Interpersonal callousness and co-occurring anxiety: Developmental validity of an adolescent taxonomy.

Anxiety

We used a measure of the likelihood of “any anxiety disorder” at age 13; this encompassed separation anxiety disorder, generalized anxiety disorder, specific phobia, social phobia, agoraphobia, and/or panic disorder. This anxiety score was generated from the Development and Well Being Assessment (DAWBA), originally developed for the British Child Mental Health Surveys ( R. Goodman, Ford, Richards, Gatward, & Meltzer, 2000 ). Using parent-reported symptoms, preliminary DSM–IV psychiatric diagnoses were generated using a well-defined computerized algorithm (see http://www.dawba.com ), producing six-level ordered-categorical “probability bands” for each disorder, ranging from <0.1% to >70% probability of diagnosis. These “bands” have functioned well as ordered-categorical measures when evaluated in two large-scale national samples, showing dose-response associations with mental health service contacts, and similar associations with potential risk factors as clinician-rated diagnoses ( A. Goodman, Heiervang, Collishaw, & Goodman, 2011 ).

Maternal psychopathology

Anxiety and depression in mothers were assessed by the Crown-Crisp Experiential Index (CCEI; Crown & Crisp, 1979 ) and the Edinburgh Postnatal Depression Scale (EPDS; Cox, Holden, & Sagovsky, 1987 ), respectively. Assessments were conducted at 18 and 32 weeks prenatally, and postnatally at 8 weeks, 8 months, 21 months, and 33 months. The anxiety subscale of the CCEI comprises eight self-reported items, measuring the frequency with which anxiety-related feelings and behaviors are experienced along a four-point scale ( never to very often ). The EPDS is a 10-item self-report questionnaire of depressive symptoms experienced in the last seven days. Latent variables combining depression and anxiety into overall “maternal psychopathology” were created for the prenatal and postnatal periods, with higher values suggesting greater psychopathology.

Harsh parenting

Disciplinary parenting practices were assessed by two items each at ages 2 and 4, asking the mother, “When you are at home with your child, how often do you do the following”: (i) shout at him/her; and (ii) slap him/her. The original response scale (1 = every day to 5 = rarely/never ) was reversed so that higher scores reflected harsher parenting. Resulting scores from both ages were combined into a single latent factor.

Emotional difficulties and low pro-social behavior

Also at ages 7, 10, and 13 years, emotional difficulties and pro-social behavior were measured using mother reports on subscales of the Strengths and Difficulties Questionnaire (SDQ), which has previously shown good reliability and validity ( R. Goodman, 1997 ). Subscales comprised five items each, rated on three-point scales ( not true , somewhat true , and certainly true ). To measure low pro-social behavior, pro-social SDQ items (“considerate of other people’s feelings,” “shares readily with other children,” “helpful if someone is hurt,” “kind to younger children,” “volunteers to help others”) were reverse coded, such that higher scores reflected the disregard for others and lack of empathy that form a key component of IC. Some of these items have previously been employed in assessment of childhood callousness, although it should be noted that commonly recognized components of broader IC are not included in this construct ( Viding, Blair, Moffitt, & Plomin, 2005 ; Whelan, Stringaris, Maughan, & Barker, 2013 ).

Child academic performance

National standardized test data were used to evaluate academic progress throughout primary education. Year-on-year progress of UK children is divided into “key stages,” with compulsory national tests at the end of each stage. For Key Stage 1, at the end of Year 2 (i.e., 6–7 years of age), English (reading, writing) and Mathematics are examined. For Key Stage 2, at the end of Year 6 (i.e., 10–11 years), tests of English, Science, and Mathematics are administered. Key Stage I and 2 scores were created by summing the national curriculum level scores (Levels 1–8) achieved for each subject.

Attrition and Missing Data

Participants with complete IC and anxiety data at age 13 were selected for analysis, resulting in a sample of 6,791 (49.99% boys). Using multivariate logistic regression, we tested the extent to which study variables predicted exclusion from the analytic sample. Odds ratios (ORs) showed that mothers excluded from the present analysis were more likely to experience postnatal family adversity ( OR = 1.05, 95% CIs [1.02, 1.09]). However, mothers included in analysis were more likely to experience greater adversity ( OR = 1.33, 95% CIs [1.25, 1.41]) and anxiety/depression during pregnancy ( OR = 1.07, 95% CIs [1.01, 1.14]), with their children more likely to show conduct disorder symptoms at age 7 ( OR = 1.22, 95% CIs [1.06, 1.40]). On all other study variables, included and excluded participants did not differ.

Statistical Analyses

Analyses were performed using SPSS Statistics version 22, Mplus version 7.11 ( Muthén & Muthén, 2012 ), and the mclust package in R version 3.2.1 ( Fraley, Raftery, Murphy, & Scrucca, 2012 ). Given the relatively large sample size (and high statistical power), we applied stringent significance thresholds throughout analyses; specifically, p < .001 for main effects and p < .01 for interactions. In reporting results, we first present significant three-way interactions, followed by two-way interactions, and finally main effects. Given the hierarchical nature of interaction terms, we refrained from discussing significant lower-ordered terms in the presence of significant higher-ordered terms; that is, we did not discuss two-way interactions that were nested in significant three-way interactions, and did not discuss main effects that were nested in significant two-way interactions. Effect sizes were interpreted using Cohen’s (1988) suggested guidelines. Cohen’s d (small = 0.2; medium = 0.5; large = 0.8) is reported for differences between two groups, and (partial) eta squared (η 2 ; small = 0.01; medium = 0.06; large = 0.14) is given for ANOVA main and interaction effects. Analysis comprised three steps:

Consistent with past adolescent studies ( Docherty et al., 2016 ; Kimonis et al., 2012 ; Kimonis et al., 2011 ; Tatar et al., 2012 ), we performed model-based cluster (MBC) analysis on IC and anxiety scores at age 13. Though not uncommon in this field (see Docherty et al., 2016 ; Euler et al., 2015 ), our cluster analysis included only two variables: one for IC and one for anxiety. A data-driven approach, MBC tests the relative fit of 10 models, which vary in their assumptions about the distribution of clusters (spherical, diagonal, or ellipsoidal), and whether clusters have equal or variable size, shape, and orientation in space ( Skeem et al., 2007 ). Within each of these models, the number of clusters is varied from one to nine; thus, 90 different cluster solutions are examined. Goodness-of-fit is determined by the Bayesian Information Criterion (BIC). Generally, the model with the lowest BIC value is preferred. Further conventions around BIC values in MBC are discussed elsewhere (see Raftery, 1995 ).

Step 2: Early risk exposure

We compared mean differences between resulting IC/anxiety groups on prenatal and early postnatal measures of family adversity, maternal psychopathology, and harsh parenting, using a multivariate analysis of variance (MANOVA) where all five risk exposures were entered simultaneously. We also investigated potential interactions based on sex.