Adult separation anxiety: Personality characteristics of a neglected clinical syndrome

Abstract

Over the past 2 decades, interest in the relationship between personality and psychopathology has resurged. However, the clinical problem of adult separation anxiety (ASA) has been largely excluded from this endeavor due to the age-of-onset criterion in older editions of the DSM that prohibited first-onset diagnoses in adulthood. This study tests relationships between ASA symptoms and higher- and lower-order personality traits in a community sample of 565 women. It accounts for systematic error by utilizing informant report, 2 personality inventories, and data from 2 time points over 3 years, and by adjusting for mood state. It also tests longitudinal ASA–personality models. Results indicate that ASA is robustly associated with negative emotionality and its facet of stress reaction, as well as to aggression, alienation and absorption to somewhat lesser degrees. These relationships are not due to overlap with other traits (except in the case of alienation), or mood-state biases, and they are verified by informants. Moreover, negative temperament predicts greater levels of ASA 3 years later, adjusting for baseline ASA. Neither positive emotionality or temperament, nor positive emotionality’s lower-order scales, were uniquely related to ASA in multi-trait models, whereas relationships between ASA and disinhibition and constraint were inconsistent. These findings lay the groundwork for future research testing the mechanisms and causal links between these personality traits and ASA and may help clinicians anticipate traits that are associated with ASA in order to tailor treatments to patients’ personalities.

General Scientific Summary:

Identifying how personality traits relate to various forms of psychopathology can advance treatment efforts and lay the groundwork for future research on mechanisms and causal links. This study suggests that adult women with higher levels of separation anxiety are more prone to experiencing negative emotions and reacting strongly to stress, and that they may also tend towards aggression, feeling alienated, and being absorbed in their sensory experiences. The study also suggests that the tendency to experience negative emotions may lead to increased levels of separation anxiety three years later.

Over the last 2 decades, there has been a resurgence in efforts to join the fields of adult psychopathology and personality research (Clark, 2005; Kotov, Gamez, Schmidt, & Watson, 2010; Krueger, Caspi, Moffitt, Silva, & Mcgee, 1996). One clinical problem that has been excluded almost entirely from this endeavor is separation anxiety. A recent meta-analysis on personality-psychopathology relationships included 175 studies and 11 disorders but did not mention adult separation anxiety (ASA) (Kotov et al., 2010). The exclusion of ASA from such research does not reflect its clinical significance or prevalence; indeed, the lifetime prevalence of ASA is surprisingly high, ranging from 7% in a representative population-based survey (Shear, Jin, Ruscio, Walters, & Kessler, 2006) to 23% (Silove, Marnane, Wagner, Manicavasagar, & Rees, 2010) or 42% (Pini et al., 2010) in clinical samples, and nearly one third of adults with only ASA and no comorbid disorders report severe role impairment (Shear et al., 2006). Rather, ASA was excluded from personality-psychopathology research in adults because it was considered a childhood disorder and past editions of the DSM prohibited new diagnoses in adulthood. In part due to evidence that a substantial proportion of patients and the vast majority of adults in the community with ASA report first experiencing symptoms in adulthood (Shear et al., 2006; Silove et al. , 2010), this criterion was lifted in DSM-5. Clinical descriptions of prototypical cases (e.g., Bögels, Knappe, & Anna, 2013; Milrod et al., 2014) highlight how individuals with ASA are “separation-sensitive” (Milrod et al., 2014, p. 35) across social contexts.

In the few studies that have examined ASA-personality associations, researchers have utilized clinical samples only, which are prone to bias due to high levels of disorder severity and comorbidity. They found that adults with separation anxiety disorder have higher levels of neuroticism compared to other anxiety disorder patients (Manicavasagar, Silove, Curtis, & Wagner, 2000; Silove et al., 2010). Others using Cloninger’s model of personality reported that the personality profiles of ASA patients were similar to those of panic disorder patients, except that ASA patients scored higher on harm avoidance and lower on self-directedness, and, in comparison to healthy controls, ASA patients scored higher on reward dependence and self-transcendence (Mertol & Alk, 2012). It remains an open question whether ASA is linked to personality traits in non-treatment-seeking community populations.

