|Home | About | Journals | Submit | Contact Us | Français|
To examine maternal parenting behaviors, child temperament and their potential interactions in families of obese children and demographically similar families of nonoverweight children.
A total of 77 obese youth (M body mass index (BMI) z-score values, zBMI = 2.4; ages 8–16, 59% female, 50% African American) and their parents were recruited from a pediatric weight management clinic and compared to 69 families of nonoverweight youth (M zBMI = − 0.03). Comparison youth were classmates of each obese participant matched on gender, race and age.
Maternal report of child temperament, parenting style and anthropometric assessments were obtained.
Compared to nonoverweight youth, mothers of obese youth described their child as having a more difficult temperament and their parenting style as lower in behavioral control. A logistic regression model indicated that difficult temperament, lower behavioral control and the interaction of low maternal warmth and difficult child temperament were associated with increased odds of a child being classified as obese.
Treatment-seeking obese youth and their parents are characterized by different parent and child factors when compared to nonoverweight comparison families. These findings direct investigators to test more complex models of the relation between parent and child characteristics and their mutual role in the weight-related behavior change process.
Pediatric obesity prevalence rates continue to increase to epidemic proportions worldwide. However, intervention efforts have been hindered by a lack of knowledge of specific behavioral and environmental correlates,1 making the identification of child and family risk factors imperative. The purpose of the current study was to examine parenting style (for example, maternal warmth, control) and child temperament in families of treatment-seeking obese youth compared to normal-weight comparison children.
Parents serve as one of the primary socializing influences on child and adolescent health,2 including the development of health risk behaviors (for example, poor diet, inactivity). Although parents are seen as a critical agent of change in pediatric obesity intervention models,3,4 only a limited number of studies have examined parenting in the context of pediatric obesity. For example, investigators have linked pediatric overweight/obesity to specific maternal feeding practices.5–7 However, when researchers have examined broader aspects of parenting (for example, warmth, control) that are less domain specific, the findings have been more inconsistent.8–10 Moreover, these studies utilized community samples or were limited by small sample size.
In the developmental literature, child adjustment is believed to result from interactions between parenting and child characteristics such as temperament. Children with difficult temperament (that is, biological arrhythmicity, negative mood, high intensity and withdrawal) have increased risk for negative interactions with adults and greater psychosocial difficulties.11,12 In addition, the ‘fit’ between child temperament and parenting behaviors may have important implications for child psychosocial adjustment.13 Interestingly, some studies indicate that more temperamentally difficult children are at increased risk for becoming obese,9,14,15 perhaps because child temperament may predispose some children to poor eating habits and more sedentary activity, or negatively impact their ability to change these health behaviors.14 For example, Pliner and Loewen16 found that children rated by parents as having a temperament high on shyness and emotionality (that is, child gets upset easily) had less variety of foods in their diet and were less likely to want to try new foods based on both maternal report and observations of child eating. These authors explain this result using Kagan’s theory17 that children who are ‘inhibited to the unfamiliar’ (that is, new foods) are also likely to be more shy and fearful. However, critical gaps remain in our understanding of the relation between child temperament and parenting within an obesigenic family environment.
The current study compared maternal parenting style (for example, warmth and control), child temperament and their interaction in a group of families of treatment-seeking obese youth and families of nonoverweight comparison youth using psychometrically sound measures. In previous work with this sample, mothers of obese children reported significantly greater psychosocial distress and family conflict relative to mothers of nonoverweight youth.18 Given the broader literature suggesting less optimal parenting behaviors for parents reporting distress,19,20 we hypothesized that mothers of obese children would exhibit lower levels of warmth and higher levels of control relative to mothers of nonoverweight youth. Additionally, it was hypothesized that obese children would be described by mothers as having a more difficult temperament. Exploratory analyses considered whether group differences varied by child race, gender or age. Finally, after controlling for maternal body mass index (BMI), it was expected that maternal parenting style and child temperament would be significant indicators of child/adolescent obesity status.
The current study is part of a larger, two phase investigation of obese youth (BMI ≥95th percentile) who began treatment at a pediatric weight management clinic and their parents along with demographically similar nonoverweight comparison peers and their parents. A detailed description of the study has been previously published.18 During phase 1, each obese child’s classroom was visited to collect peer relationships data and to select a comparison peer based on corresponding gender, race and age, and nonoverweight status based on visual assessments.21 The present study focused only on data collected during phase 2, which included data regarding individual, parent and family psychosocial functioning for families of obese youth and comparison peers.
