Developmental Trajectories of Anxiety Subtypes from Childhood to Early Adolescence: the Role of Parenting Practices and Maternal Distress

Previous studies investigating developmental trajectories of anxiety symptoms have largely focused on trajectories within childhood or within adolescence. Trends suggest that anxiety symptoms generally remain stable from early to late childhood (e.g., Ahlen & Ghaderi, 2019); Broeren et al., 2013; de Lijster et al., 2019). In a population-based sample of 7499 children and their caregivers from the Generation R Study based in the Netherlands, anxiety symptoms remained low from age 1.5 to 10 years for a substantial proportion (82.4%) of children (de Lijster et al., 2019). Across adolescence, several longitudinal studies have utilized latent growth curve modelling to examine trajectories of anxiety subtypes with conflicting results (e.g., McLaughlin & King, 2015; Murray et al., 2021; Nelemans et al., 2013; Ohannessian et al., 2017). In an ethnically diverse U.S. sample of 1437 10–15 year-olds, McLaughlin and King (2015) demonstrated, on average, declining generalized anxiety, social anxiety, separation anxiety, and physical anxiety symptoms across three timepoints over one year. In contrast, Nelemans et al. (2013) identified diverse developmental trajectories for five different anxiety disorder symptoms in a community sample assessed annually over 8 years from early to late adolescence. Investigations that account for the vulnerable transition from childhood to adolescents could shed light on this equivocality. Further, while emerging work has assessed the course of anxiety symptoms from childhood to adolescence, these studies predominantly evaluated overall or generalized anxiety symptoms (e.g., LoParo et al., 2024; Morales-Muñoz et al., 2023; Tseliou et al., 2024), thus warranting further investigation into subtype-specific developmental trajectories from childhood to adolescence.

Prior work has also documented significant sex differences in anxiety levels and their developmental courses. In cross-sectional investigations, adolescent girls consistently report a higher lifetime prevalence of anxiety disorders (Kessler et al., 2012) and more severe anxiety symptoms (Hale et al., 2008; McLaughlin & King, 2015) in comparison to similarly-aged boys. However, the nature of sex differences in trajectories of specific anxiety subtypes, particularly from childhood to adolescence, remains unclear. In studies that examined sex-specific trajectories of anxiety subtypes from childhood to adolescence (e.g., Steinsbekk et al., 2022; Xu et al., 2021), some work observed no significant differences in growth rates of generalized anxiety, social anxiety, and separation anxiety symptoms from early childhood to adolescence (Steinsbekk et al., 2022). In contrast, other investigations found that girls are more likely to be identified in trajectories classes marked by initially high or increasing, compared to consistently low, social anxiety symptoms (Xu et al., 2021). Other investigations have explored sex differences in developmental courses of specific anxiety symptoms within adolescence, similarly yielding equivocal results. Girls have been found to exhibit a gradual increase of symptoms from early to mid/late-adolescence (i.e., from approximately 10 to 19 years of age¹) while boys show declining symptoms, particularly for generalized anxiety and social anxiety symptoms (Hale et al., 2008; Nelemans et al., 2013). Conversely, other studies indicate declining generalized anxiety, panic disorder, and social anxiety symptoms from middle to late adolescence for girls and more stable trajectories for boys (Ohannessian et al., 2017). Taken together, these mixed findings point to the need for more careful examination of specific anxiety subtypes alongside demographic characteristics, such as age and sex, to clarify the developmental trajectories of anxiety symptoms from childhood to adolescence.

The present work integrates Brofenbrenner’s (1979) ecological systems framework with Rapee et al.’s (2019) model of adolescent development and social-emotional disorders. Drawing on prior theoretical work, parenting practices and maternal mental health represent the most salient factors in the early home environment that shape the onset and progression of anxiety symptoms (Rapee et al., 2019, 2022). These caregiving factors are theorized to provide a crucial foundation for children’s coping strategies, decision-making, and socioemotional functioning (Sirois & Bernier, 2018). Hence, while early adolescence is marked by a magnified importance of peer relationships for social support, self-esteem, and self-concept development (Sebastian et al., 2008; Westenberg et al., 2007), pre-adolescent caregiving factors establish a foundational layer of risk or resilience that interacts with characteristics of adolescent development to shape developmental trajectories of anxiety.

