Impact of Moderating Demographic Variables on a Health Intervention for People with Serious Mental Illness

Abstract

Purpose: Healthy lifestyle interventions can improve the health of people with serious mental illness (SMI). Little is known whether demographic variables moderate the effectiveness of these interventions on health outcomes. Method: Data from an effectiveness trial of a peer-led healthy lifestyle intervention (PGLB) for people with SMI examine whether age, racial/ethnic minoritized status, and gender moderated the effectiveness of PGLB compared to usual care (UC) in achieving clinically significant improvements in weight, cardiorespiratory fitness, and cardiovascular disease (CVD) risk reduction. Results: Compared to UC, PGLB was most beneficial for participants age 49 and younger for achieving clinically significant weight loss and for racial/ethnic minoritized communities for achieving clinically significant weight loss and reductions in CVD risk. Conclusions: These findings suggest the impact of healthy lifestyle interventions for people with SMI may not be uniform and adaptations may be needed to make these interventions responsive to the needs of diverse populations.

Keywords

serious mental illness peer-led interventions healthy lifestyle interventions supportive housing

Serious mental illness (SMI) encompasses a cluster of mental disorders, such as schizophrenia spectrum disorders, bipolar disorder, major depression, and many others that interfere with major life activities, such as work, education, and social relationships resulting in serious functional limitations (Interdepartmental Serious Mental Illness Coordinating Committee, 2017; Kessler et al., 2003). Improving the physical health of people with SMI is a public health priority given the deadly health inequities experienced by this historically marginalized population (Firth, Siddiqi, Koyanagi, Siskind, Rosenbaum, 2019). Compared to the general population, people with SMI die approximately 10–25 years earlier largely due to elevated rates of preventable health conditions like obesity, type-2 diabetes, and cardiovascular disease (CVD) (Firth et al., 2019). Healthy lifestyle interventions that use behavioral strategies (e.g., self-monitoring, goal setting) to support healthy dietary habits and increase physical activity can be beneficial for people with SMI. Clinical trials in the U.S. indicate that between 30% to 40% of people with SMI participating in healthy lifestyle interventions achieve clinically significant weight reductions (i.e., weight loss of at least 5% from baseline) and improvements in cardiorespiratory fitness (CRF; i.e., the circulatory and respiratory systems ability to supply oxygen to muscles during physical activity) (Aschbrenner, Naslund, Gorin, Mueser, Browne, 2021; Bartels, Pratt, Aschbrenner, Barre, Jue, 2013; Bartels, Pratt, Aschbrenner, Barre, Naslund, 2015; Daumit, Dickerson, Wang, Dalcin, Jerome, 2013; Green, Yarborough, Leo, Yarborough, Stumbo, 2015; Vancampfort, Rosenbaum, Probst, Soundy, Mitchell, 2015). Both outcomes independently reduce the risk for CVD and premature mortality (Vancampfort et al., 2015; Wise & Brown, 2005).

Despite these promising results, existing literature indicates that little is currently known as to whether the effectiveness of healthy lifestyle interventions for people with SMI differ by key participant characteristics associated with weight loss in the general population, like age, gender, and minoritized racial/ethnic status (Cabassa et al., 2010; 2017; McGinty, Baller, Azrin, Juliano-Bult, Daumit, 2015). Studies focusing on people without SMI have found that being male, older, and non-Hispanic white were associated with greater weight loss compared to their counterparts (Kumanyika, Espeland, Bahnson, Bottom, Charleston, 2002; Wadden, West, Neiberg, Wing, Ryan, 2009; West, Elaine, Bursac, & Felix, 2008). Identifying which subgroups of people with SMI benefit most from healthy lifestyle interventions is critical as it can inform targeted efforts for implementation and help clarify if adaptations are needed for certain subgroups (Alexander, McGinty, Wang, Dalcin, Jerome, 2019).

Evidence of subgroup differences in people with SMI participating in healthy lifestyle interventions is scant. In one of the few studies to systematically examine subgroup differences among people with SMI, none of the participant characteristics examined (e.g., gender, age, or racial/ethnic minoritized status) moderated the impact of an 18-month healthy lifestyle intervention on weight loss (Alexander et al., 2019). These findings are encouraging as they suggest that this behavioral weight loss intervention worked similarly across a diverse group of people with SMI.

Yet, key gaps remain. No study to date has replicated these findings in other large effectiveness trials of healthy lifestyle interventions for people with SMI. Moreover, no study has examined how age, gender, and racial/ethnic minoritized status moderate the effects of healthy lifestyle interventions on other health outcomes beyond overall weight loss for people with SMI, such as achieving clinically significant improvements in weight loss, CRF, and CVD risk. All are key outcomes that can help reduce the risk of premature mortality in people with SMI. Lastly, no study to date has examined these moderators in a peer-led healthy lifestyle intervention delivered by peer specialists - people with lived experiences recovering from SMI - in community settings like supportive housing (Cabassa et al., 2017). Supportive housing programs are a critical setting for serving people with SMI providing housing along with general medical, mental health, and social services (Nelson & Laurier, 2010). Delivering healthy lifestyle interventions in supportive housing agencies helps move these interventions closer to the community, reducing access barriers by bringing these interventions to “people’s doorsteps” (Henwood, Cabassa, Craig, & Padgett, 2013). Further, utilizing peer providers who can build off a foundation of shared lived experiences to lead intervention delivery may further expand the reach and responsiveness of health interventions for people with SMI (Bochicchio, Stefancic, Gurdak, Swarbrick, Cabassa, 2019).

