Predictors of Overweight and Obesity in American Indian Families with Young Children

Abstract

Objective:

To describe sociodemographic factors and health behaviors among American Indian (AI) families with young children and determine predictors of adult/child weight status among these factors.

Design:

Descriptive, cross-sectional baseline data

Setting:

One urban area and 4 rural American Indian reservations nationwide

Participants:

450 American Indian families with children ages 2–5 participating in the Healthy Children, Strong Families 2 intervention

Intervention:

Baseline data from a healthy lifestyles intervention

Main Outcome Measures:

Child BMI z-score and adult BMI, multiple healthy lifestyle outcomes

Analysis:

Descriptive statistics and step-wise regression

Results:

Adult and child overweight/obesity rates were high-−−82% and 40%, respectively. Food insecurity was high (61%), and multiple lifestyle behaviors, including fruit/vegetable and sugar-sweetened beverage consumption, adult physical activity, and child screen time, did not meet national recommendations. Adult sleep was adequate, but children had low overnight sleep duration of 10hrs/day. Significant predictors of child obesity included: more adults in the household, an adult AI caregiver, high adult BMI, gestational diabetes, high child birth-weight, and family activity/nutrition score.

Conclusions and Implications:

Multiple factors influence early childhood obesity in AI children, who experience disproportionately high risk for obesity development. Interventions to mitigate modifiable early life influences, home environments, and health behaviors are needed.

INTRODUCTION

Obesity rates in American Indian (AI) children are among the highest of all races and ethnicities.¹ Moreover, AI children are the only group of children for whom an increase in obesity prevalence has been reported since 2003.² Development of obesity in young AI children is of particular concern not only because it persists into later life and greatly increases the risk of chronic disease but also because of other health risk factors unique to these families, including historical trauma, extremely high rates of poverty, and inconsistent access to health care.^3,4 Despite evidence for the importance of early years in determining individual factors related to obesity risk, little is known about these risk factors among American Indian children, who experience disproportionately high risk for development of obesity. Specifically, there is scant literature on obesity and related health behaviors in very young AI children (2–5 years), and no studies to date have assessed this age group concurrently with information about an adult caregiver or included data on health behaviors of this age group within the family home environment.

The Healthy Children, Strong Families 2 intervention was designed to target this critical window by focusing on early childhood and the family home environment in AI communities. Healthy Children, Strong Families 2 consisted of monthly mailed toolkits delivering lessons focusing on child obesity prevention and health promotion coupled with text messages and FaceBook, and is described in detail elsewhere.^5,6 The current paper presents cross-sectional child-, adult-, and family-level data at baseline enrollment into Healthy Children, Strong Families 2, which offers a critical opportunity to assess obesity and related risk factors among AI families with young children. The dataset includes 450 adult/child dyads from four rural and one urban AI community ranging in population density from approximately 3.5 to 3,000 people per square mile. This geographic diversity is noteworthy as very few studies have included urban AI families. The study also explored relationships between multiple factors influencing child body mass index (BMI) z-score and adult BMI within these households, which may provide important insights into future intervention design with these communities.

METHODS

Study Design and Participants

This study describes cross-sectional baseline data from Healthy Children, Strong Families 2, a randomized controlled trial of an obesity prevention intervention (methods described in full elsewhere).⁶ Community stakeholders were engaged in all phases of the study, from participant recruitment through discussion and contextualizing of study findings. Enrollment occurred in five communities from February 2013 to April 2015. Informational brochures were sent home with children at Head Start programs for the reservation-based sites, and information was displayed at a health center for the urban site. Inclusion criteria consisted of enrolling a child between the ages of 2 and 5 years, the ability to attend data collection visits, and a working cell phone (due to delivery of some intervention components via text messaging). Identifying as AI was not required, as race, ethnicity, and culture are often complicated issues among families in these communities. For example, a parent may not identify as AI but will identify their child as AI or vice versa, leading to perceived discrepancies in numbers of AI participants. However, all families were recruited either from within reservations or from organizations providing services to American Indian families. Institutional Review Board (IRB) approvals were obtained from University of Wisconsin, participating tribal councils, and tribal IRBs upon their request. Adult participants provided signed informed consent for themselves and their participating child.

