Purpose: Large administrative electronic databases now make it possible to conduct cross-sector longitudinal analyses of child abuse and neglect cases (e.g., Jonson-Reid, Drake, Chung, & Way, 2003), however, existing methods restrict hypothesis testing to specific pathways. This is problematic when questions concern the service system as a whole. Individuals can cycle through services and interactions emerge between services and pathways. New methods that capture these complexities are needed in order for researchers to advance empirical knowledge of service systems. The purpose of this research is to demonstrate the feasibility of a new method for identifying and analyzing cross-sector service pathways.

Methods: A procedure was developed for analyzing sequences of events and states from date variables in an administrative data base. The resulting sequences represented individuals going through multiple transitions and service systems (e.g., child abuse or neglect report, receiving income support, entering special education). These were then analyzed to identify and characterize cross-sector service pathways that cycled. That is, service pathways where an individual returned to some previous state. The procedure was tested on a merged database containing information on approximately 10,250 children and 199 date variables (e.g., health, mental health services, social services, runaway shelters, special education, and family courts). Cycle metrics (total number of cycles, number of distinct cycles, total time to travel a cycle, number of transitions in a cycle etc.) were then calculated for each child. General linear models were used to test hypotheses of who would be more likely to cycle and associations between child characteristics and cycle metrics.

Results: A total of 770 unique states were generated contributing to 103,734 transitions between services and events. This led to 18,126 identified cyclic service pathways. Of the 10,250 children in the data set, 71% had one or more cycles with 45% having two or more cycles. Children entering the system with reports of child abuse and neglect were more likely to have cycles than their AFDC only comparison group. Child's sex or substantiation of first reports did not increase the risk of cycling. Of the children with two or more cycles, 41% had non-repeating cycles, meaning that their service pathways changed over time, while the other 59% had static cycles. Boys and children entering the system from child abuse or neglect reports also tended to have an increased number of distinct cycles. The procedure was computationally efficient and could be run incrementally, making it feasible to conduct regular analyses of large administrative databases for decision support.

Implications: This method will allow researchers to develop better empirically driven models of service systems that do not exclude cyclic cross-sector service pathways. Such models could be used by administrators and policy makers to characterize the overall state of a service system, study the implications of policy changes, and potentially used in real-time decision support for caseworkers to track the trajectory of clients through multiple services.

Jonson-Reid, M., Drake, B., Chung, S., & Way, I. (2003). Cross-type recidivism among child maltreatment victims and perpetrators. Child Abuse & Neglect, 27, 899-917.