Present Study

The present study addresses these critical gaps by using data from an effectiveness trial of a peer-led healthy lifestyle intervention (PGLB) for people with SMI to examine whether age, racial/ethnic minoritized status, and gender moderated the effectiveness of PGLB compared to usual care (UC) in achieving clinically significant improvements in weight, cardiorespiratory fitness, and cardiovascular disease (CVD) risk reduction (Cabassa et al., 2015). The main results of this trial indicated that although a larger proportion of participants randomized to PGLB versus UC achieved clinically significant weight loss, clinically significant increases in CRF and clinically significant reductions in CVD risk at 18 months, none of these differences were statistically significant (Cabassa, Stefancic, Lewis-Fernández, Luchsinger, Weinstein, 2021). Despite these null findings, PGLB achieved outcomes comparable to those of other U.S. based healthy lifestyle trials of people with SMI (Bartels et al., 2015; Daumit et al., 2013; Green et al., 2014) and those reported in a recent meta-analysis (Naslund, Whiteman, McHugo, Aschbrenner, Marsch, 2017). The present study extends the existing work of this trial by examining whether the impact of a peer-led healthy lifestyle intervention and usual care on clinically significant improvements in weight loss, CRF, and CVD risk reduction was moderated by demographic variables. This is the next logical step in in this line of research as it can help identify who benefitted most from PGLB compared to UC. Due to existing racial/ethnic health inequities as well as findings of differential intervention effectiveness from studies in the general population, the hypothesis for this study is that non-Hispanic white, male, and older participants would benefit most from the peer-led healthy lifestyle intervention compared to usual care in achieving these three clinically significant health outcomes.

Method

Trial Overview

The full protocol for this effectiveness trial is published elsewhere providing methodological details about study site selections, randomization, sampling approach and study measures (Cabassa et al., 2015). Briefly, this trial was conducted in three supportive housing agencies located in two Northeastern cities in the U.S. All participants gave written informed consent, and the study was approved by the institutional review boards of Columbia University, the Philadelphia Department of Public Health, and Washington University in St. Louis. The trial is registered in clinicaltrials.gov (Clinical Trials # NCT02175641).

Sample

Study inclusion criteria were minimal (Ford & Norrie, 2016). Eligible participants were residents of their respective supportive housing agency, male/female, 18 years of age or older, English or Spanish speaking with a chart diagnosis of SMI (e.g., schizophrenia; bipolar disorder), and a BMI ≥ 25 (kg/m²) assessed by a trained research assistant (RA). Participants randomized to PGLB also had to obtain a medical clearance letter from a primary care physician. Participants were excluded if at the time of recruitment needed substance use detoxification services, posed a danger to self or others, failed a capacity-to-consent questionnaire (Zayas et al., 2005), or self-reported medical conditions that contraindicated participation in a weight loss program (e.g., active cancer treatment). Participants 65 years of age or older who screened positive for cognitive impairment on the Mini-Cog clock test were also excluded (Palmer & Meldon, 2003).

Four hundred and forty-eight people were screened, 340 were eligible, and 314 were enrolled and randomly assigned to PGLB or usual care (UC). The sample for these analyses consisted of participants randomized to UC (N = 157) and those randomized to PGLB who attended at least one PGLB session (N = 131). As reported elsewhere, 26 (16.6%) participants randomized to PGLB did not attend any sessions and were excluded from the current analyses (Tuda, Stefancic, Hawes, Wang, Guo, 2021).

Intervention

The Peer-led Group Lifestyle Balance (PGLB) is a 12-month manualized healthy lifestyle intervention adapted from the Group Lifestyle Balance (GLB) intervention to meet the needs of people with SMI living in supportive housing and to be delivered by peer specialists (Kramer, Kriska, Venditti, Miller, Brooks, 2009; O’Hara et al., 2017). PGLB consisted of weekly core sessions for 3 months, bi-monthly transition sessions for 3 months, and monthly maintenance sessions for the remaining 6 months, for a total of 22 sessions over the course of a year. Sessions lasted 60 minutes, on average, and were delivered to a group of three to six participants in their housing agency with the option of receiving individual make-up sessions if sessions were missed. The intervention focused on improving dietary habits and increasing physical activity by using behavioral techniques (e.g., self-monitoring, problem solving).

Certified peer specialists delivering PGLB were employed at their respective housing agencies and were trained and supervised by the study team. This training and supervision approach has been described in detail elsewhere (Stefancic, Bochicchio, Tuda, Harris, DeSomma, 2021; Vélez-Grau et al., 2019). Briefly, our training included: 1) a 2-day GLB certification program delivered by a GLB master trainer, and 2) a 3-month intensive session-by-session training that included peer-specialists using intervention elements in their everyday lives (e.g., using food logs, practicing calorie counting) and delivering mock sessions to the study team and agency supervisors prior to facilitating the intervention. The study team monitored intervention fidelity throughout the trial by reviewing session audio recordings and rating the degree to which key elements of PGLB were present in these sessions (Cabassa et al., 2021). Weekly supervision meetings took place in-person or via telephone during which the study team provided constructive feedback to peer specialists on session delivery and performance to avoid intervention drift.

Usual Care

Usual Care (UC) for physical health at each supportive housing site consisted of health promotion groups (e.g., tobacco cessation, cooking, understanding diabetes, yoga), linkages to medical care and community resources (e.g., gym). These services were not manualized interventions and focused mostly on health education. Agency staff (e.g., case managers, nurses) at study sites helped clients connect with primary care and specialized health services as needed based on participants’ health needs.

Outcome Measures

Three outcome measures were examined: clinically significant weight loss, clinically significant improvements in CRF, and clinically significant reductions in CVD risk. The proportion of participants who achieved clinically significant weight loss, defined as weight loss equal to or larger than 5% from baseline at 6-, 12-, and 18-months. Weight (lb.) was measured with a calibrated digital scale with participants wearing indoor clothing without shoes (Cabassa et al., 2015). CRF is “the ability of the circulatory and respiratory systems to supply oxygen to working muscles during sustained physical activity” (Vancampfort et al., 2015, p. 132). CRF was measured with the 6-minute walk test (6MWT), an objective measure of functional exercise capacity that captures the distance (in meters) that participants walk at a normal pace along a flat and straight course for 6 minutes (Rasekaba, Lee, Naughton, Williams, & Holland, 2009). The 6MWT is a reliable and valid measure for obese adults and has been used in several trials of people with SMI (Bartels et al., 2013). Clinically significant improvement in CRF was defined as an increase of 50 m or more in the 6MWT from baseline (Bartels et al., 2013, 2015). This level of improvement is associated with reductions in CVD risk (Rasekaba et al., 2009; Wise & Brown, 2005). Consistent with previous trials, clinically significant reduction in CVD risk was defined as either weight loss of ≥5% from baseline or an increase of ≥50 m on the 6MWT from baseline (Bartels et al., 2015).

