Home-Delivered Pediatric Weight Management for Low-Income Families: A Randomized Controlled Trial

Children from lower-income households are almost twice as likely to have obesity as those from higher-income households.^1,2 They are also less likely to benefit from pediatric weight management interventions,³ which may stem from barriers associated with socioeconomic disadvantage that interfere with accessing and adhering to treatment.⁴ For example, households of lower socioeconomic status have lower rates of attendance in pediatric weight management programs,^5,6 and attendance is predictive of weight loss outcomes.^7,8 

Some interventions have incorporated home visits in which intervention staff deliver some or all of a weight management program in the family’s home.^9–14 In addition to mitigating transportation, childcare needs, and other barriers to attendance, home delivery may improve outcomes by enabling interventionists to provide tailored recommendations based on direct observations of the child’s home environment. This could be a powerful enhancement, as many pediatric weight management interventions recommend restructuring the child’s physical home environment and daily routines in ways that promote healthy eating and physical activity.^15–20 

Prior studies have been limited to comparisons of home-delivered pediatric weight management interventions with qualitatively different and less intensive comparator interventions.²¹ Therefore, it remains unclear whether any superior effects of these interventions are attributable to home delivery per se or are an artifact of their greater treatment intensity.

This manuscript reports the primary findings from the Creating Healthy Environments for Chicago Kids (CHECK) trial, which compared otherwise identical standard-of-care pediatric weight management interventions delivered in home versus clinic settings. The trial included children from lower-income households aged between 6 and 12 years with overweight/obesity (BMI percentile >85). The primary clinical outcome was 12-month change in child BMI z-scores (zBMI). The proportion of children achieving clinically significant weight loss of at least .25 zBMI^22,23 was a secondary clinical outcome. Group differences in intervention session attendance and contact time and changes in accelerometer-based moderate-vigorous physical activity, obesity-related household routines, and child weight-related problem behaviors were examined as intermediate and secondary outcomes. The impact of the COVID-19 pandemic on trial conduct and weight loss outcomes is described.

The CHECK trial was a 2-arm randomized controlled trial conducted in Chicago, Illinois (clinicaltrials.gov: NCT03195790). Enrollment occurred between August 2017 and August 2021. Intervention delivery and data collection concluded in September 2022. COVID-19 pandemic stay-at-home orders at the study site took effect on March 16, 2020, and continued through late July 2020. Additional restrictions and remote schooling remained in effect through February 2022. A description of the CHECK trial’s methodology has been published.²⁴ 

Children were recruited from 2 large primary care practices at major Chicago medical centers through provider referrals and telephone outreach to families identified through electronic health record queries. Eligible households had annual household incomes less than or equal to 200% of the federal poverty guideline at the time of enrollment and included at least 1 child aged between 6 and 12 years with a BMI percentile of at least 85.²⁵ Key exclusion criteria included 1) a child or primary caregiver who was not fluent in either English or Spanish, 2) a child or caregiver who was unwilling or unable to attend home-based and clinic-delivered treatment, 3) a medical contraindication or barrier to weight loss (eg, long-term oral steroid use or syndromal obesity), 4) a major medical or uncontrolled psychiatric condition likely to interfere with treatment, 5) families who lived or planned to relocate more than 15 miles from the study site, 6) families living in temporary or group housing (as the intervention targeted the home environment), or 7) unsafe conditions in the home (eg, extreme unsanitary conditions or gang or illegal activity). Study procedures were approved by the Institutional Review Boards of Rush University Medical Center (protocol 15120306) and University of Illinois Chicago (protocol 20170452). Informed consent and child assent were obtained.

Both arms received identical versions of a standard-of-care, family-based pediatric weight management program with 18 in-person sessions and 12 monthly telephone support calls (30 total planned contacts). The intervention was informed by empirically supported socioecological models that emphasize the role of children’s physical and social environments in shaping behaviors that influence energy balance.^15,26 Caregivers were trained to complete a weekly “home monitoring checklist” to identify and address obesity-promoting aspects of the physical home environment or household routines. Checklists were reviewed with interventionists at each contact and referenced when collaboratively setting behavioral goals. Trained interventionists with backgrounds in nutrition or nursing presented a 10-module educational curriculum focused on key topics such as beverage choices, portion control, healthy eating on a budget, and use of positive reinforcement to promote weight control behaviors. Child and other adult attendance at intervention contacts was encouraged but not required.^27,28 The CHECK intervention is described in greater detail elsewhere.²⁴ 

Arms differed only in the location of in-person intervention sessions. Households assigned to home-delivered treatment received in-person sessions in their home, whereas households in the clinic-delivered arm only received sessions in an outpatient primary care clinic. Parking or transportation to clinic sessions was reimbursed. To maintain equivalence, sessions in both arms were scheduled during typical clinic hours.

