Pediatric-Onset and Adult-Onset Separation Anxiety Disorder Across Countries in the World Mental Health Survey

Although separation anxiety disorder traditionally has been a diagnosis assigned to children and adolescents, DSM-5 removed the 18-year age-at-onset restriction on diagnosis because studies had found later onset to be common. It is timely therefore to examine key aspects of the epidemiology of separation anxiety disorder as a condition with onsets that span the life course.

The importance of adult-onset separation anxiety disorder is indicated by the fact that 20%−40% of adult patients with mood and anxiety disorders have been found to have symptoms of the disorder, and between one-third and one-half of these patients reported onsets after 18 years of age (1–3). Patients with adult separation anxiety disorder experience high levels of functional impairment and show a poor response to conventional treatments used for other anxiety subtypes (4).

There is a dearth of epidemiologic data focusing on separation anxiety disorder across the lifespan. A longitudinal study commencing in childhood recorded a 5% lifetime prevalence of separation anxiety disorder by the time the cohort reached early adulthood (5). The National Comorbidity Survey Replication in the United States found a lifetime separation anxiety disorder prevalence of 6.6% after the pediatric age-at-onset requirement was removed, with two-thirds of case subjects having onsets after 17 years of age (6), but comparable data in other samples or populations have yet to be reported.

The relationship of separation anxiety disorder with other mental disorders also remains to be clarified. A longstanding theory has posited a specific developmental relationship between childhood-onset separation anxiety disorder and adult panic disorder and/or agoraphobia (7, 8), an association supported by findings from twin and laboratory studies (9, 10). Yet a meta-analysis of community studies has indicated that separation anxiety disorder may represent a generic risk factor for a range of anxiety disorders and other forms of psychopathology in adulthood (11) rather than primarily for panic disorder and/or agoraphobia.

Consistent with the principles of attachment theory (12), a recent developmental model has suggested that separation anxiety symptoms may mediate the associations between early family adversity and trauma and subsequent onset of common adult mental disorders (13). Family dysfunction and exposure to major disasters appear to be associated with subsequent onset of separation anxiety in childhood (13–15). Overall, however, data are limited concerning associations involving early adversity, exposure to trauma, onset of separation anxiety disorder, and a range of later psychopathologic outcomes.

In the present study, we analyzed data from the cross-national World Health Organization (WHO) World Mental Health surveys (16) to assess the following key aspects of separation anxiety disorder in the general population: 1) lifetime and 12-month prevalence overall and by gender and country income grouping; 2) the proportions of case subjects experiencing childhood and adult onsets; 3) patterns of comorbidity and temporal relationships with onset and persistence of other common DSM-IV disorders; 4) associations of other predictors (sociodemographic variables, childhood adversities, lifetime traumatic events) with separation anxiety disorder onset and persistence; and 5) severity of role impairment associated with 12-month separation anxiety disorder.

Method

Samples

Separation anxiety disorder was assessed in 18 World Mental Health surveys: nine in high-income countries (Belgium, France, Germany, Italy, the Netherlands, Northern Ireland, Portugal, Spain, and the United States), five in upper-middle income countries (Brazil, Bulgaria, Lebanon, Mexico, and Romania), and four in low-/lower-middle income countries (Colombia, Nigeria, the People’s Republic of China, and Peru). A total of 38,993 respondents were assessed. All surveys used probability sampling based on multistage area clustered household survey designs with no substitution for nonparticipants. The majority of surveys were based on nationally representative samples, the remainder on samples representative of particular urban areas (Sao Paulo in Brazil, Beijing and Shanghai in the People’s Republic of China), all nonrural areas in the country (Colombia, Mexico), or major regions of the country (Nigeria). Response rates ranged from 45.9% to 90.2% and averaged 70.3%. More details on the World Mental Health Survey sampling are presented elsewhere (17).

The World Mental Health Survey interview was administered in two parts to reduce respondent burden. All respondents completed part I, which assessed core disorders. All respondents with a part I disorder plus a probability subsample of other part I respondents were administered part II, which assessed additional disorders and correlates. The part I data were weighted for differential probabilities of selection and to match population distributions on sociodemographic and geographic variables. The part II data were additionally weighted to adjust for differential probabilities of selection from part I into part II. Separation anxiety disorder was typically assessed in part II but in some countries was included in part I. In one-half of the surveys, the assessment of separation anxiety disorder was restricted to respondents in the age range 18–39 or 18–44, while in the other surveys, there was no such age restriction.

