Health for Hearts United Longitudinal Trial: Improving Dietary Behaviors in Older African Americans

Abstract

Introduction:

Church-based interventions have been shown to improve the dietary health of underserved populations, yet few studies have examined sustainability of health behavior change over time. This paper examines dietary outcomes over a 24-month period (baseline and 6, 18, and 24 months) for fruit and vegetable (F/V) and fat consumption behaviors of African American participants in the Health for Hearts United church-based intervention in North Florida.

Study Design:

This quasi-experimental, longitudinal trial was conducted from 2009 to 2012. Data were analyzed in 2018.

Setting/participants:

Six churches in a two-county area (three treatment, three comparison) were selected for the study using community-based participatory research approaches. Participants were African American adults (aged ≥45 years; n=211 at baseline) randomly selected from the churches, stratified by age and sex.

Intervention:

Health for Hearts United intervention was developed by the three treatment churches. The 18-month intervention was implemented in three 6-month phases, framed around three conceptual components, which included four types of programs and four key messages.

Main outcome measures:

F/V consumption was assessed using a single item (F/V intake) and the National Cancer Institute F/V Screener. Fat consumption was determined using a single item (fat intake) and National Cancer Institute Fat Screener. Background characteristics included age, sex, educational level, and marital status.

Results:

Significant time effects only were found for daily F/V intake (p<0.001), fat intake (p<0.001), and the Fat Screener (p<0.001) with dietary improvements in both treatment and comparison groups across the intervention phases. F/V Screener results showed that time (p<0.001) and the interaction between time and treatment (p<0.01) were significant, with increases in F/V consumption over time for both treatment and comparison groups and with the increase differing between groups. Post-hoc analysis revealed that the treatment group had greater increases in F/V consumption than the comparison group between Phases 1 and 3 (p=0.03).

Conclusions:

Dietary behaviors of mid-life and older African Americans can be improved and sustained over 24 months using a church-based heart health intervention, with similar improvements noted for both comparison and treatment participants.

Trial registration:

This study is registered at www.clinicaltrials.gov .

INTRODUCTION

As the leading cause of death in the U.S., cardiovascular disease (CVD) disproportionately affects African Americans, who lead all racial/ethnic groups for mortality rates.¹ Sedentary lifestyle, diet, and excess body weight are key risk factors for CVD and managing a healthy lifestyle to a large extent can reduce risk of this disease.¹ To reach African Americans and other underserved populations, community-based interventions are recommended, especially those with participatory approaches.² There is considerable evidence from effectiveness studies that church-based health interventions can improve health behaviors including fruit/vegetable (F/V) consumption and physical activity, lower BMI, and decrease waist circumference.^3–7 Yet, there are several ongoing challenges in church-based research, including weak evaluation data and lack of longitudinal designs.^3,5–8 In particular, there is lack of multiphase longitudinal designs that examine sustainability of health behavior change over time.⁷ Considering the need to reduce CVD and improve related overall health in African Americans, studies are needed that incorporate strong evaluation components and longitudinal designs that measure long-term results.

This article describes the evaluation of Health for Hearts United (HHU), an 18-month longitudinal church-based intervention to reduce CVD risk in mid-life and older African Americans developed using community-based participatory research (CBPR) approaches.^9,10 Specifically, this article examines outcomes over a 24-month period (baseline and 6, 18, and 24 months) related to selected dietary behaviors (F/V and fat consumption) for HHU participants from churches in a two-county area of North Florida.

This study was guided by socioecological theory and the Transtheoretical Model of Change (TTM). Different levels of factors reflected in socioecological theory include intrapersonal characteristics that influence health behavior, interpersonal characteristics (group characteristics such as social networks and social support that help to support healthy behaviors), organizational characteristics (policies, facilities, and organizational structures), and environment/policy (community or government resources, policies, advocacy).¹¹ Consistent with the intrapersonal level, TTM is based on the premise that people move through a series of changes in their attempt to change a behavior.¹² In this study, individual dietary behaviors were determined across the four phases of HHU, with the intrapersonal level (TTM, socioecological theory, individuals) examined in relation to behavioral outcomes with implications for the interpersonal and organizational levels (socioecological theory, health structures, and the church itself).