Participants were randomized to clinic-delivered or home-delivered treatment in a 1:1 ratio, in permuted blocks of size 4 and 6. Randomization was stratified by referring site and weight status (BMI percentile: 85–94.9 vs ≥95). Child sex was later added as an additional stratification factor to correct for an imbalance that emerged midway through enrollment. Randomization was implemented centrally by the trial biostatistician, and the sequence remained concealed until treatment allocation occurred. In cases where multiple children per household met eligibility criteria, 1 child was randomly selected as the “index case” for outcome assessment purposes (eg, accelerometry/sleep actigraphy protocol, child-specific surveys). Interventionists were blinded to the identity of the index case in households with multiple children. Households were also randomized to an interventionist who worked with families in both arms and delivered all intervention components to mitigate any confounding influences of interventionist characteristics on between-arm comparisons. Outcomes assessors were blinded to treatment allocation.

Baseline assessment activities took place at both the study site and participants’ homes to ensure that, at least in principle, families had the capacity to attend intervention sessions in either treatment arm. Subsequent follow up assessments at 3, 6, 9, and 12 months after randomization occurred at participants’ homes. All outcomes were collected at the 6- and 12-month assessments, whereas the 3- and 9-month assessments were limited to anthropometric measurements and documentation of adverse events.

The primary outcome of 12-month changes in child zBMI, based on the Centers for Disease Control and Prevention norms,²⁵ was derived from height and weight measurements taken in light clothing with research-grade equipment.

The proportions of planned in-person (18) and telephone (12) visits attended were determined for each household. Session duration (in minutes) was calculated from session start and end times documented by intervention staff. Total intervention contact time was calculated for each participant.

Daily minutes of moderate-vigorous intensity physical activity were measured with a 7-day accelerometer protocol. Children wore the device (Actigraph GT3x+, ActiGraph LLC) at the right hip during waking hours. Accelerometer scoring rules are described in Supplemental Table 1.

The Family Nutrition and Physical Activity screening tool includes 20 items rated on a 4-point scale reflecting household routines linked to pediatric obesity risk.^29,30 Items are summed to a total score (20–80 possible), with lower values reflecting greater engagement in household routines associated with pediatric obesity.

The 24-item Lifestyle Behavior Checklist³¹ was completed by caregivers in reference to the index child’s experience of obesity stigma, misbehavior with food (eg, hiding or sneaking food), overeating tendencies, and physical inactivity during the preceding month. Higher scores reflect more problematic behavior in each domain.

Caregivers reported their household size, household income, and children’s age, sex, and ethnicity/race for the purpose of describing the sample.

Following the issuance of pandemic-related public health guidelines and stay-at-home orders in March 2020, in-person intervention visits in both arms were held via videoconference or telephone (n = 46 [1.9%] of all in-person visits). As restrictions eased in June 2020, in-person visits resumed under COVID-19 transmission precautions (eg, masking and social distancing). Monthly telephone support calls continued throughout the pandemic. Data collection procedures requiring physical interaction, such as anthropometric measurements, were conducted with modifications between March and June 2020, including training caregivers to collect physical measurements with video monitoring by research staff. Other data collection procedures were largely unaffected.

An analytic sample of N = 212 (n = 106 per arm) was determined to provide at least .80 power to detect a between-arm difference in zBMI change as small as −0.10, with α = .05 (two-tailed) and assuming (from prior studies) a standard deviation of zBMI change ranging from 0.35 to 0.43 and a correlation between repeated zBMI measurements of ρ = 0.82. An enrollment target of N = 266 accounted for up to 20% attrition.

Variable distributions were examined with descriptive statistics and diagnostic tests. Multiple imputation was used to conduct intention-to-treat analyses while accounting for cases with missing values on the primary and secondary outcomes. Sensitivity analyses using full information maximum likelihood to account for missing data and including only those who completed the final assessment yielded similar results and are not reported. The primary outcome analysis used linear regression models to compare the home-delivered and clinic-delivered arms on zBMI change. In these models, 12-month zBMI values were predicted from treatment group and baseline zBMI, both with and without additional adjustment for child sex, age, and ethnicity/race. To enable comparisons with other trials, similarly structured models examining change in BMI (kg/m²) and BMI percentile are also reported. Unadjusted and adjusted treatment group comparisons on the secondary outcome of the proportion of children achieving a clinically significant weight loss of at least .25 zBMI were conducted using logistic regression. Group differences in treatment attendance and total contact time, and changes in BMI, BMI percentile, average daily minutes of moderate-vigorous physical activity, pediatric obesity-related household routines, and weight-related problem behaviors were examined using similarly adjusted and unadjusted regression models. Associations of treatment attendance and contact time with zBMI change were tested in regression models including only patients who completed the final assessment (n = 241, 90%) to avoid conflating effects of study attrition and attendance.