In the current study, we test these associations in a community sample using the Big Three personality framework (Tellegen, 1985; Watson & Clark, 1992), which consists of three higher order traits: negative emotionality, positively emotionality, and constraint. To mitigate systematic error that can bias self-report personality research, we utilized two different measures of personality, tested associations concurrently and across a three-year period, and gathered reports from target participants and informants. We expected the longitudinal ASA-personality associations to be present across a three-year period because personality traits tend to be enduring, particularly in middle and later adulthood (Costa & McCrae, 1998; Johnson, McGue, Krueger, 2005), and because the magnitude of personality-psychopathology relationships tends to be fairly stable over time (e.g., Jeronimus, Kotov, Riese, & Ormel (2016). We also used a dimensional assessment of ASA because dimensional assessments of psychopathology are more reliable and have higher power compared to categorical classifications (Markon, Quilty, Vagby, & Krueger, 2013). Finally, we tested how personality predicts ASA three years later, adjusting for baseline levels of ASA, to inform models of ASA development and elucidate trait vulnerabilities for early-identification and intervention.

Method

Participants

The participants are mothers from 609 families in an ongoing longitudinal study of children’s temperament and psychopathology. Families from the local community who had a 3-year-old child at the first wave of the study (W1) and at least one English-speaking biological parent were eligible to participate; families were excluded if the child had a developmental delay. Of the 609 families, 559 were recruited at W1 and 50 additional minority families were recruited at the second wave (W2) three years later to increase racial and ethnic diversity (see Bufferd, Dougherty, Carlson, Rose, & Klein, 2012 for recruitment details). Parents provided written informed consent after receiving a description of the study. The study (“Temperamental emotionality in preschoolers and depression risk”) was approved by the human subjects review committee at Stony Brook University, and families were compensated. The current study includes reports from three waves (W1, W2, and W3), which were approximately three years apart.

The sample for the current study includes 565 women who had self- or informant-reports of personality at waves 1 or 2; 99 of them (17.5%) had only self-reports of personality and 5 (0.9%) had only informant-reports. Only the primary caretaker completed the measure of ASA at W2; for the current sample, this included 378 women. For predictive models, we also included mothers’ W3 ASA (n = 467) in the current study.

The majority of women were married or living with their child’s father at W2 (86.4%) and at W3 (84.3%). They were 39.0 years old on average (SD = 4.9; range: 22.8–51.8) at W2 and 41.8 years old on average (SD = 4.8; range: 25.9–53.5) at W3. Approximately half had at least a 4-year college degree at W2 (56.7%) and at W3 (56.7%). The larger study’s main focus is on children’s psychopathology. Thus, parents’ race and ethnicity were not collected and participating children’s race and ethnicity are used instead as proxies; 89.0% of mothers had children who were White and 12.7% had children who were Hispanic or Latino.

Missingness Analysis

Women included in the study sample were compared on the demographic variables and the ASA measure to the women from the larger study who were excluded due to missing all personality assessment data. Participants did not differ on demographic variables or on the ASA measure (Supplemental Table 1). Missing data on demographic variables were due to failure to respond.

Measures

See Table 1 for item counts and Cronbach’s α for each measure.

Table 1.

Concurrent and longitudinal associations between personality traits and adult separation anxiety, and descriptive statistics and scale information for personality variables

