The Title IV-E Prevention Services Clearinghouse: Fatally flawed like the law that created it

by Marie Cohen

In Home Visiting: More Hype than Hope, I wrote about the failure of decades of studies to prove that home visiting programs have real-life outcomes for children and families. In this post, I explain how the federal government promotes these and other programs as interventions to prevent foster care despite the lack of meaningful evidence for their success. A “Prevention Services Clearinghouse” established by federal legislation approves programs for reimbursement based on the results of outcome studies, with virtually no attention to the quality of the measures used, the lack of corroboration, the total weight of the evidence, or the logic and size of the impacts. There is little evidence to support the effectiveness of these programs in making it possible for children who are maltreated (or at risk of foster care for another reason) to remain safely in their homes. Congress should consider eliminating the Clearinghouse and changing the type of services services eligible for federal funding.

The stated purpose of the Family FIrst Prevention Services Act (FFPSA),¹ which was signed by President Trump in 2018, was to “provide enhanced support to children and families and prevent foster care placements through the provision of mental health and substance abuse prevention and treatment services, in-home parent skill-based programs, and kinship navigator services.” To achieve this purpose, FFPSA expanded the use of Title IV-E funds from foster care and adoption to what it called “foster care prevention services and programs” designed to prevent the removal of children from their homes.² These programs were nothing new; at the state level, they are often called in-home services, family preservation services, or intact family services. Under FFPSA, these services are available to children who are “candidates for foster care” [meaning the child is at imminent risk of entering foster care but can safely remain at home or with a relative with the provision of services provided under the Act³] and to pregnant or parenting foster youth and the parents or kin caregivers of these children.⁴

Under FFPSA, these “foster care prevention services” must be provided “in accordance with practices that meet the requirements for being promising, supported, or well-supported practices.” Among the requirements, the components of the practice must be described in a book or manual, there must be no evidence suggesting the program does more harm than good, and when there are multiple studies of one program, the “overall weight of the evidence” must support the benefits of the practice. Each practice must be found to be “superior to an appropriate comparison practice using conventional standards of statistical significance in terms of demonstrated meaningful improvements in validated measures of important child and parent outcomes, such as mental health, substance abuse, and child safety and well-being.”

To receive the highest rating of “well-supported,” a practice must have data from at least two randomized controlled trials (RCT’s), or quasi-experimental research designs if RCT’s are not available. At least one of the studies must demonstrate that the practice has a sustained effect lasting at least a year. The requirements are somewhat lower to be a “supported” practice and lower still to be a “promising” practice. In order for a program to be rated as “supported” or “well-supported,” the supporting studies must have been conducted in a “usual care or practice” setting, which is defined as an existing service provider that delivers substance abuse, mental health, parenting or kinship navigator services as part of its typical operations.

At least 50 percent of a state’s foster care prevention services funding must be for “well-supported” programs in order to claim federal support under Title IV-E. Thus it is not surprising that states have chosen to include mostly “well-supported practices” in their Title IV-E Prevention Plans.

In order to further define the requirements and apply them to existing programs, FFPSA created the Title IV-E Prevention Services Clearinghouse (“the Clearinghouse” from now on), which is run under contract by Abt Global, formerly Abt Associates. The Clearinghouse has published a Handbook of Standards and Procedures describing the methodology it has developed to evaluate whether a program meets the requirements established by FFPSA.⁵ Program ratings are based on what the handbook calls “contrasts.” A contrast is defined as “a comparison of an eligible intervention condition to an eligible comparison condition on a specific outcome for a specific posttest measurement.” A “contrast” might be a comparison of the number of substantiated child abuse reports for the experimental vs the control group in one study, for example. A contrast must be statistically significant based on conventional standards in order to be considered favorable.⁶

The Clearinghouse staff has interpreted the requirements written by Congress in a way that sets a very low bar for a program to be judged “well-supported,” “supported,” or “promising.” To be “well-supported,” the Clearinghouse requires no more than two contrasts from different studies that show favorable effects.” At least one of the contrasts must demonstrate a “sustained favorable effect” of at least 12 months. This is a low standard for many reasons.

A program can be rated “well-supported” with as few as two positive contrasts, even if they are wildly outnumbered by contrasts that fail to show an impact. Even within a specific category (such as parental drug abuse, for example), one favorable contrast is enough, no matter how many other related indicators from the same or other studies show no effect.
Any positive contrast is counted, regardless of whether it is based on objective indicators or subjective ones like self-reports. Improvements on subjective measures are counted even if objective measures fail to find impacts.
Any pattern of impacts (or contrasts) is accepted, regardless of whether it is predicted by the program’s theory or logic model. If a program is supposed to work by changing parental behavior, but it instead changes children’s behavior directly, the outcome is counted. There is no requirement that a program impact be explained by its logic model or theory.
There is no requirement that a specific contrast reported by one study be corroborated by another study. Many of the individual impacts that are reported are supported by only one study, with each study cited providing evidence of a different impact.
Any contrast that is statistically significant is accepted, regardless of whether it is of practical significance. The American Statistical Association has stated that “[S]tatistical significance is not equivalent to scientific, human, or economic significance.” A tiny effect can be statistically significant if the sample size or measurement precision is large enough.
Studies are counted even when the specific populations studied are very different from those eligible for Title IV-E Prevention services.

