Poverty and Child Maltreatment Reports As Risk Factors for Childhood Injury Mortality

Background/Purpose:

Recent population-based research in California has indicated that children reported to Child Protective Services (CPS) have an increased risk of preventable injury death, one that is independent of poverty (e.g. Putnam-Hornstein, 2011). The conclusion from this line of research is that contact with the child welfare system serves as a secondary/tertiary prevention opportunity for children at high-risk of mortality. This research, however, has relied only on individual-level measures of poverty; other studies have indicated that community-level measures of poverty also contribute to children’s risk of injuries and mortality.  Thus, the specific aims of our study were to 1) Replicate the findings of Putnam-Hornstein (2011) in a different US state, and 2) To identify the “most likely” set of risk factors for childhood injury mortality, with a specific focus on the comparison of models with and without child maltreatment and poverty/deprivation variables.

Methods:

Data:  A prospective birth cohort method was employed including all children born in Washington State between 1999 and 2012.  Birth records were linked with CPS and death records, using deterministic and probabilistic methods.  Injury death before age five was indicated by ICD-10 codes and sociodemographic covariates matching the California study were included in the analysis.  We constructed a community-level deprivation measure according to the Association of Maternal & Child Health Programs measure for Concentrated Disadvantage using Census Block Group American Community Survey (ACS) data. 

Analysis:  We used Bayesian Information Criterion (BIC) to test model fits for models that include the CPS maltreatment report and the two poverty/deprivation measures.  Descriptive statistics using chi-square tests assessed differences in the distribution of the sociodemographic covariates among children who were fatally injured.  We then used event history analysis, specifically Cox Regression models, to test the association between deprivation and injury mortality.

Results:

In regards to the first aim, in the models that include the exact set of covariates used by Putnam-Hornstein (2011), the effects of CPS reports and covariates are very similar.

In terms of our second aim, we find that the model including both poverty measures but without CPS report (BIC: 8,689.27) was a better fit than the other models with different measures of deprivation/poverty and CPS report.  The best fitting model that included CPS report had a BIC of 12,069.37.  Both measures of poverty increased the risk of a child dying from an injury, with the individual measure at 1.63 (CI: 1.24, 2.15) and the community-level measure at 1.37 (CI: 1.09, 1.73) when controlling for the other covariates.

Conclusions & Implications:

Our findings suggest that poverty, both at the individual and community levels, are important risk factors for childhood injury mortality.  Further, our comparisons of model fit, indicate that at least in Washington State, a statistical model, which includes poverty and no CPS variables, is a better fit than a model that includes reports to CPS. Since the risk factors in this model are all known to practitioners at birth, these findings suggest earlier and community-targeted interventions could be developed as primary prevention strategies for childhood injury mortality.