In previous work, two key dimensions of parenting—sensitivity and autonomy support—have been identified to predict unique portions of children’s psychosocial adjustment (Sirois & Bernier, 2018). Parental sensitivity (i.e., attunement and responsiveness to child’s cues; Ainsworth, 1979) is posited to facilitate healthy caregiver attachments and scaffold socioemotional skills (e.g., emotion regulation) which set a foundation for healthy psychosocial functioning and fewer internalizing symptoms across childhood and adolescence. Parental autonomy support, a related but distinct concept, refers to parents’ encouragement of children’s independent problem-solving and decision-making. These practices are theorized to provide children with socioemotional resources, such as self-esteem, that promote psychosocial adjustment and protect against the development of anxiety symptoms (Sowislo & Orth, 2013). Accordingly, higher levels of responsive and autonomy-granting parenting practices in middle-to-late childhood are longitudinally associated with lower internalizing problems, such as anxiety symptoms, into early adolescence (Gao et al., 2022; Matte-Gagné et al., 2015). Additionally, maternal mental health particularly in early childhood has been identified as a salient predictor of children’s psychopathology outcomes during childhood and adolescence. Poor maternal mental health in early childhood contributes to young children’s greater exposure to stressful home environments and mothers’ dysfunctional cognitions and emotions (Burstein & Ginsburg, 2010), which are risk factors for socio-emotional difficulties including anxiety symptoms (Beesdo-Baum & Knappe, 2012). Empirical evidence also shows that maternal depression and anxiety symptoms in early childhood have enduring relations with internalizing symptoms into early adolescence (Bailey et al., 2021; Daundasekara et al., 2021) and predict persistent trajectories of emotional problems from early childhood to adolescence (4 to 17 years of age; Tseliou et al., 2024). In line with an ecological systems approach, substantial evidence points to the pertinent role of parenting and maternal mental health in the onset and maintenance of anxiety symptoms from childhood to adolescence.

While the aforementioned studies shed light on trajectories of anxiety symptoms through childhood and adolescence and the role of caregiving factors, the extant literature has a few but key limitations. First, greater empirical attention is required to understand how core anxiety subtypes may evolve from childhood to adolescence. Nascent studies have assessed the course of composite anxiety or internalizing symptoms from childhood to adolescence (e.g., LoParo et al., 2024; Morales-Muñoz et al., 2023; Tseliou et al., 2024), which may mask the variability of trends across specific subtypes. Developmental psychopathology models (Beesdo-Baum & Knappe, 2012) note that the predominant expressions of anxiety symptoms are tied to normative developmental periods and tasks. Alongside greater time with peers and transitions to more complex social environments, the shift from childhood to adolescence is marked by the increased salience of peer relationships, heightened sensitivity to social evaluation, and vulnerability of self-concept development to social feedback (Sebastian et al., 2008; Silk et al., 2012; Westenberg et al., 2007). These socioemotional changes are theorized as developmental risk factors that can trigger social anxiety symptoms (Rapee et al., 2019, 2022). In contrast, separation anxiety symptoms appear to predominate in early-to-middle childhood and decline from late childhood as children assert greater autonomy from their parents (e.g., McLaughlin & King, 2015). The transition from childhood to adolescence thus provides a critical window to examine subtype-specific trajectories of anxiety symptoms.

