Feasibility and Benefits of a Parent-Focused Preschool Child Obesity Intervention

Abstract

Objectives. This field study tested the feasibility and benefits of a program to promote 6 targeted parental behaviors to prevent obesity in children served by the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).

Methods. Two WIC sites participated in a nonrandomized, controlled 1-year prospective study to assess parents’ self-reported behavior changes.

Results. Statistical analyses of preintervention and postintervention assessments of parental behavior demonstrated significant changes in 2 behaviors: frequency of offering the child water and frequency of engaging in active play with the child. In both cases, the intervention proved effective in increasing the desired behavior.

Conclusions. The findings demonstrate the feasibility of changing parental behaviors through multidimensional education in a WIC clinic setting.

The prevalence of children and adolescents with a body mass index (BMI) greater than the 95th percentile has doubled in the past 2 decades1; African American and Hispanic children have higher rates of overweight and risk for overweight than White children.2 In 2001, the United States Department of Agriculture (USDA) reported that 26% of the nation’s 3.8 million preschool children served by Special Supplemental Nutrition for Women, Infants, and Children (WIC) programs in state health departments were overweight.3 Because many serious health problems in both adults and children are associated with obesity, effective prevention strategies are a priority for public health.4–10 Given the paucity of high-quality data to confirm the effectiveness of overweight prevention programs for children, effective interventions to prevent overweight in preschool children must be identified.11

Parents can be influential in helping to prevent overweight in their children and may benefit from specific education supporting their efforts.5,12,13 Between 5% and 25% of weight variance has been attributed to genetic influences14,15; however, the increasing prevalence of obesity during the past 20 years appears to owe mostly to environmental factors such as home milieu and parental influence. The contribution of these factors can be significant in encouraging healthy eating behaviors and physical activity, both of which may be protective against child overweight.16–19

WIC agencies serve high percentages of low-income children in the United States and are organizationally positioned to teach obesity prevention to parents.20 In 1999, the USDA competitively funded 5 WIC programs in state health departments to work collaboratively to determine the feasibility of developing and implementing model childhood obesity prevention programs. The goal of each state’s model program was to target low-income mothers whose children were between 2 and 4 years of age and to develop a program designed to prevent overweight in these children.

METHODS

Intervention

WIC staff from the Virginia Department of Health, in cooperation with university faculty members in prevention research, developed a state-specific “Fit WIC” childhood overweight prevention program grounded in social cognitive theory, self-efficacy theory,21–23 and relevant applied research studies.24–30 Fit WIC also incorporated the practical developmental guidelines outlined in “Bright Futures in Practice,” a comprehensive health supervision tool developed from age-specific anticipatory guidance protocols for parental feeding practices and role modeling to promote healthy weight in infants and in children from birth to 4 years of age.31

In the Virginia WIC program, clients attend educational groups once every 2 months and an individual session with a WIC nutritionist every 6 months. The educational groups are led by nutritionists or nutrition assistants. The Fit WIC intervention modified the educational groups’ content and provided activities to reinforce the educational message but did not modify the frequency or duration of the bimonthly and semiannual contacts.

For the Fit WIC intervention, the educational groups were used over a 1-year period to introduce WIC participants to 6 key messages: (1) increase physical activity, (2) monitor mealtime behavior, (3) limit household television viewing, (4) drink water instead of sweetened beverages, (5) consume 5 fruits or vegetables daily, and (6) increase family activities to promote fitness. These messages were presented with a set of “Fit WIC Virginia Guidance Cards.” The cards were bound together in a notebook that could be set up between the presenter and the participant so that each saw a different side. The side facing the WIC participant featured an attractive picture and a simple message (e.g., “Active kids are healthy kids. Encourage your child to get moving every day.”) in Spanish and English. The side of the notebook facing the presenter contained detailed instructions under the headings “What to Expect,” “Discussion Points,” “Setting Goals,” “Handouts,” and “Referrals.” WIC parents were encouraged to serve as role models for their children, with the expectation that long-term health behaviors of parents and children could be sustained.