Moderators and Sample Characteristics

All moderators and sample characteristics were collected during structured face-to-face interviews at baseline prior to randomization (Cabassa et al., 2015). Three demographic moderators were examined. Participants’ age was dichotomized using the median split yielding a group consisting of those 49 years old or younger and another group of those 50 years old or older. Participants’ gender was self-reported and dichotomized as male or female. Racial/ethnic minoritized status was self-reported and dichotomized into two groups: non-Hispanic Whites and participants from racial/ethnic minoritized groups (e.g., Blacks, Hispanics). These two groups were created since most participants in the trial were from racial/ethnic minoritized groups (81.47%), mostly non-Hispanic Blacks (56.64%). The small sample size of Hispanics and other racial/ethnic minoritized groups (e.g., Asians) enrolled in this trial prevented the examination of more granular analyses for these groups. The rationale for examining these three demographic variables as moderators is that studies in the general population have documented differential effectiveness in healthy lifestyle interventions by participants’ gender, age, and race/ethnicity (Kumanyika et al., 2002; Wadden et al., 2009; West et al., 2008). Participants’ health and mental health conditions at baseline were collected via self-reported lifetime physician confirmed items listing common medical conditions (e.g., diabetes, high cholesterol) and psychiatric disorders (e.g., schizophrenia, bipolar disorders). Research assistants measured participants’ weight using a digital scale, and height using an anthropometric tape. Body mass index (BMI; kg/m2) was calculated from the participants’ height and weight.

Data Analysis

A treated-sample approach rather than an intent-to-treat approach was used to examine how participants’ gender, racial/ethnic minoritized status, and age moderated the impact of receiving either PGLB or UC on the trial’s main outcomes. This treated approach ensures that PGLB participants included in these analyses were exposed to the intervention while those randomized to PGLB who did not attend any sessions and thus not exposed to PGLB throughout the trial (N = 26) were excluded (Tuda et al., 2021). Baseline differences in demographic variables between PGLB and UC participants were examined using Chi-square test and t-test for categorical and continuous variable, respectively.

Associations between the binary outcomes and their predictors were examined using mixed-effects logistic regression models via Stata (Version 15.1) due to the longitudinal structure of the study data (i.e., same subjects observed across four time points) (Hedeker & Gibbons, 2006). For each outcome, three mixed-effects logistic models were conducted, one for each moderator variable (i.e., gender model, minority/non-minority model, age model). Each model included the main effects of site, baseline weight, the two conditions (PGLB vs. UC), time (e.g., 6, 12, and 18 months), and our three moderator variables. The following two-way interaction terms were included in each of our models: moderator variable × time, moderator variable × treatment condition, and time × treatment condition. To facilitate the interpretation of statistically significant moderator variable × treatment condition interaction terms, the “margins” procedure in Stata Version 15.1 was used to estimate the predicted marginal means of achieving the respective outcome by the specific moderator variable. These findings are displayed in bar graphs stratified by treatment condition. The “margin” procedure utilizes the parameters of the main effects and the significant interaction term included in the respective models and trims from the calculation any non-significant interaction terms when estimating the predicted marginal means (Dawson & Richter, 2006). The predicted marginal means displayed in the bar graphs are proportions ranging from 0 to 1 representing the probability of participants achieving the respective outcome. Finally, three-way interaction terms (treatment condition × moderator variables × time) were examined for each of our three outcomes to determine whether the interaction effect between treatment condition and each moderator variable varied by time.

Results

Sample Characteristics

Two hundred and eighty-eight participants were included in these analyses (See Table 1). Participants were mostly male (55.75%) and from a racial/ethnic minoritized status (81.47%), largely non-Hispanic blacks (56.64%), with a mean age of 48.91 years (SD = 11.12). At baseline, participants had a mean BMI of 33.65 (SD = 7.18). The most common psychiatric diagnoses at baseline were major depression (76.49%), schizophrenia/schizoaffective disorder (58.6%), and bipolar disorder (47.72%). The most common medical conditions at baseline were hypertension (54.93%), high cholesterol (37.32%), and diabetes mellitus (33.57%). PGLB and UC groups did not differ significantly in any baseline characteristics except for racial/ethnic group, with a larger number of Hispanics in UC than in PGLB, and for baseline BMI and weight, which were both higher for the PGLB group (See Table 1).

Table 1.

Sample Characteristics.

	Total Sample (N = 288)		Usual Care (N = 157)		PGLB (N = 131)
	N (%)	Mean (SD)	N (%)	Mean (SD)	N (%)	Mean (SD)	P
Age (Continuous)		48.91 (11.12)		48.81 (11.6)		49.03 (10.58)	0.87
Age
49 or younger	130 (45.1)		72 (45.86)		58 (44.27)		0.79
50 and older	158 (54.86)		85 (54.14)		73 (55.73)
Gender
Male	160 (55.75)		90 (57.32)		70 (53.43)		0.47
Female	127 (44.25)		66 (42.04)		61 (46.56)
Racial/Ethnic groups
Non-Hispanic Whites	53 (18.53)		29 (18.47)		24 (18.32)		0.02
Non-Hispanic Blacks	162 (56.64)		77 (49.04)		85 (64.89)
Hispanics	33 (11.54)		23 (14.65)		10 (7.63)
Others	38 (13.29)		26 (16.56)		12 (9.16)
Racial/Ethnic Minoritized Status
Non-Minority	53 (18.53)		29 (18.47)		24 (18.32)		0.93
Minority	233 (81.47)		126 (80.25)		107 (81.68)
Self-Reported Physician Confirmed Mental Health Disorder
Schizophrenia/Schizoaffective Disorders	167 (58.6)		85 (54.14)		82 (62.6)		0.16
Depression	218 (76.49)		115 (73.25)		103 (78.63)		0.32
Bipolar Disorders	136 (47.72)		72 (45.86)		64 (48.85)		0.64
Self-reported Physician Confirmed Medical Condition
Hypertension	156 (54.93)		89 (56.69)		67 (51.15)		0.29
High Cholesterol	106 (37.32)		60 (38.22)		46 (35.11)		0.54
Diabetes Mellitus	96 (33.57)		51 (32.48)		45 (34.35)		0.80
Baseline Body Mass Index (BMI)		33.65 (7.18)		32.87 (6.72)		34.58 (7.63)	0.04
Baseline Weight (Pounds)		217.93 (54.27)		211.83 (48.83)		225.23 (59.52)	0.04
Baseline 6 Minutes Walking Test (Meter)		320.99 (97.79)		328.71 (100.77)		311.68 (93.6)	0.49
Clinically Significant Health Outcomes
Clinically Significant Weight Loss from Baseline^a
6-Month (n = 273)	42 (15.38)		25 (16.78)		17 (13.71)		0.48
12 Month (n = 245)	62 (25.31)		33 (24.26)		29 (26.61)		0.68
18 Month (n = 236)	73 (30.93)		40 (30.77)		33 (31.13)		0.95
Clinically Significant Improvement in Cardiorespiratory Fitness from Baseline^b
6-Month (n = 262)	56 (21.37)		26 (18.31)		30 (25)		0.19
12 Month (n = 231)	64 (27.71)		34 (25.95)		30 (26.61)		0.50
18 Month (n = 228)	60 (26.32)		31 (24.6)		29 (28.43)		0.51
Clinically Significant Reduction in Cardiovascular Disease Risk from Baseline^c
6-Month (n = 273)	90 (32.97)		46 (30.87)		44 (35.48)		0.42
12 Month (n = 245)	107 (43.67)		56 (41.18)		51 (46.79)		0.38
18 Month (n = 236)	113 (47.88)		63 (48.46)		50 (47.17)		0.84