As prior aggregated analyses showed that weight loss outcomes in this trial were substantially altered following the onset of the COVID-19 pandemic,³² post hoc analyses examined pandemic-related and intervention-related temporal trends in zBMI in each arm. Participants were coded as prepandemic, peripandemic, or postpandemic based on whether their timeline of participation had concluded prior to the pandemic (reference date of March 16, 2020³³), spanned the onset of the pandemic, or occurred entirely after pandemic onset, respectively. Mixed-effects linear models with random intercepts and unstructured covariance estimated zBMI change across 5 time points. The 3-way interaction of pandemic group (reference group: prepandemic), treatment arm (reference group: clinic-based), and time (visits 1–5, modeled continuously), was tested to determine whether treatment group difference in zBMI change varied with the timing of the pandemic.

Participant flow through the trial is depicted in Figure 1. The characteristics of enrolled children, primary caregivers, and households at baseline are summarized in Table 1. No sociodemographic or anthropometric differences between arms were present at baseline.

Table 2 provides estimates of change in weight outcomes from baseline to 12 months by treatment arm. Mean zBMI change in the home-delivered arm (−0.031, SD = 0.26) and clinic-delivered arm (−0.002, SD = 0.30) was not significantly different in adjusted models (coeff = −0.020, SE = 0.036, P = .58). The pattern of results was unchanged when an outlying value of zBMI change (−2.26) was excluded. The percentage of children that demonstrated clinically meaningful weight loss of at least 0.25 zBMI was 13.7% in the home-delivered arm and 8.6% in the clinic-delivered arm, which was not a significant group difference (adjusted OR = 1.53, 95% CI [0.61–3.81]; P = .36).

Data on the duration, attendance, and total contact time for in-person treatment sessions and telephone support calls are shown in Table 3. Families in the home-delivered arm attended a greater number of CHECK’s 18 planned in-person treatment sessions than those in the clinic-delivered arm (median = 11 vs 6.5 sessions, respectively; P < .0001). Both groups completed a similar number of telephone support calls (P = .46). The duration of in-person sessions was virtually identical between arms (median = 45 minutes in both groups), whereas telephone support calls were approximately 1 minute longer in the clinic-delivered arm (median of 11 vs 10, respectively; P = .0006). Total in-person contact time was substantially higher in the home-delivered arm than the clinic-delivered-arm (median = 500.0 vs 315.5 minutes, respectively; P = .0001). Across arms, the number of in-person treatment sessions attended (coeff. = −0.008, SE = 0.003, P = .006) and total in-person contact time (coeff. = −0.0001, SE = 0.0001, P = .046) were significant predictors of 12-month zBMI change. Neither association was moderated by treatment group.

The 12-month changes in daily minutes of moderate-vigorous physical activity and obesity-related household routines did not differ between treatment groups (Table 2). Children randomized to home-delivered treatment demonstrated a greater overall reduction in weight-related problem behaviors than those randomized to clinic-delivered treatment (coeff. = 5.36, SE = 2.19, P = .01), although group differences on the individual subscales of this measure were not statistically significant.

Exploratory analyses examined whether patterns of change in zBMI, measured at 5 time points across 12 months of intervention, were moderated by the COVID-19 pandemic (prepandemic [n = 103], peripandemic [n = 93], and post-pandemic [n = 73]). Baseline zBMI significantly differed across pandemic periods; children entered the trial at significantly higher (P = .042) levels of adiposity after the pandemic (n = 73, M = 2.22, SD = 0.46) compared with before the pandemic (n = 103, M = 2.07, SD = 0.49). Baseline zBMI during the peripandemic period (n = 93, M = 2.14, SD = 0.48) was higher than prepandemic and lower than postpandemic levels but not significantly different from either. A significant 3-way interaction of time point by pandemic period by treatment group (F = 6.03, P = .0025) indicated that the treatment group difference in patterns of zBMI change was moderated by the pandemic phase. Before the pandemic, there was a significant treatment group by time point interaction (coeff. = −0.03, SE = 0.008, P = .0004) in which children randomized to home-delivered treatment exhibited larger average zBMI reductions across time points than those randomized to clinic-delivered treatment. The treatment group by time point interaction was not significant in the peripandemic (coeff. = 0.008, SE = 0.008, P = .32) or postpandemic (coeff. = 0.02, SE = 0.014, P = .28) periods. Figure 2 depicts the distributions of 12-month zBMI change in each arm during the pre-, peri-, and postpandemic periods.

