Comparative Effectiveness of Clinical and Community-Based Approaches to Healthy Weight

The Clinic and Community Approaches to Healthy Weight Trial was a randomized trial conducted in 2 FQHCs and 2 YMCAs serving the same 2 communities. The study design has been previously described (Fig 1).⁹  All study activities were approved by the Institutional Review Board at the Massachusetts Department of Public Health. The trial has been recorded in the clinicaltrials.gov national registry of randomized trials (NCT03012126). There were no adverse events.

Eligibility criteria were (1) age 6.0 to 12.9 years at referral, (2) BMI ≥85th percentile, and (3) a parent who could speak English or Spanish. We excluded (1) children who did not have at least 1 guardian who was able to follow study procedures; (2) families who planned to leave the FQHC within the study time frame; (3) families for whom the intervention was inappropriate, as determined by the primary care clinician (eg, because of emotional or cognitive difficulties that would not allow them to participate; (4) children who were taking medications that substantially interfere with growth; and (5) children who had a sibling enrolled.

Recruitment began in December 2016 and continued through February 2018; data collection ended in September 2019. Children were referred to the study by their primary care provider at a visit in which a height and weight was obtained. After receiving the referral, research assistants called parents and/or guardians to confirm eligibility, obtain verbal consent, and complete a survey. Participants were randomly assigned at the end of the survey. Families were mailed a $25 gift card for completing each of the baseline, 6-month, and 12-month surveys, as well as up to 3 $25 gift cards as incentives for attending 1 visit and completing one-third and completing two-thirds of the offered visits. The Consolidated Standards of Reporting Trials flow diagram is displayed in Fig 2.

We randomly assigned participants using 2 randomization lists (1 for each FQHC) created by the data analyst. Lists were organized into blocks of 4 to keep the sizes of intervention groups similar, and assignment was concealed in numbered envelopes from those conducting the random assignment. Participants were randomly assigned in the order in which their consent was obtained. Follow-up surveys were completed by a different research assistant to ensure those assessing outcomes were blinded to intervention-arm assignment.

We had planned to enroll 400 children into the study and were able to enroll 407. We were powered at 91%, with N = 388 at follow-up to detect a clinically meaningful difference in BMI change.

Children in both interventions were exposed to quality-of-care improvements in their FQHC, which included primary care provider weight management training and text messages to guardians for self-guided behavior-change support. Automated text messages were sent 2 to 3 times per week with tips on how to make lifestyle changes and messages that supported social and emotional wellness and promoted community-resource use and resources for mitigating unmet social needs. In addition, we mailed everyone enrolled in the interventions community resources for mitigating unmet social needs.

The HWC provided a comprehensive, multidisciplinary team intervention and offered 30 hours of contact to the parent and/or guardian and child. Siblings were allowed to attend but were not included in the trial analysis. The team included a pediatrician or advanced practice clinician, community health worker, and dietitian, with access to mental health providers as needed. The HWC team used motivational interviewing for counseling and goal setting. For the first 6 months, participants were offered monthly group visits with other families, alternating with monthly individual visits with the multidisciplinary team. Each HWC visit, whether individual or group, was 1.5 hours in length. The group-visit curriculum included understanding health, healthy eating, healthy drink choices, physical activity, bullying, sleep and screen time, food label reading, and eating outside the home. In the second 6 months, participants were offered monthly individual visits with the multidisciplinary team. Participants received twice-monthly health-coaching calls (∼10 minutes in duration) from the community health worker or dietitian in the first 6 months and monthly health-coaching calls in the second 6 months.

