Introduction

In New Mexico, where the present study was conducted, rates of substantiated child maltreatment is a significant public health issue (NM-IBIS, 2017). A recent report conducted by Child Trends found that New Mexico ranks dead last in the percentage of youth who experience multiple adverse childhood experiences (childtrends.org), which encompass child abuse and neglect, placing them in a category of particularly high risk. Moreover, New Mexico ranks below the national average on several key indicators of child welfare and has higher referral, substantiation and response rates compared to the national average. For example, in 2017, about 25 children per 1,000 were substantiated for abuse and neglect, compared to 9.4 substantiations per 1000 children nationally. Moreover, despite the relatively higher share of youth who are involved with Child Protective Services (CPS) in New Mexico, only 40% percent of victims receive post-response services – 24 percentage points less than the US average (https://www.nmlegis.gov/lcs/lcsdocs/197264.pdf). Adverse childhood experiences translate into a broad array of negative adult outcomes including sexually transmitted diseases, anxiety and depression, delayed cognitive development, involvement in sex trafficking and future violence perpetration and victimization (Reed, 2017).

This paper contributes to a growing body of research that uses Bayesian geostatistical modeling to link neighborhood characteristics and processes to the spatiotemporal risk of child maltreatment. Specifically, this paper contributes to the literature by:

• examining small area spatio-temporal trends in child maltreatment substantiation risk over a 9-year period;
• evaluating the role of multiple measures of neighborhood vulnerability on relative risk of substantiated maltreatment controlling for spatial and temporal heterogeneity; and
• highlighting the utility of region-specific surveillance for high risk subpopulations.

Bayesian Space-Time Modeling

Bayesian space-time models produce measures of uncertainty associated with risk estimates at each stage of the modeling process. Moreover, they are particularly useful when standard assumptions of stochastic independence are violated, i.e., if the data are spatially or temporally dependent. The INLA library of the R Statistical Package was used to perform the analysis.

Data

Data for this study came from the New Mexico Department of Public Health and the CDC’s Social Vulnerability Index.

ResearcherID (developed by Thomas Reuters and used in Web of Science) and Scopus Author ID (developed by Elsevier and utilized in Scopus) are two examples of these efforts. Whereas ORCID is “a platform-agnostic identifier,” ResearcherID and Scopus Author ID are connected to proprietary, subscription-based systems. (Quoted from the Library of University of Chicago)

Remains to mention: ORCID is free, platform independent and Open Source under the MIT License.

Results

The map below shows the probability exceedance > 2 for the study region.

Conclusion

To learn more about Bayesian statistics, watch this video!