Measures

Overview.

World Mental Health Survey interviews were conducted face-to-face by lay interviewers. Consistent translation, back-translation, and harmonization procedures were used in adapting the interview for local administration (18). Consistent interviewer training and field quality-control procedures were applied across sites (19). All respondents provided informed consent according to the requirements of local institutional review boards before being interviewed.

Diagnostic assessment.

The diagnostic interview was the WHO Composite International Diagnostic Interview (CIDI) (20), a fully-structured interview that assessed lifetime and 12-month prevalence of DSM-IV mood (major depressive disorder and/or dysthymia, bipolar disorder), anxiety (panic disorder and/or agoraphobia, specific phobia, social phobia, generalized anxiety disorder, posttraumatic stress disorder [PTSD]), and externalizing (intermittent explosive disorder, alcohol and drug abuse with or without dependence) disorders. A special probing strategy shown to yield improved age-at-onset reports of individual disorders was used (21). A blinded clinical reappraisal study using the Structured Clinical Interview for DSM-IV (SCID) (22) as the gold standard found good diagnostic concordance between core CIDI and SCID diagnoses, although the module for separation anxiety disorder was not included (23). The CIDI module for separation anxiety disorder departed from DSM-IV in assessing lifetime onset of symptoms not only as of age 18 but also for onsets that occurred at ages 19 or later (6). A separation anxiety disorder diagnosis required endorsement of at least three of the eight criterion A DSM-IV symptoms, having symptoms for at least 1 month, and experiencing associated clinically significant distress or role impairment.

Sociodemographic variables.

Sociodemographic variables considered here include age at interview (18–34, 35–49, 50–64, and ≥65 years), gender, education (student, and among nonstudents, low, low-average, average-high, and high-level of education based on country-specific distributions), and marital status.

Functional Impairment.

A modified version of the Sheehan Disability Scales (24) was used to assess severity of role impairment associated with separation anxiety disorder in the previous year. Respondents were asked to quantify severity of role impairment in home management, work, social life, and personal relationships using a 0–10 self-anchoring scale with the following response categories: none (0), mild (1–3), moderate (4–6), severe (7–9), and very severe (10). Respondents rated the Sheehan Disability Scales for the month in the previous year when separation anxiety disorder was most severe. Scores for the Sheehan Disability Scales were dichotomized into severe (range 7–10 for any domain of the scales) or not severe (scores lower than 7–10 for all domains).

Childhood family adversities.

World Mental Health Survey respondents were asked retrospectively about exposure to a wide range of childhood family adversities. As reported previously (25), exploratory factor analysis of responses found one factor for experiences indicative of childhood maladaptive family functioning (parental mental illness, substance misuse, criminal behavior, domestic violence, and child physical abuse, sexual abuse, and neglect), while other childhood adversities were combined into a second scale that included parental death, parental divorce, other long separations from a parent, serious illness of a close family member, and family economic adversity.

Traumatic events.

The CIDI assessed 29 lifetime traumatic events aggregated into the seven domains of accidents and natural disasters, war events, intimate and sexual violence, death of a loved one, other interpersonal violence, network events, and other traumas participants decided not to disclose (26). Dichotomous measures were created for one or more events in each of these domains.

Analysis Procedures

Cross-tabulations were used to estimate lifetime and 12-month separation anxiety disorder prevalence separately for men and women in each survey. Age-at-onset reports were analyzed using the two-part actuarial method to estimate survival curves (27). The proportion of lifetime cases with onsets after age 18 was calculated for each country and country income group. Discrete-time survival analysis with person-year as the unit of analysis and a logistic link function (28) was used to examine cross-lagged associations of temporally primary separation anxiety disorder with subsequent first onset of other disorders and reciprocal associations of other temporally primary disorders with subsequent first onset of separation anxiety disorder. Survival coefficients and standard errors were exponentiated to generate odds-ratios with 95% confidence intervals.