These messages were reinforced by WIC staff members and collaborating organizations in the surrounding community. All clinic staff with client contact received a set of the educational cards. In addition, clinic staff, including support staff, were encouraged to participate in 6 “staff wellness challenges.” The staff wellness challenges ran concurrently with the group education classes. The nutrition and physical-activity messages presented to WIC staff mirrored those used in the nutrition education classes so that staff members could understand what challenges the participants were experiencing. WIC staff also were encouraged and reminded to model healthy lifestyle habits to WIC participants through behaviors that clients might witness while waiting in the clinic (e.g., eating homemade lunches).

Every other month, members of the local coalition of community services received from WIC health education materials to share with their clients. These materials were identical to the materials being used in the WIC clinic. Thus, WIC clients who used other community services received the same health message in a variety of venues, reinforcing the Virginia Fit WIC key messages. These venues included recreation centers, Department of Parks and Recreation facilities, the public library, a food bank, parenting classes, and a multicultural center, as well as other venues for family-oriented community services.

Randomized clinical trials provide the most reliable assessments of program effectiveness32; however, such a design for this project was not possible in the local health departments. Because a number of risk-reduction programs implemented in everyday settings have provided useful data for health education and promotion efforts, an evaluation design similar to designs reported in recent studies was selected as the most appropriate method for determining effectiveness.33–37 Two WIC clinics in Northern Virginia participated in a field study with a pretest–posttest, nonequivalent (nonrandomized) control group design38 to evaluate the Virginia Fit WIC. The 2 WIC clinics were chosen by experienced WIC professional staff who had reviewed the current WIC statistics and were familiar with how the clinics operated. The rates of childhood overweight at the 2 clinics were significantly higher than the state average. More than 24% of the children at both sites were overweight, compared with the state average of 17%. The Fit WIC program was implemented in 1 clinic, and the second clinic served as a comparison site. Like the intervention site, the comparison site provided the standard WIC interventions (nutrition education classes once every 2 months and an individual session with a nutritionist once every 6 months) but lacked the intervention’s specific content and the supportive staff and community activities. The sites had similar community resources and were both located in Northern Virginia, but they were sufficiently distant from each other to minimize crossover in the use of community resources. Table 1 ▶ shows the racial/ethnic distribution of clients at each site. The primary evaluation research issue was whether it is feasible to implement a parent-focused child obesity prevention program in a WIC clinic. The corresponding hypothesis was that clients (i.e., parents) at the clinic offering the Virginia Fit WIC intervention program would be significantly more likely than clients at the comparison clinic to report 1 or more of the 6 targeted behavioral changes.

TABLE 1—

Racial/Ethnic Distribution of Intervention and Control Groups

Race/Ethnicity, No. (%)	Intervention (n = 2171)	Control (n = 1827)
White	291 (13)	494 (27)
African American	207 (10)	330 (18)
Hispanic	1375 (63)	740 (40.5)
Asian	276 (13)	260 (14)
Other	22 (1)	3 (< .5)

Informed consent and confidentiality.

The study protocol was approved by the Virginia Department of Health and was determined to be in compliance with the policies of the University of Virginia’s internal review board. Participants provided written informed consent in which they agreed to pretest and posttest completion of measures. Participants gave consent to be contacted at their homes by telephone for posttest completion if they were no longer receiving WIC services; this was important because children are no longer eligible to receive WIC services the month they reach 5 years of age.

Participants.

At baseline, 336 WIC parents with 2- to 4-year-old children were consecutively recruited from the WIC clinics by staff during a 2-month period. Of these 336, 185 received services at the site offering the Fit WIC intervention and 151 at the comparison site. An analysis of recruitment patterns showed that among all eligible participants, 10% were not approached by staff members and 5% refused participation, resulting in an 85% participation rate. At the posttest following the 12-month intervention, 65% of the intervention group (n = 121) completed the assessment, compared with 43% of the comparison group (n = 65) (χ²₁ = 4.94; P = .026). The most common reasons for becoming lost to follow-up were that the client was no longer enrolled in WIC, had moved, or had provided incorrect contact information. Parents were more likely to complete the posttest if the target child in the family was female (59% completion vs 50% completion for girls and boys, respectively; χ²₁ = 10.75; P = .005) and if the parent’s preferred language was Spanish (61% completion vs 48% completion for Spanish and English, respectively; χ²₁ = 5.27; P = .014).