Previous research that focuses on African American dietary behaviors shows low F/V intake and high intake of fat.^13–16 In addition, research shows that church-based health interventions can improve food choice and dietary quality.^3,17–19 With regard to longitudinal church-based findings, the review by Lancaster et al.⁶ of faith-based obesity studies published prior to July 2012 (N=27 studies) showed that 60% of participants reported a significant increase in F/V intake and that intervention length ranged from 14 weeks to 2 years. Since 2012, seven additional faith-based studies with dietary components were identified.^20–26 Of these studies, the interventions ranged from 6 weeks to 17 months with one single group design, five RCTs, and one implementation design. Three studies found significant improvements in dietary intake (including increases in F/V consumption, decreases in fat consumption, or increases in general healthy eating) and four indicated no significant change in dietary behaviors.

Based on the empirical findings of previous studies and theoretical perspectives of socioecological theory and TTM, it is hypothesized that participants in the treatment group of the present church-based intervention will have sustained dietary improvements (i.e., increased F/V consumption, decreased fat consumption) from baseline to post-intervention (24 months). In addition, this study controls for background characteristics (age, educational level, sex, and marital status) that may influence these outcomes.

METHODS

Study Population

Data for this study came from the Reducing Cardiovascular Risk in African Americans Study, a quasi-experimental, longitudinal trial to determine the effectiveness of a church-based intervention developed using CBPR approaches.⁹ A quasi-experimental design was used because the church setting does not allow for true random assignment of participants to treatment. The data were collected from participants in six churches (three treatment, three comparison) in two North Florida counties, one more urban and the other more rural. Advisory groups were established in each county that assisted in identifying churches for the study. Selected churches were to have at least 150 members, no health ministry or an inactive health ministry, and pastors who supported evidenced-based health. The six selected churches included two larger churches in the urban county and four smaller churches, two each in communities in the rural county. Consistent with CBPR, the community advisors in the two counties made recommendations regarding which churches should be selected for the treatment versus comparison group to control for variability in church size and to avoid possible contamination within communities, especially in the rural county.⁹ Final decisions were made using an iterative process between community advisors and staff, based on church size and community as well as readiness to begin working on project activities. The three churches selected for treatment included a large church in the urban county with >500 members and two smaller churches located within the same community in the rural county with approximately 200 members per church.⁹ The comparison churches selected included a large church in the urban county and two smaller churches, both located in a different community in the rural county. Comparison churches were similar in size to their respective treatment counterparts.

Participants for the study were identified through lists obtained from the pastors for use by the researchers that included church members aged ≥45 years who attended church at least twice a month.⁹ Figure 1 presents the participant flow diagram for the overall study. Based on the church lists, a total of 576 participants met the criteria for the study, 70 were excluded after declining to participate, and 506 (278 treatment, 228 comparison) were eligible for the study. A total of 300 participants (143 treatment, 157 comparison) were then randomly selected from the church lists stratified by age (45–64 years, ≥65 years) and sex, with 244 consenting to participate (112 treatment, 132 comparison).⁹ To encourage participation, participants could enter the study at baseline or at Time 2; however, those entering at Time 2 only are not included in this paper. During the course of the study, 36 participants (11 treatment, 25 comparison) did not receive all phases of the intervention after declining to participate or because they were found to be mentally challenged or incapacitated, and 23 (13 treatment, 10 comparison) were lost to follow-up owing to relocation, being deceased, being non-responsive, or declining to continue to the next phase. The total analytic sample (n=226) represents those who completed the study, indicating an overall retention rate of 92.6% based on those who were randomly selected and provided consent (n=244) (Figure 1).

Figure 1.

CONSORT flow diagram.