Examining the Clearinghouse evidence on some of the most popular programs clearly reveals the flaws of its methodology, providing multiple examples of the problems noted above.

Parents as Teachers is the most popular program rated by the Clearinghouse, having been included by 31 states in their Title IV-E Prevention plans as of August 2024. PAT is a home visiting program that describes its mission as “To promote the optimal early development, learning and health of children by supporting and engaging their parents and caregivers.” The rating of “well-supported” was based on only three studies, one of which was not conducted in a “usual care and practice setting” and another of which was conducted in Switzerland–with a very different population from that of the U.S. The Clearinghouse found three favorable contrasts (out of six total contrasts) on child social functioning, all from one study, along with one unfavorable contrast and two showing no effect. They found two favorable contrasts on child cognitive abilities (one from the Swiss study) compared to 10 findings of no effect). And they found two small favorable contrasts on child welfare administrative reports (with two showing no effect) from a comparison group study that was not done in a “usual care or practice setting,” Only one study reviewed looked at the quality of parenting or the home environment, and it found no favorable effects. The authors of that study said that the results “raised questions about the underlying premise of PAT that focusing services on parents to improve parenting knowledge, attitudes, and behaviors is an effective way to benefit children.” They also concluded that their results (including the three positive contrasts on child social functioning and one on child cognitive functioning) “are consistent with the overall research base for family-focused early childhood programs, which have produced ‘modest and inconsistent effects.'” (ee Appendix A for more detail on all the contrasts discussed here).

Functional Family Therapy (FFT) is second in popularity only to Parents as Teachers, being included in the plans of 25 states as of August, 2024. As described by the Clearinghouse, FFT “aims to address risk and protective factors that impact the adaptive development of 11 to 18 year old youth who have been referred for behavioral or emotional problems.” The Clearinghouse rated FFT as “well-supported” based on the results of six studies. The Clearinghouse reported two favorable contrasts, 23 contrasts showing no effect, and one unfavorable contrast on Child Behavioral and Emotional Functioning–the core goal of the program. It showed two favorable contrasts, two unfavorable contrasts, and 16 contrasts showing no effect on child delinquent behavior. No contrasts showed an effect on positive parenting practices and only two out of 13 contrasts showed an favorable effect on family functioning. Ten contrasts showed a favorable impact on children’s substance abuse, compared to eight findings of no effect. But all of these favorable contrasts came from one study of alcohol-abusing youths who resided in a shelter after running away–a very specific population that may not be generalizable to children abusing other substances or those who are living at home.

Like FFT, Motivational Interviewing (MI) is included by 25 states in their Family First plans. MI, according to the Clearinghouse, is a” method of counseling clients designed to promote behavior change and improve physiological, psychological, and lifestyle outcomes.” The Clearinghouse reviewed studies of MI focused on illicit substance and alcohol use or abuse among youth and adults, and nicotine or tobacco use among youth under the age of 18. These programs are typically delivered in one to three sessions with each session lasting about 30 to 50 minutes. The Clearinghouse rated MI as “well-supported” based on results from 21 studies. The contrasts reviewed showed no effects on child substance use, caregiver mental health, caregiver criminal behavior, family functioning, parent/caregiver physical health, or economic and housing stability. The only favorable effects were on parent/caregiver substance abuse, for which there were sixteen favorable contrasts, two unfavorable contrasts, and 91 showing no effect. Eleven of the favorable contrasts came from one study of heavy-drinking college students–a very different population from parents of children at risk of foster care placement; presumably few if any were parents at all. It was also not conducted in a “usual care or practic setting.” It is no surprise that a program of one to three sessions would not lead to major changes in parents’ and children’s lives, but it is strange that serious analysts would accept these results as proof that such a minimal program would change lives.

The Clearinghouse rated the Healthy Families America (HFA) home visiting program (included by 22 states in their Title IV-E Prevention Plans) to be “well-supported” based on the results of six studies. Healthy Families America was developed as an intervention to prevent child maltreatment and is now the signature program of an organization called Prevent Child Abuse America, as I discussed in an earlier post. Yet, HFA’s effects on child maltreatment are decidedly underwhelming. The Clearinghouse found five positive contrasts on self-reported measures of maltreatment and no favorable contrasts on 99 more objective measures like CPS reports, hospitalizations and injuries. Of the 16 remaining favorable contrasts for the entire program, five come from improvements in parental reports of child behavioral and emotional functioning (a self-reported item with possible bias), another was on the child’s self reports about skipping school “often,” and none were corroborated by another study.

