Introduction
It is well known that anxiety disorders run in families (Lawrence, Murayama, & Creswell, Reference Lawrence, Murayama and Creswell2019; Merikangas, Avenevoli, Dierker, & Grillon, Reference Merikangas, Avenevoli, Dierker and Grillon1999). For childhood and adolescent anxiety, this aggregation reflects parent-to-child and child-to-parent effects (Ahmadzadeh et al., Reference Ahmadzadeh, Eley, Leve, Shaw, Natsuaki, Reiss and McAdams2019; Knappe et al., Reference Knappe, Lieb, Beesdo, Fehm, Ping Low, Gloster and Wittchen2009), as well as shared genetic liability (Boomsma, Van Beijsterveldt, & Hudziak, Reference Boomsma, Van Beijsterveldt and Hudziak2005). But the familial patterning of one particular anxiety disorder – separation anxiety – is less well characterized. This is because separation anxiety has not historically been recognized as a problem that occurs in adulthood, instead conceptualized and classified until recently as something that children experience in relation to their parents or primary caregivers (Bowlby, Reference Bowlby1973). Consequently, studies on separation anxiety in parents, or adult separation anxiety (ASA), and in turn how separation anxiety clusters in families, are limited compared to other anxiety disorders.
The surprisingly high prevalence and clinical significance of ASA suggest that understanding the familial patterning of separation anxiety is important. Lifetime rates in the community are 7% (Shear, Jin, Ruscio, Walters, & Kessler, Reference Shear, Jin, Ruscio, Walters and Kessler2006) and rates in clinical samples are substantially higher (Pini et al., Reference Pini, Abelli, Shear, Cardini, Lari, Gesi and Cassano2010; Silove, Marnane, Wagner, Manicavasagar, & Rees, Reference Silove, Marnane, Wagner, Manicavasagar and Rees2010b). It predicts poorer treatment outcomes (Aaronson et al., Reference Aaronson, Shear, Goetz, Allen, Barlow, White and Gorman2008; Kirsten, Grenyer, Wagner, & Manicavasagar, Reference Kirsten, Grenyer, Wagner and Manicavasagar2008; Miniati et al., Reference Miniati, Calugi, Rucci, Shear, Benvenuti, Santoro and Cassano2012) and is associated with functional impairment, even beyond the effects of comorbid disorders (Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006), and more so than other anxiety and mood disorders (Pini et al., Reference Pini, Abelli, Shear, Cardini, Lari, Gesi and Cassano2010).
The only study on separation anxiety in families to date suggests that the association between ASA and children's separation anxiety (CSA) is strong and highly specific. However, this study used a small sample of 54 parent–child dyads (Manicavasagar, Silove, Rapee, & Waters, Reference Manicavasagar, Silove, Rapee and Waters2001). We expand on this past work by examining the associations between parents' and CSA and other internalizing psychopathology using a larger community sample and multiple assessment points. The timing of our assessments captures middle childhood and early adolescence, a period of high CSA (Copeland, Angold, Shanahan, & Costello, Reference Copeland, Angold, Shanahan and Costello2014), which also covers the median age of anxiety disorder onset (11 years; Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005) and the beginning of the rise in rates of other internalizing disorders (Avenevoli, Knight, Kessler, & Merikangas, Reference Avenevoli, Knight, Kessler, Merikangas, Abela and Hankin2008). We also richly represent children's psychopathology as latent variables that capture what is in common across multiple reporters (mothers, fathers, and children) and multiple methods (symptom reports and a semi-structured diagnostic interview).
Given that ASA, like all psychiatric problems, is highly comorbid with other disorders (Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006), we also test the specificity of ASA-CSA associations by adjusting for other forms of parental psychopathology, in contrast to the past study on familial transmission of separation anxiety, which only tested bivariate relationships (Manicavasagar et al., Reference Manicavasagar, Silove, Rapee and Waters2001). We also examine associations with multiple of children's internalizing problems simultaneously in order to determine whether ASA is associated with children's separation anxiety in particular or internalizing problems more generally. This is important because previous research suggests that disorders like depression and anxiety are better construed as specific manifestations of a higher-order internalizing factor and that familial transmission may occur at this level rather than at the level of specific disorders (Lahey, Van Hulle, Singh, Waldman, & Rathouz, Reference Lahey, Van Hulle, Singh, Waldman and Rathouz2011; Starr, Conway, & Brennan, Reference Starr, Conway and Brennan2014).
