Making Meaning of Cumulative Child Welfare System Involvement

A new study by Putnam-Hornstein et al., in this issue of AJPH (p. 1157), is the latest of many to estimate cumulative incidence rates for child welfare system (CWS) involvement. Whereas previous work relied on synthetic life tables produced from the National Child Abuse and Neglect Data Systems (NCANDS) and the Adoption and Foster Care Analysis and Reporting Systems (AFCARS), this latest work uses linked administrative data from California. It is an important effort to assess the plausibility of previous estimates, and the scientific community will benefit from similar replications in other states. I focus on two issues: (1) discrepancies in estimated cumulative investigation rates, and (2) possible interpretations of the overall rates of CWS involvement.

The authors’ estimated prevalence rates for back-end CWS involvement (substantiated CWS investigations, foster care entry, and termination of parental rights) are roughly similar to previous national estimates, thus providing reasonable confidence about the sufficiency of NCANDS and AFCARS data for tracking those outcomes. However, Putnam-Hornstein et al. find a substantially lower rate of exposure to CWS investigations (∼26%) than the NCANDS estimates (∼37%).¹ They emphasize one possible explanation for the difference in estimates: that NCANDS cannot track children across states (and thus double counts children investigated in multiple states). By contrast, this study is limited to only California-based reports (and thus cannot account for California-born children who are investigated in other states). This explanation seems unsatisfactory for explaining an 11-point gap (a relative difference of 42%). California also does not differ greatly from the nation on investigation rates,² so the discrepancy in estimates raises important questions about whether NCANDS identifiers for children in unsubstantiated investigations are reliable (unique) in each state. Further assessment of this issue is warranted, given the attention that cumulative investigation estimates have received to date.

Notwithstanding, the new study likely provides a more accurate estimate of cumulative exposure to a CWS investigation and affirms previously estimated rates of downstream CWS involvement. The question, then, is what do we make of these findings?

Putnam-Hornstein et al. conclude that CWS has limited specificity—implying that levels of investigation are disproportionate to the occurrence of maltreatment and ensnare a large number of families for whom maltreatment did not actually occur or who do not require CWS involvement. We cannot draw this conclusion based on available evidence. Cumulative rates of CWS investigation may astonish many, but we should not discount the possibility that it is a reasonable approximation of the prevalence of child maltreatment or imminent risk thereof and that the comparatively low rates of substantiation and subsequent formal intervention indicate limited sensitivity.

Some calls to CWS are made simply because reporters do not know how else to connect families to services. Many of the 40% of referrals that are screened out without investigation each year² perhaps fall into this category, as, likely, does some proportion of referrals that are investigated. Predictive risk modeling, an area in which Putnam-Hornstein is a leader,³ is a promising strategy for identifying referrals that can be appropriately and safely diverted to voluntary community resources without an investigation.

Yet the sum of existing evidence indicates that investigations are a reasonable approximation of maltreatment exposure, regardless of the substantiation determination. Estimated rates of child maltreatment exposure derived from surveys and other study methodologies are not consistently lower than the rate of CWS investigation,⁴ and a CWS investigation is predictive of a range of adverse outcomes regardless of substantiation and after controlling for socioeconomic factors.^5–7 CWS decisions also occur in a context of high uncertainty, with conflicting or vague child disclosures, lack of direct physical evidence, limited powers to compel cooperation, inexperienced and poorly trained staff, political pressure to avoid foster care even in the highest risk cases, and limited resources to provide services. These factors imply a tendency not to substantiate or formally intervene even when maltreatment has occurred.

In contextualizing their findings, Putnam-Hornstein et al. further assert that CWS is designed to surveil many but serve few. However, is it true that most investigations result in no intervention? A key distinction here is informal versus formal intervention. CWS frequently provides, refers to, or arranges services or other interventions to children it declines to count as victims. For example, California reportedly provided services to nearly two thirds of children with unsubstantiated investigations in 2019.² CWS also relies heavily on informal kinship care as an explicit strategy for avoiding foster care.⁸

Are informal interventions limited to cases in which no maltreatment occurred or in which risk of harm is low? Probably not. Formal interventions (e.g., supervised in-home services, court involvement, and foster care) are comparatively rare by design, as the authors claim. Indeed, it is the expected result of a system that is required to investigate maltreatment but is legally and politically pressured to avoid usurping parental authority. Reliance on informal kinship care, voluntary services, and noninvestigative “assessments” to address maltreatment were originally the demands of advocates for CWS reform who wanted a more family-friendly entity.⁹ Acquiescence to these demands also serves the interests of government officials, who claim success in reducing child maltreatment on the basis of declining numbers of substantiated victims and foster care caseloads. Whether relying on informal interventions serves the interests of children is not clear. These estimates imply that a substantial proportion of children are born to parents who are, at some point, unable to provide minimally adequate care and for whom existing systems are inadequately resourced to respond.

Lastly, the study also finds disproportionate rates of CWS involvement for Black and Indigenous children, with larger disparities for back-end than front-end involvement. It is worth noting that the landscape of foster care changed drastically during the study period: Black children were 38% of the foster care population in 1999¹⁰ versus 23% in 2019.¹¹ Regardless, racial disparities certainly remain in various levels of CWS involvement. Because discourse on CWS generally, and foster care and termination of parental rights in particular, often centers on parents rather than children, overrepresentation is often equated to harm. But it matters that disparities in CWS investigation are generally consistent with estimated disparities in child maltreatment,¹² child fatalities,² and other metrics indicative of risk to children. It matters because making CWS the problem allows society to ignore or minimize the causes of these disparities—namely, the wholly inadequate efforts to overcome the legacy of de jure segregation, state-sanctioned violence, economic isolation, and discrimination that harms the health and wealth of Black and Indigenous families in the present. Reducing or eliminating CWS will not rectify these broader inequalities and may worsen them, given the intergenerational implications of child maltreatment.

This is not an argument that CWS is performing well—it is not. CWS appears to be frequently ineffective at preventing revictimization or mitigating its effects. Putnam-Hornstein et al. emphasize a lack of specificity in front-end functions as a critical issue undermining CWS effectiveness. Although that is likely at least somewhat true, it is not a complete picture. On the whole, evidence points to both some degree of overinvestigation for comparatively low-risk cases and substantial levels of underintervention in response to high-risk cases.