		Associations with Wave 2 ASA
Wave	Independent Variable	β	[95% CI]	M (SD)	Cronbach’s α	# of Items
1	Negative Emotionality	0.48***	[0.39, 0.57]	8.47 (6.07)	.85	40
	Stress Reaction	0.43***	[0.33, 0.52]	5.26 (3.94)	.84	15
	Stress Reaction-I	0.25***	[0.13, 0.37]	4.41 (3.43)	.84	15
	Alienation	0.26***	[0.14, 0.37]	1.57 (2.12)	.74	13
	Aggression	0.32***	[0.18, 0.46]	1.64 (1.94)	.62	12
	Positive Emotionality	−0.10^	[−0.21, 0.00]	32.28 (8.55)	.87	52
	Well Being	−0.16*	[−0.29, −0.03]	9.33 (3.20)	.80	14
	Well Being-I	−0.04	[−0.18, 0.09]	7.9 (3.48)	.86	14
	Social Potency	0.00	[−0.12, 0.12]	6.62 (3.40)	.81	14
	Achievement	−0.02	[−0.13, 0.10]	7.67 (3.00)	.78	12
	Social Closeness	−0.14*	[−0.26, −0.01]	8.66 (3.06)	.78	12
	Social Closeness-I	−0.08	[−0.22, 0.07]	8.44 (3.42)	.84	12
	Constraint	−0.13*	[−0.25, −0.02]	26.08 (5.24)	.69	37
	Control	−0.16*	[−0.29, −0.04]	9.79 (2.64)	.72	13
	Harm Avoidance	−0.03	[−0.15, 0.09]	8.81 (2.68)	.66	12
	Harm Avoidance-I	0.02	[−0.10, 0.13]	8.54 (3.00)	.71	12
	Traditionalism	−0.03	[−0.17, 0.10]	7.48 (2.40)	.67	12
	Absorption	0.30*	[0.19, 0.40]	5.32 (3.05)	.76	12
2	Negative Temperament	0.44***	[0.35, 0.53]	4.23 (3.69)	.81	14
	Negative Temperament-I	0.13*	[0.00, 0.25]	9.31 (3.93)	.78	14
	Positive Temperament	−0.16*	[−0.29, −0.03]	9.03 (3.14)	.82	13
	Positive Temperament-I	−0.13^	[−0.26, 0.00]	8.4 (2.29)	.58	9
	Disinhibition	0.23**	[0.10, 0.35]	2.97 (2.35)	.63	16
	Disinhibition-I	0.11^	[−0.01, 0.22]	7.08 (3.34)	.61	9

Notes:

^{^}p < .10

^*p < .05

^**p < .01

^***p < .001.

n = 362–475. Higher-order traits bolded. I = informant-report; all others are self-report. ASA = adult separation anxiety.

Adult Separation Anxiety Questionnaire (ASA-27).

The ASA-27 is a self-report measure of separation anxiety symptoms experienced over age 18 (Manicavasagar, Silove, Wagner, & Drobny, 2003). It has 27 items, which are rated on a 4-point scale (1 = This happens very often; 4 = This has never happened). For scoring, items were reverse coded and subtracted by 1. The questionnaire was based on the Adult Separation Anxiety Semistructured Interview (ASA-SI; Manicavasagar, Silove, & Curtis, 1997), which drew questions from DSM-IV criteria for juvenile separation anxiety disorder, literature on attachment theory and research, the authors’ clinical impressions, and a qualitative study of patients with suspected ASA. The ASA-27 shows good test-retest reliability over approximately three weeks (r = 0.86, p < .001; Manicavasagar et al., 2003). In our sample, Cronbach’s α was .89 at W2 and .91 at W3.

Several studies support the validity of the ASA-27. The measure’s developers found that, among patients with panic disorder, ASA-27 scores are correlated with anxious and avoidant styles of attachment, and not with secure attachment (Manicavasagar, Silove, Marnane, & Wagner, 2009). In an independent study (Cyranowski et al. 2002), the ASA-27 was strongly related to scores on the Structured Clinical Interview for Separation Anxiety Symptoms (SCISAS; r = .84), an interview that closely resembles the questions on the structured interview used in the National Comorbidity Study-Replication (Shear et al., 2006). The ASA-27 was also related to retrospective self-reports and interview assessments of separation anxiety experienced during childhood (r = .77 and r = .70, respectively; Cyranowski et al. 2002). In our sample, the measure has shown partial strict measurement invariance across gender (Finsaas, Olino, Hawes, Mackin, & Klein, 2020).

Personality.

The brief form of the Multidimensional Personality Questionnaire (MPQ-BF; Patrick & Curtin, 2002) is a 155-item self-report inventory. Items are rated using a dichotomous response format. The MPQ assesses the broad trait of negative emotionality, with facets of stress reaction, alienation, and aggression; positive emotionality, with facets of well-being, social potency, social closeness, and achievement; and constraint, with facets of control, harm avoidance, and traditionalism. Additionally, it assesses the primary trait of absorption, which does not load onto any of the higher order factors.