^aTotal of 300 was randomly selected for the study.

^bTotal of 244 consented and received either the treatment or comparison intervention.

^cParticipants could enter the project at either Time 1(t1) (n=244) or Time 2 (t2) (n=56) to encourage participation.

^dTotal of 36 declined to continue or were excluded for other reasons.

^eTotal of 23 were lost to follow-up.

^fTotal of 226 represent the final sample for overall project at Time 4. Based on number consented and randomly selected, retention rate was 92.6%.

^gSample size varies for individual studies based on completion of phases and missing data.

For this paper, those who completed the study variables at baseline ranged from n=130 to n=219. Those included in the analysis who completed study variables at all phases ranged from n=82 to n=154 (Appendix Table 1). Based on those who were randomly selected and provided consent (n=244), the retention rate for study variables ranged from 33.6% to 63.1%. The percentage of missing data by variable comparing overall sample to sample analyzed at baseline was 30.1% for F/V intake (i.e., 219 – 153 = 66); 29.4% for National Cancer Institute (NCI) F/V Screener (60), 29.6% for fat intake (65), and 36.9% for the NCI Fat Screener (48) (Appendix Table 1). Attrition analysis showed that missingness was most likely random, and less likely to bias estimates. Thus, data were not imputed for outcome variables.

Table 1.

Baseline Characteristics of Participants

Background characteristics	Treatment participants, n (%), n=101a	Comparison participants, n (%), n=110b
Age, years
45–49	17 (17.0)	23 (20.9)
50–56	17 (17.0)	27 (24.5)
57–63	24 (24.0)	30 (27.3)
64–70	15 (15.0)	13 (11.8)
71–77	14 (14.0)	9 (8.2)
78–84	7 (7.0)	5 (4.5)
≥85	6 (6.0)	3 (2.7)
Sex
Female	73 (72.3)	80 (72.7)
Male	28 (27.7)	30 (27.3)
Marital status
Single	11 (11.0)	14 (12.7)
Married	48 (48.0)	47 (42.7)
Divorced/separated	18 (18.0)	34 (30.9)
Widowed	22 (22.0)	12 (10.9)
Other	1 (1.0)	3 (2.7)
Education
Some high school	8 (7.9)	12 (11.0)
High school graduate	29 (28.7)	32 (29.4)
Some college	34 (33.7)	31 (28.4)
Bachelor’s degree	14 (13.9)	15 (13.8)
Master’s degree or above	13 (12.9)	17 (15.6)

^aTreatment sample ranged from n=98 to n=101 due to missing data.

^bComparison sample ranged from n=107 to n=110 due to missing data.

The authors tested mean differences in background characteristics (age, sex, marital status) and selected study variables (F/V intake, NCI F/V Screener, fat intake) at Time 1 between completers and non-completers at Time 4. The differences were not significant for any of the variables except sex. The comparison group showed that ≥15% more female than male participants stayed in the study until Time 4.

Follow-up procedures included a rigorous team-based protocol that included telephone calls and confidential contacts with pastors or health leaders to determine possible reasons for participant non-responsiveness. These procedures were tracked by staff and discussed in weekly meetings where updates were shared.

All participants received a $25 discount store gift card for each completed study phase.⁹ The study was approved by the Florida State University IRB.

Measures

Using CBPR approaches, the HHU intervention was developed by the three treatment churches.^9,10 HHU was implemented in 6-month phases, framed around three conceptual components. Awareness building involved individual knowledge development, clinical learning focused on individual and small group educational sessions, and efficacy development involved recognition and sustainability.¹⁰ Key messages were identified for the intervention including eating better, moving around more, reducing stress, and taking charge of one’s health. Four types of programming were used throughout the intervention to emphasize the key messages. First, church-initiated programming included CVD awareness kickoff events implemented by the treatment churches. Second, joint programming encompassed educational sessions planned jointly with health leaders and staff and held at the treatment churches. Third, standard programming was composed of culturally tailored post cards and newsletters developed by staff with input from health leaders and sent to the treatment participants. Finally, additional health promotion activities were provided for both treatment and comparison participants, which consisted of distributing generic materials on reducing CVD risk during data collection and providing individualized counseling sessions with a registered dietitian to review dietary and clinical outcomes. Within the three 6-month phases, CVD awareness kickoff events were implemented in Phase 1 (awareness building), followed by joint programming in Phase 2 (clinical learning) and Phase 3 (efficacy development). Standard programming was implemented in all three phases.¹⁰ A tracking tool was implemented in Phase 2 to encourage health behavior change.²⁷