The YMCA of the United States of America worked with 2 local YMCAs to train staff to implement a modified version of the HWYC intervention that included 16 weekly sessions, followed by 4 sessions every other week, ending with 5 monthly sessions, for a total of 25 sessions offered to the parent and/or guardian and child over 1 year. Siblings were allowed to attend but were not included in the trial analysis. Each session was 2 hours long. In total, the M-HWYC offered 50 contact hours. This modified intervention was different from the HWYC that is being implemented nationally, which contains 25 sessions delivered over 4 months versus over 12 months, as was offered in this intervention. Additionally, the national HWYC intervention is offered only to children with a BMI in the ≥95th percentile. Two group leaders provided support, education, and activities during each session, including a family huddle, which incorporates goal setting, a parent discussion, and 60 minutes of physical activity for the children, the last 30 minutes of which is for the whole family.

Eight demographically matched, comparison community health centers in Massachusetts were chosen. All BMI data from children aged 5.1 to 14.1 years were obtained from these health centers from December 2015 to September 2019 (N = 4037). This mirrored the ages and date ranges of electronic health record (EHR) BMI measurements we obtained for our intervention participants.

The primary outcomes were change in BMI and %BMIp₉₅ per year, calculated on the basis of the Centers for Disease Control and Prevention growth curves. Height and weight were collected by clinical staff per their clinical protocols at routine visits. We extracted age and longitudinal BMI data from the electronic records of all 10 health centers for 4425 children, aged 5.1 to 14.1 years, who received their primary care at these health centers. This provided a BMI trajectory preintervention for both the intervention participants and the comparison sites.

Secondary outcomes included lifestyle changes, including changes in screen time, physical activity, sleep duration, and sugary beverages for those enrolled in either the HWC or M-HWYC at baseline and 1 year.^10–16 

Race and/or ethnicity for intervention participants was collected via survey for comparison sites, and for any intervention participants missing race and/or ethnicity on the survey, it was collected from their EHR data. Income and parent BMI were collected via self-report on the baseline survey.

We performed a 2-tailed, upper-bound 90% confidence interval (CI) noninferiority test of child %BMIp₉₅ to examine if %BMIp₉₅ change difference for the M-HYWC was appreciably inferior to that of the standard HWC intervention.⁹  Noninferiority was supported if the bounds of the 90% CI for the difference in %BMIp₉₅ changes did not contain what we considered a minimally clinically important difference, on the basis of previous data (0.87).¹⁷  We stratified these results by severe obesity versus overweight or obesity.

We then examined changes in BMI and %BMIp₉₅ and compared changes among each intervention arm to the comparison sites. To assess changes per year in BMI and %BMIp₉₅, we used indicator variables for time and the intervention arm (HWC, M-HWYC, or comparison site). We performed an intention-to-treat analysis for children enrolled in the interventions with BMI measurements at follow-up (N = 388 [95%]). We used the MIXED procedure in SAS 9.4 (SAS Institute, Inc, Cary, NC) to fit mixed linear regression models with random intercepts and slopes. The models accounted for clustering of observations within individuals and within sites. There was a larger prevalence of Hispanic patients and higher baseline BMIs in the intervention groups compared with the comparison sites, so models were adjusted for these covariates. Given the increased prevalence of severe obesity enrolled in the intervention groups compared with the comparison sites, we examined severe obesity as an interaction term that was significant, and thus we present stratified results.

We examined changes in lifestyle change from baseline to 1 year for those enrolled in the intervention arms for those who completed the follow-up survey (N = 330 [81%]). We examined guardian-reported change in sugary beverages, fast-food intake, sleep duration, screen time, and physical activity assessed via survey at baseline and 1 year. Models were adjusted for community, self-reported parent BMI category (lean or overweight or obesity) and household income (≤$20 000 vs >$20 000).

Figure 2 reveals the participant flow of the study. Clinicians referred 770 children; we attempted to contact 673 children. We enrolled 407 children: 201 to the HWC and 206 to the M-HWYC. At follow-up, we obtained BMI from 388 (95%) children. A total 4037 children were identified from the comparison sites, as detailed above. Table 1 reveals baseline characteristics of the study sample. Parental BMI and income were not balanced at random assignment between the HWC and M-HWYC. Compared with children in the comparison sites, those in the intervention groups had a higher percentage of Hispanic children and children with severe obesity.