Informant-reports using the MPQ were collected from the mother’s partner. To reduce burden while assessing aspects of each of the 3 broad domains, informants completed a subset of the narrow-band scales. Consistent with other informant report measures of personality, questions were reworded in the third person (e.g., Bagby et al., 1998; Markon, Quilty, Bagby, & Krueger, 2013). The informant questionnaire included the well-being, social closeness, stress reaction, and harm avoidance scales. These scales were selected so that there was at least 1 scale for each superfactor, providing partial coverage of the 3 higher-order domains, and because they were relevant to the broader project goals.

The Schedule of Nonadaptive and Adaptive Personality (SNAP; Clark, 1993) includes three temperament scales covering negative emotionality, positive emotionality, and disinhibition. Items are rated using a dichotomous response format. These scales were derived originally from the General Temperament Survey (Watson & Clark, 1992) and later used as the adaptive personality scales in the SNAP.

As with the MPQ, informant-reports of personality were collected from the mothers’ partners. The SNAP-Other Description Rating Form (SNAP-ORF; Harlan & Clark, 1999) assesses negative emotionality, positive emotionality, and disinhibition using pairs of vignettes that describe the high and low end of each trait. Items are rated on 6-point Likert scales.

Mood State.

The Diagnostic Inventory for Depression (DID; Zimmerman, Sheeran, & Young, 2004) is a self-report measure designed to assess depressive symptoms over the past week; it was used as a measure of mood state in the current paper. It has 38 items, which each consist of 5 statements that increase in severity or frequency. The DID has high test-retest reliability (r = 0.91; Zimmerman et al. 2004). In our sample, Cronbach’s α was .87.

Data Analysis

All regression models were estimated in Mplus 8 (version 1.6; Muthen & Muthen, 2012–2018) using full-information maximum likelihood on both the predictor variables and the outcome variables (Enders, 2013). Robust standard errors were specified. We also included demographic variables as auxiliaries using a saturated correlates approach (Graham, 2003) to recapture information in the outcome variable and reduce bias (Enders, 2010). An example of the Mplus syntax for a single-trait model can be found in the supplemental materials. The R package MplusAutomation was used to extract model fit statistics and parameters (Hallquist & Wiley, 2018) and descriptive statistics were computed in R Studio (version 1.2.1335; R Core Team, 2016).

Single-Trait Models.

We first estimated a set of linear regression models testing the relationships between each W1 and W2 personality trait with W2 ASA. Analyses using W2 personality traits are cross-sectional; those using W1 personality traits are longitudinal but should be viewed simply as associations—we estimated regression models as opposed to correlations in order to incorporate the demographic variables as auxiliary variables. Longitudinal predictive models adjusting for prior levels of the dependent variable are tested later.

Multi-Trait Models.

Next, we tested for unique relationships between W1 and W2 personality and W2 ASA by simultaneously entering the personality traits from each wave that significantly predicted ASA in the single-trait models as predictors in a series of multi-trait regression models; one included the W1 higher-order traits as predictors, another included the W1 lower-order traits as predictors, and a third included the W2 higher-order traits as predictors.

Auto-Regressive Models.

Finally, in three separate models, we tested whether W2 negative temperament, positive temperament, and disinhibition predicted ASA at W3 beyond the effects of W2 baseline ASA.

Results

Descriptive statistics for personality can be found in Table 1. At W2, the mean of ASA-27 was 11.86 (SD = 9.28); at W3, it was 10.34 (SD = 9.86). The mean of W2 DID was 5.02 (SD = 5.72).

Single-Trait Models.

Self-reported W1 negative emotionality, all three of its lower order facets (stress reaction, alienation, and aggression), and absorption were positively related to W2 ASA. In contrast, W1 well-being and social closeness, which are lower order facets of positive emotionality, were negatively related to W2 ASA, as were constraint and its lower order facet of control (Table 1). Of the W2 self-reported traits, negative temperament and disinhibition were positively related to W2 ASA and positive temperament was negatively related to W2 ASA. When mood state (as assessed by the DID) was included as a covariate in the models testing concurrent relationships between W2 personality and ASA, negative temperament was still related to higher levels of ASA (β = .30, SE = .05, 95% CI [.21, .40], p < .001) as was disinhibition (β = .14, SE = .06, 95% CI [.03, .26], p = .02), whereas positive temperament was no longer related to ASA (β = −.01, SE = .06, 95% CI [−.12, .10], p = .83). Using informant-reported personality, W1 stress reaction and W2 negative temperament were positively related to ASA. In contrast, W1 well-being, social closeness, and harm avoidance and W2 positive temperament and disinhibition were not significantly related to ASA.