^aClinically significant weight loss was defined as achieving ≥5% weight loss from baseline.

^bClinically significant improvement in cardiorespiratory fitness was defined as a ≥ 50 m increase from baseline on the 6-minute walking test.

^cClinically significant reduction in cardiovascular disease risk was defined as either achieving ≥5% weight loss from baseline or ≥ 50 m increase from baseline on the 6-minute walking test.

Trial Main Outcomes

At 18-months, 30.93% of participants achieved clinically significant weight loss, 26.32% achieved clinically significant improvements in CRF, and 47.88% achieved clinically significant reductions in CVD risk. As reported elsewhere and shown in Table 1, there were no statistically significant differences between PGLB and UC on any of these clinically significant outcomes throughout the trial (Cabassa et al., 2021).

Moderating Effects

The impact of PGLB and UC on achieving clinically significant weight loss throughout the trial was moderated by participants’ racial/ethnic minoritized status and age, but not gender (See Table 2). For this outcome, the 2-way interaction terms for treatment conditions (e.g., PGLB vs. UC) and participants’ racial/ethnic minoritized status was significant (Odds Ratio [OR] = 1.74, 95% Confidence Intervals [CI} 1.69, 81.50). As seen in Figure 1, for PGLB participants the probability of achieving clinically significant weight loss was higher for racial/ethnic minority participants than for non-Hispanic whites, while for UC participants non-Hispanic whites had a higher probability of achieving this weight loss outcome than racial/ethnic minorities. Similarly, the 2-way interaction term for treatment conditions and participants’ age (e.g., 49 and below vs. 50 and above) was significant (OR = 0.16, 95% CI 0.03, 0.77). As displayed in Figure 2, for PGLB participants the probability of achieving clinically significant weight loss throughout the trial was higher for younger participants than for older participants; while for UC, older participants had a higher probability of achieving this weight loss outcome than younger participants (See Figure 2).

Table 2.

Logistic Mixed Effects Models for Achieving Clinically Significant Weight Loss.

	Gender Model		Minority/Non-Minority Model		Age Model
	OR	95% CI	OR	95% CI	OR	95% CI
Site (Ref: Site 1)
Site 2	0.30	0.11, 0.85	0.29	0.10, 0.82	0.27	0.09, 0.75
Site 3	0.30	0.11, 0.84	0.33	0.12, 0.91	0.28	0.10, 0.78
Baseline Weight (in Pounds)	1.00	0.99, 1.00	1.00	0.99, 1.00	1.00	0.99, 1.00
Treatment (Ref: Usual Care)
PGLB	0.70	0.14, 3.48	0.10	0.01, 0.83	2.11	0.37, 12.06
Time	1.96	1.22, 3.13	1.65	0.87, 3.12	2.20	1.26, 3.84
Age (Ref: 49 and below)
50 and above	1.39	0.62, 3.07	1.37	0.62, 3.00	4.61	0.85, 24.83
Race/Ethnicity (Ref: Non-Hispanic White)
Racial/Ethnic Minority	0.34	0.13, 0.89	0.07	0.01, 0.49	0.36	0.14, 0.92
Gender (Ref: Male)
Female	1.06	0.21, 5.39	1.12	0.51, 2.45	1.09	0.49, 2.39
Two-Way Interactions Terms
Gender (Ref: Male)*Time	1.02	0.57, 0.1.83
Gender (Ref: Male)*Treatment (Ref: Usual Care)	0.91	0.19, 4.30
Race/Ethnicity (Ref: Non-Hispanic White)*Time			1.27	0.64, 2.53
Race/Ethnicity (Ref: Non-Hispanic White)*Treatment (Ref: Usual Care)			11.74	1.69, 81.50
Age (Ref: Young 49 and below)*Time					0.85	0.47, 1.54
Age (Ref: Young 49 and below)*Treatment (Ref: Usual Care)					0.16	0.03,0.77
Time*Treatment (Ref: Usual Care)	1.29	0.72, 2.31	1.27	0.71, 2.28	1.25	0.69, 2.25

Note. Statistically significant two-way interaction terms at p<0.05 are bolded.

Figure 1.

Predicted Marginal Means of Achieving Clinically Significant Weight Loss by Treatment Condition and Racial/Ethnic Minoritized Status. Note. The predicted marginal means displayed in the graph are proportions ranging from 0 to 1 that represent the probability of participants achieving clinically significant weight loss throughout the trial. Higher predictive marginal means indicate higher probability of achieving this outcome.

Figure 2.

Predicted Marginal Means of Achieving Clinically Significant Weight Loss by Treatment Condition and Age. Note. The predicted marginal means displayed in the graph are proportions ranging from 0 to 1 that represent the probability of participants achieving clinically significant weight loss throughout the trial. Higher predictive marginal means indicate higher probability of achieving this outcome.