Adverse events were monitored for all child and adult household members. Twenty-four serious adverse events were adjudicated (6 in the home-delivered arm, 18 in the clinic-delivered arm). No serious adverse events were related to trial participation, and most (67%) involved adult participants. No unanticipated problems occurred.

The CHECK trial examined whether delivering family-based pediatric weight management in the home setting improves weight loss outcomes relative to the traditional clinic-delivered intervention among children from lower-income households. Home delivery substantially increased session attendance and in-person contact time. While higher attendance and contact time were associated with better weight loss outcomes in the entire sample, the primary comparison between home-delivered and clinic-delivered treatment was non-significant. Exploratory analyses suggested that home-delivered treatment had a small but statistically significant beneficial impact on weight loss prior to the COVID-19 pandemic. Children receiving home-delivered treatment showed greater reductions in weight-related problem behaviors, but change in physical activity and obesity-related household routines were similar in both arms.

To our knowledge, this is the first trial that directly compared home-delivered and clinic-delivered interventions with identical behavioral treatment targets and planned treatment intensity. Overall, weight losses were variable and the average weight loss across participants was small. Only 11% of children demonstrated clinically meaningful weight loss of at least 0.25 zBMI. The small overall intervention effect is partially attributable to the COVID-19 pandemic, during which many enrolled children exhibited significant increases in adiposity.³² Additionally, the sample was composed entirely of children from lower-income households, with about 95% of children and caregivers identifying as Hispanic or Black/African American, populations in which prior trials have reported minimal weight losses or even overall weight gain.^3,34 Very few studies have specifically focused on households of lower socioeconomic status, which has been cited as a barrier to developing more generalizable pediatric obesity treatment guidelines.³⁵ The present study suggests that home delivery reduces the burden of attending intervention sessions, which may be more acute in families of lower socioeconomic status. However, the lack of a significant weight loss difference across the entire trial period and the small overall treatment effect suggest that home delivery alone is insufficient to achieve meaningful improvements in weight loss outcomes.

The COVID-19 pandemic began midway through this trial, which temporarily affected trial conduct and precipitated a large population-level increase in children’s obesogenic behaviors and weight gain^36–39 that was particularly pronounced in economically and racially marginalized groups.^40–42 This represents a type of confound known as a history effect.³³ Sensitivity analyses examining patterns of weight change over the duration of the trial indicated that children entered the trial at higher levels of adiposity following the pandemic, consistent with the population-level increase. Prepandemic weight loss outcomes were significantly better for children randomized to home-delivered versus clinic-delivered treatment, whereas children in both arms exhibited adiposity increases in the peripandemic phase. Children participating in the trial after the pandemic lost weight to a similar degree in both arms. Of note, many pandemic-related societal changes persisted through much of the postpandemic phase of the CHECK trial, including Chicago Public Schools’ exclusive use of remote online instruction through August 2021. Therefore, it is unclear whether any apparent weight loss advantage of home-delivered treatment observed before the pandemic generalizes to the present postpandemic context.

The methodological strengths of the CHECK trial included the use of essentially identical interventions that differed solely in the location of delivery, inclusion of English- and Spanish-speaking participants, detailed documentation of intervention receipt, and strong retention across follow-up (90%). The use of sensitivity analyses to examine treatment effects before and after the COVID-19 pandemic is a recommended approach,^43,44 but these analyses involved smaller sample sizes, were underpowered, and should be considered exploratory. The trial’s focus on children from urban lower socioeconomic households has high public health significance but limits generalizability to rural settings.

Delivering family-based pediatric weight management interventions in the home setting can facilitate treatment attendance among lower-income households, which is itself related to better weight loss outcomes. However, a beneficial effect of home delivery on weight loss outcomes was only apparent prior to the onset of the COVID-19 pandemic. Developing strategies to identify families that will benefit most from home delivery, and understanding how weight management interventions may need to evolve in a postpandemic world, are priorities for future research.

BMI

body mass index

CHECK

Creating Healthy Environments for Chicago Kids

FPG

federal poverty guideline

MVPA

moderate-vigorous physical activity

zBMI

body mass index z-score.

We are grateful for the assistance of Michelle Li, Eileen Cameron, Katerina Newman, Katey Feit, Melanie Battaglia, Tami Olinger, Serina Silvestry, Kelly Wagner, Barb Mascitti, Elizabeth Avery, Dan Lindich, and Alyssa Cho.