Multi-Trait Models.

Of the W1 higher-order traits, negative emotionality was uniquely related to W2 ASA, whereas constraint was not (Table 2). Of the W1 lower-order traits, stress reaction, aggression, and absorption were positively and uniquely associated with ASA. Finally, of the W2 higher-order traits, both negative temperament and disinhibition were positively and uniquely associated with ASA.

Table 2.

Multiple regression models testing unique associations between personality traits and adult separation anxiety

Wave 2 ASA
Wave	Predictor	β	95% CI	Fit	95% CI
1	Negative Emotionality	0 47***	[0.38, 0.56]
	Constraint	−0.10^	[−0.20, 0.00]
	R 2			.24***	[.15, .34]
	Stress Reaction	0.30***	[0.19, 0.41]
	Alienation	0.05	[−0.08, 0.17]
	Aggression	0.17*	[0.03, 0.30]
	Well Being	−0.05	[−0.21, 0.10]
	Social Closeness	−0.03	[−0.14, 0.09]
	Control	−0.07	[−0.18, 0.05]
	Absorption	0.22***	[0.10, 0.35]
	R 2			.30***	[.20, .39]
2	Negative Temperament	0.42***	[0.33, 0.50]
	Positive Temperament	0.03	[−0.09, 0.14]
	Disin hibition	0.15**	[0.04, 0.27]
	R 2			2.2***	[.13, .30]

Notes:

^{^}p < .10

^*p < .05

^**p < .01

^***p < .001.

Higher-order traits bolded. ASA = adult separation anxiety.

Auto-Regressive Models.

Negative temperament predicted significantly higher levels of ASA at W3 than would have been predicted given W2 ASA scores, whereas positive temperament and disinhibition did not (Table 3).

Table 3.

Longitudinal auto-regressive models testing relationship between adult separation anxiety and self-reported personality

	Wave 3 ASA
Predictors	β	95% CI	Fit	95% CI
ASA	0.61***	[0.51, 0.71]
Negative Temperament	0.14**	[0.05, 0.23]
R 2			.47***	[.37, .58]
ASA	0.66***	[0.58, 0.74]
Positive Temperament	−0.06	[−0.13, 0.02]
R 2			.46***	[.35, .56]
ASA	0.65***	[0.56, 0.74]
Disinhibition	0.09^	[0.00, 0.18]
R 2			.46***	[.35, .57]

Notes:

^{^}p < .10

^*p < .05

^**p < .01

^***p < .001.

Personality traits from wave 2 assessment. ASA = adult separation anxiety. CI = Confidence interval.

Discussion

We found that relationships between ASA and negative emotionality and temperament were substantial, robust to the effects of time, mood state, informant, and personality measure, and persisted in multivariate models including other higher-order traits as simultaneous independent variables. These findings are consistent with previous work in clinical samples that showed that patients with ASA have higher levels of neuroticism than patients with other anxiety disorders (Silove et al., 2010). Negative temperament also predicted higher levels of ASA than expected given baseline levels three years later, consistent with a vulnerability model of psychopathology-personality relationships.

A finer-grained analysis revealed that the lower-order facet negative emotionality of stress reaction was a particularly important feature of ASA. Aspects of stress reaction, like being prone to worry, map onto aspects of separation anxiety, such as fears that harm will befall the loved one. The tendency to feel vulnerable also aligns with core cognitive dysfunctions proposed for juvenile separation anxiety disorder that involve overestimating the danger of being left and underestimating the capacity for independent functioning (Bögels, Snieder, & Kindt, 2003; Bögels & Zigterman, 2000). On the other hand, past work suggests that stress reactivity non-differentially relates to affective and other anxiety disorders, as well as to substance use and conduct disorder (Krueger et al., 1996), so the cognitions and behaviors captured by stress reaction are not likely to be unique to ASA.