Comparison churches participated in health ministry development activities during HHU and then received the intervention on a delayed basis.^9,10 Activities, which are described in a published paper,¹⁰ included trainings on how to establish and plan health ministries, development of strategic plans, and presentation of these plans by health leaders in each church. Data were collected (2009–2012) from participants in both the treatment and comparison churches prior to the initiation of the intervention (baseline, October–November 2009), at 6 months (following Phase 1 on awareness building, June–August 2010), at 18 months (following Phase 2 on clinical learning and Phase 3 on efficacy development, August–October 2011), and at 24 months (6 months following the intervention (January–March 2012). Comparison church participants received the comparable intervention following 24 months, once all data in the treatment churches had been collected. A figure presenting the HHU intervention and timeline is published elsewhere.¹⁰

The food habits and lifestyle questionnaire was developed with assistance of the community advisors in each county, including reviewing, pilot testing and revising of items, as needed.⁹ The questionnaire was then administered to participants at sessions held at the churches. Trained staff provided assistance and conducted interviews when needed. Items from the questionnaire germane to this study included the following.

Daily F/V intake was assessed by the item: On average, what is the number of fruit/vegetable servings that you eat daily? Possible responses ranged from zero to six or more. This single item measure (F/V intake), used extensively in previous dietary studies, is positively correlated overtime with the 24-hour dietary recall (r =0.45 baseline, 0.50 follow-up).^23,28,29 Also, this measure has an inter-measure reliability of r =0.56 when correlated with mean servings based on a 61-item food frequency questionnaire.³⁰ Overall F/V intake was assessed using the NCI F/V Screener, a ten-item instrument that includes respondent’s estimate of portion size. This instrument has been validated with actual intake of F/V in a national culturally diverse sample (r =0.66 for men and 0.51 for women).³⁰ The daily F/V intake item was administered at all four phases (baseline and 6, 18, and 24 months) whereas the F/V Screener was administered at only three phases (baseline and 18 and 24 months).

Fat consumption was assessed using the single item: Overall, when you think about the foods you ate over the past 12 months, would you say your diet was high, medium, or low in fat? Also, the broader multi-item NCI Fat Screener, which gauges fat consumption over time, was used to assess fat intake. Estimated correlations between true intake and the NCI Fat Screener were r =0.64 for men and r =0.58 for women.³¹ The fat intake item was administered at all four phases (baseline and 6, 18, and 24 months) whereas the NCI Fat Screener was administered at only three phases (baseline and 18 and 24 months).

With regard to background characteristics, age was determined using categories, ranging from 45–49 years to 85 years and older, coded as 1–7. Categories instead of actual age were used to increase the response rate for this item, based on prior experiences with the study population. Sex was coded as female (1) and male (2). Respondents provided their education level using five categories ranging from some high school to master’s degree or above (coded 1–5 respectively). Marital status was determined by the item: What is your marital status? Responses included single (including separated, widowed, divorced, other) (coded as 0) and married (coded as 1).