Obese children and their parents were recruited from a pediatric weight management clinic requiring a BMI greater than the 95th percentile for age and gender.22 Study eligibility included children who were (1) 8–16 years of age, (2) not home schooled or in full-time special education and (3) without genetic syndromes for which obesity is a comorbidity. Parents of consecutive eligible clinic patients were asked to participate in a research study about the psychosocial health of obese youth and their families.
In total, 90 of 107 clinic families agreed to participate in phase 1. Of which 86 families (95.5%) were retained for phase 2. For the present study, we excluded one family in which the mother was unable to participate, and two families of children who had a BMI reduction from obese to overweight. In addition, for the purpose of the analyses, six families of children whose race/ethnicity was not identified as African American or non-Hispanic White were also excluded. The final sample resulted in 77 obese youth (31 boys, 46 girls) and their mothers. Of significance, although obese participants were recruited at their initiation of weight management treatment, 47% had withdrawn from the program at the time of data collection, similar to published rates of attrition within this program.23
Of the 90 potential nonoverweight peers who participated in phase 1, 3 were not contacted given their obese classmate declined participation in phase 2. In one classroom, no comparison peer agreed to participate in phase 2. In addition, we excluded 12 comparison peers who were marginally overweight (BMI>85th percentile) as well as 3 comparison peers whose race/ethnicity was not identified as African American or non-Hispanic White. The final sample of comparison youth included 69 children (32 boys, 37 girls) and their mothers.
For both obese and comparison participants, data were collected by two research assistants at a location independent of the weight management clinic convenient to the family (for example, research space and family home). After obtaining informed consent/assent, children and parents separately completed questionnaires, and height and weight measurements were obtained. Families received $100 for participation. Adult participants were the female primary caregivers in the home. Although a majority was the biological mother, 7% of female primary caregivers were a grandmother or an adult female sibling. For simplicity, they are referred to as mothers. We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during this research and institutional review boards approved the protocols.
This instrument assesses family characteristics including parent marital status, level of education and family income and size. Sufficient data are available to determine family socioeconomic status (SES) using the Revised Duncan (TSE1224,25), with higher scores representing greater occupational attainment.
The Revised Dimensions of Temperament Survey26 is a 54-item, parentreport instrument of child temperament that asks participants to respond on a 4-point scale (usually false, more false than true, more true than false, usually true). Ten aspects of temperament are measured: activity level-general, activity level-sleep, approach-withdrawal, flexibility-rigidity, mood quality, attentional focus-distractibility, persistence and three aspects of rhythmicity (sleep, eating, daily habits). Rhythmicity is the regularity of a child’s biological functions (for example, waking, becoming tired, hunger and bowel movements). Higher scores indicate higher levels of each temperamental characteristic (for example, higher activity, greater approach, greater flexibility, more positive mood, greater attentional focus, greater persistence and greater rhythmicity). This measure has demonstrated adequate reliability12 and validity.27
In addition, a more global measure of difficult temperament was used.27 Briefly, this score was computed by dichotomizing each temperament scale. Children received a ‘1’ each time they scored above the 70th percentile for general activity level, and below the 30th percentile for approach-withdrawal, flexibility-rigidity, mood, rhythmicity (if any of the three rhythmicity scales were below the 30th percentile) and attention (if either persistence or attentional focus-distractibility was below the 30th percentile). Scores could range from 0 to 6 with higher scores indicating a more difficult temperament.
