Previous studies have well-documented the significant linkage between family context and child psychological development (Hetherington and Martin, 1986), particularly in families with children having potential affective and behavioral problems (e.g., Cox and Paley, 2003; Smeekens et al., 2007; Lavigne et al., 2014). A wealth of literature has identified numerous family factors that placed children at increased risk of developing ODD, including poor family function, low marital quality, parental maladaptive behavior, paternal substance abuse, and low quality parent-child relationship (Burke et al., 2002; Greene et al., 2002; Marmorstein et al., 2009; Matthys and Lochman, 2010).

Both cross-sectional and longitudinal studies also proved the robust associations between family factors and child ODD symptoms. For example, Lavigne et al. (2014) found that higher scores on family risk factors (family conflict, parent hostility in parenting, child emotion temperament) were positively associated with child ODD symptoms in a cross-sectional study. Similarly, Smeekens et al. (2007) found that multiple family factors, like parent-child interaction, parent-child attachment, and various parental, child, and contextual characteristics, served as predictors of child later externalizing behavior problem in a longitudinal study.

Although previous studies have identified a wide range of family factors linked to child ODD symptoms, the majority of these studies focused almost exclusively on family factors at either one level or mixed levels. Different associations of multi-level family factors and child ODD symptoms remained unclear. Cox and Paley (2003) proposed that family was a dynamic and interactive system consisting of interdependent subsystems, including whole-family subsystem, parent-child subsystem, co-parenting subsystem, and marital subsystem. Factors at these subsystems served as important contexts in understanding child development (Cook and Kenny, 2006). Based on this theory, Lavigne et al. (2012) investigated the relations between family subsystems and child ODD symptoms. They proposed a multi-domain model of family factors for ODD symptoms encompassing contextual factors, parental depression, parenting, and child characteristics. Furthermore, they assessed the pathways of these multi-domain factors on child ODD symptoms. However, this model was too complex to clearly demonstrate the hierarchy of these family factors. Considering the complexity of Lavigne's model, Lin et al. (2013) proposed a concise three-level model on the basis of the Family System Theory (Cox and Paley, 2003) and the Bioecological Model (Bronfenbrenner, 2009), which delineated the mechanism of ODD development. This model divided family factors into three levels, including the whole level, the dyadic level (including couple dyadic and parent-child dyadic levels) and the individual level (including parental individual and child individual levels). In the present study, we adopted this comprehensive model to test the associations between family factors and two dimensions of ODD symptoms.

Previous research highlighted the crucial role of family function at the whole level in predicting children's affective and behavioral symptoms (Slee, 1996; Lucia and Breslau, 2006). Family cohesion, as one aspect of family function, was negatively correlated with child internalizing and externalizing problems (Lucia and Breslau, 2006). Children in less cohesive families were more likely to develop conduct disorder and delinquent symptoms (Leflore, 1988; Slee, 1996). In cohesive and well-adapted families, members were prone to interact with each other in a harmonious manner, which further promoted parental and children's emotion regulation abilities. Thus, children in such families were less likely to develop affective and behavioral problems.

Among the numerous dyadic level factors, marital quality, and parent-child relationship were noted as the most influential factors on children's psychological problems within the family (Lin et al., 2013). A considerable body of research has linked marital quality directly and indirectly to children's internalizing and externalizing problems (Cummings and Davies, 2002). Lower marital quality was directly related to subsequent child internalizing problems. Exposure to marital conflict was upsetting to children and appeared to elicit child maladjustment in both direct and indirect ways (Zimet and Jacob, 2001; Erath and Bierman, 2006). Additionally, poor parent-child relationship appeared to be a robust risk factor of children's behavioral adjustment (Masten and Garmezy, 1985). Negative parent–child relationships were significantly associated with child externalizing disorders (Waschbusch, 2002), namely attention-deficit/hyperactivity disorder (ADHD), conduct disorder (CD), and ODD (Burt et al., 2005). Secure parent-child relationships protected children from adverse developmental outcomes (Groh et al., 2014), whereas insecure parent-child relationship was a risk factor of child anxiety and other internalizing problems (Kerns and Brumariu, 2014).

