Evaluating maternal depression-related bias in reports of child temperament: An investigation of measurement invariance

• In Event: 1-210 - Improving the Quality and Efficiency of Measurement in Early Life

Integrative Statement

Numerous studies have examined the influence of parental psychopathology, particularly maternal depression, on child characteristics. This literature has focused on the influence of maternal depression on child temperament in early development. Maternal depression is often assessed using semi-structured or fully structured diagnostic interviews and child temperament is frequently assessed using parent-report measures. A frequently noted limitation of this work is that the experience of maternal depression is associated with mothers’ systematic biased for identifying behaviors that may place their children at risk for psychopathology. This assumes that the items reflecting these dimensions of temperament may function differently between mothers with and without a history of depression. All previous studies have examined mean-level differences in youth of mothers with and without depression without empirically examining whether the measurement properties of the items differ. To examine this possibility, the present work conducts tests of measurement invariance (MI) across reports of child temperament between mothers with and without a lifetime history of depression.
Data for this work come from two community-based studies of the development of internalizing problems. One site was from London, Ontario, Canada and the other Stony Brook, New York, USA. Each study site used the Structured Clinical Interview for DSM-IV to assess lifetime history of depression and the Child Behavior Questionnaire (CBQ) to assess child temperament. Complete data were available on 906 mothers and their children. There were 271 mothers with and 635 mothers without a lifetime history of a unipolar depressive disorder.
Our analyses examined each individual CBQ subscale individually. We examined overall model fit in the combined sample to evaluate the appropriateness of fitting single factor models to the scales. For models that achieved at least adequate model fit (Comparative fit index > .90 and root mean square error of approximation ≤.09), we continued to test MI across mothers with and without a lifetime history of depressive disorders.
Of the 16 primary CBQ scales, two (Impulsivity and Sadness) failed to demonstrate adequate unidimensional model fit. Thus, they were not subjected to tests of MI. We evaluated configural (common form), metric (equal factor loadings), and scalar (equal intercepts) invariance for the remaining 14 scales. In these models, the Approach Anticipation and High Intensity Pleasure scales failed to demonstrate adequate fit for their configural models; thus, further tests of invariance were not conducted. However, for the remaining 12 scales, imposing constraints on factor loadings and on intercepts demonstrated that metric and scalar invariance were supported.
Our results provide good evidence that reports of youth temperament via the CBQ are not biased due to maternal experience of depressive disorders. Our finding that MI is generally supported, broadly, indicates that mean-level comparisons between mothers with and without depression reflect true differences, as opposed to reflecting bias in responses. These results strengthen conclusions of previous studies relying on maternal reports of child temperament. Future work should examine the specific roles of more insidious (e.g., chronic, recurrent) forms of depression and contemporaneous dimensional levels of depression as additional influences on reports of child temperament.

Authors

• Thomas M Olino, Temple University, Department of Psychology
  Presenting Author

• ELizabeth P Hayden, Western University
  Non-Presenting Author

• Dan Klein, Stony Brook University
  Non-Presenting Author