Moreover, because ASA is a newly-recognized clinical problem, it has not been well-validated as a clinical entity. According to Robins & Guze's seminal work (Reference Robins and Guze1970), determining whether psychiatric disorders run in families is one of the five phases for establishing the validity of clinical problems. Family studies of psychiatric problems contribute to this aim by revealing meaningful patterns of associations across generations, regardless of the mechanisms of transmission. Thus, linking parents' ASA to the better-established construct of CSA would increase the diagnostic validity of ASA.
In sum, we conducted a family study on the intergenerational patterning of separation anxiety in a predominantly white community sample from Long Island, NY in order to understand how ASA relates to CSA in particular and internalizing psychiatric problems more generally, and also to validate ASA.
Methods
Participants
The study sample is from an ongoing longitudinal study of children's temperament and psychopathology; it includes 515 adult women and 510 adult men who completed a self-report measure of separation anxiety symptoms or a diagnostic interview, and 515 children who completed a diagnostic interview at one or two occasions. Two adult men had total ASA-27 scores which were over 8 s.d. above the mean (i.e. 79 and 81; range for the remainder of sample = 0–59), so they were excluded from analyses. Families with a 3-year-old child living within 20 miles of Stony Brook, New York, were eligible to participate in the larger study. Children with severe and disabling physical or developmental disorders were excluded. Participants were recruited from the community using commercial mailing lists (see Bufferd, Dougherty, Carlson, Rose, and Klein, Reference Bufferd, Dougherty, Carlson, Rose and Klein2012 for details). According to census data, the sample is demographically similar to the surrounding community. Parents provided written informed consent after receiving a description of the study. The study was approved by the human subjects review committee at Stony Brook University, and families were compensated. Families returned for assessments every 3 years. In the current study, we report on the wave 3 and wave 4 assessments.
Measures
Adult separation anxiety
The Adult Separation Anxiety Questionnaire (ASA-27) is a 27-item self-report measure of separation anxiety symptoms experienced as an adult (over age 18) (Manicavasagar, Silove, Wagner, & Drobny, Reference Manicavasagar, Silove, Wagner and Drobny2003). This measure was administered to parents at wave 3. Items are rated on a four-point scale (0 = This has never happened; 3 = This happens very often). The measure includes the same items as the Adult Separation Anxiety Semi-Structured Interview (ASA-SI) from the same group (Manicavasagar et al., Reference Manicavasagar, Silove, Wagner and Drobny2003; Manicavasagar, Silove, & Curtis, Reference Manicavasagar, Silove and Curtis1997). Items come from clinical impressions, attachment theory, open-ended interviews, and versions of the DSM-IV separation anxiety criteria for youth. Case scores of ⩾16 or ⩾22 are recommended by the scale developers, the former for capturing all cases in the community and the latter being the intersection point between sensitivity and specificity. Using the latter cutoff and diagnoses from the ASA-SI as the criterion, sensitivity = 81%; specificity = 84%; misclassification rate = 17%; positive predictive power = 76%; and negative predictive power = 88% (Manicavasagar et al., Reference Manicavasagar, Silove, Wagner and Drobny2003). The ASA-27 also had high internal consistency (Cronbach's alpha = 0.95) and good test-retest reliability (r = 0.86, p < 0.001). Convergent validity of the ASA-27 is supported by an independent study that found that ASA-27 scores were strongly related to scores on the Structured Clinical Interview for Separation Anxiety Symptoms (SCI-SAS; r = 0.84), an interview that closely resembles questions on the structured interview used in the National Comorbidity Study-Replication (Cyranowski et al., Reference Cyranowski, Shear, Rucci, Fagiolini, Frank, Grochocinski and Cassano2002; Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006). The measure has shown partial strict measurement invariance across gender (Finsaas, Olino, Hawes, Mackin, & Klein, Reference Finsaas, Olino, Hawes, Mackin and Klein2020). See Table 1 for Cronbach's α and online Supplemental Table S1 for item descriptions and correspondence to DSM criteria.
Table 1. Reliability statistics for all study measures
a Excludes separation anxiety.