The same survival analysis approach was used to estimate associations of sociodemographic variables, childhood adversities, and lifetime traumatic events (28) with first onset of separation anxiety disorder and to examine variation in strength of prediction as a function of age at onset (childhood [up through age 12], adolescence [ages 13–17], early adulthood [ages 18–29] and later onsets [ages ≥30]). The associations of the same predictors with persistence of separation anxiety disorder, defined as 12-month prevalence among lifetime cases, were examined using person-level logistic regression analysis controlling for age at onset and time since onset. Finally, cross-tabulations were used to examine joint associations of country income level, separation anxiety disorder age at onset, and 12-month comorbidities with 12-month severe separation anxiety-related role impairments. Relative fit of additive and interactive models was evaluated using the Akaike information criterion and Bayesian information criterion (29).

Standard errors of estimates were based on the Taylor series linearization method implemented with SUDAAN software (30, 31) to adjust for the weighting and geographic clustering of World Mental Health data. Multivariate significance tests were carried out with Wald chi-square tests based on Taylor series coefficient variance-covariance matrices. Statistical significance was evaluated using 0.05-level two-sided tests.

Results

Prevalence

Lifetime prevalence of the expanded CIDI definition of DSM-IV separation anxiety disorder that allowed adult onsets was 4.8% in the total sample but with a much wider interquartile range (25th−75th percentiles: 1.4–6.4) and range (0.2%−9.8%) across countries than previously found for most other DSM-IV/CIDI disorders (32) (Table 1). Lifetime prevalence was higher among women than men in 15 of the 18 countries and significantly so in the total sample (5.6% compared with 4.0%; χ²=4.0, df=1, p<0.05). Twelve-month prevalence was considerably lower than lifetime prevalence (1.0% in the total sample; interquartile range: 0.2%−1.2%; range: 0.0%−2.7%), higher among women than men in 14 of 18 countries and significantly higher among women than men in the total sample (1.3% compared with 0.8%; χ²=12.0, df=1, p<0.001).

Age at Onset

The age-at-onset distribution of separation anxiety disorder was quite similar across the three country income groups (Figure 1). Median age at onset was in the late teens in high- and upper-middle income countries and in the mid-20s in low-/lower-middle income countries. The interquartile range of the age-at-onset distribution was wider (15–35 years of age in low-/lower-middle income countries; 9–35 in high- and upper-middle income countries) than previously found for most other DSM-IV/CIDI disorders (32). A total of 43.1% of respondents with lifetime separation anxiety disorder had onsets in adulthood (ages ≥18). The proportion of lifetime cases with adult onsets was significantly higher in low-/lower-middle income countries (53.8%) than in upper-middle income countries (39.1%; χ²=98.0, df=1, p<0.001) or high-income countries (41.6%; χ²=52.8, df=1, p<0.001). Although there were only two low/low-middle income countries with sizable numbers of lifetime separation anxiety disorder cases (N=359 in Colombia; N=154 in Peru), the proportion of these cases with adult onsets was comparable in the two surveys (55.7% compared with 51.8%; χ²=1.5, df=1, p=0.22).

Persistence

Comparison of individual-level age at onset and recency reports showed that the majority of people with lifetime separation anxiety disorder remitted within a decade of onset (Figure 2). However, the recovery curves became much less steep after approximately 10 years. As with age at onset, the distribution of time to remission was quite consistent across country income groups. This relatively rapid remission of separation anxiety disorder is consistent with the low 12-month/lifetime prevalence ratio shown in Table 2.

Comorbidity

Lifetime separation anxiety disorder was significantly comorbid with 13 of the other 14 DSM-IV/CIDI disorders assessed in the World Mental Health surveys, the exception being substance dependence with abuse. Survival analysis pooled across countries showed that these associations were a result of 1) significant associations between temporally primary separation anxiety disorder and subsequent first onset of 10 other disorders (odds ratios in the range of 1.3–2.8) coupled with 2) significant associations of nine other disorders with subsequent first onset of separation anxiety disorder (odds ratios in the range of 1.3–2.1) (Table 2). The vast majority of cross-lagged odds ratio pairs between separation anxiety disorder and other disorders were reciprocal in significance and comparable in magnitude. The major exceptions were that temporally primary separation anxiety disorder was a significantly stronger predictor of subsequent attention deficit hyperactivity disorder (ADHD) (odds ratio=2.8) than ADHD was of subsequent separation anxiety disorder (odds ratio=1.1) and that there were no significant associations between temporally primary PTSD, generalized anxiety disorder, or agoraphobia and subsequent-onset separation anxiety disorder in contrast to the reciprocal relationships. Disaggregation by country income group (results available upon request from Silove) showed considerable consistency in these patterns, with the vast majority of significant odds ratios in the total sample also significant in at least two of the three country income groups.