Statistical Analysis

The data, collected between 2009 and 2012, were analyzed in 2018 using SPSS, version 25.0. Descriptive statistics were used to determine background characteristics of the sample (age, sex, educational level, and marital status). Correlation analyses were performed to determine any bivariate relationships between the background characteristics and study variables (F/V and fat consumption). Repeated measures ANOVAs were used to compare mean scores for study variables for the treatment and comparison groups across time (i.e., the data collection phases) to determine any change in dietary behaviors over measurement occasions. In addition, the analysis tested for interaction effects between time and treatment to examine the difference in change in dietary behaviors between the treatment and comparison groups over measurement occasions. Finally, post-hoc analyses were conducted to determine where differences occurred between groups.

This study did not control for possible church-level effects because six churches are not adequate to incorporate a multilevel analytical approach to take any clustering into account. Further, according to the authors’ observations and community advisor input, these churches were similar in terms of church-level characteristics (e.g., size, presence/absence of a health ministry) and unlikely to have church-level variance or effects on outcomes. Finally, intra-class correlations (church-level variance/total variance) in relation to outcome variables were negligible (intra-class correlation <1.5%), suggesting negligible clustering. Thus, church-level clustering was not taken into account in the analysis.

RESULTS

Table 1 presents the background characteristics for the treatment and comparison participants at baseline. The treatment and comparison groups were similar in terms of age (56% of treatment and 63.6% of comparison participants were aged 50–70 years), sex (72.3% vs 72.7% were female), marital status (48% vs 42.7% were married), and education (62.4% vs 57.8% were high school graduates or had some college). Correlational analyses showed no significant relationships between dietary behavior variables and the background control variables.

Table 2 presents the results for the repeated measures ANOVA for the dietary behavior variables in the study. The results showed significant time effects (an increase over measurement occasions) only for daily F/V intake (p<0.001), indicating daily servings of F/V increased in both treatment and comparison groups. However, non-significant interaction between treatment and time indicated that the magnitude of increases did not differ between groups. The results for the NCI F/V Screener showed that both the time effect (p<0.001) and the interaction between time and treatment (p<0.01) were significant, indicating that although F/V consumption increased over time, the increase over measurement occasions differed between treatment and comparison groups. The post hoc analyses of the results for the NCI F/V Screener revealed that the treatment group, compared with the comparison group, significantly increased F/V consumption between Phases 1 and 3 (p<0.03).

Table 2.

Dietary Behaviors of Treatment and Comparison Groups Over Four Phases

Dietary behavior	Means (SD)				Time effect	Treatment effect	Time X treatment interaction
	t1	t2	t3	t4	F	F	F
Daily F/V intakea
Treatment	2.2 (1.0)	2.5 (1.2)	2.6 (1.0)	3.0 (1.3)
Comparison	2.4 (1.3)	2.6 (1.3)	2.6 (1.4)	2.9 (1.6)	8.23**	0.17	0.40
NCI F/V Screenerb
Treatment	3.3 (2.5)	–	5.4 (4.2)	5.5 (5.4)
Comparison	4.1 (4.6)	–	4.4 (4.3)	4.2 (4.4)	7.40**	0.38	4.77*
Fat intakec
Treatment	2.0 (0.6)	1.8 (0.4)	1.8 (0.5)	1.7 (0.6)
Comparison	2.0 (0.5)	1.8 (0.5)	1.8 (0.6)	1.7 (0.4)	12.92**	0.003	0.20
NCI Fat Screenerd
Treatment	33.2 (3.0)		30.3 (3.4)	30.7 (4.4)
Comparison	32.3 (4.0)		29.6 (3.9)	30.2 (2.7)	4.66*	0.23	0.03

Notes: Boldface indicates statistical significance (*p<0.01, **p<0.001). NCI F/V Screener had significant time X treatment interaction between t1 and t3, (F(2, 154 ) = 3.451, p=0.03).

^an=93 treatment, n=60 comparison.

^bn=85 treatment, n=59 comparison.

^cn=93 treatment, n=61 comparison.

^dn=50 treatment, n=32 comparison.

t1, Time 1 (baseline); t2, Time 2 (6 months); t3, Time 3 (18 months); t4, Time 4 (24 months); F, F statistic; NCI, National Cancer Institute; F/V, fruit and vegetable.