Besides the factors at the whole and dyadic levels, specific individual characteristics of parents and children also accounted for the development of ODD symptoms (Eisenberg and Fabes, 2006; Cunningham et al., 2009). Smith (2010) suggested that parental emotion regulation played a vital role in child developmental outcomes. Parental dysregulated emotion contributed to inappropriate emotional expressions or behaviors toward children, which further predicted children's maladaptive emotional outcomes (Muhtadie et al., 2013). Moreover, parents with difficulty in emotion regulation exhibited poor emotional coaching, which would positively predict children's disruptive behavior, especially when children demonstrated high levels of emotion liability/negativity (Dunsmore et al., 2013). Previous studies also found that children's deficits in emotion regulation contributed to the manifestation of externalizing and internalizing symptoms (Southam-Gerow and Kendall, 2002; Kim and Cicchetti, 2010), whereas children with better emotion regulation demonstrated less subsequent internalizing and externalizing problems (Blandon et al., 2010; Kim and Cicchetti, 2010).

Previous studies indicated that ecological context played a crucial role in shaping family functioning, parenting styles, and child development (Schleyer-Lindenmann, 2006). Currently, the majority of studies examining the associations between parenting styles and child ODD symptoms were based on Western samples. The information on how parenting in non-Western societies, such as Mainland China, affect children's psychological development remains scarce. Moreover, the implementation of the one-child policy before 2015 introduced changes into parenting styles. Some families adopted the “child-centered” approach into child-rearing, particularly among well-educated parents (Chang et al., 2003). The “child-centered” parenting style linked children and parents together closely, but it also increased the possibility that children become spoiled. Meanwhile, most parents put high expectations on their children, expecting children to be obedient, respectful, and excellent in schoolwork. Consequently, parents would be under a lot of pressure, particularly when their children misbehaved. In the present study, we examined how the specific family context and parenting styles in China contribute to children's ODD symptoms.

We proposed a comprehensive framework concerning how multi-level family factors are differently related to children's ODD symptoms. Specifically, we included family function as the whole level family factor, marital quality, and parent-child relationship as the dyadic level family factors, and emotional regulation of both parent and child as the individual level family factors. Our study aimed to specify the different associations of the three-level family factors with child affective and behavioral oppositional defiant disorder symptoms. We hypothesized: (1) Among families with children identified with ODD, there are significant associations between three levels of family factors and children's ODD symptoms; (2) Three levels of family factors exhibit different associations with ODD symptoms. Specifically, family function, as the whole level factor and the most distal factor, is less correlated with ODD symptoms than family factors at the dyadic and individual levels. (3) Multi-level family factors exhibit different associations with two dimensions of children's ODD symptoms in that all family factors are more closely linked to affective ODD symptoms than to behavior ODD symptoms.

Data in this study was derived from a large research project on family risk and protective factors of ODD children in China. Between 2013 and 2014, 14 elementary schools in northern (Beijing), eastern (Shandong province), and southwestern (Yunnan province) parts of China participated in this study. School psychologists in these 14 primary schools invited all the class master teachers who taught first grade through fifth grade to nominate the children in their classes who exhibited ODD symptoms. The nomination was based on the ODD symptoms assessment checklist, which was derived from the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR; American Psychiatric Association, 2000). In total, 187 class master teachers nominated 360 students from 7,966 children in the participating schools.

After the initial nomination, two clinical psychologists from Beijing Normal University further confirmed the assessment, using a semi-structured interview guide in interviewing class master teachers. The interview was based on the DSM-IV-TR's ODD diagnostic criteria. After the interview, two clinical psychologists discussed their assessments for each child to ensure accordance. Only children with both clinical psychologists' diagnoses of ODD were recruited into the current study. Ultimately, 305 children identified as displaying the symptoms of ODD were invited to participate in the research (3.8% of the children in the participating schools), and 282 parent-child dyads agreed to join the project. Thirty-three families were excluded from the analyses because more than 20% of the items on one questionnaire were missing.

The final ODD sample consisted of 249 parent-child dyads, including 80 father-child dyads and 169 mother-child dyads. Ages of parents ranged from 25 to 53 (paternal age M = 38.35, SD = 5.08; maternal age M = 36.64, SD = 4.28). All children (180 boys and 69 girls) aged between 6 and 13 (M = 9.59, SD = 1.59). Among these children, 80% were the only child in their families. Approximately, 141 families (56.6%) had a monthly income over 5,000 Chinese Yuan (the average monthly income for Chinese urban families is about 5,485 Chinese Yuan; National Health and Family Planning Commission of the PRC, 2015). More than half of the participants, including 151 fathers (60.6%) and 141 mothers (56.6%), reported that they completed junior college education or higher education.