This study is the first to report a link between aggression and ASA, bivariately and beyond the effects of other personality traits. Features of aggression, like being physically aggressive, victimizing others, and enjoying upsetting and frightening others, are absent in accounts of prototypical ASA cases (Bögels, Knappe, & Anna, 2013; Milrod et al., 2014). They are, however, congruent with Bowlby’s attachment behavioral system, which suggests that aggression in the context of separation anxiety may be “anger born of fear” (p. 247; Bowlby, 1973). This aggression may be a desperate effort to coerce attachment figures to stay when other means have failed and separation is too difficult to bear due to unmet attachment needs (Dutton & White, 2012). Indeed, some theorize a link between ASA and domestic violence perpetration (Dutton & White, 2012). Angry or aggressive behavior can also be seen in children with separation anxiety when facing separation, and the DSM-5 describes these behaviors as characteristic of the disorder in youth but does not mention them in regard to adults (American Psychiatric Association, 2013). Aggression may also contribute to the interpersonal impairment experienced by adults with higher levels of separation anxiety (Shear et al., 2006). Notably, in past research, MPQ aggression was not associated with other anxiety (and related) disorders (Krueger et al., 1996).

We also found that women with higher levels of separation anxiety reported higher levels of alienation. However, this association did not persist in the multi-trait model, which suggests that it is better explained by covariance between other personality traits and ASA. Moreover, like stress reaction, this trait is associated with a range of other forms of psychopathology, including other anxiety disorders (Krueger et al., 1996). Our results also suggested that disinhibition and constraint, and positive emotionality and temperament, are less important for characterizing and understanding ASA compared to negative emotionality.

Absorption was positively related to ASA in the single-trait models and uniquely in the multi-trait model. This ASA-absorption link is consistent with Mertol and Alk’s (2012) study on ASA patients; they found that the subscale elevating self-transcendence scores (a trait that is strongly correlated with MPQ absorption at r = 0.64; Laidlaw, Dwivedi, Naito, & Gruzelier, 2005) was “transpersonal identification,” or a sense of unity with objects outside the individual self. According to Tellegen and Atkinson (1974), absorption is an allocentric process. All perceptual, motoric, and cognitive resources are committed to experiencing an attentional object, which, in the case of ASA, may be another person, rather than to representing the self, such that the other “…acquire[s] an importance and intimacy that are normally reserved for the self and may, therefore, acquire a temporary self-like quality” (p. 274). The complete engagement of these representational resources in turn precludes meta-cognitions like, “This is only my imagination,” as well as distractions by external events that may otherwise temper the absorptive experience. For someone with separation anxiety, when an attachment figure leaves, it may feel like a part of the self is leaving and thus be particularly threatening and painful.

The results of this study must be considered in light of limitations. The reliability of some of the self- and informant-reported disinhibition and constraint scales were in the acceptable range, and the reliability of the informant-reported positive temperament scale was in the poor range (see Table 1). This may account for the mixed findings across reporters or the nonsignificant results for these variables (the alphas for the SNAP informant-report scales are within the range reported by the scales’ developers; Harlan & Clark, 1999). Furthermore, the sample included women only who were primarily White, middle class, and in their late 30s and early 40s with young children, which limits the generalizability of the findings, particularly because having young children may exacerbate ASA. We also administered a self-report measure of ASA; interview and informant reports of ASA may be useful. Finally, we tested temporal effects of personality on ASA but did not have the data to test temporal effects of ASA on personality, and, consistent with the scar model, it is possible that ASA influences personality over time.

ASA is characterized by negative emotionality and its facet of stress reaction, as well as to somewhat lesser degrees by aggression and absorption. These relationships are not due to co-occurring psychopathology, overlap with other traits, or mood-state biases, and they are verified by informants. Moreover, negative temperament predicts greater ASA 3 years later, adjusting for baseline ASA, which suggests that negative temperament may be a precursor of ASA, or predispose to it. This work may help clinicians anticipate traits that are associated with ASA in order to tailor treatments to patients’ personalities and lays groundwork for future research testing the mechanisms and causal links between these personality traits and ASA.