Understanding the Effect of the Opioid Epidemic on Child Maltreatment

The purpose of this project was to examine the association between county-level rates of opioid overdose events and child maltreatment indicators. Specifically, the project examined associations between county-level rates of opioid overdose events in Washington State, as measured by hospitalizations and deaths, with rates of Child Protective Services (CPS) (1) reports, (2) substantiations, (3) removals, and (4) maltreatment-related hospitalizations. The project built on and extended previous research in two ways. First, it focused on opioid overdose events specifically. Second, it used repeated and nuanced measures of opioid overdoses events, CPS involvement, and child maltreatment over an extended period of time to capture a more precise and accurate picture of temporal associations (Rebbe et al. 2020).

• Controlling for other factors in Washington State resulted in no relationship between opioid overdose events and child maltreatment indicators, despite numerous news reports stating there is a relationship between the opioid epidemic and child maltreatment. CPS reports and maltreatment-related hospitalizations increased over time, whereas CPS substantiations and CPS removals decreased in the project time period. 
• As the rates of county opioid overdose events increased (as measured by hospitalizations and deaths), children were less likely to go to their birth home from their birth hospitalization. Similarly, opioid-exposed infants were more likely to go to out-of-home care (foster care) in counties with higher rates of opioid overdose events.

To answer Research Question 1, the project staff used Bayesian Model Averaging to find the variables that were the best fit for each model with the four outcome measures (CPS report rates (regardless of screening decision), CPS substantiation rates, CPS removal rates, and maltreatment-related hospitalization rate). After identifying the variables to include in each model, the team ran an ordinary least squares model, followed by a fixed effects panel data model with year and county as the fixed effects. Project staff tested to see if the panel data model was a better fit than the ordinary least squares model, then ran a random effects panel data model and used a Hausman test to identify if the random or fixed effects model was a better fit.

To answer Research Question 2, the project identified children diagnosed at birth with opioid exposure through either the maternal or child ICD-9 diagnostic codes. Project staff ran a multistate survival model identifying three placement outcomes: (1) birth home, (2) out-of-home placement, and (3) hospitalization or death. They included a county rate of opioid-related hospitalization or death, concentrated disadvantage, presence of a CPS report, child sex, birth payment, maternal race, maternal age at birth, and parity as covariates. They completed chi-square analysis on a descriptive distribution table.

Mienko, J.A. and R. Rebbe. “Understanding the effect of the opioid epidemic on child maltreatment.” OPRE Report #2021-255, Washington, DC: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services, 2021.

Rebbe, R., A.S. Bishop, J. Ahn, and J.A. Mienko. “Opioid Overdose Events and Child Maltreatment Indicators: Differential County-Level Associations.” Children and Youth Services Review, 119, 2020, pp. 1—7. https://doi.org/10.1016/j.childyouth.2020.105671 .