Second, further research is needed to examine sex differences in the developmental courses of anxiety subtypes from childhood to early adolescence. While a few studies investigated how the course of overall anxiety or emotional symptoms vary between boys and girls from childhood to adolescence (e.g., LoParo et al., 2024; Tseliou et al., 2024), subtype-specific sex differences are less understood. Other studies assessed sex differences in the courses of specific anxiety subtypes, particularly social anxiety and obsessive-compulsive symptoms, but yielded conflicting findings (e.g., Jagannathan et al., 2024; Steinsbekk et al., 2022; Vivan et al., 2014; Xu et al., 2021). These discrepancies point to the need for targeted research to examine sex differences in the trajectories of core anxiety subtypes, especially in the transition from childhood to adolescence marked by significant sex differences in hormonal, physiological, and socioemotional changes. This knowledge can inform the development of sex-specific prevention and intervention programs, enhancing their timing and effectiveness by addressing distinct developmental patterns.

Third, drawing on an ecological systems approach to developmental psychopathology (Brofenbrenner, 1979), few studies have assessed modifiable caregiving-related risk factors in relation to trajectories of specific anxiety symptoms from childhood to adolescence. Developmental trajectories of anxiety problems from childhood to early adolescence have primarily been examined in relation to psychiatric and/or functional outcomes (see Liang et al., 2022; Marçal, 2020; Tseliou et al., 2024; Xu et al., 2021). Other studies have investigated systemic risk factors of childhood or adolescent anxiety symptoms, including early socioeconomic conditions (Ochi et al., 2014), in relation to trajectories or cross-sectional assessments of anxiety symptoms during childhood or adolescence. In contrast to systemic risk factors that are challenging to modify, greater empirical attention is needed to identify modifiable risk factors for childhood-to-adolescent anxiety, i.e., factors that are potentially within a parent’s capacity to alter or intervene on. A closer examination of risk factors such as parenting practices and maternal mental health would provide proximal and measurable targets for family-oriented prevention and intervention strategies that mitigate anxiety symptoms in childhood and adolescence.

Finally, previous research primarily examined developmental trajectories of anxiety subtypes in Western populations, including the United States, United Kingdom, and Norway (e.g., LoParo et al., 2024; Marçal, 2020; Morales-Muñoz et al., 2023; Tseliou et al., 2024; Steinsbekk et al., 2022). This highlights the need for relevant empirical work in other sociocultural contexts. Singapore, a multi-ethnic East Asian city-state in which youths’ clinical anxiety problems are highly prevalent (Wong et al., 2024), offers the opportunity to examine the transportability of results obtained in Western populations. Rising mental health problems, potentially driven by intense academic pressures and national ‘survivalist’ ideologies (Sim & Tham, 2025), highlight the need to examine subtype- and sex-specific developmental trajectories of anxiety symptoms within this sociocultural context. Culturally-informed frameworks of parenting (Bornstein, 2012) suggest that the home environment, particularly parenting practices, shapes children’s socioemotional adjustment in culturally-specific ways (e.g., Gao et al., 2021). These relationships should thus be understood within their sociocultural context. Singapore offers a unique sociocultural landscape: a majority ethnically Chinese population, Western influences due to rapid globalization, and the use of English as the lingua franca (Cheung & Lim, 2022). As a result, parenting practices in Singapore reflect a blend of interdependence-oriented and independence-oriented socialization goals, encouraging children to manage their emotions independently while maintaining familial interdependence. This multi-ethnic, East Asian sociocultural context provides a valuable setting for generating novel insights on the trajectories of anxiety subtypes and the relevance of caregiving factors.

Considering these gaps, the current study sought to address three key research aims. First, we sought to characterize the trajectories of subtype-specific anxiety symptoms from age 8.5 to 13 years in a multi-ethnic Asian cohort based in Singapore. Next, we sought to compare these trajectories in boys and girls. Finally, we sought to investigate whether early caregiving factors (i.e., at age 6 years), including parental sensitivity, parental autonomy support, and maternal mental health, were associated with changes in subtype-specific anxiety symptoms. Understanding sex-specific trajectories of holistic anxiety subtypes enables more precise identification of developmental windows for heightened risk, as well as identification of children prone to specific symptoms at different ages. Furthermore, by identifying proximal vulnerabilities and resilience factors related to the caregiving environment, our findings enable the development of targeted prevention and intervention strategies that reduce long-term emotional risks beyond childhood.