Implementation fidelity.

A process evaluation was conducted to assess the activities of the implementation protocol to determine its feasibility. In addition to using the guidance cards, staff members documented all contacts with clients in records that are subject to regular quality assurance review. The quality assurance review for the intervention clinic incorporated a documentation check on the presentation of Virginia Fit WIC information to clients. This information was maintained at the clinic to protect client confidentiality and to minimize interference of the process evaluation with standard operating procedures. Parents also completed 2 short assessments of their exposure to other intervention components. The first of these was a checklist of community resources collaborating with the Virginia Fit WIC intervention that parents had used. The second assessment contained 2 open-ended questions to ascertain which health behaviors parents recalled having seen clinic staff members model. Staff focus groups and in-service meetings provided feedback and confirmation of the results of the process evaluation.

Measures

At the time of the study, no published reliable, validated instruments were available for assessment of changes in physical activity in preschool children. It was not possible to use physical-activity monitoring devices with the children owing to budget constraints and concerns about possible liability issues (e.g., a 2-year-old child might chew or swallow the device). Consequently, reports from the parent (usually the mother) on child behaviors observed within time windows of the past 24 hours or 7 days were used to assess the child’s activity level and the frequency of other target behaviors. Items on the questionnaire either were drawn from existing child and adolescent questionnaires adapted for use with preschool-aged children (Virginia state form, WIC–305WC, 596; National Center for Health Statistics Forms, 1989, assessing physical activity) or were grounded in social-cognitive and self-efficacy theory. The frequency of most behaviors—engaging in active play with the child, watching television while eating, offering fruits and vegetables daily, and modeling family activity—was assessed with a 5-point scale (1 = none/very inactive to 5 = always/very active). The frequency of offering the child water to drink instead of sweetened drinks was assessed according to a 6-point scale (1 = none to 6 = 5 or more times per day). Mealtime behavior was assessed with 2 items: a 5-point scale for the frequency of planning meals together (1 = never to 5 = always) and a 2-point scale for the frequency of eating the evening meal together (1 = not usually, 2 = usually). A summary index of togetherness around meals was scored by taking the product of these 2 items.

Data Collection Protocol

Questionnaires were developed in English and Spanish. The Spanish version was prepared (with back-translation) by a nutritionist whose native language is Spanish, in consultation with a graduate student in Spanish literature at the University of Virginia. Parents completed pretest questionnaires onsite at the time of enrollment and posttest assessments at the conclusion of the 12-month intervention period (or by telephone from their homes if they had left the program). Staff helped clients with poor reading skills to complete the instrument.

Statistical Analyses

Baseline differences between sites in demographic characteristics and postintervention scores on the intervention exposure measures were analyzed with multivariate analysis of variance for continuous variables and χ² analysis for discrete variables. The samples being compared were convenience samples rather than randomly assigned samples; as a result, inferential statistics should be viewed as more suggestive than demonstrative of significant differences. Changes in outcome variables were tested with the general linear model analysis of variance for repeated measures. Sample sizes were too small for statistical analysis of a single fully interactive model with all 6 outcome variables. Therefore, for the purposes of the repeated-measures analysis, the primary outcome variables were divided into 2 groups. The 3 variables related to activity level (active play with the child, family activity level, and watching television while eating) were analyzed together, and the 3 variables related to nutrition (water consumption, fruit and vegetable consumption, and mealtime behavior) were analyzed together with full factorial models. All outcome analyses included 2 between-participant factors: site and language. Time (pretest to posttest) was the single within-participants factor. Statistical analyses were conducted with SPSS version 10.0 (SPSS Inc, Chicago, Ill).

RESULTS

Description of Sample at Baseline

Table 2 ▶ summarizes the demographic information for those members of each group who completed both the pretest and the posttest questionnaires.

