9 (PHQ-9 ≥5),¹⁸ the Geriatric Depression Scale-15 (GDS-15 ≥5),¹⁹ or the Geriatric Anxiety Inventory (GAI ≥8);²⁰ and mild to moderate mobility impairment defined by SPPB scores between 3 and 11.²¹ SPPB cut-off scores ≤11 have been identified as better predictors of disability in community-dwelling older adults.²² Key exclusion criteria were: (1) current (upcoming month) or recent (past 3 months) specialty mental health treatment (excluding psychopharmacology); (2) lack of capacity to consent using a validated screener; (3) current substance use problems; (4) history of psychosis, mania, or psychotic symptoms; (5) current suicidal risk (‘yes’ to items 4 or 5 on the Paykel suicide scale),²³ which resulted in connection to emergency services; or (6) previous participation in PMSB. Participants with medical comorbidities were included, unless their condition results in screening scores that led to exclusion.

Interventions

PMSB-E is an enhanced version of two interventions given concurrently, a psychosocial (Positive Minds, PM) and exercise training (Strong Bodies, SB) intervention, whose effectiveness was previously tested.¹⁴ In the parent study,¹⁴ PM was a 10-session, one-on-one, mostly in-person intervention, adapted from cognitive-behavioral therapy (CBT) and delivered by a trained CHW over 6 months. SB consisted of a 12-to-14-week group exercise training (1–3 participants), with up to 36 sessions delivered in person by a trained exercise trainer (ET). PMSB-E was extended to 12 months and delivered in two parts: a one-on-one phase in months 1–6 and a group maintenance phase in months 7–12. While PMSB-E was originally designed for in-person administration, this was unfeasible during the COVID-19 pandemic. We adapted PMSB-E for virtual (Zoom) or phone administration for PM and Zoom for SB. Because SB involved demonstration of physical exercises and performance under ET observation, it could only be offered by Zoom (see Supplemental Materials).

PM Enhanced

In months 1–6, trained CHWs administered up to 10 one-on-one, 1-hour sessions, virtually or by phone, to identify and address negative cognitions, engage in relaxing and breathing exercises, mindfulness, promote behavioral activation, and encourage supportive relationships. PM enhanced included adaptations to address diverse cultural beliefs and presentation of therapeutic strategies, including expanding interpersonal components with practical communication skills, an increased number of role-play practices, and further attention to nutrition and sleep habits. In months 7–12, we offered 1-hour virtual group sessions as a drop-in option for additional support and practice (up to 10 participants per group and 24 sessions). Further details, including CHWs’ training and demographic characteristics, are included as Supplemental Material.

SB Enhanced

In months 1–6, ETs delivered up to 24 one-on-one, 1-hour sessions of functional exercises twice weekly. An additional 12 sessions were performed independently, using an instructional video for safe at-home use. SB was designed to increase muscle power (strength × velocity), a determinant of successful performance of daily living activities and a better predictor of mortality than muscle strength.²⁴ Participants performed exercises wearing a weighted vest, with a progressive overload of 2% body weight. ETs monitored safe positioning, balance, posture, and form.²⁵ In months 7–12, ETs offered 1.5-hour, twice-a-week, drop-in virtual group sessions to help maintain gains (up to 4 participants per group and 48 sessions). Further details, including ETs’ training and demographic characteristics, are included as Supplemental Material.

Randomization and Masking

Study-eligible participants were randomized on a ratio of 1:1, using computer-generated blocks of size 2, to PMSB-E or the control condition (characterized as enhanced usual care [EUC]). Randomization was stratified by site to achieve balance in the total sample and within site (i.e., about half of participants were PMSB-E and half EUC in each site). At the end of month 6, only PMSB-E intervention participants were randomized again (1:1) to group maintenance with safety assessments (questionnaires monitoring symptom severity and assessing need for behavioral health or suicidality referrals) or safety assessments alone. Thus, one-fourth of the total trial participants were in the PMSB-E intervention group (one-on-one, months 1–6), plus group maintenance sessions and safety assessments (months 7–12). Another one-fourth were in a PMSB-E intervention group (one-on-one, months 1–6) plus safety assessments only (months 7–12).

The remaining one-half of the total participants were in the EUC group, receiving an NIH booklet in their language on depression and anxiety (months 1–6 only). Control participants were involved in clinical ‘usual care’ as available for their conditions through primary or specialty care. As such, EUC participants potentially accessed services simultaneously. Research assistants who evaluated outcomes were blind to study condition, but participants and data analysts were not.

Intervention Fidelity

PM sessions were audio recorded. Licensed clinical supervisors reviewed each CHW’s first two sessions, plus a random 10%, using a session activities checklist. Supervisors provided CHWs and ETs with weekly feedback on fidelity and issues of clinical significance. PM average session fidelity was 80.4% (95% CI, 75.0% to 85.8%). With the onset of the COVID-19 pandemic, SB session quality was monitored, but fidelity checks were unfeasible (see Supplemental Materials, page 10).

Assessments and Outcomes

Participant demographic and clinical characteristics were collected at screening and baseline. The Hopkins Symptom Checklist-25 (HSCL-25)²⁶ assessed past 2-week self-reported symptoms of depression and anxiety. The HSCL-25 includes 25 items, each rated from 1 (not at all) to 4 (extremely). Total scores are calculated by averaging all items (range 1 to 4, with higher scores indicating more severe symptoms).²⁶ The SPPB¹⁹ objectively assessed physical functioning through balance, gait, and chair-rise timed tests. Because time is an essential determinant of successful performance, these tests reflected muscle power and were more specific to our exercise intervention. A summary performance score is calculated by summing categorical rankings on the three tests (range 0 to 12, with higher scores indicating better performance).¹⁹ The Function Component of the Late-Life Functioning and Disability Instrument (LLFDI)²⁷ assessed self-reported physical functioning, with 32 items assessing difficulty (1 ‘cannot do’ to 5 ‘none’) to perform discrete actions or activities without the help of someone else or using assistive devices. Total scores are calculated by summing all items (range 32 to 160, with higher scores indicating fewer difficulties).²⁷ The 12-item version of the WHO Disability Assessment Schedule (WHODAS 2.0)²⁸ assessed past 30-day self-reported disability to capture difficulty performing activities in six domains of life (1 ‘none’ to 5 ‘extreme or cannot do’). Total scores are calculated by summing all items (range 12 to 60, with higher scores indicating greater disability).²⁸

The HSCL-25, LLFDI, and WHODAS 2.0 were evaluated at baseline and in months 3, 6, and 12 in-person, virtually, or by phone. Trained raters assessed in-person SPPB at approximately the same time (median time between SPPB and other outcome measures ranged from 7 to 9 days). A small number of enrolled participants (20 [4.7%]) were administered a virtual version of the SPPB at baseline because of social distancing measures related to the COVID-19 pandemic.

Statistical Analysis

Sample size was selected to provide at least 80% power, assuming a 12-month effect size of PMSB-E based on results of the parent study.¹⁴ The primary endpoint was change from baseline to month 12 in HSCL-25, SPPB, and LLFDI scores. The key secondary endpoint was the change from baseline to month 6. We conducted our analyses in the intention-to-treat population and included all randomized participants irrespective of their actual participation. All tests were 2-sided. The threshold for statistical significance was set to P <0.05. P-values for the primary and key secondary endpoints were adjusted for 6 multiple comparisons (3 outcome scores [HSCL-25, SPPB, and LLFDI] at 2 follow-up points [months 12 and 6]) using the Benjamini-Hochberg (BH) procedure.²⁹ Confidence Intervals (CIs) for the primary and key secondary endpoints were also adjusted for these 6 multiple comparisons using the Benjamini-Yekutiele (BY) procedure,³⁰ which is the analog of the BH procedure. In sensitivity analyses, we also estimated a single model on an aggregate variable capturing the three outcome scores (HSCL-25, SPPB, and LLFDI) to test global effectiveness (see Supplemental Material). In prespecified subgroup analyses, we evaluated primary and key secondary endpoints based on participants’ race and ethnicity (non-Latino Black, Asian, Latino), which resulted in 6 omnibus interaction tests (race/ethnicity × 3 outcome scores × 2 follow-up points). We did not adjust for the multiple comparisons in our prespecified subgroup analysis, and thus conservatively expected one statistically significant interaction test based on chance alone (ceiling of 6 × 0.05). Change from baseline to months 6 and 12 in WHODAS 2.0 scores was a secondary outcome. Change from baseline to month 3 in all outcomes was an exploratory endpoint.

We compared changes in outcome scores using repeated measures linear mixed-effects models with a participant-level random intercept. All models were adjusted for baseline scores and included time since baseline, time since month 6 (to account for possible change in slope between months 6–12), intervention condition, and time × intervention interactions. Model estimates evaluated differences between intervention and control at months 3, 6, and 12. We presented differences as point estimates and standardized effect sizes (Cohen’s d). The standardized differences represent a common metric that allows comparison of intervention effects across all primary and key secondary endpoints. Prespecified subgroup analyses added interaction terms between intervention and race/ethnicity and time × intervention × race/ethnicity interactions. Under the missing at random assumption, missing values were handled using model-based multiple imputation in BLiMP.³¹ As shown in Supplemental Material Tables S1–S3, participants with missing outcome follow-up data had similar baseline outcome scores compared to those with complete data.

