Executive functions as a mediator of childhood maltreatment on adult psychopathology: A longitudinal mediation analysis comparing maltreatment factor models

Abstract

Background: Childhood maltreatment (CM), executive functions (EFs), and psychiatric disorders all correlate highly. Changes in EFs during adolescence related to CM present a possible mediating mechanism for the development of psychiatric disorders, yet no study has analyzed this longitudinally while comparing predictive capacity of different CM factor structures. We hypothesized that changes in EFs from adolescence to adulthood would mediate, in part, associations between CM, internalizing disorders (INT), and anti-social personality disorder (ASPD) while different subtypes of CM would differentially predict INT and ASPD.

Objective: This study longitudinally examined the mediating effects of EFs on associations between CM, INT, and ASPD while comparing prediction of two CM factor structures.

Participants: High-risk subjects selected for drug use in adolescence (N = 658) from mean ages 16 to 23.

Methods: A Bayesian structural equation model was deployed to analyze change in EFs as a mediator of the relationship between CM and adult INT and ASPD. CM was measured using two factor structures: a single overall factor and four correlated factors representing CM subtypes.

Results: CM significantly predicted INT and ASPD but there was no evidence that the relationship was substantially mediated through EFs. High correlations among subtypes of CM limited the unique predictions of each subtype on INT and ASPD.

Conclusion: In this high-risk sample, the collinearity of CM subtypes obscured their predictions of outcome measures supporting the use of one CM factor. EFs did not significantly mediate associations between CM and psychiatric disorders, but further research on these relationships is warranted.

Keywords: Bayesian structural equation model; Childhood maltreatment; Executive functions; Internalizing disorders; Mediation.

Fig. 1.

Shows the underlying structural equation model for the mediation analysis of Model 1 and 2 along with path coefficients. Boxes around variable names indicate observed variables, circles indicate latent variables. P-IQ = Performance Intelligence Quotient, CARI-Q = the Colorado Adolescent Rearing Inventory Questionnaire, which is our maltreatment measure, CM = Child Maltreatment, NEG = neglect, PSYC = psychological abuse, PHYS = physical abuse, INT = internalizing disorders, w1 = wave-1, w2 = wave-2. Covariates of sex and age at each wave were included in the models but are not shown so the figures are not cluttered. The dots between CARI-Q 1 and CARI-Q 20 represent all the CARI-Q questions which weren’t able to be shown. The loadings of the four neglect and abuse factors are replaced by their correlations for visualization purposes. NEG takes CARI-Q 1–3, PSYC takes CARI-Q 4–10, PHYS takes CARI-Q 11–15, and SEX abuse takes CARI-Q 16–20. Solid lines and bold font mean p < .001. Dashed lines and regular font mean the path was not significant.