None of the other post hoc analyses were found to be significant. Only significant time effects were noted for both fat intake (p<0.001) and the NCI Fat Screener (p<0.001), indicating decreases in fat consumption in both groups. However, the non-significant interaction between treatment and time indicated that the magnitude of these decreases did not differ between groups. Figure 2 presents the graph for the analysis related to the NCI F/V Screener.

Figure 2.

National Cancer Institute Fruit and Vegetable Screener outcomes over three phases.

DISCUSSION

This study examined outcomes over a 24-month period related to selected dietary behaviors (F/V and fat consumption) for mid-life and older African Americans who participated in HHU, an 18-month heart health church-based health intervention. The results, which were shared with community advisors and church health leaders for input at a dissemination symposium at the end of the project, showed significant improvements in dietary behaviors for both F/V and fat consumption in both treatment and comparison groups. The post-hoc analysis revealed that the treatment group had greater increases in F/V consumption than the comparison group between Phases 1 and 3.

These findings for dietary improvements as a result of a church-based intervention are consistent with previous studies, including recent longitudinal studies.^{3,17–19,23,24} The fact that both treatment and comparison groups significantly improved within the 24-month period is perhaps the most noteworthy finding. There are several possible reasons for these results. First, the HHU intervention included some nutrition-related activities, albeit limited, for the comparison group. For example, as a part of the data collection process for both treatment and comparison groups, general health promotion materials were distributed and an individualized session with a registered dietitian was conducted at Times 2 and 4 to review selected dietary data and discuss areas of needed improvement. These activities may have provided participants with sufficient knowledge and motivation to improve dietary behaviors. Another possible reason is that behavior change may have been a result of participants contemplating the dietary messages and taking action without the need for more-extensive activities. To the authors’ knowledge, brief interventions have not been explored in church-based health research, although they are well established in the alcohol and other addiction literature and also explored in dietary studies with mixed results.^32–35 The findings of this study suggest that brief interventions should be explored in the church-based settings as well.

The notable exception in the similar findings for both treatment and comparison participants are the results related to the NCI F/V Screener, which showed that there were significant differences in increase in F/V consumption between treatment and comparison groups between Phases 1 and 3. These findings could be related to the attention given to F/V consumption as a part of the intervention. For example, in Phase 1, the kickoff events implemented by the treatment churches were focused on nutrition and four of the seven postcards sent to participants focused on dietary health. Of particular note is how these behavior trends were sustained between Phase 3 and Phase 4 for the treatment participants.

The findings of this study support both socioecological theory and TTM in that health behavior changes were noted at the intrapersonal level in both treatment and comparison groups and these behaviors, for the most part, persisted over time. Future church-based studies with longitudinal designs can examine how factors in these theoretical frameworks are associated with the change process and outcomes over longer durations than presently in the literature.^6,20–26

There are several strengths in this study. First, to the authors’ knowledge, this is the first longitudinal study that examined dietary health of African Americans over a period of 24 months with four data collection phases. Another strength is that the study included validated and reliable instruments for both F/V and fat consumption.

Limitations

There are, however, several limitations, including a limited geographical area, non-random assignment of churches, small sample size, and attrition and missing data issues over the phases of the study. In addition, with all of the instruments being self-reported, the improvement in both the treatment and comparison groups may have been a result of the participants wanting to please the researchers. Finally, theoretical construct variables (socioecological theory intrapersonal, TTM dimensions) were not included in this study. Thus, future longitudinal studies will need to take these limitations into consideration.

CONCLUSIONS

This study shows that dietary behaviors of mid-life and older African Americans can be improved and sustained over 24 months using a church-based heart health intervention. Future results from the broader study will reveal whether these changes in dietary behaviors may have resulted in improved clinical outcomes for participants. However, this study suggests that the HHU intervention made a difference for these church members, regardless of whether they were treatment or comparison participants, which has implications for the adoption of brief interventions in church-based settings.