This study was embedded within the Growing Up in Singapore Towards healthy Outcomes (GUSTO; Soh et al., 2014) pre-birth cohort study, a population-based longitudinal cohort study in Singapore. Between June 2009 and December 2010, pregnant women in their first trimester (M_age = 30.6 years, range = 18–46 years) were recruited from two public hospitals. Eligibility criteria included being Singapore citizens or permanent residents of Chinese, Malay, or Indian ethnicity. In Singapore, the three major ethnic groups include Chinese, Malay, and Indian (Department of Statistics, Ministry of Trade and Industry, Singapore, 2024). Participants also needed to have homogenous parental ethnic background, intentions of delivering in either of the two public hospitals, and intentions to remain in Singapore for the next 5 years. The study was approved by the National Healthcare Group Domain Specific Review Board (D/2014/00414), the SingHealth Centralised Institutional Review Board (2018/2767/D), and the A*STAR Review Board (2020-011). Written informed consent was obtained from all mothers at recruitment, and all children provided written assent to the study when they turned 7 years of age from 2017 to 2018.

Embedded within the larger GUSTO cohort, a sub-cohort of mother-child dyads were invited to participate in neuropsychological assessments when children were ages 6 years, 8.5 years, 10 years, and 13 years. Data on parent-child relationships and maternal mental health was available at ages 6 years (n = 533 parent-child relationships, n = 551 maternal depression, n = 547 maternal anxiety, n = 539 maternal stress), while data on children’s anxiety symptoms were available at ages 8.5 years (n = 340), 10 years (n = 575), and 13 years (n = 676). As previously detailed by Cai et al. (2015), we excluded subjects who had a last recorded Apgar score of < 9, birth weight < 2500 g or > 4000 g, were born preterm (< 37 weeks), or were part of a twin pregnancy or pregnancy with complications. A total of 547 subjects who underwent at least one neuropsychological assessment at age 6 years and one follow-up assessment for anxiety symptoms met the eligibility criteria and were included in our analytic sample (see Figure S1 for a participant flowchart). Compared to the initial sample of 1247 mothers at recruitment, those excluded from the analytic sample only significantly differed in their ethnic proportions, χ²(3) = 12.28, p =.006, Cramér’s V = 0.09. There were lower proportions of Indian mother-child dyads and higher proportions of Malay mother-child dyads in the analytic sample compared to those excluded. Aside from ethnicity, there were no significant differences between the full and analytic samples on other demographic variables measured at recruitment, including child sex (p =.817), maternal education (p =.460), and household income (p =.164).

Within the analytic sample (n = 547), Little’s (1988) Missing Completely at Random (MCAR) test based on all age-invariant and age-varying variables revealed a normed chi-square (\(\:{\chi\:}^{2}/df)\) of 1.03 (p =.210). This indicated that the pattern of missing data was not materially different from a random pattern (Bollen, 1989). As missingness was not related to any covariates or variables of interest, all mixed effects models were analyzed using Mplus 8.11 with the MLR estimator, accounting for missing data using Full Information Maximum-Likelihood estimation (FIML). Null models without predictors assessed intraclass correlations (ICC) which indicated the proportion of variance in anxiety symptoms attributable to between-person variation. Mixed effects modelling was then conducted to investigate longitudinal trajectories of anxiety from 8.5 to 13 years. Mixed effects models allowed the nesting of multiple anxiety assessments within children, and accommodated unequal assessment intervals so that all measurement occasions could be retained. Child age in years was rescaled such that 0 represented 8.5 years to ease interpretation of the intercept. In unconditional growth models, child age was added as a random effect allowing both intercepts and slopes to vary across children. As an example, a random-intercept-and-time model is as follows for child i at age j: β_o represents the average anxiety level at 8.5 years, and v_oi refers to child i’s deviation from β_o. β₁ represents the average rate of change in anxiety symptoms, and v_1i refers to child i’s deviation from β₁. E_ij captures the residual term or unexplained variance in each growth model.