Anthropometrics and Surveys

Height and weight were collected according to standardized protocols and converted to age- and sex-specific BMI percentiles for children⁷ and adult BMI for adults.⁸ Adult participants who were pregnant provided self-reported pre-pregnancy weight. Participating adults also completed a packet of validated self-report surveys: screeners based on NHANES 2009–10 (adults)⁹ and the 2010 National Youth Physical Activity and Nutrition Survey (children)¹⁰ assessed foods eaten during the previous week; the Godin Leisure‐Time Exercise Questionnaire (adults)¹¹ and Netherlands Physical Activity Questionnaire (children)¹² assessed physical activity; the Family Nutrition and Physical Activity Survey assessed family activity and the home environment¹³ (with two validated questions on food security from the USDA Household Food Security survey¹⁴); the Physical Activity and Nutrition Self-Efficacy Scale (PANSE),¹⁵ Perceived Stress Scale (PSS),¹⁶ and 12-Item Short Form Health Survey (SF-12)¹⁷ assessed stress and other psychosocial measures, and a validated survey assessed cultural identity.¹⁸ The research group created or adapted surveys to assess health history, sleep, screen time use, social media use (adult only), and readiness to change in six health behaviors (adult only, assessed on a 5-point scale with higher scores indicating greater readiness).

Analysis

Descriptive statistics were performed on physical measures and surveys. The study also examined the possible association of sociodemographic variables and health behaviors and attitudes with adult BMI and child BMI z-score using a step-wise multiple regression approach (SPSS Statistics v.23, IBM Corporation, Armonk, NY), with the demographic and health behavior variables at baseline entered as predictors. The step-wise procedure selected a subset to include in the final model that explained the maximal variance.^19,20 For adults, all adult variables listed in Table 3 were included, in addition to household income, education, ethnicity (AI vs. non-AI), urban vs. rural, and baseline PANSE and cultural involvement scale scores. For children, all child variables from Table 3 were included, in addition to adult BMI, the adult variables for stress and the SF-12, and the variables concerning pregnancy and birth known to be relevant to obesity (e.g., birthweight, breastfeeding duration). Child predictors were entered in three stages outlined in Table 3: first, “Sociodemographic Variables,” followed by “Genetic/Early Developmental Variables,” and finally “Family Lifestyle Variables” in the third stage. Significance level was set at p<0.05.

Table 3.

Factors Predicting Adult BMI and Child BMI z-Score in American Indian Adult-Child Dyads as Determined by Logistic Regression.

Predictors of Adult BMI
	B	95% confidence levels		Beta	Sig.
FNPA nutrition score	−1.512	−4.335	1.312	−0.067	0.293
FNPA activity score	−1.838	−4.082	0.406	−0.107	0.108
Income	0.103	−0.27	0.476	0.037	0.588
Highest education	0.185	−0.534	0.904	0.033	0.613
Food insecurity present	0.457	−1.573	2.487	0.028	0.658
In urban setting	−0.413	−2.311	1.485	−0.026	0.669
Adult total screen time	0.001	−0.004	0.006	0.034	0.561
Adult fruit and vegetable/day	0.044	−0.031	0.119	0.068	0.248
Adult soda pop per week	−0.075	−0.151	0	−0.119	0.049
SF-12 mental health composite	−0.041	−0.161	0.078	−0.053	0.498
SF-12 physical health composite	−0.273	−0.387	−0.158	−0.285	<.001**
Adult sleep weekdays	−0.008	−0.018	0.002	−0.084	0.136
PANSE total score	−0.019	−0.093	0.056	−0.03	0.623
PSS total score	−0.033	−0.228	0.161	−0.026	0.737
Adult is AI	0.339	−1.692	2.37	0.019	0.743
Cultural involvement total score	0.094	−0.294	0.482	0.029	0.634
Predictors of Child BMI z-Score
Sociodemographic Variables
Highest adult/caregiver education	−.078	−.164	.008	−.105	.076
Number of adults	.180	.061	.299	.153	.003**
Number of children	−.033	−.116	.050	−.042	.434
In urban setting	−.141	−.388	.105	−.064	.261
Income	−.078	−.172	.016	−.101	.104
Adult is AI	.279	.039	.519	.116	.023*
Food insecurity present	−.052	−.302	.198	−.023	.681
Genetic/Early Developmental Variables
Weeks of breastfeeding	−.001	−.004	.001	−.053	.306
Adult BMI	.024	.010	.039	.176	.001**
Gestational diabetes present	.389	.031	.747	.110	.033*
Mother smoked in pregnancy	.265	−.085	.615	.078	.138
Birthweight in ounces	.013	.008	.019	.247	<.001**
Child age in months	.002	−.006	.011	.026	.617
Family Lifestyle Variables
Child total screen time	.000	−.001	.001	.015	.776
Child fruit and vegetable/day	−.052	−.136	.031	−.065	.220
Child soda pop per week	−.021	−.067	.024	−.051	.357
FNPA nutrition score	.208	−.173	.588	.069	.284
FNPA activity score	.297	.019	.574	.130	.036*
SF-12 mental composite	.010	−.006	.025	.087	.215
Child prefers vigorous games	−.073	−.168	.021	−.077	.129
PSS total score (adult)	.008	−.015	.032	.050	.486
Child sleep minutes, weekdays	−.002	−.003	.000	−.095	.067

^*p<0.05;

^**p<0.01

RESULTS

Participants.