Maternal self-report of parenting styles was obtained using the PRPBI.28 Three factors have been identified, each containing 10 items rated as being ‘like’, ‘somewhat like’ and ‘not like’ the parent. These factors include: (1) acceptance versus rejection—Warmth. Higher scores suggest greater caregiver appreciation of the child and greater focus on their needs (for example, ‘gives child a lot of care and attention,’ ‘believes in showing love for the child’); (2) psychological control versus psychological autonomy– Psychological Control. Higher scores reflect caregiver use negative control practices such as guilt and manipulation (for example, ‘reminds child of all the things she has done for them,’ ‘avoids looking at child when they have disappointed her’) and (3) firm control versus lax control–Behavioral Control. Higher scores indicate caregiver use of structure and order (for example, ‘believes in having a lot of rules and sticking with them’). Internal consistency is considered adequate for each of the three scales (α’s ranging from 0.69 to 0.82) as well as convergent and discriminant validity.29
Measurements of both child and maternal weight and height were obtained. Weight was measured (0.1 kg) on a portable SECA digital scale (SECA, Hamburg, Germany). Standing height was measured with a calibrated custom portable stadiometer (Creative Health Products, Plymouth, MI, USA). Measurements were taken in triplicate with participants without shoes and in street clothing. These data were used to derive BMI (kg/m2). BMI z-score values (zBMI) were calculated for children using age- (to the nearest month) and sex-specific median, standard deviation and power of the Box-Cox transformation (LMS method) based on national norms from the Centers for Disease Control.30
Groups of families were compared using two-tailed independent t-tests on demographic variables, child temperament and parenting styles. Exploratory analyses were conducted to identify possible interactions of child obesity status with sex, race and age using 2 (obese versus comparison) by 2 (sex or race or age using a median split at 12.46 years) analyses of variance (ANOVAs) (exploratory analyses with age measured continuously revealed similar findings to our 2×2 ANOVA comparisons). Holm’s correction was used to control type I errors on comparisons not hypothesized a priori.31 Hierarchical logistic regression analyses were completed to determine the odds of being classified as obese or nonoverweight based on child temperament, parenting styles and their interaction after controlling for maternal BMI.
No significant differences were identified between families of obese youth and comparison families on general demographic variables including family SES, family composition and maternal characteristics (Table 1). However, mothers of obese youth had significantly higher mean BMIs relative to comparison mothers, with 67% of mothers of obese youth also being obese (BMI ≥ 30 kg/m2), as compared to 33% of comparison mothers.
Relative to mothers of comparison peers, mothers of obese youth described their parenting style as significantly lower in behavioral control, although reported warmth and psychological control were similar for both groups (Table 2). In addition, Pearson’s correlations were used to examine the associations between maternal BMI and parenting style in families of obese children. Maternal BMI was not significantly associated with maternal warmth (r=0.08, P=0.36 ), behavioral control (r=−0.04, P=0.59) or psychological control (r=−0.05, P = 0.51).
Mothers of obese children described their children as more temperamentally difficult relative to mothers of comparison children (Table 3). When examining specific dimensions of temperament, mothers perceived obese children as significantly lower in flexibility, positive mood, approach and persistence than comparison peers. In addition, obese children were reported as having less biological rhythmicity in eating and in daily habits, as well as more activity while sleeping.
Exploratory analyses were conducted to examine if significant interactions occurred between obesity status (obese versus comparison) and gender, race or age for all parenting styles and temperament variables. No significant interactions were identified.
Hierarchical logistic regression analyses were completed to examine the odds of a child being classified as obese versus nonoverweight based on maternal report of children’s difficult temperament, maternal self-report of parenting style (for example, warmth, psychological control or behavioral control) and the interaction of difficult temperament with each parenting style variable. Maternal BMI was included as a control variable, but child sex, race and age were not used given the lack of significant interactions with obesity status in exploratory ANOVAs. All independent variables were standardized and centered before entry, and each interaction term was formed by multiplying the centered variables.33 Maternal BMI was entered in step 1 of the model, all main effects (difficult temperament and parenting style) were entered simultaneously in step 2 and all interactions between difficult temperament and each parenting style variable were entered in step 3.