In per-protocol analyses, we evaluated treatment effects when completing the recommended treatment dose (≥6 PM and ≥24 SB sessions)¹⁴ using methods to handle post-treatment confounding (see Supplemental Material). We also conducted post-hoc exploratory analyses for depression and anxiety symptoms (HSCL-25 scores) by the following subgroups: baseline severity (HSCL-25 ≥1.75), baseline number of social determinants, baseline cognitive status,³² and mental health service use post-randomization (see Supplemental Material). In another post-hoc exploratory analysis, we also investigated whether treatment effects varied as a function of mode of administration of the PM sessions: phone (46.3%) versus in-person/Zoom (53.7%). All analyses were conducted in Stata version 18 (StataCorp)³³ and RStudio version 2024.12.0.467 (Posit team).³⁴

Results

Participants

We conducted PMSB-E between November 20, 2020, and October 31, 2024. Of 1287 individuals assessed for eligibility, 427 were study-eligible and randomized to PMSB-E (n =211) or EUC (n =216) (Figure 1). There was a low contamination risk because PMSB-E included multifaceted interventions with a significant behavioral component that cannot easily be transferred between participants. The randomized sample had a mean (SD) age of 71.3 (7.7) years, and 337 (78.9%) were female. Participants self-identified as non-Latino White (19 [4.4%]), non-Latino Black (90 [21.1%]), American Indian (1 [0.2%]), Asian (117 [27.4%]), or Latino (200 [46.8%]). Baseline demographics, clinical characteristics, and screening and outcome scores were similar between PMSB-E and EUC groups (Table 1).

Figure 1.

CONSORT Diagram

Abbreviations: PMSB-E, Positive Minds Strong Bodies-Enhanced; EUC, Enhanced Usual Care. a Because of low participant engagement, the maintenance phase of PMSB-E was discontinued in the last 9 months of the study.

b Groups were closed when there was an insufficient number of participants.

Table 1.

Demographic and clinical characteristics of participants at baseline

	No. (%)
Baseline measure	Total (N = 427)	EUC (n = 216)	PMSB-E (n = 211)	Test statistic (df); P
Survey language
English	123 (28.8%)	64 (29.6%)	59 (28.0%)	χ2(3) = 0.50; P = 0.92
Spanish	194 (45.4%)	96 (44.4%)	98 (46.4%)
Mandarin	66 (15.5%)	35 (16.2%)	31 (14.7%)
Cantonese	44 (10.3%)	21 (9.7%)	23 (10.9%)
Study location
Massachusetts	269 (63.0%)	136 (63.0%)	133 (63.0%)	χ2 (3) = 0.10; P = 0.99
Puerto Rico	88 (20.6%)	45 (20.8%)	43 (20.4%)
New York	53 (12.4%)	26 (12.0%)	27 (12.8%)
Maryland	17 (4.0%)	9 (4.2%)	8 (3.8%)
Age, mean (SD), y	71.3 (7.7)	70.8 (7.6)	71.9 (7.7)	t(425) = −1.53; P = 0.13
Gender
Male	90 (21.1%)	38 (17.6%)	52 (24.6%)	χ2 (1) = 3.19; P = 0.07
Female	337 (78.9%)	178 (82.4%)	159 (75.4%)
Race and ethnicity
Non-Latino White	19 (4.4%)	12 (5.6%)	7 (3.3%)	χ2(4) = 2.33; P = 0.68
Non-Latino Black	90 (21.1%)	46 (21.3%)	44 (20.9%)
American Indian	1 (0.2%)	0 (0.0%)	1 (0.5%)
Asian	117 (27.4%)	59 (27.3%)	58 (27.5%)
Latino	200 (46.8%)	99 (45.8%)	101 (47.9%)
Birthplace
Foreign born	246 (57.6%)	124 (57.4%)	122 (57.8%)	χ2(1) = 0.01; P = 0.93
US born	181 (42.4%)	92 (42.6%)	89 (42.2%)
Education
Less than high school	153 (35.8%)	75 (34.7%)	78 (37.0%)	χ2(1) = 0.27; P = 0.60
High school and above	273 (63.9%)	141 (65.3%)	132 (62.6%)
Missing	1 (0.2%)	0 (0.0%)	1 (0.5%)
Marital status
Married or cohabitating	140 (32.8%)	65 (30.1%)	75 (35.5%)	χ2(2) = 2.92; P = 0.23
Widowed, divorced, or separated	236 (55.3%)	120 (55.6%)	116 (55.0%)
Never married	46 (10.8%)	28 (13.0%)	18 (8.5%)
Missing	5 (1.2%)	3 (1.4%)	2 (0.9%)
Insurance status
Uninsured	1 (0.2%)	0 (0.0%)	1 (0.5%)	χ2(1) = 1.03; P = 0.31
Insured	421 (98.6%)	214 (99.1%)	207 (98.1%)
Missing	5 (1.2%)	2 (0.9%)	3 (1.4%)
Socioeconomic statusa
Living very well or comfortably	141 (33.0%)	74 (34.3%)	67 (31.8%)	χ2(2) = 0.86; P = 0.65
Living month-to month or check-to-check	213 (49.9%)	104 (48.1%)	109 (51.7%)
Living almost poor or poor	66 (15.5%)	36 (16.7%)	30 (14.2%)
Missing	7 (1.6%)	2 (0.9%)	5 (2.4%)
Cognitive impairment, MoCA 5-minute protocol
No cognitive impairment	213 (49.9%)	104 (48.1%)	109 (51.7%)	χ2(2) = 0.84; P = 0.66
Mild cognitive impairment	128 (30.0%)	69 (31.9%)	59 (28.0%)
Major cognitive impairment	86 (20.1%)	43 (19.9%)	43 (20.4%)
No. of chronic conditions, mean (SD)	4.3 (1.8)	4.3 (1.8)	4.2 (1.8)	t(425) = 1.02; P = 0.31
Screening measures, mean (SD)
Depressive symptoms, PHQ-9	9.5 (4.8)	9.5 (5.1)	9.5 (4.5)	t(425) = 0.02; P = 0.98
Depressive symptoms, GDS-15	5.4 (3.4)	5.3 (3.5)	5.5 (3.4)	t(425) = −0.69; P = 0.49
Anxiety symptoms, GAI	9.0 (6.0)	8.7 (5.9)	9.3 (6.0)	t(425) = −0.96; P = 0.34
Primary outcomes, mean (SD)
Depression and anxiety symptoms, HSCL-25	1.6 (0.4)	1.6 (0.4)	1.6 (0.4)	t(425) = −0.15; P = 0.88
Physical functioning, SPPB	7.9 (2.1)	7.8 (2.2)	8.0 (2.1)	t(405) = −0.67; P = 0.51
Physical Functioning, LLFDI	115.4 (26.6)	114.2 (28.1)	116.6 (25.0)	t(425) = −0.91; P = 0.36
Secondary outcomes, mean (SD)
Functioning, WHODAS 2.0	23.4 (8.2)	23.1 (8.1)	23.7 (8.3)	t(425) = −0.69; P = 0.49