TABLE 2—

Demographic Comparison of Intervention and Comparison Groups

	Intervention Group (n = 121)	Comparison Group (n = 65)	Statistical Test Results	P
Household characteristics, mean (SD)
Parent age, ya	32.88 (6.3)	33.06 (6.36)	F1,173 = 0.13	NS
No. of persons in householda	2.67 (1.84)	2.52 (1.9)	F1,184 = 0.26	NS
No. of childrena	3.04 (0.7)	3.09 (0.64)	F1,182 = 0.3	NS
Target child characteristics
Age, y, mean (SD)a	3.06 (0.85)	3.15 (0.9)	F1,183 = 0.81	NS
Male gender, no. (%)b	55 (46)	32 (49)	χ21 = 0.24	NS
Body mass index, mean (SD)a	17.2 (1.93)	17.38 (1.688)	F1,184 = 0.37	NS
Parent characteristics, no. (%)
Relationship to target childb			χ22 = 3.8	NS
Biological mother	105 (87)	47 (74)	. . .	. . .
Biological father	7 (6)	8 (12)	. . .	. . .
Grandmother	3 (2)	3 (4)	. . .	. . .
Other	6 (5)	7 (10)	. . .	. . .
Race/ethnicityb			χ23 = 24.9	<.001
African American	10 (8)	15 (23)	. . .	. . .
Hispanic	85 (70)	24 (37)	. . .	. . .
White	18 (15)	12 (18)	. . .	. . .
Other	8 (7)	14 (22)	. . .	. . .
Educational attainmentb			χ21 = 3.4	NS
High school or less	77 (64)	35 (54)	. . .	. . .
At least some college	31 (26)	26 (40)	. . .	. . .
Missing data	13 (10)	4(6)	. . .	. . .
Type of public assistanceb
TANF	119 (97)	65 (100)	χ21 = 1.09	NS
Medicaid	47 (38)	27 (40)	χ21 = 0.13	NS
Food stamps	113 (92)	65 (100)	χ21 = 4.5	.03
Language of questionnaireb			χ21 = 21.9	<.001
Spanish	86 (71)	9 (16)	. . .	. . .
English	35 (29)	56 (84)	. . .	. . .

Note. NS = not statistically significant (i.e., P > .05); TANF = Temporary Assistance for Needy Families. All P values are 2-tailed.

^a1-way analyses of variance, with single factor (site).

^bχ² analyses for categorical data.

Hispanic participants were overrepresented in the intervention group (69% vs 35% of control group; χ²₃ = 24.9; P < .001). Table 3 ▶ presents the baseline (preintervention) scores for the outcome variables among both groups. Multivariate analyses with 2 factors—site and language—revealed no significant interaction effect of site by language for any outcome variable and no significant direct effect of site for any outcome variable. By contrast, a direct effect of language was found across both sites for 2 variables: viewing television and consuming water. Spanish-speaking participants had significantly lower scores at baseline than did English-speaking participants for frequency of watching television while eating (mean = 2.41 [95% confidence interval (CI) = 2.18, 2.64] vs mean = 2.94 [95% CI = 2.69, 3.20], respectively) and frequency of offering the child water during the day (mean = 3.60 [95% CI = 3.31, 3.89] vs mean = 4.43 [95% CI = 4.11, 4.75], respectively). Results of the multivariate tests of statistical significance are shown in Table 3 ▶. Statistical tests were organized to test for interaction effects with language.

TABLE 3—

Pretest Results of Outcome Variables for Intervention and Comparison Groups

	Mean (SD)
	Intervention Group (n = 121)	Comparison Group (n = 65)	MANOVAa Results for Language Effect	P
Primary outcome measures
Active play itemb: no significant effect	3.57 (1.41)	3.7 (1.24)	. . .	. . .
Family activity itemc: no significant effect	3.48 (0.99)	3.3 (0.9)	. . .	. . .
Television viewing itemd: language effect	2.64 (1.21)	2.74 (1.09)	F1,153 = 7.65	.01
Mealtime behavior iteme: no significant effect	8.13 (2.02)	8.21 (1.64)	. . .	. . .
Water-consumption itemf: language effect	4.14 (1.43)	3.77 (1.24)	F1,153 = 19.52	< .001
Fruit and vegetable itemg: no significant effect	3.48 (1.15)	3.37 (1.17)	. . .	. . .

Note. MANOVA = multivariate analysis of variance.

^aMANOVA with 2 factors (site and language): significant multivariate statistic for language only: F_6,148 = 4.88; P < .001; no significant interaction effect for language by site. All P values are 2-tailed.