Table 1 provides sociodemographic information on participants. The study recruited 210 (46.6%) urban-based and 240 reservation-based dyads. Of the caregivers, 91.1% were a biological parent (392 mothers, 18 fathers), 7.3% were a biological grandparent, and 1.6% identified as aunt or foster parent. Compared to reservation-based participants, urban-based participants had lower income levels (p = 0.008) and more children were identified as non-AI (44.7% vs. 29.1%, p=0.001). Twice as many urban families consisted of an AI adult with a non-AI child (17% vs. 8%), most of whom (57%) were identified as Hispanic. In the overall sample, 30% of adults reported total annual income of <$5,000, and only 20% reported >$35,000 per year. Food insecurity was reported for 61% of households,²¹ and 80.9% of adults reported participation in WIC (Special Supplemental Nutrition Program for Women, Infants and Children).

Table 1.

Demographic Information for 450 American Indian Adult-Child Dyads Recruited from Four Rural Communities and One Urban Health Clinic.

Age			Mean	SD
Adult age, years (n=450)			31.4	8.45
Child age, months (n=450)			45.0	12.98
Gender			N	%
Adult (% female)			426	94.7%
Child (% female)			226	50.2%
Education			N	%
High school equivalent or less			169	37.6
Some college/assoc. degree			235	52.2
College degree or post-graduate			46	10.2
Household income			N	%
<$5,000			132	30
$5,000–19,999			124	28.2
$20,000–34,999			94	21.4
>$35,000			90	20.5
WIC participation			80.9%
Percent food insecure			61.0%
Adults in household (mean, high)			2.1, 8
Children in household (mean, high)			2.6, 11
Ethnicity*	Adult		Child
	N	%	N	%
American Indian	368	81.8	390	86.7
Asian	4	0.9	6	1.3
Native Hawaiian/Pacific Islander	2	0.4	2	0.4
Black/African American	4	0.9	14	3.1
Hispanic/Latino	51	11.3	80	17.8
White	77	17.1	89	19.8
Other	1	0.4	2	0.4

^*Ethnicity percents total to more than 100 due to multiple choices. WIC, Special Supplemental Program for Women, Infants, and Children.

Obesity and related behaviors.

Table 2 presents descriptive results for the adult BMI and child BMI z-score and health behaviors related to obesity risk. For adults, 60% were classified as obese, 22% as overweight, and 18% as healthy weight. Twenty-seven adults were excluded because of current or recent pregnancy. Among children, 22% were classified as obese, 18% as overweight, and 60% as healthy weight. The median adult readiness to change score was 4 (“have actively started making changes”) on all health behaviors except for screen time, which had a median score of 3 (“no, but have been making specific plans to start in the next 30 days”). Adults reported engaging in moderate to vigorous activity 3.8 ± 3.8 times per week (in 15-minute bouts). The child activity survey does not translate into metabolic equivalents or other comparable measures, so results are not reported here. Families reported that a television was on in the home 5. ± 0.5 hours per day, and 114 ± 120 and 92 ± 87 minutes of television were watched for adults and children, respectively. When all screens were included (e.g., television, phones, computers, video games, tablets), average watch times were 188 ± 187 minutes/day for adults and 124 ± 120 minutes/day for children. For diet variables, adults consumed 14.5 ± 12.9 servings/week of sugar sweetened beverages compared to 9.2 ± 9.8 for children; combined fruit and vegetable servings per week were 15.9 ± 12.5 and 16.7 ± 11.9 for adults and children, respectively.

Table 2.

