Putting Agency Back in Child Welfare: A Multilevel Analysis of Agency Effects on Youth Sexual Initiation.

Introduction: Compared to other adolescents, youth involved with the child welfare system experience a disproportionate burden of the fully preventable morbidity and/or mortality resultant from youth sexual risk taking, such as adolescent pregnancies and/or sexually transmitted infections. Although the role of individual or familial factors in contributing towards sexual health disparities between child welfare system (CWS) involved youth and their peers has been well documented, it is not yet known whether the characteristics of the local child welfare system and/or child welfare agencies independently influence the sexual or reproductive health outcomes experienced by CWS involved youth, i.e., is there an agency effect?

This research study thus makes the following original contributions to the fields of child welfare and public health: it represents the first known multilevel analysis of the contextual effects of child welfare agency characteristics upon sexual risk behaviors among a nationally representative sample of CWS involved youth; it also provides the first known empirically derived typology of child welfare agencies in the United States, as well as an exploration of whether child welfare agency type is independently related to youth sexual initiation.

Ultimately, this study demonstrates that child welfare agencies have agency to significantly alter the sexual risk trajectories of the youth they serve.

Methods: In the first part of this study, data from the National Survey of Child and Adolescent Well-Being (NSCAWII) were used to perform hierarchical generalized linear regression modeling (HGLM) in SAS to determine whether child welfare agency characteristics were independently related to the likelihood of youth sexual initiation between Wave 1 and Wave 3 of the NSCAWII survey. The final multilevel sample consisted of 80 child welfare agencies matched to 4,436 children. For the purpose of subpopulation analysis, the child level sample for this study was further statistically restricted, using the zero-variance method, to all children aged 11 years old or older at Wave 3, who did not previously report being sexually active at Wave 1 (n=996 children).

In the second part of the study, exploratory factor analysis was performed to first operationalize the construct of a child welfare agency, and all NSCAWII participating agencies were then grouped into unique clusters on the basis of these dimensions. Discriminant analysis was then used to validate the final obtained clusters (i.e., agency types). Finally, HGLM analysis was repeated, but this time to examine whether child welfare agency type (as opposed to specific agency level variables) was significantly related to the likelihood of sexual initiation among the youth in this study.

Results: Nearly thirty percent of the youth included in this study became sexually active between Wave 1 and Wave 3 of the NSCAWII survey (n=249). Of them, 66% reported having multiple sexual partners (n=159), and 16% became pregnant between Wave 1 and Wave 3 (n=246). In this study, child welfare agency level factors independently accounted for 19% of the overall variation in the likelihood of sexual initiation among youth by Wave 3 of the NSCAWII survey (p-value <=.01). In other words, there was an ‘agency effect.’ In the first part of this study, HGLM analysis revealed that youth served by child welfare agencies that did not make referrals for tutoring services had nearly three times higher odds than other youth of becoming sexually active between Wave 1 and Wave 3 of the NSCAWII survey (OR=2.71, CI [1.18-6.18], p-value <=.02).

In the second part of this study, the results of the factor analysis indicated the presence of four latent dimensions essential for operationalizing the construct of a child welfare agency: a) Post-Investigation Services b) Child Welfare Casework c) Organizational and Social Context and d) Agency Operations. Once NSCAWII participating agencies were grouped on the basis of these dimensions, the use of cluster analysis indicated the presence of three distinct types of child welfare agencies: Type 1- Rigid and in Flux (n=42 agencies), Type 2 - Close-Knit and Centralized (n=12 agencies), and Type 3 -Decentralized and Collaborative (n=26 agencies). This study found that youth served by Type 1 agencies (n=460 youth) had three times the odds of becoming sexually actively between Wave 1 and Wave 3 of the NSCAWII survey as compared to youth served by Type 3 agencies (n=400 youth, OR=3.23, CI [1.05-9.92], p-value <=.04).

As compared to other types of agencies, Type 3 agencies were more likely to be: a) large, with moderate turnover b) highly de-centralized and collaborative c) had a moderate level of worker reported co-worker and supervisor support and d) had moderate levels of worker reported organizational rigidity. Of particular note, Type 3 agencies were different from other types of agencies not only with respect to the extent of their collaboration but, also, with their greater propensity to involve former or current CWS involved biological parents in their policy making groups or in the training of agency caseworkers.

Conclusion: The single most important message of this study is that child welfare agencies have agency to proactively protect the youth they serve from experiencing the disproportionate burden of morbidity and/or mortality resultant from youth sexual risk behavior. As such, child welfare agencies are strongly encouraged to include adolescent sexual health outcomes and youth sexual risk behaviors as additional metrics in their existing definitions of essential child well-being indicators, and to continually monitor and evaluate agency efforts with respect to these metrics on a routine basis.

Additional research is also recommended to determine whether child welfare agency effects also exist with respect to other important youth outcomes, such as homelessness, educational attainment, or adolescent physical or mental health. Finally, child welfare agencies are also strongly advised to build upon the findings of this study by engaging in increased partnerships with CWS involved biological parents who, despite their involvement with the child welfare system, may still have great potential to serve as important collaborative partners in improving youth outcomes.