Notes: SCID, Structured Clinical Interview for DSM-IV, Non-Patient Version. ASA-27, Adult Separation Anxiety Symptom Questionnaire. K-SADS, The Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version. SCARED, Screen for Child Anxiety Related Disorders. CDI, Children's Depression Inventory.
Parent psychopathology
Adult participants were interviewed at waves 1 and 3 using the Structured Clinical Interview for DSM-IV, Non-Patient Version (SCID-NP; First, Gibbon, Spitzer, Williams, & Benjamin, Reference First, Gibbon, Spitzer, Williams and Benjamin1997; Endicott & Spitzer, Reference Endicott and Spitzer1978). The SCID is considered the gold standard diagnostic interview for psychopathology in adults; for example, in the National Comorbidity Survey Replication Study, the validity of a different diagnostic interview was determined by comparison to the SCID-NP (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005). Masters-level clinicians and advanced clinical psychology graduate students conducted the interviews by telephone, which yield similar results as face-to-face interviews (Rohde, Lewinsohn, & Seeley, Reference Rohde, Lewinsohn and Seeley1997). The wave 1 interview assessed lifetime diagnoses and the wave 3 interview the interval between the two assessments. Results from both interviews were combined to create lifetime diagnosis variables. When participants were unavailable, family history interviews were conducted with co-parents. To assess inter-rater reliability, second-raters independently rated audiotaped interviews. See Table 1 for kappas. Depressive disorder included major depressive disorder and dysthymia; anxiety disorder included panic disorder, specific phobia, social phobia, agoraphobia without panic disorder, GAD, obsessive compulsive disorder, and post-traumatic stress disorder; and substance use disorder included alcohol, cannabis, and hard drug abuse or dependence.
Children's psychiatric diagnoses
The Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version (K-SADS-PL; Kaufman et al., Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci and Ryan1997) was used to assess DSM-IV diagnoses in the children at the waves 3 and 4, when they were approximately 9 and 12 years old, respectively. The K-SADS-PL is widely used as the gold standard against which other diagnostic measures of psychopathology in youth are evaluated (e.g. Dougherty, Klein, and Olino, Reference Dougherty, Klein, Olino, Hunsley and Mash2018; Kessler et al., Reference Kessler, Avenevoli, Green, Gruber, Guyer, He and Merikangas2009) and has been shown to generate valid diagnoses (Dougherty et al., Reference Dougherty, Klein, Olino, Hunsley and Mash2018; Kaufman et al., Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci and Ryan1997). Interviews were conducted separately with parents and children by advanced clinical psychology graduate students and a masters-level clinician and supervised by an experienced child psychiatrist and clinical psychologist. Further information was obtained to reconcile discrepancies as needed, and the interviewer made final ratings based on the combination of reports. To assess inter-rater reliability, second-raters independently rated videotapes. See Table 1 for kappas. Depressive disorders included major depressive disorder, dysthymic disorder, depressive disorder NOS; other anxiety disorders included specific phobia, social phobia, generalized anxiety disorders, and panic/agoraphobia; and separation anxiety disorder was considered on its own. For the current project, the variables at wave 3 cover the child's lifetime until that point (age 9) and the variables at wave 4 reflect whether a diagnosis was present at any point during the 3-year interval since wave 3.
Children's anxiety symptoms
At waves 3 and 4, children and their parents completed the 41-item youth self-report and parent-report versions, respectively, of the Screen for Childhood Anxiety Related Disorders (SCARED; Birmaher et al. Reference Birmaher, Khetarpal, Brent, Cully, Balach, Kaufman and Neer1997, Reference Birmaher, Brent, Chiappetta, Bridge, Monga and Baugher1999). Children and their parents rated the presence of anxiety symptoms in the child over the past 3 months on a three-point scale. The SCARED is made up of five factor-analytically derived subscales: panic/somatic, general anxiety, separation anxiety, social phobia, and school phobia. These subscales reflect anxiety disorder symptoms as conceptualized in the DSM-IV. To create the indicators for the other anxiety factors, all subscales except for the separation anxiety scale were summed. This measure is widely used and its validity has been supported by numerous studies (Monga et al., Reference Monga, Birmaher, Chiappetta, Brent, Kaufman, Bridge and Cully2000; Muris & Steerneman, Reference Muris and Steerneman2001; Muris, Merckelbach, Van Brakel, & Mayer, Reference Muris, Merckelbach, Van Brakel and Mayer1999; Rappaport, Pagliaccio, Pine, Klein, & Jarcho, Reference Rappaport, Pagliaccio, Pine, Klein and Jarcho2017). See Table 1 for Cronbach's α.