Cross-sectional Weight Status and Health Behaviors of American Indian Adult-Child Dyads

	Adult n=450	Child n=450
Weight Status
BMI/BMI z-score (mean ± SD)	32.0 ± 7.9	0.78 ± 1.10
Healthy weight (%)	18%	60%
Overweight (%)	22%	18%
Obese (%)	60%	22%
Early child development
Gestational diabetes, n (%)		47 (11%)
Maternal smoking during pregnancy, n (%)		62 (14%)
Birthweight, ounces (mean ± SD)		118.0 ± 21.6
Low birthweight (<88 oz), n (%)		26 (6%)
High birthweight (>144 oz), n (%)		30 (6%)
Diet variables in servings/week (mean ± SD)
Sugar-sweetened beverages	14.5 ± 12.9	9.2 ± 9.8
Fruits & vegetables	15.9 ± 12.5	16.7 ± 11.8
Screen time (all screens: TV, phone, computer, video games)
Total daily minutes (mean ± SD)	187.8 ± 187.3	124.3 ± 120.2
Physical activity (mean ± SD)
Number of 15-min bouts of moderate/vigorous activity/week	3.8 ± 3.8
Sleep (mean ± SD)
Weeknight hours	8.0 ± 1.5	10.1 ± 1.1
Weekend hours	8.6 ± 1.6	10.3 ± 1.1
Home environment (FNPA) (mean ± SD)
Total score	3.08 ± 0.37
Nutrition score	3.22 ± 0.36
Physical activity score	2.91 ± 0.48
Readiness to change (mean ± SD)a
More physical activity	3.63 ± 1.06
More fruits and vegetables	3.86 ± 0.90
Less screen time	3.04 ± 1.30
Less added sugar	3.71 ± 1.14
Manage stress	3.65 ± 1.08
Improve sleep	3.61 ± 1.08
Stress (PSS) (mean ± SD)	16.49 ± 6.27
SF-12 (mean ± SD)
Physical health component	49.18 ± 8.12
Mental health component	46.80 ± 9.91

^aScale of 1–5, with higher scores indicating greater readiness to change.

Adults reported sleeping 8.0 ± 1.0 hours per night, while children slept 10.1 ± 1.1 hours. Adults reported they fell asleep or went to bed with TV on as follows: never (58%), some of the time (22%), most of the time (14%), or always (6%). For children, responses were never (63%), some of the time (24%), most of the time (10%), or always (3%). Adults reported that 51% of children had a working television in the room where they slept.

Factors associated with obesity.

Table 3 provides coefficients for all variables entered into regression models predicting adult BMI and child BMI z-score. The adult model explained 8.5% of variance in adult BMI, most of which was attributable to the SF-12 Physical Health Composite score (PCS; 7.2% of variance). For children, the overall model explained 16.5% of variance in child BMI z-score. For the three stages of the regression, the “sociodemographic factors” explained 2.6% of the variance, the “genetic/early developmental factors” 10.5%, and the “family lifestyle factors” 3.4%. Among the sociodemographic factors, higher adult/caregiver education and living in an urban setting were associated with a lower BMI z-score, while having a primary caregiver who identified as AI was associated with higher BMI z-score. Among the genetic/early developmental factors, gestational diabetes, high birthweight, and high adult BMI were positively associated with BMI z-score. Among the family lifestyle factors, the FNPA activity total score was associated with increased child BMI z-score.

DISCUSSION

This cross-sectional cohort of AI families with young children represents AI communities from extreme rural to urban across five states. It is the first cohort of AI families to have included both child and adult data from the same household and information on multiple health habits, including the lesser-studied obesity risk factors of sleep and stress. Novel findings from these data included low child but not adult sleep, high adult readiness to change lifestyle behaviors, and that child BMI z-score was positively predicted by the number of adults in the household and a caregiver who self-identified as AI (among other factors). In addition, adult BMI was predicted by SF-12 physical scores. Our data add to the scant literature on obesity risk factors and health behaviors for AI families, who experience disproportionate risk for obesity.

Similar to many other studies,^1,22–24 high rates of early childhood overweight/obesity in young AI children (40%) were found in this study. Multiple behaviors related to obesity risk also were examined, including the first report of sleep data in AI children. Although adults in this sample reported meeting recommended levels of sleep,²⁵ these data suggest a substantial number of AI children may not be meeting current American Academy of Sleep Medicine guidelines.²⁶ Daytime nap data were not collected, which represents a limitation of this study. This cohort of children had reported screen time use of 60+ minutes more than recommended by the American Academy of Pediatrics,²⁷ and >50% of them had a working TV in the room where they slept. The prevalence of low sleep (<11 hours/night) and high use of screen time (>2 hours/day) was also seen in two other large cohorts of mixed-race (but non-AI) children.^28,29 A recent longitudinal study found inconsistent bedtimes in early childhood predicted obesity in early adolescence but found no relationship between screen time in early childhood and adolescent obesity.³⁰ The finding that more than half of the children in our study had a TV in their bedroom, however, suggests a possible interaction of these obesity risk factors in this population that warrants further investigation.