Participants’ gender, racial/ethnic minoritized status, and age did not moderate the impact of PGLB and UC on achieving clinically significant CRF throughout our trial (See Table 3). However, the impact of PGLB and UC on achieving clinically significant reductions in CVD risk was moderated by participants’ racial/ethnic minoritized status, but not their gender or age (See Table 4). For this outcome, the 2-way interaction terms for treatment conditions and participants’ racial/ethnic minoritized status was significant (OR = 7.28, 95% CI 1.57, 33.77). As shown in Figure 3, the probability of achieving clinically significant reductions in CVD risk for PGLB participants throughout this trial was higher for those who were racial/ethnic minorities compared to non-Hispanic whites; while among UC participants, the probability of achieving this outcome throughout the trial was higher for non-Hispanic whites than for racial/ethnic minorities. Lastly, none of the three-way interaction terms (e.g., treatment condition × moderator variables [gender, racial/ethnic minority status and age, respectively] × time; results not shown) were statistically significant; indicating that the differential effects of PGLB and UC on the three outcomes examined by our three moderating variables did not change significantly over the course of the trial.

Table 3.

Logistic Mixed Effects Models for Achieving Clinically Significant Cardiorespiratory Fitness.

	Gender Model		Minority/Non-Minority Model		Age Model
	OR	95% CI	OR	95% CI	OR	95% CI
Site (Ref: Site 1)
Site 2	1.60	0.49, 5.23	1.59	0.49, 5.16	1.75	0.52, 5.97
Site 3	5.25	1.63, 16.96	5.30	1.64, 17.11	5.87	1.74, 19.79
Baseline Weight (in Pounds)	1.00	0.99, 1.01	1.00	0.99, 1.01	1.00	0.99, 1.00
Treatment (Ref: Usual Care)
PGLB	3.13	0.62, 15.73	1.51	0.16, 14.43	2.27	0.39,13.07
Time	1.66	1.02, 2.71	1.72	0.84, 3.53	2.37	1.39, 4.05
Age (Ref: 49 and below)
50 and above	0.44	0.18, 1.02	0.43	0.19, 1.00	2.26	0.42, 12.18
Race/Ethnicity (Ref: Non-Hispanic White)
Racial/Ethnic Minority	0.96	0.34, 2.72	1.07	0.13, 8.65	0.97	0.33, 2.85
Gender (Ref: Male)
Female	1.87	0.36, 9.68	0.40	0.40, 2.16	0.91	0.38, 2.15
Two-Way Interactions Terms
Gender (Ref: Male)*Time	0.79	0.44, 1.41
Gender (Ref: Male)*Treatment (Ref: Usual Care)	0.61	0.12, 3.14
Race/Ethnicity (Ref: Non-Hispanic White)*Time			0.84	0.40, 1.74
Race/Ethnicity (Ref: Non-Hispanic White)*Treatment (Ref: Usual Care)			1.88	0.23, 15.52
Age (Ref: Young 49 and below)*Time					0.41	0.22, 0.75
Age (Ref: Young 49 and below)*Treatment (Ref: Usual Care)					1.27	0.23, 6.86
Time* Treatment (Ref: Usual Care)	0.78	0.45, 1.43	0.79	0.44, 1.42	0.78	0.43, 1.42

Note. Statistically significant two-way interaction terms at p<0.05 are bolded.

Table 4.

Logistic Mixed Effect Models for Achieving Clinically Significant Reductions in Cardiovascular Disease Risk.

	Gender Model		Minority/Non-Minority Model		Age Model
	OR	95% CI	OR	95% CI	OR	95% CI
Site (Ref: Site 1)
Site 2	0.49	0.21, 1.12	0.48	0.21, 1.09	0.49	0.21, 1.12
Site 3	0.83	0.37, 1.85	0.88	0.40, 1.94	0.83	0.37, 1.87
Baseline Weight (in Pounds)	1.00	0.99, 1.00	1.00	0.99, 1.00	1.00	0.99, 1.00
Treatment (Ref: Usual Care)
PGLB	2.05	0.60,6.99	0.39	0.07, 2.02	2.86	0.77, 10.62
Time	1.92	1.30, 2.82	1.82	1.04, 3.19	2.41	1.56, 3.71
Age (Ref: 49 and below)
Age (50 and above)	0.61	0.33, 1.14	0.61	0.33, 1.12	2.14	0.60, 7.57
Race/Ethnicity (Ref: Non-Hispanic White)
Racial/Ethnic Minority	0.46	0.21, 0.98	0.18	0.04, 0.84	0.47	0.22, 1.01
Gender (Ref: Male)
Female	1.56	0.45, 5.45	1.31	0.71, 2.42	1.24	0.67, 2.31
Two-Way Interactions Terms
Gender (ref: Male)*Time	0.92	0.57, 1.47
Gender (ref: Male)*Treatment(ref: Usual Care)	0.88	0.26, 2.94
Race/Ethnicity (ref: Non-Hispanic White)*Time			1.02	0.57, 1.81
Race/Ethnicity (ref: Non-Hispanic White)*Treatment (ref: Usual Care)			7.28	1.57, 33.77
Age (Ref: 49 and below)*Time					0.62	0.39, 1.01
Age (Ref: 49 and below)*Treatment (ref: Usual Care)					0.49	0.14, 1.68
Time*Treatment (ref: Usual Care)	0.82	0.52, 1.32	0.83	0.52, 1.32	0.83	0.52, 1.32

Note. Statistically significant two-way interaction terms at p<0.05 are bolded.

Figure 3.

Predicted Marginal Means of Achieving Clinically Significant Reduction in Cardiovascular Disease (CVD) Risk by Treatment Condition and Minoritized Status. Note. The predicted marginal means displayed in the graph are proportions ranging from 0 to 1 that represent the probability of participants achieving clinically significant reduction in CVD risk throughout the trial. Higher predictive marginal means indicate higher probability of achieving this outcome.

Discussion

This study used data from an effectiveness trial to examine whether participants’ age, minoritized racial/ethnic status, and gender moderated the impact of a peer-led health intervention for people with SMI who were overweight or obese in achieving clinically significant improvements in weight loss, CRF, and CVD risk reduction. Based on findings from previous trials in the general population, non-Hispanic whites, men, and older participants were hypothesized to benefit most from PGLB compared to UC in achieving these three clinically significant outcomes. The results of this study did not support this hypothesis.