Children's depressive symptoms
At waves 3 and 4, children completed the 27-item youth self-report and parents completed the 17-item parent-report versions of the Children's Depression Inventory (CDI; Kovacs, Reference Kovacs1992). Children and their parents are asked to rate the presence of depressive symptoms in the youth in the past 2 weeks on a three-point scale. The CDI is widely used and has been shown to have concurrent, discriminant, and predictive validity (Dougherty et al., Reference Dougherty, Klein, Olino, Hunsley and Mash2018; Timbremont, Braet, & Dreessen, Reference Timbremont, Braet and Dreessen2004). See Table 1 for Cronbach's α.
Data analysis
We computed descriptive statistics for demographic and study variables and conducted missingness analyses. To further characterize the sample, we estimated bivariate relationships between parent ASA scores at wave 3 and children's diagnoses at waves 3 and 4 using point-biserial correlations. Next, to determine aggregation of separation anxiety in families, while also testing the specificity of associations with children's psychopathology and addressing the issue of comorbidity in parents, we estimated models for mothers and fathers testing concurrent parent–child psychopathology relationships (parent and child variables measured at wave 3) and prospective parent–child psychopathology relationships (parent variables measured at wave 3 and child variables at wave 4). Children's latent separation anxiety, other anxiety, and depression factors were indicated using the respective K-SADS diagnoses and mother, father, and child report of anxiety and depressive symptoms. K-SADS diagnoses were specified as categorical indicators. The residual variances of indicators by the same reporter were permitted to covary, as were the K-SADS indicators. Next, parents' separation anxiety, measured by the ASA-27, and depressive, other anxiety, and substance use disorders, assessed by the SCID, were added as predictors. Models also included child ethnicity, gender, and age, and parent education, age, and marital relationship as covariates (child race was not included because empty cells with diagnosis variables caused problems with model convergence). Although cut-offs are somewhat arbitrary (Marsh, Hau, & Wen, Reference Marsh, Hau and Wen2004), current conventions suggest that adequate fit is indicated by CFI and TFI > 0.90, and RMSEA between 0.05 and 0.08 (Hu & Bentler, Reference Hu and Bentler1999; MacCallum, Browne, & Cai, Reference MacCallum, Browne and Cai2006).
Descriptive statistics and correlations were computed in R Studio (version 1.2.1335; R Core Team, 2016) and CFA models were estimated in Mplus 8 (version 1.6; Muthen & Muthen, 2012–2018) using the robust weighted least squares estimator (WLSMV; Flora and Curran, Reference Flora and Curran2004), which is suitable for categorical data, and maximum likelihood estimation on both the predictor and the outcome variables in order to utilize all available data (Enders, Reference Enders, Hancock and Mueller2013).
Results
At wave 3, children were 9.3 years old on average (s.d. = 0.42, range = 8.4–11.0); at wave 4, they were 12.7 years old on average (s.d. = 0.46, range = 11.5–14.2). About half were female (n = 224, 46.3%). Based on reports from parents about their children's race, 457 children were white (88.7%), 42 were Black (8.2%), 14 were Asian (2.7%), and 1 was Native American (0.0%); 1 parent (0.0%) selected the ‘other’ category. Regarding ethnicity, 64 children were Hispanic (12.4%) according to the parent reports. At wave 3, mothers were 41.81 years old on average (s.d. = 4.83, range = 25.9–53.5) and fathers were 44.0 years old on average (s.d. = 5.85, range = 27.4–61.1). The majority of mothers (n = 413, 80.2%) and fathers (n = 411, 80.6%) were married to or living with the child's biological parent. About half of the mothers (n = 276, 53.6%) and fathers (n = 216, 42.4%) had graduated from college with a 4-year degree. Data on demographic variables were self- and parent-reported on surveys (see online Supplemental Table S2 for description and coding of variables). Age was calculated using birthdate and date of assessment.