After signing the informed consents, each participating child was asked to deliver a parent survey package to their father/mother. Parents were invited to fill in the parent survey and return the completed surveys to the class master teachers in 1 week.

All study procedures, including informed consent and child assent, were approved by the Institutional Review Board of Beijing Normal University. Each participant received a token of appreciation to acknowledge their participation in the study. All the 305 children who met the criteria for ODD were provided with the opportunity for treatment from psychiatrists in Beijing Anding Hospital, psychological counselors, and family therapists from the Center of Family Study and Therapy at Beijing Normal University.

All data in the current study was based on self-reports from parents. To reduce concerns about common method variance, we conducted Harman's one-factor test using Exploratory Factor analysis with a principal axis factoring method of extraction. The first extracted component accounted for 16.48% of the total variance, indicating that the common method variance was not of great concern (Aulakh and Gencturk, 2000).

Then, the preliminary analyses were performed to evaluate the descriptive statistics and correlations between the study variables. Preliminary data analyses were performed using SPSS 20.0. Bivariate correlations were conducted to test initial relations between variables (see Table 1).

Finally, the structural equation model (SEM, see Figure 1) was tested using Mplus Version 7.0 (Muthé and Muthén, 2010). Model fit criteria used in this study were chi-square statistic (χ²), the goodness-of-fit index (CFI), the Tucker-Lewis index (TLI), the root mean square error of approximation (RMSEA), and the standardized root mean residual (SRMR). A model is typically considered as acceptable fit the data when the CFI and TLI values are larger than 0.90, the RMSEA value < 0.08, and the SRMR value is no > 0.08 (Hu and Bentler, 1999).

Descriptive statistics and bivariate correlational analyses were presented in Table 1. As shown in Table 1, basically each dimension of family function, marital quality, parent-child relationship, parent emotion regulation, child emotion regulation, and ODD symptoms was significantly correlated with each other (ps < 0.01).

Before testing the hypothesized structural model, the preliminary analyses examined the gender, primary school grade, number of siblings, as well as household income differences in two dimensions of child oppositional defiant disorder symptoms (see Table 2). Among all these demographic variables, only the children's gender was significantly associated with observed variables. Hence, children's gender was placed into the subsequent path model test as covariates, yet it did not show any significant effect in the models presented below.

Table 2 presented the means and standard deviations for ODD affective and behavioral symptoms across basic demographic variables. Further, ANOVA was conducted to compare two dimensions of child ODD symptoms in various demographic groups. There was no significant difference of child affective ODD symptoms [F_{(1, 247)} = 1.17, p = 0.28] and child behavioral ODD symptoms [F_{(1, 247)} = 0.06, p = 0.81] between families with only one child and families with siblings. Families with different monthly incomes did not show significant difference in reporting child affective ODD symptoms [F_{(1, 247)} = 2.38, p = 0.12] and child behavioral ODD symptoms [F_{(1, 247)} = 0.36, p = 0.55]. Children from different primary grades did not score differently on child affective ODD symptoms [F_{(2, 246)} = 2.08, p = 0.13] and child behavioral symptoms [F_{(2, 246)} = 0.59, p = 0.56]. However, it is noteworthy that there were significant differences in boy's and girl's affective and behavioral ODD symptoms. Hence, children's gender was placed into the subsequent model test as covariates, yet it did not show significant effect on either dimension of children's ODD symptoms in the model presented below. A model of proposed relationships among the study variables, controlling for child gender, was presented in Figure 1.

Model results indicated that the proposed model fit the data very well (χ² = 236.23, df = 114, CFI = 0.95, TLI = 0.93, RMSEA = 0.06). Results of the model indicated that on the whole level, family function was not significantly associated with either dimensions of child ODD symptoms. When testing the association of dyadic level factors on child ODD symptoms, we found that both marital quality (β = 0.19, p < 0.05) and parent-child relationship (β = −0.64, p < 0.001) were significantly linked to the ODD affective symptoms. However, only the parent-child relationship was significantly related to the ODD behavioral symptoms (β = −0.66, p < 0.001). At the individual level, both parental emotion regulation (β = −0.17, p < 0.05) and child emotion regulation (β = −0.25, p < 0.001) were significantly correlated with child ODD affective symptoms, but only child emotion regulation was negatively linked to ODD behavioral symptoms (β = −0.19, p < 0.01).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.