Abbreviations: EUC, Enhanced Usual Care; PMSB-E, Positive Minds Strong Bodies-Enhanced; MoCA, Montreal Cognitive Assessment; HSCL-25, Hopkins Symptom Checklist-25 (range 0 to 4); SPPB, Short Physical Performance Battery (range 0 to 12); LLFDI, Late-Life Functioning and Disability Instrument (range 32 to 160); WHODAS 2.0, WHO Disability Assessment Schedule 2.0 (range 12 to 60).

^aSocioeconomic status was measured with one self-reported item “Of the following phrases, which best describes your family’s economic status” (Living very well or comfortably, Living month-to month or check-to-check, Living almost poor or poor).

Acceptability

Intervention participants attended a mean (SD) of 8.02 (3.51) PM sessions and 21.78 (15.54) SB sessions, with no differences in attendance across sites (F_3,210 =1.50, P =0.22 for PM and F_3,210 =0.66, P =0.58 for SB). One-hundred and sixty-five (78.2%) attended ≥6 PM sessions, 120 (56.9%) attended ≥24 SB sessions, and 114 (54.0%) attended ≥6 PM and ≥24 SB sessions. At month 12, 89.9% of intervention participants were ‘very satisfied’ with the care received by their CHWs/ETs, and 91.1% were ‘very satisfied’ with the program.

Safety

Two serious adverse events (SAEs) were recorded during the trial, one related and one unrelated to the study protocol. Three deaths occurred among participants, unrelated to the protocol. A detailed description is included in the Supplemental Material.

Primary Endpoint

From baseline to month 12, PMSB-E participants had greater improvements in SPPB total score than EUC participants (0.60-point difference; BY adjusted 95% CI, 0.07 to 1.14; t(1738.58) =2.65; BH adjusted P =0.02). There were no statistically significant intervention effects in either HSCL-25 total scores (0.01-point difference; BY adjusted 95% CI, −0.06 to 0.09; t(2114.24) =0.34; BH adjusted P =0.84) or LLFDI total scores (2.70-point difference; BY adjusted 95% CI, −1.32 to 6.71; t(2203.70) =1.32; BH adjusted P =0.38) (Table 2).

Table 2.