Maternal BMI was significant in step 1, with a 1U increase in maternal BMI associated with increased odds of a child being classified as obese (P<0.001; odds ratio (OR)=2.8; 95% confidence interval (CI): 1.80–4.34). In step 2, maternal BMI remained significant (P<0.001; OR=2.85; 95% CI: 1.78–4.57), and difficult temperament (P<.05; OR=1.57; 95% CI: 1.07–2.30) and maternal behavioral control (P<0.05; OR=0.61; 95% CI: 0.41–0.90) were also significant. The addition of difficult temperament and maternal behavioral control resulted in a significant model (χ2(3, N=145)=39.73, P<0.001, Nagelkerke R 2=0.32). Two additional logistic regressions were run to examine the unique variance accounted for in child weight status (obese versus comparison) by each independent variable after controlling for the other. Maternal BMI was entered in the first step of each regression (odds ratio (OR)=0.90, P<0.001; Omnibus χ2 for step 1: (χ2(1, N=145)=27.65, P<0.001, Nagelkerke R 2=0.23)). For the first regression, difficult temperament (OR=0.65, P<0.05) was entered in step 2 (Omnibus χ2 for the step: χ2(1, N=145)=5.35, P<0.05, Nagelkerke R 2=0.27)), followed by maternal behavioral control (OR=1.64, P<0.05) in step 3 (Omnibus χ2 for the step: χ2(1, N=145)=6.73, P<0.01, Nagelkerke R 2= 0.32)). For the second regression, maternal behavioral control (OR=1.60, P<0.05) was entered in step 2 (Omnibus χ2 for the step: χ2(1, N=145)=6.33, P<0.05, Nagelkerke R 2=0.28)), followed by difficult temperament (OR=0.64, P<0.05) in step 3 (Omnibus χ2 for the step: χ2(1, N=145)=5.75, P<0.05, Nagelkerke R 2=0.32)). In step 3, only one significant interaction was found between difficult temperament and maternal warmth (P<0.01; OR=0.45; 95% CI: 0.26–0.77). The addition of the interaction term resulted in a significant model improvement (χ2(4, N=146)=51.62, P<0.001, Nagelkerke R 2=0.40)).
The final reduced model included maternal BMI entered in step 1 and the significant main effects and interaction entered in steps 2 and 3, respectively (Table 4). A test of the final model against a constant-only model revealed a statistically significant difference, indicating that the included variables discriminated between obese and nonoverweight children (χ2 (1,4)=51.62, P<0.0001)). For main effects, ORs indicated that a 1U increase in maternal behavioral control was associated with a reduction in the odds of a child being classified as obese by greater than 1/3 whereas a 1U increase in children’s difficult temperament was associated with an increase in the odds by one and a half times (Table 4).
To probe the interaction effect of difficult temperament and maternal warmth, we dichotomized these two variables based on median splits. Four groups were created based on the combination of difficult temperament and maternal warmth: (1) high difficult temperament and low maternal warmth, (2) low difficult temperament and low maternal warmth, (3) high difficult temperament and high maternal warmth and (4) low difficult temperament and high maternal warmth. Each child was classified into one of these groups based on their score of difficult temperament and maternal warmth. As shown in Table 5, four separate χ2 analyses were computed to compare the number of obese children and comparison peers within each of these four groups. Only one significant finding was identified, χ2(1, N=146)=6.29, P=0.01. Specifically, 69% (n=29 of 42) of obese youth were classified as being high on difficult temperament and low on maternal warmth as compared to 31% (n=13 of 42) of comparison youth.
The current study demonstrated that mothers of treatment-seeking obese youth report a parenting style lower in behavioral control and describe their children as temperamentally more difficult relative to mothers of nonoverweight peers. These effects were additive, as parenting style and difficult temperament contributed unique variance above the other, while controlling for maternal weight status. Further, the interaction of difficult temperament and lower maternal warmth differentiated groups of obese and nonoverweight youth. The present study is important as it expands our knowledge of parent and child factors and their interactions that characterize youth whose families present in clinical pediatric weight management settings. As such, these data have important clinical implications and provide direction for future research.
Data from the current study add to a small existing literature9,14 linking parent perceptions of difficult child temperament to childhood obesity. Interestingly, obese youth differed from nonoverweight comparison peers across multiple dimensions of temperament. Relative to comparison peers, obese youth were perceived by mothers as exhibiting lower biological rhythmicity as well as less approach, flexibility, positive mood and persistence. It is possible that aspects of more difficult temperament such as lower flexibility and persistence may result in an obese child or adolescent experiencing greater challenges with regard to making lifestyle changes that impact their daily routine, eating and activity. In addition, given that difficult temperament is predictive of greater internalizing symptoms (for example, depression and anxiety) and poorer social functioning in adolescence and young adulthood,13 obese youth who present in a clinical setting may be at increased psychosocial risk.