Mothers, fathers, and children included in the study sample were compared on all demographic variables to those who participated in the larger study but did not complete the aforementioned measures and/or interviews. Mothers and children did not differ. Fathers in the sample were more likely to be married and/or living with the child's biological mother. See online Supplemental Table S3 for missingness rates and statistical comparisons and Table 2 for complete rates on study variables.
Table 2. Descriptive statistics for children's anxiety and depressive symptoms from mother, father, and child report
Notes: For mothers, n = 515. For fathers, n = 510. For children, n = 515. SCARED, Screen for Child Anxiety Related Disorders; CDI, Children's Depression Inventory.
Descriptive statistics for children's symptoms are in Table 2 descriptive statistics for children's diagnoses are in Table 3. The average ASA score for mothers was 10.26 (s.d. = 9.37, range = 0–58) and the average for fathers was 8.10 (s.d. = 8.46, range = 0–54). Thirty-eight percent of mothers (193/514) and 19% of fathers (97/508) were diagnosed with a depressive disorder in their lifetimes, 38% of mothers (195/514) and 21% of fathers (109/506) were diagnosed with an anxiety disorder, and 22% of mothers (114/514) and 44% of fathers (224/508) were diagnosed with a substance use disorder. Generally, rates of psychiatric disorders are higher when based on cumulative prevalence across multiple assessments (Olino et al., Reference Olino, Shankman, Klein, Seeley, Pettit, Farmer and Lewinsohn2012).
Table 3. Bivariate correlations between parents' separation anxiety and children's psychiatric disorders and univariate descriptive statistics for children's psychiatric disorders
Notes: ***p < 0.001; **p < 0.01; *p < 0.05; ^p < 0.10. Age 9 = wave 3 assessment, which covered child's lifetime through age 9. Ages 9-12 = wave 5 assessment, which was done when children were 12 years old and covered the period since the age 9 assessment. Parents' adult separation anxiety assessed at wave 3. Correlations with agoraphobia and panic disorder are not reported here because of low counts (<1%) but these variables are included in structural models.
In bivariate analyses, maternal ASA was positively related to children's lifetime separation anxiety assessed at wave 3, and both maternal and paternal ASA were positively related to children's 3-year interval separation anxiety assessed at wave 4 (Table 3). Maternal ASA was also related to GAD at both waves and to lifetime depression at wave 3.
All four structural models fit reasonably well: CFI and RMSEA exceeded cutoffs for a good fit for all four models, and TFI exceeded the cutoff for 2 of the 4 (Table 4). All indicators loaded on respective factors at p < 0.001 (Figs 1 and 2). In all models, the children's latent psychopathology factors were positively intercorrelated (separation anxiety/other anxiety: β = 0.61–0.68; separation anxiety/depression: β = 0.29–0.40; other anxiety/depression: β = 0.60–0.65).
Table 4. Model fit statistics
Notes: CFI, comparative fit index; TLI, Tucker-Lewis Index; RMSEA, root mean square error of approximation.
Fig. 1. Concurrent and prospective relationships between maternal psychopathology and children's separation anxiety, other anxiety, and depression factors.
Notes: Solid lines indicate significant paths at p < 0.05. ***p < 0.001; **p < 0.01; *p < 0.05; ^p < 0.10. Dx, diagnosis; M, mother report; F, father report; C, child report; ASA-27, Adult Separation Anxiety Symptom Questionnaire; K-SADS, The Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version; SCARED, Screen for Child Anxiety Related Disorders; CDI, Children's Depression Inventory. Maternal and child diagnoses at age 9 are from lifetime assessments. Child diagnoses at age 12 cover the interval since the last assessment. Covariances between independent variables are not shown, nor are covariances between children's psychopathology factors, which were all significant at p < 0.001. The residual variances of indicators by the same reporter were permitted to covary, as were the K-SADS indicators.
Fig. 2. Concurrent and prospective relationships between paternal psychopathology and children's separation anxiety, other anxiety, and depression factors.
Notes: Solid lines indicate significant paths at p < 0.05. ***p < 0.001; **p < 0.01; *p < 0.05; ^p < 0.10. Dx, diagnosis; M, mother report; F, father report; C, child report; ASA-27, Adult Separation Anxiety Symptom Questionnaire; K-SADS, The Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version; SCARED, Screen for Child Anxiety Related Disorders; CDI, Children's Depression Inventory. Paternal and child diagnoses at age 9 are from lifetime assessments. Child diagnoses at age 12 cover the interval since the last assessment. Covariances between independent variables are not shown, nor are covariances between children's psychopathology factors, which were all significant at p < 0.001. The residual variances of indicators by the same reporter were permitted to covary, as were the K-SADS indicators.
Maternal ASA was non-specifically related to children's internalizing psychopathology, beyond other maternal disorders, concurrently (β = 0.19–0.28) and prospectively (β = 0.22–0.26; Fig. 1). Maternal depression was likewise concurrently uniquely related to all three child factors (β = 0.13–0.21) and prospectively to CSA (β = 0.14) and children's depression (β = 0.18). Maternal anxiety was not related to any child factors above and beyond other disorders concurrently but it was related to CSA (β = 0.13) and children's other anxiety (β = 0.15) prospectively. Maternal substance use was not uniquely related to child internalizing psychopathology in either model. Like maternal ASA, paternal ASA was significantly related to the child psychopathology factors in the concurrent (β = 0.26–0.34) and prospective models (β = 0.14–0.27), beyond other paternal disorders (Fig. 2). In contrast to the maternal findings, however, none of the other paternal psychopathology variables was significantly associated with any of the child psychopathology factors.
Discussion
Recent evidence suggests that separation anxiety is an impairing clinical problem that can occur during adulthood (Bögels, Knappe, & Clark, Reference Bögels, Knappe and Clark2013). We examined the familial transmission of separation anxiety using a large community sample. Overall, our results highlight that maternal and paternal ASA is associated non-specifically with children's internalizing psychopathology, including CSA, in middle childhood and early adolescence. We first showed how ASA was linked to children's disorders in bivariate descriptive analyses. Maternal ASA was somewhat non-specifically related to children's internalizing problems, whereas paternal ASA was not related to any of children's psychopathology at age 9 and only to CSA between ages 9 and 12. The non-specificity of the associations for maternal ASA contradicts a study using a small clinical sample of children and their parents which found a high degree of specificity for ASA (Manicavasagar et al., Reference Manicavasagar, Silove, Rapee and Waters2001). The studies differ in terms of samples, but given the high rate of comorbid disorders in children in the clinical sample (54%), one would expect more rather than fewer associations between ASA and children's other disorders. The differences may be due to the fact that their study combined mothers and fathers; if our samples were independent (i.e. mothers and fathers were not parents of the same children), combining mothers and fathers would likely weaken the multiple relationships observed with maternal ASA. Our study also had more power to detect associations based on larger sample size and a continuous measure of ASA.
We then explored whether ASA associations are specific to CSA or generalized to other forms of children's internalizing disorders in structural equation models, and whether the associations between ASA and CSA are due to comorbid disorders in the parents. In addition to testing multiple relationships simultaneously, these models had the advantage over the bivariate tests of greater precision as children's psychopathology was latent variables. These models are agnostic about the mechanism of transmission by which these associations arise and likely encompass environmental bidirectional parent-to-child and child-to-parent effects, as well as shared genetic liability. Concurrent models characterize familial clustering of separation anxiety and internalizing problems for youth in middle childhood when parent and child psychopathology are measured at the same time. The prospective models indicate the level of psychopathology we would expect to observe in 12-year-old children given the level of psychopathology in parents when children were 9 years old. Comparing associations between the two models facilitate inferences about the degree to which the patterning of relationships is stable across time and developmental periods.
Concurrent and prospective models in fathers and mothers show that ASA relates in a generalized, rather than specific, fashion to internalizing problems in youth. The associations are present when adjusting for the effects of maternal depressive, other anxiety, and substance disorders, suggesting they are not due to overlap with other forms of parental psychopathology; this highlights the significance of ASA in youth psychopathology, despite the generalized pattern of associations. Notably, the strength of the associations between ASA and children's internalizing were similar in the concurrent and prospective models for mothers, suggesting they are stable from middle childhood to early adolescence, but somewhat attenuated in the prospective model for fathers.