Within-subject models

Between-subject models

Thereafter, we incorporated the cross-level Child Sex*Age interaction term in each growth model to examine whether initial anxiety levels and rates of changes varied according to child sex, while retaining child sex as a fixed effect. Finally, to assess associations between early caregiving factors and anxiety trajectories from age 8.5 to 13, we included between-person predictors (i.e., early parental responsiveness, emotional availability, autonomy-support, and maternal distress indices) as fixed effects, age as a random effect, and adjusted for child sex, ethnicity, maternal education, paternal education, household income, and mothers’ marital status as age-invariant covariates. In ancillary analyses, we specified Child Sex*Predictor interaction terms to test whether predictors have differential associations with trajectories of anxiety subtypes according to child sex. A two-level model is as follows for child i at age j. β₂ represents the average association between an early predictor and baseline anxiety, and β₆ represents the average association between baseline anxiety and rate of change.

Within-subject models

Between-subject models

Table 1 outlines the demographic characteristics for our sample (see Table S2 in Supplementary Material for descriptive statistics of other key variables). Mean levels of overall anxiety symptoms were 63.57 (SD = 22.85) at 8.5 years, 59.73 (SD = 22.98) at 10 years, and 55.72 (SD = 27.05) at 13 years (out of a maximum score of 150). Compared to normative ratings from representative youths in the United States aged 8–11 and 12–15 (see March, 2013), our sample reported correspondingly higher scores for total and all subtype-specific anxiety symptoms with the exception of harm avoidance (see Table S3 in Supplementary Material for details).

In null models without predictors, intraclass correlations (ICCs) of each anxiety dimension ranged from 0.27 to 0.33, which indicated that substantial variation in anxiety symptoms was explained by between-person differences while the remaining variance was attributable to within-person change. This supported the appropriateness of mixed effects models which clustered anxiety scores at each timepoint within individual children. Following this, unconditional mixed effects models were specified for each dimension of anxiety, with initial anxiety levels and rates of change allowed to vary across children (i.e., age as a random effect). At age 8.5 years, initial anxiety scores were 11.04, 10.57, 10.15, 12.86, 15.65, and 11.55 for social anxiety, generalized anxiety, physical, obsessive-compulsive, harm avoidance, and separation anxiety symptoms, respectively (ps < 0.001). Social anxiety symptoms had an estimated yearly increase of \(\:b\) \(\:=\:\)0.24 (SE = 0.10, 95% CI [0.05, 0.43]), while physical, obsessive-compulsive, harm avoidance, and separation anxiety symptoms showed an estimated yearly decline of −.7\(\:b=\:\)6 to − 0.29. Only the random slope for generalized anxiety symptoms across ages was nonsignificant (\(\:b\) = − 0.09, SE = 0.08, 95% CI [−0.24, 0.07]). For each anxiety subtype, we found significant between-person variance in both initial symptom levels and slopes of longitudinal change. See Table 2 for details of between-person and within-person variance in intercepts and random slopes for each anxiety subtype. Thereafter, Child Sex*Age interaction terms were included and emerged significant in predicting all corresponding random slopes of anxiety symptoms across ages except for separation anxiety. However, initial levels of all anxiety subtypes did not differ by child sex. Figure 1 depicts the trajectories of anxiety symptoms in boys and girls, across dimensions of (a) social anxiety, (b) generalized anxiety, (c) physical, (d) obsessive-compulsive, (e) harm avoidance, and (f) separation anxiety symptoms. Social anxiety symptoms increased more quickly for girls than boys (\(\:b\)_sex*age= 0.79, SE = 0.19, 95% CI [0.42, 1.17]); obsessive-compulsive (\(\:b\)_sex*age = 0.41, SE = 0.19, 95% CI [0.04, 0.78]) and harm avoidance (\(\:b\)_sex*age = 0.28, SE = 0.14, 95% CI [0.01, 0.54]) declined more slowly for girls than boys. In addition, girls showed an increase in generalized anxiety (\(\:b\)_sex*age = 0.74, SE = 0.16, 95% CI [0.43, 1.04]) and physical symptoms (\(\:b\)_sex*age = 0.85, SE = 0.19, 95% CI [0.48, 1.23]) with age, while boys showed a decline in symptoms.