Comparison of our parenting data to that of previous researchers proves difficult given the use of different measures of parenting style,8–10 different methods of data collection (for example, observational at mealtimes)34 or more focused assessments of domain-specific parental feeding practices.5–7 The three studies that examined pediatric obesity and general parenting styles focused on combinations of warmth and control (for example, authoritative and authoritarian) and found either no association9,10 or evidence of authoritarian parenting (low warmth and high control) increasing obesity risk.8 However, by examining warmth as well as different aspects of control separately, the current study found that mothers of obese youth had a parenting style characterized by lower behavioral control (for example, fewer rules and less strict), but were not different in terms of warmth or psychological control (for example, guilt), relative to comparisons. Thus, parents who present for weight management treatment for their obese child likely have fewer household rules and are less consistent in discipline practices. Furthermore, while overall, mothers of obese children were similar to comparison mothers in warmth, the interaction of low maternal warmth and difficult temperament proved important in differentiating between groups. Longitudinal studies will therefore be necessary to understand the direction of effects among temperament, parenting and weight status.
Taken together these findings direct investigators to consider and test more complex models of the relation between parent and child characteristics and their mutual role in a weight-related behavior change process. For example, some researchers have suggested that general parenting style may have an indirect effect on child behavior by changing the effectiveness of parenting practices in a given context (for example, diet and activity).35 Therefore, within the context of prescribed pediatric weight loss, it is conceivable that specific parenting practices around child-feeding or reducing sedentary activity may have different effects on child outcomes for children of parents with different parenting styles or on children of different temperaments. For instance, children with difficult temperaments may be resistant to recommended changes in nutrition or activity levels, particularly if their parents have a history of ineffective or limited parenting skills, such as disciplinary strategies. Alternately, on the basis of their recent literature review, Ventura and Birch 36 speculated that child weight can influence what parenting practices are used.
The demonstration of independent effects of maternal overweight, maternal parenting style and difficult temperament suggest that taking all into account in the context of obesity treatment may result in more effective interventions. Parenting interventions aimed at enhancing positive parenting behaviors that also complement the child or adolescent’s behavioral style, and encouraging levels of behavioral control that are age-appropriate may help improve child behavior, the parent’s sense of competence and the parent–child relationship.
This study is the first to our knowledge to demonstrate the transactional nature of the parent–child relationship in the context of pediatric obesity; however, several limitations were noted with consequent directions for future research. Given the present data are cross-sectional, the study design precludes any conclusions regarding the temporal ordering of associations until longitudinal studies are performed. Longitudinal studies are necessary to allow for an understanding of potential reciprocal effects between such variables. Although the clinical relevance of these specific data are clear, it is critical to better understand whether these parenting and temperament patterns are also typical of obese youth in the broader community who do not or cannot access care. Although we did not find evidence of race, gender or age effects in our exploratory analyses, potential differences may become more evident when utilizing a larger sample. This is particularly relevant given literatures that have suggested ethnic differences in parenting style,37 and the rate of increase in prevalence of obesity for ethnic minority groups.38 Because mothers provided the information both on children’s temperament and their own parenting style, it may be argued that the findings on their associations represent merely shared method variance. In particular, given our prior findings that mothers who present their child for pediatric obesity treatment self-report greater psychological distress than comparison mothers,18 it could be argued that distress may lead these mothers to perceive their child as more difficult temperamentally. However, we demonstrated that these clinically referred obese youth are seen as problematic (for example, more withdrawn and lower acceptance) within the peer environment as well.21
Although the weight loss prescription is known (for example, eat less and move more), the pediatric intervention literature has not progressed at a rate to ease the individual, family and societal burden of the epidemic. Efficacy trials of family-based treatments have established strong empirical support,3,4,39 with targeting parent and child together in weight loss, as well as parent training in child behavior management skills as critical components of care. Interestingly, however, the vast majority of clinical trials have not included measures of parenting pre- and post-intervention, nor assessed whether parents were in fact using the parenting skills they were being taught, as measured outcomes focus on weight change. A recent efficacy trial by Stein et al.40 proves an exception, where better child weight loss outcomes were observed to be associated with an increase in child-report of father warmth in parenting. To our knowledge, no efficacy trials have targeted the ‘fit’ between child temperament and the parenting style in treatment models. Finally, how parent and child interactions lead to the development of obesity and its associated health risk behaviors (for example, physical inactivity and poor diet) is a vitally important area for future research utilizing prospective longitudinal and controlled designs.
This research was funded by a grant from the National Institutes of Health (K23 DK60031) awarded to the first author (MHZ). Additional support of effort for this research was funded by a post-doctoral training grant from the National Institutes of Health (T32 DK063929) awarded to the second author (REB). We extend our gratitude to Christina Ramey for her role in managing the data collection, and the staff and families of the HealthWorks! program at Cincinnati Children’s Hospital Medical Center.