The impact of PGLB and UC on achieving the three health outcomes examined in this trial was not moderated by participants’ gender. These results are consistent with a previous study suggesting that men and women with SMI participating in a healthy lifestyle intervention trial tended to achieve comparable health outcomes, particularly weights loss (Alexander et al., 2019), and extend this finding to two health outcomes (clinically significant improvements in CRF and reductions in CVD risk) not examined in previous studies. These findings indicate that male and female participants enrolled in our trial seemed to benefit equally from our peer-led healthy lifestyle intervention. More studies with larger sample sizes are needed to corroborate these gender findings in other healthy lifestyle interventions for people with SMI.

Participants’ age moderated the impact of PGLB and UC only on achieving clinically significant weight loss, but not in the direction that was expected. Previous studies in the general population have found that, compared to younger participants, older participants tend to benefit most from intensive healthy lifestyle interventions and report greater treatment attendance and engagement in physical activity and healthy dietary practices (Wadden et al., 2009). Yet, a different pattern was found in this study; PGLB was most beneficial for achieving clinically significant weight loss among younger participants (49 years old or younger) compared to those who were 50 years old or older. Potential differences in attendance by age do not explain this finding, given that participants’ age was not significantly related to the number of PGLB sessions attended (Tuda et al., 2021). Older age in people with SMI is associated with more sedentary behaviors and lower CRF resulting in poorer health and more comorbid health conditions (e.g., arthritis), thus potentially making it more difficult for older participants with SMI to fully benefit from a healthy lifestyle intervention, like PGLB, and achieve clinically significant outcomes (Stubbs, Firth, Berry, Schuch, Rosenbaum, 2016; Vancampfort et al., 2017). The moderating impact that age had in our trial underscores the importance of considering participants’ age when designing and delivering healthy lifestyle interventions for people with SMI. Intervention adaptations that take into consideration participants’ levels of functioning, mobility, flexibility, and strength related to their age are needed to fully engage older participants with SMI in these interventions (Cabassa et al., 2020).

Finally, participants’ racial/ethnic minoritized status moderated the impact of PGLB and UC on achieving clinically significant weight loss, and clinically significant reductions in CVD risk, but not in the directions that was hypothesized. Although racial/ethnic minoritized groups (e.g., blacks and Hispanics) without SMI benefit from healthy lifestyle interventions, these interventions tend to underperform for minoritized groups when compared to non-Hispanic whites, particularly when examining weight loss outcomes (Kumanyika et al., 2002; Wadden et al., 2009; West et al., 2008). These differences have been attributed to differential attendance and non-treatment-related factors (e.g., lack of transportation) that can affect participation in these interventions (Kumanyika et al., 2002; Wadden et al., 2009). A different pattern was found in this study; PGLB was most beneficial for achieving clinically significant weight loss and CVD risk reductions among participants from racial/ethnic minoritized status compared to participants who were non-Hispanic whites. These findings indicate that PGLB seemed to have been more responsive to participants from racial/ethnic minoritized status.

Several characteristics of PGLB could have contributed to these differential benefits. First, peer specialists delivering PGLB shared with participants not only their lived experiences with SMI, but similar cultural and racial backgrounds. Three of the four PGLB peer specialists were black (Bochicchio et al., 2019). These shared racial backgrounds enabled peer specialists to draw from their own cultural reservoir and wisdom to provide practical advice for making healthy lifestyle changes that are familiar and salient to participants’ lives; for example on how to use healthier ingredients to prepare traditional dishes without sacrificing taste and cultural traditions. Second, as reported elsewhere (O’Hara et al., 2017), peer specialists had lived in participants’ neighborhoods and under similar financial constraints enabling them to share practical advice and strategies on how to shop for healthy foods with limited income and where to find safe and affordable local places to engage in regular physical activity. Finally, beyond the technical delivery of PGLB, the interpersonal connections peer specialists made with our participants—characterized by unconditional encouragement and support, hope and motivation, and making participants’ feel comfortable -directly contributed to participants’ willingness to use intervention strategies (e.g., self-monitoring of diet and physical activity, calorie counting, mindfulness eating) and engage in healthy lifestyle changes (Bochicchio et al., 2019). Each of these PGLB characteristics could have resonated more deeply with participants from minoritized racial/ethnic backgrounds since it is still a very rare occurrence to have racial and ethnic minorities deliver these types of manualized interventions in community settings, like supportive housing.

Limitations

Several study limitations need to be considered. Due to logistical constraints, assessors blind to participants’ group assignments were not utilized to measure study outcomes, which could have biased study findings, though our outcomes were derived from objective health measures rather than subjective ratings. The treatment effects by different racial and ethnic groups (e.g., Hispanics, Asians) could not be examined due to small sample sizes of these groups in this trial. Future studies with larger samples of racial and ethnic minoritized groups are needed to examine differential treatment effects in these diverse populations. The findings of this analysis should be interpreted with caution since this trial was not properly powered to examine treatment moderators, which contributed to the wide 95% confidence intervals observed in some of the odd ratios reported in these analyses.

Despite these methodological limitations, this study addresses major gaps in the existing literature for improving the physical health of people with SMI. Few studies to date have explored how the impact of healthy lifestyle interventions in achieving clinically significant health outcomes is moderated by participants’ age, gender, and racial and ethnic minoritized status. Our findings indicated that PGLB was most beneficial for participants who were 49 years old and younger and from historically marginalized racial and ethnic communities. These differential treatment outcomes suggest that the impact of healthy lifestyle interventions for people with SMI may not be uniform across these diverse populations and careful adaptations may be needed to make these interventions more responsive to the needs and preferences of these different populations. To develop more robust health interventions for people with SMI, more studies are needed to clarify why and how these interventions work for certain group versus others.

Footnotes

Acknowledgments

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors would like to thank the study participants and community partners for their engagement in this study.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research reported in this publication was supported by grants from the National Institute of Mental Health: R01MH104574 and T32MH019960 and by the Washington University, Institute for Public Health, Center for Dissemination and Implementation Pilot Program.

ORCID iDs

Leopoldo J. Cabassa

Mark R. Hawes

Shenyang Guo

References

Alexander

McGinty

E. E.

Wang

N.-Y.