Characterizing the associations with paternal ASA is in line with a call in a recent systematic review for continued research on ‘paternal mental health disorders, particularly anxiety disorders, and adolescent anxiety and depression’ (p. 244; Wickersham, Leightley, Archer, and Fear, Reference Wickersham, Leightley, Archer and Fear2020). The magnitudes of all ASA associations were larger for fathers compared to mothers in the concurrent models and for ASA-CSA in the prospective models. Paternal ASA was also the only paternal disorder with significant relationships with any child psychopathology factor. Given that women report higher levels of ASA compared to men (Aaronson et al., Reference Aaronson, Shear, Goetz, Allen, Barlow, White and Gorman2008; Silove et al., Reference Silove, Marnane, Wagner, Manicavasagar and Rees2010b) and have higher rates of ASA disorder (Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006), it is possible that ASA in men reflects a particularly strong liability, which could place offspring at higher risk for poor outcomes. That is, the threshold for manifesting or disclosing symptoms may be higher for men compared to women due to genetic or environmental (e.g. reporter bias, stigma) factors, such that it requires more severe symptomatology for men to experience or endorse symptoms. However, in our sample, the ASA-27 demonstrated partial strict invariance across genders (Finsaas et al., Reference Finsaas, Olino, Hawes, Mackin and Klein2020); this indicates that item associations with the underlying ASA trait were mostly equivalent for men and women, so reporting differences cannot completely explain the differences in associations. The stronger associations observed for men's ASA with children's psychopathology should be replicated and explored further in future work.
Parents are a significant part of their children's environments, and their ASA symptoms may increase children's risk for separation anxiety via modeling. Parents who become outwardly distressed upon actual or anticipated separation from their children may inadvertently teach children that they are incapable of facing challenges on their own and model temporary separations as distressing and threatening events. Indeed, Bögels and Zigterman (Reference Bögels and Zigterman2000) propose that overestimating the danger of being left and underestimating independent functioning may be core cognitive dysfunctions in CSA. Alternatively, ASA may increase the risk for other internalizing disorders like depression via its links with self-criticism (Hock & Lutz, Reference Hock and Lutz1998), low self-esteem (McBride & Belsky, Reference McBride and Belsky1988), and negative self-representations (Hock & Schirtzinger, Reference Hock and Schirtzinger1992).
Parents may also engage in certain parenting behaviors intended to avoid separation. These parenting behaviors may in turn limit children's opportunities for social interactions and thereby encroach on their abilities to develop a sense of themselves as connected to, but ultimately independent, from their parents. This notion has been discussed in detail in the literature on separation anxiety specifically within the context of the parent–child relationship (e.g. Bartle-Haring, Brucker, & Hock, Reference Bartle-Haring, Brucker and Hock2002; Hock & Lutz, Reference Hock and Lutz1998). Parents' ASA may also confer risk for psychiatric outcomes via its consequence in parents. For example, ASA is associated with poor interpersonal functioning, alienation, and aggression (Finsaas & Klein, Reference Finsaas and Klein2021; Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006); these parents may be more likely to have children with poor interpersonal skills and increased risk for internalizing problems broadly.
Finally, given previous work showing child-to-mother (but not father) effects of anxiety and internalizing symptoms in children and adolescents (Ahmadzadeh et al., Reference Ahmadzadeh, Eley, Leve, Shaw, Natsuaki, Reiss and McAdams2019; Schulz, Nelemans, Oldehinkel, Meeus, & Branje, Reference Schulz, Nelemans, Oldehinkel, Meeus and Branje2021; Yirmiya, Motsan, Kanat-Maymon, & Feldman, Reference Yirmiya, Motsan, Kanat-Maymon and Feldman2021), it is likely that associations with maternal psychopathology in our study are not solely due to transmission from parent to child. Children with separation anxiety, or other psychiatric, developmental or physical problems, may require more support and attention, which could lead to increases in mothers' ASA. Alternatively, an external event, like the death of a family member, may lead to disorder onset in parents and children alike.
