Using Bayesian Space-Time Models to Understand the Substance Use Environment and Risk for Being Referred to Child Protective Services

Abstract

Parental substance misuse has often been cited as a cause of children being referred for investigation of child abuse and neglect. Research on how the substance use environment might affect this relationship is still in its infancy with primarily only cross-sectional studies finding a positive relationship of alcohol outlet density at the level of neighborhoods and alcohol prices at level of states and maltreatment. A longitudinal study shows that increasing female drug-related arrests are related to increasing rates of maltreatment in rural and urban counties. The current study incorporates three aspects of the substance use environment in a panel study of 58 California counties over 4 years (n = 232) to study this relationship for referrals to child protective services (CPS) for child abuse and neglect. We use conditionally autoregressive (CAR) Bayesian models to model the spatial and temporal structure in the data. We find that use of welfare benefits, the number of outliers per population, and the number of drug-related arrests per population are positively related to referrals while unemployment and admissions to publicly funded alcohol and drug user treatment programs are negatively correlated to referrals. Significant spatial structure and space-time relationships are also found. The findings indicate that supply of alcohol and drugs (as measured by number of alcohol outlets and arrests for drug use and sales) may increase risk for being referred to CPS, but treatment for substance use does not increase the risk for referral.

Keywords:

• alcohol outlet density
• substance use environment

Notes

*The journal's style utilizes the category substance abuse as a diagnostic category. Substances are used or misused; living organisms are and can be abused. Editor's note.

¹A third model was attempted that would assess the effects of outlet type (bars, off-premise outlets, and restaurants) on referrals but because of the large positive correlations between the variables, the model would not converge.