Dalcin

Jerome

G. J.

Miller

E. R.

Dickerson

Charleston

Young

D. R.

Gennusa

J. V.

Goldsholl

Cook

Appel

L. J.

Daumit

G. L.

(2019). Effects of a behavioural weight loss intervention in people with serious mental illness: Subgroup analyses from the ACHIEVE trial. Obesity Research & Clinical Practice, 13(2), 205–210. https://doi.org/10.1016/j.orcp.2019.02.002

Aschbrenner

K. A.

Naslund

J. A.

Gorin

A. A.

Mueser

K. T.

Browne

Wolfe

R. S.

Xie

Bartels

S. J.

(2021). Group lifestyle intervention with mobile health for young adults with serious mental illness: A randomized controlled trial. Psychiatric Services, 72(2), 141–148. https://doi.org/10.1176/appi.ps.202100047

Bartels

S. J.

Pratt

S. I.

Aschbrenner

K. A.

Barre

L. K.

Jue

Wolfe

R. S.

Xie

McHugo

Santos

Williams

G. E.

Naslund

J. A.

Mueser

K. T.

(2013). Clinically significant improved fitness and weight loss among overweight persons with serious mental illness. Psychiatric Services, 64(8), 729–736. https://doi.org/10.1176/appi.ps.003622012

Bartels

S. J.

Pratt

S. I.

Aschbrenner

K. A.

Barre

L. K.

Naslund

J. A.

Wolfe

Xie

McHugo

G. J.

Jimenez

D. E.

Jue

Feldman

Bird

B. L.

(2015). Pragmatic replication trial of health promotion coaching for obesity in serious mental illness and maintenance of outcomes. American Journal of Psychiatry, 172(4), 344–352. https://doi.org/10.1176/appi.ajp.2014.14030357

Bochicchio

Stefancic

Gurdak

Swarbrick

Cabassa

L. J.

(2019). “We’re all in this together”: Peer-specialist contributions to a healthy lifestyle intervention for people with serious mental illness. Administration and Policy in Mental Health and Mental Health Services Research, 46(3), 298–310. https://doi.org/10.1007/s10488-018-0914-6

Cabassa

L. J.

Camacho

Vélez-Grau

C. M.

Stefancic

(2017). Peer-based health interventions for people with serious mental illness: A systematic literature review. Journal of Psychiatric Research, 84, 80–89. https://doi.org/10.1016/j.jpsychires.2016.09.021

Cabassa

L. J.

Ezell

J. M.

Lewis-Fernández

(2010). Lifestyle interventions for adults with serious mental illness: A systematic literature review. Psychiatric Services, 61(8), 774–782. https://doi.org/10.1176/ps.2010.61.8.774

Cabassa

L. J.

Stefancic

Lewis-Fernández

Luchsinger

Weinstein

L. C.

Guo

Palinkas

Bochicchio

Wang

O’Hara

Blady

Simiriglia

Medina McCurdy

(2021). Main outcomes of a peer-led healthy lifestyle intervention for people with serious mental illness in supportive housing. Psychiatric Services, 72(5), 555–562. https://doi.org/10.1176/appi.ps.202000304

Cabassa

L. J.

Stefancic

O’Hara

El-Bassel

Lewis-Fernández

Luchsinger

J. A.

Gates

Younge

Wall

Weinstein

Palinkas

L. A.

(2015). Peer-led healthy lifestyle program in supportive housing: Study protocol for a randomized controlled trial. [Trials Electronic Resource], 16(1), 388– https://doi.org/10.1186/s13063-015-0902-z

10.

Cabassa

L. J.

Stefancic

Wang

Guo

N. Y.

Weatherly

(2020). Correlates of physical activity and cardiorespiratory fitness in racially and ethnically diverse people with serious mental illness in supportive housing. Community Mental Health Journal, 56(60), 1139-1152.

11.

Daumit

G. L.

Dickerson

F. B.

Wang

N.-Y.

Dalcin

Jerome

G. J.

Anderson

C. A. M.

Young

D. R.

Frick

K. D.

Gennusa

J. V.

Oefinger

Crum

R. M.

Charleston

Casagrande

S. S.

Guallar

Goldberg

R. W.

Campbell

L. M.

Appel

L. J.

(2013). A behavioral weight-loss intervention in persons with serious mental illness. New England Journal of Medicine, 368(17), 1594–1602. https://doi.org/10.1056/NEJMoa1214530

12.

Dawson

J. F.

Richter

A. W.

(2006). Probing three-way interactions in moderated multiple regression: Development and application of a slope difference test. Journal of Applied Psychology, 91(4), 917–926. https://doi.org/10.1037/0021-9010.91.4.917

13.

Firth

Siddiqi

Koyanagi

Siskind

Rosenbaum

Galletly

Stubbs

(2019). The lancet psychiatry commission: A blueprint for protecting physical health in people with mental illness. The Lancet Psychiatry, 6(8), 675–712. https://doi.org/10.1016/S2215-0366(19)30132-4

14.

Ford

Norrie

(2016). Pragmatic trials. New England Journal of Medicine, 375(5), 454–463. https://doi.org/10.1056/NEJMra1510059

15.

Green

C. A.

Yarborough

B. J. H.

Leo

M. C.

Yarborough

M. T.

Stumbo

S. P.

Janoff

S. L.

Perrin

N. A.

Nichols

G. A.

Stevens

V. J.

(2015). The STRIDE weight loss and lifestyle intervention for individuals taking antipsychotic medications: A randomized trial. American Journal of Psychiatry, 172(1), 71–81. https://doi.org/10.1176/appi.ajp.2014.14020173

16.

Hedeker

Gibbons

R. D.

(2006). Longitudinal data analysis. Hoboken: John Wiley & Sons. https://doi.org/10.1002/0470036486

17.

Henwood

B. F.

Cabassa

L. J.

Craig

C. M.

Padgett

D. K.

(2013). Permanent supportive housing: Addressing homelessness and health disparities? American Journal of Public Health, 103(S2), S188–S192. https://doi.org/10.2105/AJPH.2013.301490

18.

Interdepartmental Serious Mental Illness Coordinating Committee . (2017). The way forward: Federal action for a system that works for all people living with SMI and SED and their families and caregivers. https://store.samhsa.gov/product/PEP17-ISMICC-RTC

19.