^b5-point scale for number of times in past 7 days (1 = none to 5 = ≥ 4 days).

^c5-point scale for ordinary activity level (1 = very inactive to 5 = very active).

^d5-point scale for number of hours parent watches television Monday through Friday (1 = < 1 hour/day to 5 = > 8 hours/day).

^e10-point scale for togetherness around planning meals and eating evening meal (1 = lowest togetherness to 10 = highest togetherness).

^f6-point scale for how many times in the past day (1 = none to 6 = ≥ 5 times).

^g5-point scale for number of days 5 fruits or vegetables were eaten in the past 7 days (1 = none to 5 = every day).

Intervention Exposure

The process evaluation confirmed the feasibility of implementing Fit WIC in the intervention clinic. Intervention staff followed the protocol, replacing the standard WIC services that were provided at the control site with the Fit WIC services. Fit WIC parents were significantly more likely than comparison-group parents to report observing WIC staff engaging in a variety of healthy behaviors (F_1,148 = 10; P = .002); 52% of Fit WIC parents, compared with 6% of parents in the comparison group, reported observing staff members engaging in 3 or more of the 6 target healthy behaviors. Fit WIC parents reported significantly more use of community activity centers than did parents in the comparison group (F_1,148 = 6.7; P = .01); 72% of Fit WIC parents, compared with 44% of parents in the comparison group, reported use of at least 1 community activity center.

Outcome Analyses

Table 4 ▶ presents the postintervention mean change in scores for both groups. The interaction effect of site and language was not significant for any of the outcome variables. For the analysis of repeated measures, the multivariate tests for pretest–posttest differences by site (time × site) were statistically significant for 2 behaviors: frequency of engaging in active play with the child (F_1,161 = 7.03; P = .009) and frequency of offering the child water (F_1,145 = 8; P = .005). At the end of the intervention period, participants in the comparison group reported decreased frequency of active play with the child, whereas participants in the intervention group reported increased frequency (Table 3 ▶). Participants in both groups reported increased frequency of offering the child water; the increase was significantly greater for participants at the intervention site (Table 3 ▶). There also was a statistically significant interaction effect for time × language for 1 variable: Spanish-speaking participants reported a greater increase in frequency of offering water instead of sweetened beverages than did English-speaking participants (F_1,145 = 9.15; P = .003). Inferential statistics must be interpreted cautiously when evaluation design includes an intervention group and a nonequivalent comparison group with convenience samples instead of participants randomly assigned to groups. The study’s methodological limitations also include differential rates of follow-up at the intervention and comparison sites and reliance on parental self-report measures; however, the differential effectiveness related to the target behaviors suggests that the self-report results do not reflect simply an increase in the adult’s providing the learned, socially desirable answer, because equivalent results were not found for all 6 targeted behaviors. No specific randomization technique was used to determine the order of presentation, nor was any specific seasonal correspondence intended. The 2 behaviors that showed significant change were not presented sequentially; thus, no order effect was in operation. The evaluation design did not permit assessment of the independent contribution of each of the 3 intervention components to the observed effects.

TABLE 4—

Mean Change in Outcome Variables for Intervention and Comparison Groups

	Mean (95% CI)
	Intervention Group (n = 121)	Comparison Group (n = 65)	Statistical Testa Result
Primary outcome measures
Active play itemb	0.47 (0.14, 0.8)	−0.22 (−0.7, 0.26)	F1,161 = 7.03*
Family activity itemc	−0.3 (−0.62, 0.03)	−0.3 (−0.66, 0.06)	F1,161 = 0.33
Television viewing itemc	−1.4 (−1.67, −1.12)	−1.38 (−1.76, −1)	F1,161 = 0.42
Mealtime behavior scalec	−0.86 (−1.4, −0.31)	−1.31 (−2.13, −0.49)	F1,145 = 0.13
Water-consumption itemd	0.64 (0.19, 1.09)	0.16 (−0.16, 0.49)	F1,145 = 8*
Fruits and vegetables itemc	0.53 (0.23, 0.83)	0.46 (0.03, 0.89)	F1,145 = 0.36
Secondary outcome measures
Efficacy belief itemc	5.06 (4.37, 5.75)	5.27 (4.3, 6.24)	F1,141 = 0.05
Outcome expectancy scalec	2.69 (1.44, 3.94)	2.2 (0.43, 4)	F1,141 = 0.03
Risk perception iteme	0.09 (−0.2, 0.38)	0.11 (−0.3, 0.52)	F1,141 = 0.09
Readiness-to-change itemc	1.47 (0.56, 2.38)	1.34 (0.05, 2.63)	F1,141 = 0.64
Client satisfaction scalec	1.1 (0.27, 1.93)	1.1 (0.06, 2.28)	F1,134 = 0.03

Note. CI = confidence interval. All P values are 2-tailed.

