Purpose: The purpose of this study was to examine predictors of psychotropic use and mental health services utilization, controlling for demographic characteristics, placement variables, health service utilization, and community mental health factors, in a statewide sample of children in foster care. By accessing administrative data from three systems, we are able to simultaneously consider the contribution of child, family, and system factors.

Methods: This study is based on a statewide sample of 8631 of children in foster care at least 90 days during fiscal year 1999. Analyses employed linked child records from child welfare, mental health, and Medicaid. We identified a sub-sample of children that had, during the observation year, either 1) been assigned an psychiatric diagnosis by a medical provider or 2) were prescribed a psychotropic medication. This procedure resulted in "clinically identified" sub-sample (n=2194). After examining the distribution of psychiatric diagnoses and medications among the clinically identified children, we compared their general characteristics to the children in the main sample that did not meet the clinically identified criteria (n=6437) with differences in means and proportions identified through t-tests. Finally, we employed a set of logit models to assess the relative influence of child demographic factors (age, sex, race/ethnicity), foster care history, medical care, and psychiatric diagnosis on 1) the estimated likelihood of utilizing of any one of four types of community mental health services during the current year, and 2) the estimated likelihood of community mental health system utilization during preceding years. To adjust for differences in the effectiveness among local community mental health systems in serving foster children, we included in the logit models the general foster child utilization rate specific to each community mental health system.

Results: Bi-variate analyses revealed that the clinically identified children were older, and more often male and Caucasian. Their foster placement histories showed a pattern of earlier, longer, and less stable placements. Examining the reasons for placement, clinically identified children were more likely to have a history of sexual abuse and behavioral problems. The clinically identified children also were higher utilizers of medical care than their counterparts without identified psychiatric problems. To explore the independent effects of race and the other descriptive variables, a logit procedure was run with clinical identification as the dependent variable. This analysis showed that being male, Native American, having a history of alleged sexual abuse or being placed due to behavioral issues, having a special payment for placement, and a greater number of physician visits, were significant predictors of being clinically identified We also examined predictors of type of mental health services utilized. Controlling for regional differences in the mental health system, five factors were positively and significantly related to all types of treatment receipt of special foster care payment, emergency room care, being clinically identified, diagnosed with oppositional defiance disorder, and number of days on psychotropic medications.

Implications: The implications of the findings for mental health treatment of this highly vulnerable population are discussed with particular attention to the issues of racial disproportionality and gender.