Finally, the observed associations may also be due to shared genetic liability. Anxiety disorders and related behaviors show moderate heritability across childhood (Boomsma et al., Reference Boomsma, Van Beijsterveldt and Hudziak2005; Trzaskowski, Zavos, Haworth, Plomin, & Eley, Reference Trzaskowski, Zavos, Haworth, Plomin and Eley2012). Twin studies of child separation anxiety disorder show that genetic effects explain most individual differences (43% genetic; Scaini, Ogliari, Eley, Zavos, and Battaglia, Reference Scaini, Ogliari, Eley, Zavos and Battaglia2012), but genetically informative studies of anxiety more broadly also provide evidence of environmental effects (Ahmadzadeh et al., Reference Ahmadzadeh, Eley, Leve, Shaw, Natsuaki, Reiss and McAdams2019; Eley et al., Reference Eley, McAdams, Rijsdijk, Lichtenstein, Narusyte, Reiss and Neiderhiser2015); no studies have used the latter study design for separation anxiety in particular.
Our findings also partially support the validity of ASA as a clinical entity. ASA was associated with offspring internalizing problems, including CSA (Robins & Guze, Reference Robins and Guze1970), but parental ASA was also related to other offspring disorders, providing less support for its distinctness. However, this non-specific patterning of associations was also observed for the much better-established disorder of maternal depression, and thus it may reflect in both cases the limitations of representing psychiatric problems as discrete entities v. as specific and sometimes overlapping manifestations of underlying liabilities, a perspective that is gaining in popularity (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby and Zimmerman2017), and in line with work showing transmission at the level of internalizing liability (Lahey et al., Reference Lahey, Van Hulle, Singh, Waldman and Rathouz2011). Nonetheless, the fact that the associations between ASA and children's internalizing psychopathology were present over and above other better-established forms of adult psychopathology underscores ASA's clinical significance; the diagnosis is capturing something not already picked up by other diagnostic categories that relates to children's psychopathology.
The study findings should also be viewed in light of limitations. Although there is evidence from retrospective reports from adults that CSA precedes ASA (Pini et al., Reference Pini, Abelli, Shear, Cardini, Lari, Gesi and Cassano2010; Shear et al., Reference Shear, Jin, Ruscio, Walters and Kessler2006; Silove et al., Reference Silove, Marnane, Wagner, Manicavasagar and Rees2010b), we do not know whether ASA in our sample was a continuation of CSA or adult-onset. We also cannot distinguish parents' fear of separation from their children from fear of separating from other attachment figures. Since it is a community sample, we also have relatively low rates of disorders in youth. However, this is tempered by using measures of symptoms alongside diagnoses in structural models. In addition, the reliability of a few of the scales, particularly father's reports, were in the poor or questionable range, but modeling children's outcomes as latent factors using multiple sources of information mitigates the impact of suboptimal measurement on associations. The sample is also mostly white, with only a small proportion of Hispanic participants. Although this reflects the demographic makeup of the surrounding community, it limits the generalizability of the results. Moreover, participating fathers were more likely to be married/living with their child's parent than non-participating fathers, so whether results generalize to fathers who are unmarried or not living with their child's other parent is not known. Also, there may be unknown biases in samples derived from commercial mailing lists. In addition, while we utilized latent versions of children's psychopathology, we used a single observed indicator (diagnoses) for parents since we did not have multiple measures of each diagnosis in parents; this may have decreased precision. ASA and other parent psychopathology were also assessed using different types of instruments and covering different assessment periods, which may have impacted predictive power. Subthreshold levels of ASA are captured by the ASA-27 self-report questionnaire, whereas only clinically significant cases of the other forms of psychopathology were captured by the SCID; the higher dimensionality of the ASA variable may have given it more predictive power. At the same time, the SCID, a gold standard tool for assessing psychopathology administered by a trained interviewer, likely produced cleaner representations of psychopathology compared to the self-report ASA-27 instrument, and thus the other parent psychopathology variables may have been more predictive in this way.
In summary, ASA in mothers and fathers is non-specifically related to children's internalizing problems, including CSA in middle childhood and adolescence, beyond the effects of other common parental mental disorders. Future research should delineate the mechanisms that underlie this intergenerational transmission, particularly parent and child characteristics and parenting behaviors that modify risk.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721005249
Financial support
This research was supported by NIMH Grant: R01MH069942 (DNK). MCF was supported by NSF Graduate Research Fellowship: 2015201335 and NIMH Grant: T32MH013043.
Open access