Adults in this sample reported under-consuming fruits/vegetables³¹ in conjunction with high intake of sugar-sweetened beverages (14.5 servings/week). The younger children in this sample met recommended intake levels of fruits/vegetables but not the older children,³¹ with a reported 9.2 servings/week of sugar-sweetened beverages overall, likely substantially higher than recommended levels of <10% added sugars daily. Comparable poor dietary habits were noted for AI children in a South Dakota reservation–based study,²² and dietary patterns are known to become less optimal over time in children.³² Adults in the current study also reported levels of physical activity approximately 100 minutes below national guidelines of 150 minutes/week.³³ These findings suggest a focus on diet and activity in the context of obesity prevention remains an appropriate approach. Moreover, adults in the current study reported high levels of willingness to change diet, activity, and other health behaviors. Some studies have found that higher levels of willingness to change are associated with significant levels of subsequent behavior change,^34–36 suggesting this variable may be important in understanding health behavior.

This study was unique in providing data on a significant number of urban AI families, who often are not included in large federal-level surveys or any smaller, reservation-based studies. Of note, the research group previously found differences in food insecurity and dietary patterns between urban- and rural-based families in this sample;²¹ similar findings have been reported for the general US population.³⁷ No significant differences were observed in the other measured health behaviors (e.g., adult/child sleep, adult physical activity) in urban vs. rural families (data not shown). Although the number of urban and rural families was approximately equal in this sample, only one urban-based site was included, which limits the generalizability of the finding to other urban AI families. In addition, the four rural reservation-based sites may not be representative of all of the AI communities nationwide.

Similar to other studies, child BMI z-score was shown to be predicted by family lifestyle variables (family activity) and genetic/early developmental factors (adult BMI, report of gestational diabetes mellitus, high birthweight).^38,39 However, a positive association of child BMI z-score with maternal smoking in pregnancy was not observed as has been shown in previous studies in Indigenous children,^39,40 perhaps because maternal smoking rates were much lower in the current sample (14% vs. >40% in other studies). Interestingly, number of adults in the household and having a caregiver who identified as AI were positively associated with child BMI z-score. Household number has been inversely associated with child BMI in Colombian children, which may be related to food scarcity.⁴¹ A very high prevalence of reported household food insecurity was observed among the families in this sample, but it is unclear how this factor may be affecting child BMI. Moreover, discussions with community partners reveal that a higher number of adults in the household make lifestyle changes more challenging due to disruptions in meal and bedtimes, irregular caregiving arrangements, and the increased number of people supported by the household budget and resources. The high number of adults in households reported by the families in the current study is likely related to the high cost and low availability of housing for AI families, and a more comprehensive understanding of the interaction of these factors and their contribution to child health is warranted.

This study’s data are limited by being a baseline sample of families signing up for a two-year health and safety intervention; therefore, there may be more motivated families in this cohort. However, in several of the reservation sites, greater than 60% of the total number of families with children in this age range were recruited suggesting a wider range of families may be represented. Although a range of communities was included, these findings may not be representative of all AI communities. Other limitations include the large number of self-reported measures used and the lack of other sampling strategies for many lifestyle variables. However, the study worked to maximize the accuracy of reported information in two ways: local site coordinators assisted adults as they filled out surveys, and several surveys included similar questions. Finally, there are likely other important obesity predictors that were not included in the study. Specifically, household substance abuse, domestic violence, and maternal depression and other factors related to adverse childhood experiences and trauma may influence child weight and represent important areas of future inquiry.

Implications for Research and Practice

A more comprehensive understanding of the individual-, family-, and community-level factors influencing obesity is vital to health promotion efforts currently ongoing in many AI communities. The current findings illustrate the interaction of multiple factors influencing early childhood obesity, including complex family- and environmental-level variables. Future approaches should seek to understand the impact of additional adults in the household and food insecurity on family eating habits and their ability to make nutrition-related changes. The study also found inadequate sleep in the children and high levels of caregiver stress, perhaps relating to the high levels of food insecurity, poverty, and historic and intergenerational trauma often experienced by these families. These challenges highlight areas for future intervention and also warrant an investigation of the resiliency employed by families to overcome these challenges.