^a2-factor (site and language) general linear model analysis of variance for repeated measures.

^bOnly site effect was significant; no significant direct or interaction effect was found for language.

^cNo significant direct or interaction effects were found for site or language; statistic is for site effect.

^dSignificant effects were found for time, time by site, and time by language; statistic is for site effect.

^eSignificant direct effect for language; no significant site-by-language interaction effect; statistic is for site effect.

*P = .01.

DISCUSSION

The 3-pronged intervention strategy (educational groups, staff reinforcement, community reinforcement) successfully influenced 1 food-related behavior—frequency of offering a child water instead of sweetened drinks—and 1 activity-related behavior—frequency of active play with the child—for both English-speaking and Spanish-speaking participants. One important implication of this finding is that similar intervention strategies can change both food-related and activity-related behavior. Because weight in children is a multidetermined characteristic, interventions aimed at preventing or decreasing overweight must act upon multiple determinants. This study demonstrates the feasibility of influencing parental behavior to promote healthy eating and activity behaviors in preschool children by following an anticipatory guidance model for nutrition and fitness. However, the implementation experience also demonstrates the difficulties, costs, and limitations of such interventions. The least costly components were the development of the materials and the use of the materials in the regularly scheduled group and individual sessions. The most costly component was the additional staff time needed for contacting community partners, following up with WIC participants, educating WIC staff in the effective use of the prepared materials, and mentoring them as they participated in the staff modeling activities. The WIC program operates according to a well-established protocol with the type and quantity of foods and beverages provided to clients determined by federal regulations. The WIC professional and support staff were willing and able to make changes in provision of services to parents of preschool children by using available WIC time slots to substitute focused child obesity prevention services for standard services. The WIC staff also were able to provide valuable insight into potential problems with childhood obesity prevention programs within the guidelines of existing government policies. For example, they pointed out that the intervention goal of replacing sweetened drinks, such as juice, with water might seem contradictory to WIC parents, given than the program currently provides juice in the monthly food package. Possible causes of site differences in follow-up rates were discussed with WIC program and site administrators. From these discussions, the suggestion emerged that staff at the intervention site were more efficient in contacting parents for follow-up than were staff at the comparison site, who had to be reminded several times of the importance of contacting parents for follow-up data. This hypothesized cause illustrates the timeintensive nature of such interventions and the difficulty of maintaining necessary follow-up activities as a priority in already stressed and stretched schedules. Community coalition members and WIC nutritionists formed an active partnership to support the project by announcing area activities consistent with the Fit WIC messages. At the beginning of the project, coordination between the community coalition and the WIC clinic was poor owing to the slow start-up time. Adding an on-site nutritionist with direct responsibility for enhancing the community–WIC collaborative relationship solved this problem. By the project’s conclusion, the community coalition members demonstrated strong support for the final goals of Fit WIC. This gradual improvement illustrates the need to prioritize staff support and community outreach.

A longitudinal, randomized clinical trial with change in child BMI as an outcome is needed to confirm the association of health behaviors of parents and preschool child health behaviors and weight status. If such an association is confirmed, interventions similar to WIC Fit may have practical applications in other settings that provide services to low-income mothers with preschool children, such as the Head Start program. More research—with adequate sample sizes and validated outcome measures—is needed to further examine cultural differences in response to preschool child obesity prevention programs. In addition, prevention research is needed to investigate the relative impact of different groups on parental behaviors and the relative appeal of different prevention messages targeting the preschool child. Finally, effective replication of childhood obesity prevention programs across agencies must be assessed within the framework of feasibility, acceptability, and cost analysis studies.