Summary of Primary, Key Secondary, and Exploratory Endpoints

	Intent-to-treat
	Estimated mean (SE)		Difference (FDR adjusted CI)a	t-test [df] (FDR adjusted P)b	Cohen d
Outcome	EUC	PMSB-E
HSCL-25
3-monthc	1.52 (0.03)	1.45 (0.03)	−0.07 (−0.14, 0.01)	t(3632.34) = −1.83; P = 0.07	ns
6-monthc	1.50 (0.03)	1.49 (0.03)	−0.01 (−0.08, 0.07)	t(3394.47) = −0.21; P = 0.84	ns
12-monthc	1.50 (0.03)	1.51 (0.03)	0.01 (−0.06, 0.09)	t(2114.24) = 0.34; P = 0.84	ns
SPPB
3-monthc	8.03 (0.15)	8.78 (0.17)	0.75 (0.31, 1.20)	t(1909.70) = 3.32; P = 0.001	0.35
6-monthc	8.08 (0.15)	8.81 (0.16)	0.73 (0.21, 1.25)	t(2452.23) = 3.29; P = 0.006	0.34
12-monthc	8.19 (0.15)	8.80 (0.17)	0.60 (0.07, 1.14)	t(1738.58) = 2.65; P = 0.02	0.28
LLFDI
3-monthc	113.39 (1.37)	118.39 (1.42)	5.00 (1.12, 8.89)	t(3903.80) = 2.52; P = 0.01	0.19
6-monthc	113.03 (1.34)	114.52 (1.44)	1.50 (−2.40, 5.40)	t(3642.65) = 0.75; P = 0.68	ns
12-monthc	112.13 (1.40)	114.83 (1.47)	2.70 (−1.32, 6.71)	t(2203.70) = 1.32; P = 0.38	ns
WHODAS 2.0d
3-monthc	22.54 (0.52)	21.04 (0.54)	−1.50 (−2.99, −0.02)	t(2643.60) = −1.98; P = 0.047	0.18
6-monthc	23.01 (0.51)	21.88 (0.56)	−1.13 (−2.62, 0.37)	t(2492.44) = −1.48; P = 0.14	ns
12-monthc	22.86 (0.53)	22.69 (0.57)	−0.17 (−1.71, 1.37)	t(1578.35) = −0.21; P = 0.83	ns
	Per-protocole
	Estimated mean (SD)		Difference (95% CI)	Test statisticf	Cohen d
Outcome	EUC	PMSB-E
HSCL-25
3-monthc	1.58 (0.05)	1.40 (0.03)	−0.18 (−0.29, −0.06)	NA	0.43
6-monthc	1.52 (0.05)	1.43 (0.03)	−0.09 (−0.21, 0.02)	NA	ns
12-monthc	1.53 (0.07)	1.50 (0.04)	−0.03 (−0.19, 0.12)	NA	ns
SPPB
3-monthc	7.92 (0.24)	9.26 (0.17)	1.35 (0.78, 1.91)	NA	0.63
6-monthc	8.50 (0.32)	9.15 (0.20)	0.65 (−0.06, 1.38)	NA	ns
12-monthc	8.14 (0.47)	9.11 (0.20)	0.97 (0.02, 1.98)	NA	0.45
LLFDI
3-monthc	115.46 (1.76)	120.45 (2.13)	4.99 (0.36, 9.49)	NA	0.19
6-monthc	116.29 (3.61)	116.05 (2.54)	−0.24 (−8.36, 6.96)	NA	ns
12-monthc	118.19 (2.39)	114.93 (2.47)	−3.26 (−9.01, 3.00)	NA	ns
WHODAS 2.0d
3-monthc	22.67 (0.78)	19.79 (0.73)	−2.88 (−4.73, −0.89)	NA	0.35
6-monthc	22.75 (0.95)	20.41 (0.69)	−2.34 (−4.63, −0.13)	NA	0.29
12-monthc	22.90 (1.48)	22.47 (0.93)	−0.42 (−3.92, 2.54)	NA	ns

Abbreviations: EUC, Enhanced Usual Care; PMSB-E, Positive Minds Strong Bodies-Enhanced; FDR, False Discovery Rate; HSCL-25, Hopkins Symptom Checklist-25 (range 1 to 4); SPPB, Short Physical Performance Battery (range 0 to 12); LLFDI, Late-Life Functioning and Disability Instrument (range 32 to 160); WHODAS 2.0, WHO Disability Assessment Schedule 2.0 (range 12 to 60); NA, not applicable; ns, not significant. Figure S1 of the Supplemental Material provides a graphical representation of the intent-to-treat results.

^aConfidence intervals are adjusted for FDR using the Benjamini-Yekutieli correction, which is the analog to adjusting P-values for FDR using the Benjamini-Hochberg correction.

^bP-values are adjusted for FDR using the Benjamini-Hochberg correction.

^c3-month was an exploratory endpoint, 6-month a key secondary endpoint, and 12-month the primary endpoint.

^dWHODAS 2.0 total score was a secondary outcome.

^ePer-protocol analyses used propensity score genetic matching to create an appropriate counterfactual in the control group (i.e., participants who would have completed the recommended treatment dose if assigned to the intervention).

^fConfidence intervals in per-protocol analyses are calculated using the 2.5% and 97.5% quantiles of the posterior distribution of the difference between EUC and PMSB-E groups within a Bayesian framework. Thus, no test statistic is associated with this difference between EUC and PMSB-E groups.

