The Role of County Poverty in Mental Health Need for Urban and Rural Children in a Child Welfare Sample Using Multilevel Modeling

Purpose:The purpose of this study is to determine how county level variables influence an individual child or youth’s need for mental health services in a child welfare population.  This study extends previous work on the variation in mental health service need and utilization by adding important county-level explanatory variables, building a multilevel model that allows the interaction between individual and county level variables to be explored and uses a large national, rather than regional, dataset to explore these interactions.   Multilevel modeling allows for the interactions between characteristics of the location and people to be explored rather than controlling for the geographic similarities that result from the sampling frame. 

Methods:  Utilizing data from Wave I of the National Survey of Child and Adolescent Well-Being II (NSCAW II), a sample of children age 5 and older who remained in the home following a report to child protective services due to maltreatment (n=1,380)  were examined.  The Primary Sampling Unit (PSU) was the Level 2 factor, individual children were the Level 1 factor.  Analyses were conducted using R 2.14.1 (package, lme4).  A cross-sectional generalized multilevel model was constructed that predicts the need for mental health services, as measured by a Child Behavior Checklist (CBCL) score at or above the borderline cutoff, based on the child’s gender, percent of county residents below the poverty line, the child’s family was below vs above the poverty line, the child was in a rural vs urban county and the interaction between county poverty and rural/urban residence.

Results:  Of the children and youth in this study, 20% had CBCL scores above the borderline cutoff of 60.  Nearly one-fifth (18%) were from rural areas, 43% were Caucasian, 25% Hispanic, 22% African American.  52% of the sample were female, 56% lived below the poverty level.  County poverty rates ranged from 4% to 28%.  

The intraclass correlation coefficient (ICC) indicated that approximately 5% of the variation in the odds of having a mental health need was a result of the second level of the model.  The final model included all of the above variables except for race and age.   The interaction between county poverty and county was rural vs urban, child’s gender and urban vs. rural residence were significant (p<.1) indicating a differential risk for a child depending on the specific combination of the variables for the child’s home county.  The effect of county poverty on the odds of having a high CBCL score was much stronger for urban children than rural regardless of whether the child’s family was above or below the poverty line.

Conclusion and Implications:  Most research on child welfare system focuses on individual or family level variables.  These results identify that the child’s environment (poorer vs. wealthier, urban vs. rural) impacts his or her mental health needs and that the characteristics of the child may interact with the county factors to change the mental health risk. Future research should incorporate techniques that can examine geographic characteristics, such as availability of resources and concentrated poverty.