Analysis of Extant Macro-Level Data on Child Maltreatment: County Maltreatment Rates in Texas As an Example

Background/Purpose: Efforts to prevent child maltreatment are underway across the nation, receiving attention from public and private organizations and funding from a limited pool of resources. These efforts are complicated by the difficulties inherent in measuring rates of maltreatment and by the limited availability of program data. The current project explores a variety of ways that publically accessible macro-level data can be used to augment more traditional analysis of maltreatment programs (e.g., qualitative interviews or self-report measures completed by caregivers) and offer a more comprehensive view of maltreatment that includes “high-level” variables.

A brief overview of extant data analysis in general will be given before an example of the process is detailed using county-level data from the state of Texas. Analyses include an exploration of the validity of using rates of use of diagnostic codes for child abuse in emergency room admissions as a proxy for child maltreatment rates at the county level and identifying matched control counties via propensity score for an intervention implemented to prevent maltreatment.

Methods: Data was accessed via the internet and cleaned using the statistical programming language R. Processes of data selection, variable manipulation, propensity score matching, and analyses will be detailed in the presentation.  Publically accessible data, including county population level, Medicaid enrollment, education and employment statistics, and emergency room admissions, were sourced from appropriate internet archives or through public data requisition procedures. Proprietary and private data provided by the state of Texas Department of Health and Human Services is also included as a comparison measure to demonstrate the validity of publically accessible data.

Results: Results from this project can be divided in to two portions: process and outcomes. Process results primarily indicate support for the possibility of using extant macro-level data to inform interventions and evaluations. More specifically, the process revealed lessons about data selection (e.g., where to find trustworthy data and how to get it), model design (e.g., how to best select variables for the set of research questions at hand), and presentation (how to effectively combine macro-level data with more traditional results for maximum impact). Outcomes from the example using county-level data from the state of Texas reveal both the promise of meaningful impact of county-level data and some important shortcomings.

Conclusions and Implications: Extant macro-level data has demonstrated potential utility in the realm of child maltreatment and may be a low-cost method of understanding the etiology of maltreatment and improving the evaluation of intervention efforts. Such analyses are possible with far fewer resources than are required for an original study, and methods and techniques for conducting them can be readily learned.