Key Secondary Endpoint

The pattern of change in the key secondary endpoint was consistent with that of the primary endpoint. Six months after baseline, PMSB-E participants had greater improvements than EUC participants in SPPB (0.73-point difference; BY adjusted 95% CI, 0.21 to 1.25; t(2452.23) =3.29; BH adjusted P =0.006). No significant intervention effects were observed in either HSCL-25 (−0.01-point difference; BY adjusted 95% CI, −0.08 to 0.07; t(3394.47) =−0.21; BH adjusted P =0.84) or LLFDI (1.50-point difference; BY adjusted 95% CI, −2.40 to 5.40; t(3642.65) =0.75; BH adjusted P =0.68) (Table 2). Our joint test of global effectiveness also found a significant effect of the intervention on SPPB scores at 6 and 12 months at the 0.05 level (see Supplemental Material). Subgroup analyses by race and ethnicity were generally consistent with the primary and key secondary endpoint results, with no significant intervention × race/ethnicity interactions (Figure 2).

Figure 2.

Model Implied Coefficient Estimates for Hopkins Symptom Checklist-25 (HSCL-25), Short Physical Performance Battery (SPPB), Late-Life Functioning and Disability Instrument (LLFDI), and WHO Disability Assessment Schedule 2.0 (WHODAS 2.0) According to Treatment Groups and Race and Ethnicity

Abbreviations: EUC, Enhanced Usual Care; PMSB-E, Positive Minds Strong Bodies-Enhanced; Coef., coefficient; CI, confidence interval. HSCL-25 scores range from 1 to 4; SPPB scores range from 0 to 12; LLFDI scores range from 32 to 160; WHODAS 2.0 scores range from 12 to 60.

* Confidence intervals are adjusted for False Discovery Rate using the Benjamini-Yekutieli correction, which is the analog to adjusting P-values using the Benjamini-Hochberg correction.

** WHODAS 2.0 was a secondary outcome

Secondary Outcome and Exploratory Endpoint

Participants in the PMSB-E group did not show greater improvements in WHODAS 2.0 total scores from baseline to month 6 or month 12. Except for HSCL-25, PMSB-E showed greater improvements vs. EUC on all other outcomes at the exploratory 3-month endpoint: point-difference SPPB: 0.75 (95% CI, 0.31 to 1.20; t(1909.70) =3.32; P < 0.001), LLFDI: 5.00 (95% CI, 1.12 to 8.89; t(3903.80) =2.52; P = 0.01), and WHODAS 2.0: −1.50 (95% CI, −2.99 to −0.02; t(2643.60) =−1.98; P = 0.047) (Table 2).

Per-Protocol and Post-Hoc Analyses

In per-protocol analyses, PMSB-E participants who completed ≥6 PM and ≥24 SB sessions had larger improvements vs. intention-to-treat in SPPB from baseline to month 12 (per-protocol point-difference, 0.97; intention-to-treat, 0.60). Larger improvements were also observed at the exploratory 3-month endpoint in HSCL-25 (per-protocol point-difference, −0.18; intention-to-treat, not-significant), SPPB (per-protocol point-difference, 1.35; intention-to-treat, 0.75), and WHODAS 2.0 (per-protocol point-difference, −2.88; intention-to-treat, −1.50) (Table 2).

In post-hoc analyses, intervention effects on HSCL-25 remained non-significant (Supplemental Material). The lack of improvement in mental health symptoms (HSCL-25) might have affected self-perceived physical functioning (LLFDI), which also did not improve despite the significant intervention effect in observer-rated physical functioning (SPPB). In additional post-hoc analyses, we evaluated whether individual differences in HSCL-25 were associated with individual differences in LLFDI at 12 months (adjusting for intervention condition). Indeed, a clinically significant improvement in HSCL-25 at 12 months (from ≥1.75 to <1.75) was associated with a 17.5-point increase in LLFDI (t(315) =4.98; P <0.001), exceeding the 5-point minimal detectable substantial change in LLFDI scores.³⁵ Regarding PM session mode of administration, results from our post-hoc analyses indicated that some self-reported outcomes improved significantly when sessions were delivered in-person/Zoom compared to phone (Table 3). Most notably, significant improvements were observed at all endpoints (months 3, 6, and 12) for the secondary outcome of self-reported disability (WHODAS 2.0 scores) when completing the PM sessions in-person/Zoom, but not over the phone. PM mode of administration did not influence effectiveness on the objectively measured SPPB scores since SB sessions were done in person/Zoom.

Table 3.

Summary of Primary, Key Secondary, and Exploratory Endpoints by Positive Minds Sessions Mode of Administration

	Phone				In-person/Zoom
	Estimated mean (SD)		Difference (95% CI)c	Cohen d	Estimated mean (SD)		Difference (95% CI)c	Cohen d
Outcome	EUC	PMSB-E			EUC	PMSB-E
HSCL-25
3-montha	1.66 (0.07)	1.52 (0.05)	−0.15 (−0.32, 0.02)	ns	1.51 (0.04)	1.41 (0.03)	−0.10 (−0.19, −0.01)	0.25
6-montha	1.63 (0.07)	1.59 (0.05)	−0.03 (−0.21, 0.14)	ns	1.49 (0.05)	1.40 (0.03)	−0.09 (−0.19, 0.04)	ns
12-montha	1.66 (0.04)	1.61 (0.06)	−0.05 (−0.19, 0.08)	ns	1.39 (0.11)	1.41 (0.04)	0.02 (−0.22, 0.21)	ns
SPPB
3-montha	7.15 (0.39)	8.46 (0.25)	1.31 (0.50, 2.16)	0.61	7.51 (0.30)	8.97 (0.23)	1.46 (0.75, 2.15)	0.68
6-montha	7.19 (0.43)	8.53 (0.28)	1.34 (0.37, 2.30)	0.63	7.47 (0.36)	8.88 (0.26)	1.41 (0.57, 2.24)	0.66
12-montha	6.95 (0.48)	8.33 (0.24)	1.38 (0.33, 2.38)	0.65	7.55 (0.45)	9.02 (0.30)	1.47 (0.46, 2.53)	0.69
LLFDI
3-montha	110.13 (2.59)	117.25 (3.11)	7.12 (1.29, 13.17)	0.27	112.96 (2.07)	116.98 (2.51)	4.02 (−0.87, 9.05)	ns
6-montha	112.02 (2.63)	113.77 (2.96)	1.76 (−4.40, 7.90)	ns	113.25 (2.38)	114.91 (2.70)	1.66 (−3.86, 7.28)	ns
12-montha	108.17 (2.68)	113.53 (3.26)	5.36 (−1.60, 12.29)	ns	111.43 (2.50)	114.67 (3.13)	3.24 (−3.18, 9.68)	ns
WHODAS 2.0b
3-montha	22.63 (1.47)	19.80 (1.03)	−2.83 (−6.33, 0.20)	ns	22.23 (0.80)	20.00 (0.82)	−2.23 (−4.22, −0.21)	0.27
6-montha	25.20 (1.34)	21.55 (1.10)	−3.65 (−7.10, −0.36)	0.45	22.28 (0.83)	19.59 (0.69)	−2.69 (−4.51, −0.71)	0.33
12-montha	23.24 (1.46)	23.05 (1.20)	−0.18 (−4.14, 3.18)	ns	22.63 (1.10)	19.68 (0.87)	−2.95 (−5.43, −0.32)	0.36

Abbreviations: EUC, Enhanced Usual Care; PMSB-E, Positive Minds Strong Bodies-Enhanced; HSCL-25, Hopkins Symptom Checklist-25 (range 1 to 4); SPPB, Short Physical Performance Battery (range 0 to 12); LLFDI, Late-Life Functioning and Disability Instrument (range 32 to 160); WHODAS 2.0, WHO Disability Assessment Schedule 2.0 (range 12 to 60); ns, not significant. Figure S1 of the Supplemental Material provides a graphical representation of the intent-to-treat results.

^a3-month was an exploratory endpoint, 6-month a key secondary endpoint, and 12-month the primary endpoint.

^bWHODAS 2.0 total score was a secondary outcome.

^cConfidence intervals in analyses by mode of administration are calculated using the 2.5% and 97.5% quantiles of the posterior distribution of the difference between EUC and PMSB-E groups within a Bayesian framework. Thus, no test statistic is associated with this difference between EUC and PMSB-E groups.

Discussion and Conclusions

This study examined whether PMSB-E slowed functional decline and improved depression and anxiety symptoms in racially, ethnically, and linguistically diverse older adults ≥60. Results provided mixed findings on PMSB-E effectiveness, with clear improvement in individuals’ objective mobility and functionality but no effect on self-reported physical functioning or mental health.

Like the parent study,¹⁴ we found significant improvements in objectively measured physical functioning (SPPB scores) through balance, walking speed, and muscle strength at three, six, and 12 months. According to the available literature, the observed improvement of 0.6 points is an indication of a clinically significant change.³⁶ Further, 6-month intervention effects were larger in the enhanced PMSB-E than in the parent study (Cohen’s d, 0.34 vs 0.27). These results underline the importance of including exercises that emphasize muscle power (velocity of movement), an essential determinant of activities of daily living. Our findings are consistent with another RCT with significantly improved SPPB scores.³⁷ Addressing physical health through exercise is paramount, given its preventive effects on disability,³⁸ biological aging,² and dementia.³⁹

Contrary to our hypotheses that, like in the parent study, PMSB-E would significantly improve mental health,¹⁴ our results show no reductions in symptoms of depression and anxiety at the primary 12-month or key secondary six-month endpoint, when compared to the control group. Even so, significant 3-month improvements in LLFDI and WHODAS 2.0 scores were observed. The lack of impact on depression and anxiety symptoms remains puzzling, given parent study results showing a small effect size at six and 12 months (Cohen’s d, 0.28 and 0.24, respectively).¹⁴ This result is also in contrast with other studies reporting significant impacts of exercise interventions on anxiety and depressive symptoms in older persons.⁴⁰ However, a recent systematic review of CHW-led interventions also found mixed results.⁴¹

Several possible explanations exist. One is that PMSB-E was conducted during COVID-19, which affected functioning and psychological aspects of the older population. Older adult participants may have experienced reduced social interactions and heightened levels of depression and anxiety, which were challenging to address.^42,43 Furthermore, the parent study was mostly in person (85.9% vs 14.1% by phone). In PMSB-E, 46.3% of PM sessions were done by telephone, and 53.7% by Zoom/in person because of COVID-19. Not all interventions implemented by phone are effective.^44,45 Older adults may have vision, hearing, and memory issues that can affect engagement with digital interventions.⁴⁶ While the SB component could use visual cues over Zoom, telephone-delivered PM was auditory only, limiting demonstration of abstract concepts.