Kessler

R. C.

Barker

P. R.

Colpe

L. J.

Epstein

J. F.

Gfroerer

J. C.

Hiripi

Howes

M. J.

Normand

S. T.

Manderscheid

R. W.

Walters

E. E.

Zaslavsky

A. M.

(2003). Screening for serious mental illness in the general population. Archives of General Psychiatry, 60(2), 184-189. https://doi.org/10.1001/archpsyc.60.2.184.

20.

Kramer

M. K.

Kriska

A. M.

Venditti

E. M.

Miller

R. G.

Brooks

M. M.

Burke

L. E.

Siminerio

L. M.

Solano

F. X.

Orchard

T. J.

(2009). Translating the diabetes prevention program. American Journal of Preventive Medicine, 37(6), 505–511. https://doi.org/10.1016/j.amepre.2009.07.020

21.

Kumanyika

S. K.

Espeland

M. A.

Bahnson

J. L.

Bottom

J. B.

Charleston

J. B.

Folmar

Wilson

A. C.

Whelton

P. K.

TONE Cooperative Research Group . (2002). Ethnic comparison of weight loss in the trial of nonpharmacologic interventions in the elderly. Obesity Research, 10(2), 96–106. https://doi.org/10.1038/oby.2002.16

22.

McGinty

E. E.

Baller

Azrin

S. T.

Juliano-Bult

Daumit

G. L.

(2015). Interventions to address medical conditions and health-risk behaviors among persons with serious mental illness: A comprehensive review. Schizophrenia Bulletin, 42(1), 96–124. https://doi.org/10.1093/schbul/sbv101

23.

Naslund

J. A.

Whiteman

K. L.

McHugo

G. J.

Aschbrenner

K. A.

Marsch

L. A.

Bartels

S. J.

(2017). Lifestyle interventions for weight loss among overweight and obese adults with serious mental illness: A systematic review and meta-analysis. General Hospital Psychiatry, 47, 83–102. https://doi.org/10.1016/j.genhosppsych.2017.04.003

24.

Nelson

Laurier

(2010). Housing for people with serious mental illness: Approaches, evidence, and transformative change. Journal of Sociology & Social Welfare, 38(4), 123–146. https://scholarworks.wmich.edu/jssw/vol37/iss4/7

25.

O’Hara

Stefancic

Cabassa

L. J.

(2017). Developing a peer-based healthy lifestyle program for people with serious mental illness in supportive housing. Translational Behavioral Medicine, 7(4), 793–803. https://doi.org/10.1007/s13142-016-0457-x

26.

Palmer

R. M.

Meldon

S. W.

(2003). Acute care. In Hazzard

W. R.

(Ed.), Principles of geriatric medicine and gerontology (pp. 157–168). New York: McGraw-Hill.

27.

Rasekaba

Lee

A. L.

Naughton

M. T.

Williams

T. J.

Holland

A. E.

(2009). The six-minute walk test: A useful metric for the cardiopulmonary patient. Internal Medicine Journal, 39(8), 495–501. https://doi.org/10.1111/j.1445-5994.2008.01880.x

28.

Stefancic

Bochicchio

Tuda

Harris

DeSomma

Cabassa

L. J.

(2021). Strategies and lessons learned for supporting and supervising peer specialists. Psychiatric Services, 72(5), 606–609. https://doi.org/10.1176/appi.ps.202000515

29.

Stubbs

Firth

Berry

Schuch

F. B.

Rosenbaum

Gaughran

Veronesse

Williams

Craig

Yung

A. R.

Vancampfort

(2016). How much physical activity do people with schizophrenia engage in? A systematic review, comparative meta-analysis and meta-regression. Schizophrenia Research, 176(2–3), 431–440. https://doi.org/10.1016/j.schres.2016.05.017

30.

Tuda

Stefancic

Hawes

Wang

Guo

Cabassa

L. J.

(2021). Correlates of attendance in a peer-led healthy lifestyle intervention for people with serious mental illness living in supportive housing. Community Mental Health Journal, 58(4), 761–769. https://doi.org/10.1007/s10597-021-00881-y

31.

Vancampfort

Rosenbaum

Probst

Soundy

Mitchell

A. J.

De Hert

Stubbs

(2015). Promotion of cardiorespiratory fitness in schizophrenia: A clinical overview and meta-analysis. Acta Psychiatrica Scandinavica, 132(2), 131–143. https://doi.org/10.1111/acps.12407

32.

Vancampfort

Rosenbaum

Schuch

Ward

P. B.

Richards

Mugisha

Probst

Stubbs

(2017). Cardiorespiratory fitness in severe mental illness: A systematic review and meta-analysis. Sports Medicine, 47(2), 343–352. https://doi.org/10.1007/s40279-016-0574-1

33.

Vélez-Grau

Stefancic

Cabassa

L. J.

(2019). Keeping the peer in peer specialist when implementing evidence-based interventions. Health & Social Work, 44(1), 57–60. https://doi.org/10.1093/hsw/hly037

34.

Wadden

T. A.

West

D. S.

Neiberg

R. H.

Wing

R. R.

Ryan

D. H.

Johnson

K. C.

Foreyt

J. P.

Hill

J. O.

Trence

D. L.

Vitolins

M. Z.

Look

A. R. G.

(2009). One-year weight losses in the look AHEAD study: Factors associated with success. Obesity, 17(4), 713–722. https://doi.org/10.1038/oby.2008.637

35.

West

D. S.

Prewitt

T. E.

Bursac

Felix

H. C.

(2008). Weight loss of black, white, and Hispanic men and women in the diabetes prevention program. Obesity, 16(6), 1413–1420. https://doi.org/10.1038/oby.2008.224

36.

Wise

R. A.

Brown

C. D.

(2005). Minimal clinically important differences in the six-minute walk test and the incremental shuttle walking test. COPD: Journal of Chronic Obstructive Pulmonary Disease, 2(1), 125–129. https://doi.org/10.1081/COPD-200050527

37.

Zayas

L. H.

Cabassa

L. J.

Perez

M. C.

(2005). Capacity-to-consent in psychiatric research: Development and preliminary testing of a screening tool. Research on Social Work Practice, 15(6), 545–556. https://doi.org/10.1177/1049731505275867