Adam C. Carle, MA, PhD, Statistical Research Division, U.S. Census Bureau, Center for Survey Methods Research, Washington, DC 20233-9100, 301-763-1863, adam.c.carle@census.gov

The possibility exists that a measurement instrument may not be equally valid across groups. Measurement bias, a form of non-sampling error, occurs when individuals equivalent on true levels of a variable, but from different groups, do not have identical probabilities of observed scores. In public health, it is important to establish a measure’s equivalence across groups before comparing groups on the instrument. Bias can lead to inaccurate diagnoses, attenuate or accentuate group differences, and affect the validity and reliability of epidemiological studies. For example, previous research has suggested a differential course of depressive symptomatology across the developmental lifespan of boys and girls; despite similar rates through late-childhood, by young adulthood, females are nearly twice as likely to be depressed. However, data addressing the cross-gender validity of the instruments used in these studies are lacking, leaving unclear whether findings reflect true similarities and differences, or measurement bias. Latent variable models (e.g., confirmatory factor analysis, item response theory) are modern quantitative tools to examine bias, yet no studies have employed these methods in childhood measures of depression. As such, data from a larger, longitudinal study examining depression and anxiety among children were used to assess the validity of the Children’s Depression Inventory (CDI) across boys and girls in late-childhood, employing three latent variable models. Results supported the use of the CDI in cross-gender comparisons generally, and provided evidence that similarities in depression for this age group likely reflect true similarities. Further implications with respect to public health, prevention, and intervention are discussed.

Learning Objectives: