Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Cooperating University

University of Kentucky


Social Work

Committee Chair

Van Zyl, Michiel A.

Author's Keywords

Item response theory; Externalizing behavior problems; Early childhood; Primary care; Screening; Children; Pediatric


Behavior disorders in children--Diagnosis


Externalizing behavior problems in very young children are associated with an array of negative and costly long-term outcomes. Pediatric primary care is a promising venue for implementing screening practices to improve early identification of this social and public health problem. In this setting, screening requires a brief, easily scored instrument which can detect sub-clinical to clinical levels of the latent construct within the context of early childhood development. Further, items used should perform consistently with children of all sociodemographic backgrounds. This study applied item response theory analyses to investigate the precision, utility, and differential item functioning (DIF) of items measuring externalizing behavior problems in two caregiverreport questionnaires: the PSC-17 (Gardner et al., 1999) and the BPI (Peterson & Zill, 1986; Zill, 1990). Caregivers (N = 900) of children ages 3 to 5 responded to both instruments and a sociodemographic questionnaire in the waiting rooms of four pediatric primary care clinics. Sociodemographic characteristics of the children were diverse: 47% were female, 50% were of minority race, and 43% were of low socioeconomic status (SES). Eighteen items comprising the instruments' combined externalizing subscales were evaluated for (a) levels of externalizing behavior problems best measured, and (b) DIF exhibited by child sex, race, and SES. Samejima's (1969) graded response model was fit to the data, and two methods of DIF-detection were employed. Estimation of item parameters allowed consideration of the levels of externalizing behavior problems at which each item was most informative. Five items were found to measure only low to average levels of externalizing problems in the target population, while the remaining 13 were informative at sub-clinical to clinical levels. Significant DIF was detected in 8 of 18 items by child sex, race, or SES. A set of 4 items was identified which (a) provided the most information at sub-clinical to clinical levels of externalizing behavior problems, and (b) exhibited the least amount of DIF by child sex, race, and SES. These items may constitute a promising tool for screening purposes with preschool-aged children in the primary care setting, potentially improving early identification of very young children with externalizing behavior problems.