The mean difference in %BMIp₉₅ between the M-HWYC and the HWC (considered the standard intervention) was 0.75 (90% CI: 0.07 to 1.43). Because the CI contained 0.87, we were unable to establish noninferiority. When examining stratified data by overweight or obesity versus severe obesity category, we were again unable to establish noninferiority for the M-HWYC for those with overweight or obesity, but for those with severe obesity, the M-HWYC was noninferior (Fig 3, Table 2).

Table 3 reveals participants’ unadjusted and covariate-adjusted changes per year in BMI and %BMIp₉₅. In the fully adjusted model, HWC participants’ BMI decreased by −0.23 (95%CI: −0.36 to −0.10) per year and %BMIp₉₅ decreased by −1.03 (95% CI: −1.61 to −0.45) per year compared with the comparison sites. In the fully adjusted model, the M-HWYC participants’ BMI increased by 0.02 (95%CI: −0.11 to 0.16) per year and %BMIp₉₅ decreased by −0.22 (95% CI: −0.83 to 0.38) per year compared with the comparison sites.

In Table 4, we present the unadjusted and covariate-adjusted BMI changes per year, stratified by baseline BMI category. In fully adjusted models, BMI decreased by −0.28 (95% CI: −0.51 to −0.05) for those with overweight, −0.36 (95% CI: −0.56 to −0.16) for those with obesity, and increased by 0.04 (95% CI: −0.23 to 0.32) for those with severe obesity in the HWC compared with the comparison sites. We saw similar findings for %BMIp₉₅ change for those in the HWC. In the M-HWYC program, BMI-stratified results for change in BMI and %BMIp₉₅ were not statistically significant.

In Supplemental Table 5, we outline the changes in lifestyle from baseline to 1 year. For those in the HWC, in fully adjusted models, sugary-beverage intake decreased by −2.59 servings per week (95% CI: −3.68 to −1.50), fast-food intake decreased by −0.61 servings per week (95% CI: −0.91 to −0.31), sleep duration increased by 0.53 hours per day (95% CI: 0.24 to 0.82), screen time decreased by −1.11 hours per day (95% CI: −1.67 to −0.55), and being physically active for ≥60 minutes per day increased by 1.01 days per week (95% CI: 0.56 to 1.46) from baseline to 1 year. For the participants in the M-HWYC intervention, in fully adjusted models, participants decreased their sugary beverages per week (−2.12 servings [95% CI: −3.10 to −1.14]) and increased the frequency of being physically active for ≥60 minutes per day (0.74 days [95% CI: 0.29 to 1.20]).

We found that a PWMI delivered in HWCs resulted in improvements in BMI; %BMIp₉₅ when compared with 8 comparison sites; and sugary-beverage and fast-food intake, sleep, screen time, and physical activity among a predominantly Hispanic and low-income population. We were unable to establish noninferiority in %BMIp₉₅ change for the M-HWYC overall and for those with overweight or obesity_, but participants in the M-HWYC did show improvements in sugary-beverage intake and physical activity. When BMI and %BMIp₉₅ results were stratified by weight category, neither intervention was effective in reducing BMI for children with severe obesity.

In the US Preventive Services Task Force evidence report and systematic review, 2 PWMIs reviewed offered 26 to 51 contact hours and reported BMI change.^17–19  The magnitude of BMI effect in the HWC compared with comparison sites in our study per year was −0.23 (95% CI: −0.36 to −0.10), which was lower but was significant, as compared with the Kalarchian et al²⁰  study in which researchers saw a 12-month BMI change of (SE) −0.61 (0.31) (P = .11) compared with usual care. Our BMI change was more modest than was found by Nemet et al,¹⁹  who demonstrated a −1.7 ± 2.3 vs 0.6 ± 0.9 in usual care. However, this study had only 46 children enrolled.¹⁹  Compared with our previous HWC intervention, change in %BMIp₉₅ was more robust in this current study in the HWC group and decreased by −1.03 (95% CI: −1.61 to −0.45) compared with our previous finding of a %BMIp₉₅ change of −0.87 (95% CI: −1.82 to 0.09).¹⁷  Our estimate was larger than the median changes found in the Pediatric Obesity Weight Evaluation Registry sites for children aged 6 to 11 years, with %BMIp₉₅ of −0.16 (interquartile range, −1.40 to 0.87) at 12 months.²¹  Finally, any decrease in BMI is likely of clinical benefit, and a strength of our study is that we saw this BMI decrease preserved up to 2 years postintervention.

This study was designed with the hypothesis that the M-HWYC intervention would be noninferior to the HWC in reducing %BMIp₉₅. Our findings did not support this, and there are several potential reasons. First, the M-HWYC was delivered at a different intensity and length compared with the original MEND program. The YMCA has returned to this shorter, more intense model. Second, it is possible that the presence of a pediatric medical provider in the HWC plays a role because past work has revealed that lifestyle change is often improved when supported by a pediatrician.^22,23  In addition, both HWCs had access to mental health, which may have improved outcomes. Our clinician champions at the HWCs played a large role in the implementation of the HWC: participated in biweekly technical assistance calls to share lessons learned and adapted the intervention collaboratively. The M-HWYC may have led to other health benefits that were not measured, such as quality of life or other cardiometabolic effects, as have been seen in other successful YMCA programs.^24,25  It is also possible that if there had been more engagement in the M-HWYC, we would have seen a BMI effect, as has been shown in previous YMCA interventions.^24,25  Finally, the comparison sites sample had a lower baseline BMI and proportion of Hispanic patients served compared with our intervention sites.

There are several reasons why neither intervention was effective in reducing BMI or %BMIp₉₅ for children with severe obesity. First, we may be underpowered for this subgroup. Furthermore, the net daily energy intake decrease necessary for children with severe obesity to achieve a healthy BMI is likely larger than what can be delivered in a primary care–based model.²⁶  Earlier referral to a HWC when a child has a BMI in the overweight or obese range or to a tertiary center for children with severe obesity may be preferable. From our estimates, it appears that children with severe obesity in all 3 groups (HWC, M-HWYC, and comparison sites) had flattening of their BMI, suggesting that they were receiving more weight management care from their pediatrician or elsewhere. Children with severe obesity may need to be in the program longer and receive a higher dose or additional social supports given the higher prevalence among Hispanic and Black populations who suffer from issues associated with racial inequity.²⁷ 

As with any study, this one is subject to limitations. First, our population was largely Hispanic, and our results may not be generalizable; however, in this study, we appropriately focus on those who are disproportionately affected. Finally, none of our participants met the recommended 26 hours of contact: HWC participants received a mean (SD) of 4.8 (5.7) hours, and the M-HWYC participants received a mean (SD) of 9.9 (12.7) hours. It is possible that if adherence had been higher, effects would have been more robust.²⁸ 

A comprehensive HWC PWMI that included a pediatric provider, dietitian, and community health worker with text messages and community-resource support resulted in improved BMI and %BMIp₉₅ for a low-income, largely Hispanic population. We were unable to establish noninferiority for the M-HWYC. Obesity is a complex disease that requires investment in high-quality interventions, and the HWC may offer a solution for children disproportionately affected.

We thank the families, institutions, faculty, research staff, and students who participated in the study.

%BMIp₉₅

percentage of the 95th percentile

CI

confidence interval

EHR

electronic health record

FQHC

federally qualified health center

HWC

healthy weight clinic

HWYC

Healthy Weight and Your Child

MEND

Mind, Exercise, Nutrition…Do it!

M-HWYC

modified Healthy Weight and Your Child

PWMI

pediatric weight management intervention

YMCA

Young Men’s Christian Association