The associations of other temporally primary disorders with subsequent separation anxiety disorder persistence (i.e., 12-month prevalence among lifetime cases controlling for age at onset and time since onset) were not significant. The same was generally true for the associations of temporally primary separation anxiety disorder with subsequent persistence of other disorders, with the exceptions being that temporally primary separation anxiety disorder was a predictor of persistence of agoraphobia (odds ratio=2.6), and to a lesser extent, PTSD. As with the analyses of first onset, disaggregation by country income group (results available upon request from Silove) showed consistency in the nonsignificance of these reciprocal associations between lifetime comorbidity and persistence.

Other Predictors of Separation Anxiety Disorder Onset and Persistence

Controlling for lifetime comorbid DSM-IV/CIDI disorders, age, and country, survival analysis showed that lifetime separation anxiety disorder was significantly associated with being female (odds ratio=1.1–1.4), having low through high-average (compared with high) education (odds ratio=1.5–1.7), maladaptive family functioning childhood adversities (odds ratio=1.7–2.8), other childhood adversities (odds ratio=1.2–1.8), and a variety of lifetime traumatic events (odds ratio=1.3–1.6) (Table 3). It is noteworthy that the associations involving maladaptive family functioning childhood adversities and other lifetime traumatic events predicted not only pediatric-onset but also adult-onset separation anxiety disorder, while other childhood adversities predicted only childhood-onset separation anxiety disorder.

Disaggregation by country income group (detailed results available upon request from Silove) showed considerable consistency in these patterns, with the vast majority of significant odds ratios in the total sample also significant in at least two of the three country income groups. As with the analysis of comorbidity, the predictors considered here were not significantly related to separation anxiety disorder persistence either in the total sample or in country income groups (detailed results available upon request from Silove). For example, the associations of maladaptive family functioning childhood adversities with separation anxiety disorder persistence were nonsignificant both in the total sample (χ²=2.5, df=5, p=0.78) and in each of the three country income groups (χ²=2.9–10.2, df=5, p=0.07–0.72). In addition, we found that age at onset (defined in categories of childhood <13 years old, adolescence 13–17 years old, young adulthood 18–29 years old, and later adulthood ≥30 years old) was not significantly related to persistence either in the total sample (χ²=4.3, df=3, p=0.23) or in any of the three country income groups (χ²=3.0–5.4, df=3, p=0.15–0.39).

Role Impairment of 12-Month Separation Anxiety Disorder

More than one-third (35.3%) of respondents with 12-month separation anxiety disorder reported severe separation anxiety-related role impairment in the year before interview (Table 4). The rate of severe role impairment was considerably lower in low-/lower-middle income countries (17.6%) than upper-middle or high-income countries (38.2%−41.4%; χ²=12.6–18.1, df=1, p<0.001) and considerably higher in the presence than absence of 12-month comorbid DSM-IV/CIDI disorders (42.4% compared with 18.3%; χ²=21.6, df=1, p<0.001). Separation anxiety-related role impairment did not differ markedly as a function of separation anxiety disorder age at onset (χ²=2.9, df=3, p=0.40). A model that assumed additive effects of country income level, comorbidity, and age at onset predicted severe separation anxiety-related role impairment with no evidence of significant interaction between the factors.

Discussion

The findings concerning the prevalence of separation anxiety disorder across countries are summarized in Figure 3. The 4.8% overall lifetime prevalence estimate found in this study suggests that separation anxiety disorder is a relatively common lifetime disorder, although with a much wider range of prevalence across countries (0.2%−9.8%; interquartile range: 1.4–6.4) than found for most other DSM-IV/CIDI disorders assessed in the World Mental Health Surveys (32). Temporally prior separation anxiety disorder was found to be associated with significantly elevated odds of subsequent first onset of a wide range of other disorders, including not only internalizing disorders (major depression, bipolar disorder, specific and social phobias, panic disorder, and/or generalized anxiety disorder) but also externalizing disorders (ADHD, oppositional defiant disorder, and conduct disorder). This finding is consistent with a recent meta-analysis concluding that separation anxiety disorder represents a generic risk factor for a range of common mental disorders (11). Importantly, therefore, our data do not support the hypothesis of an exclusive link between separation anxiety disorder and panic disorder and/or agoraphobia (7, 8), although the existence of a significant association between temporally primary separation anxiety disorder and subsequent persistence of agoraphobia has potential prognostic importance in relation to the latter disorder.