There was also a mismatch in improvement of functioning between the SPPB (administered) and the LLFDI (self-report). Previous research has suggested that there are discrepancies between objective and subjective measures of physical health.⁴⁷ While the SPPB measure offers an objective assessment of physical health, the LLFDI can offer a complementary perspective, capturing how participants experience perceived changes in physical functioning. Unlike the parent intervention, the enhanced intervention did not improve mental health symptoms, which we showed possibly affected participants’ self-perception of functioning. Finally, PMSB-E enhancements suggested by parent trial participants did not improve effectiveness. Participants expressed treatment fatigue for group maintenance (particularly through Zoom) and insufficient time to attend coordinated groups (also described in Pywell et al.⁴⁸).

Limitations

Our sample, although diverse, was predominantly female (78.9%) and Latino (46.8%). Our findings may not be generalized to all cultural groups but highlight the need to improve outreach to males and other minoritized groups. Despite rigorous translation and adequate psychometrics across languages, differences in understanding items in diverse languages could have impacted results, although randomization should have balanced this. Furthermore, the average participant in our study was classified as “young-old” (ages 65–74), rather than “middle-old” (ages 75–84) or “oldest-old” (ages 85+).⁴⁹ Previous research has suggested that levels of physical activity and severity of mental health symptoms change as older adults transition to different age groups.⁴⁹ It is unclear whether the PMSB-E intervention would have different effects for older adults across various age groups.

Only more than half of participants attended 24+ SB sessions. Anecdotally, our care managers reported participant challenges, like medical (e.g., surgery recovery) or technological issues (e.g., poor Wi-Fi). Without a multifactorial design, it was not possible to determine whether the combined intervention was more effective than the exercise intervention alone. Although scalability and reach are greater, mental health virtual interventions might not work well for older adults. Complex terminology around cognitive restructuring or mindfulness might require in-person prompts. Group coordination difficulties may have limited PM maintenance effectiveness. Formal CHW language proficiency testing, which we did not administer, might ensure that language miscommunication is not a barrier.

Previous research results which were grounded in established theory of mental health therapeutic effects led us to expect that the PMSB-E intervention would be effective, despite changes in study period and modifications of the intervention administration method. Our results showed that intervention effectiveness was not resilient to the changes required in response to the study period pandemic context. It is important, however, not to interpret the negative results as a definitive challenge to the theory and method tested. Just as a single RCT cannot establish a general claim of effectiveness, a single replication RCT cannot definitively discredit the promise of PMSB-E in other contexts.⁵⁰

Conclusion

The present study is one of few to combine mental and physical health components to address disability in racial, ethnic, and linguistic minoritized older adults. Our results demonstrate the importance of replicating past RCTs to confirm under what conditions interventions might or might not work. They also raise questions about what mode of administration researchers should use to deliver these interventions to older patients. To better address the needs of minoritized older adults, future research should address needed technology training for virtual interventions to be effective, not just availability. Similarly, increased use of visuals over complex terminology for mental health concepts could communicate skills in an accessible and responsive manner, allowing interventions for older adults to be administered in a more scalable format.

Highlights.

1). What is the primary question addressed by this study?

• The following multisite clinical trial tested whether the enhanced version of Positive Minds-Strong Bodies (PMSB-E), a culturally responsive disability prevention intervention, can significantly improve physical functioning and mental health symptoms among diverse older adults.

2). What is the main finding of this study?

• PMSB-E significantly improved physical functioning among diverse older adults. However, there were no significant improvements in depressive and anxiety symptoms, possibly because of the impact of COVID and the need to shift the method of administration to virtual, with a high proportion of participants doing the POSITIVE MINDS component through phone.

3). What is the meaning of the finding?

• PMSB-E can significantly enhance physical functioning, but in-person administration might be required for participants to benefit from the mental health component.

Data Sharing

The data collected for this study will not be made publicly available to others due to the sensitivity of the data involving a racial and ethnic minoritized sample, which includes undocumented immigrants, and our agreements with the Institutional Review Board. In our consent forms, we promised that none of the information would be shared, except among research staff. We cannot provide the data unless they were highly de-identified, which would take substantial time.

Data Statement

The data has not been previously presented orally or by poster at scientific meetings.