Children whose parents granted them greater autonomy-support showed a lower rate of increase in social anxiety symptoms, with a relatively small effect size (\(\:b\) = − 0.29, SE = 0.11, 95% CI [−0.49, − 0.08]) (Ferguson, 2009). Other predictors associated with age-related change in anxiety subtypes included maternal symptoms of anxiety and stress. Specifically, early maternal anxiety was associated with small increases in generalized anxiety symptoms (\(\:b\) = 0.03, SE = 0.01, 95% CI [0.01, 0.06]), physical symptoms\(\:\:(b\:=\) 0.03, SE = 0.01, 95% CI [0.01, 0.05]), and obsessive-compulsive symptoms \(\:(b\) = 0.04, SE = 0.01, 95% CI [0.01, 0.07]). Similarly, maternal stress was associated with small increases in physical symptoms (\(\:b\) = 0.02, SE = 0.01, 95% CI [0.01, 0.03]) across ages (see Tables S4 and S5 in Supplementary Material for full unstandardized estimates and full standardized estimates, respectively). Additional sensitivity analyses were also conducted adjusting for depressive symptoms as a time-varying covariate (see Supplement A and Table S6 in Supplementary Material for details). In ancillary analyses, interaction effects between child sex and caregiving factors largely did not predict changes in anxiety symptoms (see Table S7 in Supplementary Material for details).

The current study provides novel insights into the developmental courses of subtype-specific anxiety symptoms during a critical transitional period, while considering sex differences and early caregiving-related factors. Additional strengths include the multi-ethnic, East Asian cohort comprising both girls and boys, our longitudinal design leveraging repeated measures of anxiety symptoms, a cross-informant approach that accounted for early factors of parenting practices and maternal mental health, and the application of mixed effects modeling to better understand sex differences in the developmental course of anxiety symptoms. Nonetheless, our findings should be considered in light of key limitations. First, our longitudinal model encompasses a relatively small age span (i.e., 5 years) across three measurement waves. Hence, we note that differences in trajectories of anxiety symptoms, including sex-specific trajectories, were driven by levels of anxiety symptoms at age 13. This precluded investigating how specific forms of anxiety symptoms may progress over a longer window (e.g., early childhood to late adolescence). Future studies would benefit from including more measurement waves, as well as longer follow-up intervals, to elucidate longer-term anxiety symptom trajectories. Second, our sample predominantly comprised Chinese, Malay, and Indian mother-child dyads from nuclear families with married parents. As this could limit the generalizability of our findings to other populations, future research is needed to explore developmental trajectories of anxiety subtypes and their associations with early caregiving factors in other ethnic groups, countries, high-risk or clinical samples to verify the applicability of our findings. Finally, the validity of the BPI coding system may be limited by potential subjectivity in rater’s coding and the forced-choice format. The coding of responses may be biased by aspects of children’s personality and temperament (e.g., extroversion) that influence their perceptions of parental behavior as well as their responses to BPI items. Additionally, the BPI’s forced-choice format, which presents children with two opposing statements, may restrict the measurement of children’s perceptions by oversimplifying parent-child dynamics. It would be beneficial to incorporate additional data sources, such as parent-reports or behavioral observations of parent-child interactions, when examining parenting practices in relation to children’s socioemotional outcomes.