We also found reciprocal associations of a similarly wide range of temporally primary disorders with the subsequent onset of separation anxiety disorder, a result that is broadly consistent with the suggestion that these comorbidities are due to common causes more than to direct effects of particular early-onset disorders on particular later-onset disorders (33). Indeed, an earlier World Mental Health Survey analysis found that separation anxiety disorder had a high factor loading on the internalizing factor of a two-dimensional model and that controls for summary internalizing-externalizing dimensions accounted for the significant cross-lagged associations of separation anxiety disorder with most comorbid disorders (34). Questions therefore remain whether separation anxiety disorder has any specificity as a risk factor for onset of particular secondary disorders either in childhood or adulthood (13).

Maladaptive family functioning childhood adversities and exposure to traumatic life events were found to be associated with separation anxiety disorder onset but not persistence, both in the total sample and, separately, in low-/lower-middle, upper-middle, and high-income countries. Importantly, these associations were found for separation anxiety disorder onsets across the entire life course. It is possible that more in-depth future analyses will find significant specifications in the associations of particular types of childhood adversities or traumatic events at specific life course stages with subsequent separation anxiety disorder onset. Nevertheless, in such analyses, it will be important to account for more general associations that may exist involving a wide range of childhood adversities and traumatic events as a backdrop against which more specific specifications are considered. The mechanisms responsible for the ongoing liability to separation anxiety disorder and other mental disorders arising from early adversity and trauma, including the possible neurobiological mediators of these effects, are the focus of ongoing inquiry (35, 36).

Our findings suggest that separation anxiety disorder is more likely to be seriously impairing in higher-income than low-/lower-middle income countries. The explanation of this specification is unclear, although it is possible that a culture of individuality and independence in higher-income countries results in low personal and social tolerance of separation anxiety, whereas a collectivist culture in low-/lower-middle income countries accepts a degree of separation anxiety as normative or even adaptive. A related unanswered question is why the range of prevalence across countries is so large.

Limitations of the study include variation in response rates across countries, use of a fully-structured diagnostic interview that did not allow clinical probing to confirm diagnoses, and a cross-sectional design in which lifetime prevalence, childhood adversities, and lifetime traumatic events were assessed retrospectively. Although we used the DSM-IV 1-month duration requirement rather than the DSM-5 6-month recommended (although not mandated) duration criterion for adults in defining separation anxiety disorder, overestimation of prevalence in relation to the latter diagnostic system is likely to be slight given that the median persistence of disorder was 4–8 years. Recall bias is a more serious concern, since this might have led to an underestimation of lifetime prevalence among remitted cases and an overestimation of persistence, even though the World Mental Health Surveys used special probing strategies designed to improve recall accuracy (21) and in some countries limited duration of recall by assessing separation anxiety disorder only among respondents younger than ages 40–45. If biases exist, however, this means that separation anxiety disorder is an even more prevalent lifetime disorder with an even lower persistence than suggested by the present results. Retrospectively reported age at onset was unrelated to 12-month persistence among lifetime cases, suggesting that there was no tendency for respondents to overstate onset of separation anxiety disorder in adulthood. The high prevalence of adult-onset cases therefore supports the decision to remove the separation anxiety disorder age-at-onset restriction in DSM-5. An issue worthy of further investigation is that the ratio of 12-month/lifetime separation anxiety disorder prevalence is much lower than that of most other World Mental Health disorders (1.0% 12-month prevalence; range: 0.0%−2.7%; interquartile range: 0.2%−1.2%).

Our findings have important implications for clinical practice. The results challenge the long-established view that separation anxiety disorder should be reserved for diagnosis among children and adolescents by indicating that adult onset is prevalent across countries and that adult-onset separation anxiety disorder is equally persistent and impairing as the pediatric-onset form of this disorder. Clinicians should be alerted to the need to consider separation anxiety disorder in the differential diagnosis of patients of all ages presenting not only with anxiety but with a wide variety of internalizing and externalizing disorders. As yet, existing psychological and pharmacological treatments used for the other anxiety disorders have proven to be ineffective for adult separation anxiety disorder (4, 37). There is consequently an urgent need to devise and test novel treatments for this disorder, particularly as it manifests in adulthood.