Some readers might observe that the standard critiques of the clearinghouse focus on the difficulty, not the ease, of meeting its standards. Critics like The Child Welfare League of America have stated that the “high evidentiary standards for the clearinghouse require rigorous, costly research that many states, Tribes and programs are unable to fund. This barrier is particularly onerous for programs that focus on underserved populations, such as Black and Native families, children of more than one race, and rural communities.” This assertion is not inconsistent with my conclusions about the Clearinghouse. That is because there are two types of standards. The standards for the rigor of the studies themselves may be too high for many programs to meet. But once a study meets the requirements to be reviewed, the requirements for being “well-supported” are almost laughable.

Another problem with the Clearinghouse is that many of the programs included are generally paid for by Medicaid, private insurance, or other state and federal programs, such as the federal home visiting program. And Title IV-E is required to be the payer of last resort for these services so it cannot pay if another source is available. The idea was apparently that Family First could be used to supplement these sources for families without private health insurance or where these funds are not available, but this does not seem to be occurring on a large scale, as I explain below.⁷

Finally, the Clearinghouse does not include what may be the most important foster care prevention service of all–case management. All of the services in the clearinghouse depend on relationships, but it is the relationship with the social worker managing the in-home case that may be the most important intervention for a maltreating parent. It is the case manager who refers the client to the parenting, substance abuse and mental health programs prescribed by the Act and who maintains contact with the programs to monitor the parent’s participation and progress. The case manager is responsible, through home visits, for monitoring the safety of the child or children who have not been placed in foster care. Without such monitoring, there can be no foster care prevention services.

Currently, case management is treated as an administrative cost under Title IV-E, which means it cannot be paid for unless the client is receiving other services that are supported by the Clearinghouse. At least one agency, the District of Columbia’s Child and Family Services Agency, has found a way to adapt one of the “well-supported practices” listed in the Clearinghouse as a case management model, allowing it to claim Title IV-E funding for case management for all children receiving foster care prevention services. CFSA adapted the practice of Motivational Interviewing (MI, discussed above), which is in the Clearinghouse as a substance abuse intervention, as a model for case management for all in-home services. As one account puts it, the approval of this use for MI was “particularly notable because while [MI] was approved by the Title IV-E Clearinghouse for Family First reimbursement only as a substance abuse service, DC received approval to implement and claim for it as an integral component of CFSA’s case management practice for all families.” But it should not be necessary for states to go through this type of charade in order to obtain Title IV-E funding for their case management services.

All of the problems mentioned above may explain why states are not drawing down large amounts of Title IV-E funding for foster care prevention programs. The federal government spent only $182 million reimbursing states for Title IV-E prevention services in FY2024, serving only about 18,300 children per month. A recent federal report showed that reported reimbursement claims on Title IV-E prevention services constituted less than two percent of overall Title IV-E program reimbursement claims in FY 2023, serving about two percent of the children receiving Title IV-E funded services.

What can be done?

Raising the standards for “promising,” “supported,” ‘well-supported” to be meaningful is simply not a viable option. Adjusting the three ratings to incorporate the quality of the measures, the need for corroboration, the weight of the evidence and the logic and size of the impacts, would probably mean that few if any programs would be classified in the top tier. So there is not really a way to classify programs based on evidence that will work well for the purpose of funding foster care prevention programs in Title IV-E.

Perhaps this is not surprising. The concept of evidence-based practice was adapted from medicine. When applied to social services, the concept has many limitations. Research conducted under controlled, small-scale, well-funded conditions, even if conducted in a “usual care and practice” setting is often not applicable to the messy, underfunded world of social services practice. It is well-known that effect sizes often shrink when a small pilot program is expanded to cover a larger population. Moreover, a study conducted on one population may not be generalizable to other groups. The population of parents at risk of losing their children to foster care is distinct from many other populations included in the studies mentioned above. The antecedents of child maltreatment, substance abuse and mental illness are extremely complex and may go back for generations. The idea that a three-session, three-month or even a three-year program can eliminate these problems may be unrealistic.

Finally, in relationship-intensive services like all of those included in the Clearinghouse, the characteristics of the practitioner probably matter more more than the specific model. In psychotherapy, multiple studies have concluded that “who your therapist is matters more than the model they use.” Ordinary people dealing with mental illness or addiction do not usually look for an “evidence-based practice.” They look for the best provider they can find–the one they have heard by word-of-mouth, or by scanning the internet for ratings. As mentioned above, the quality of the therapist matters more than the model, and the same applies to program staff. When I was a foster care social worker, and we had a child who had complex mental health needs, we used available funds to reimburse a top-notch provider who did not accept Medicaid. I don’t think we ever talked about what “model” they offered, and they met with clients for much longer than than program manuals prescribe. Medicaid itself does not require proof that a program is evidence-based.

In light of these issues, Congress should consider eliminating the Title IV-E Clearinghouse, as proposed by the Bipartisan Policy Center’s child welfare working group in its recent Blueprint for Child Welfare Financing and Accountability Reform.” The group of 15 child welfare experts representing different professional experiences and perspectives agreed that the clearinghouse process for approving programs as evidence-based “is cumbersome, is idiosyncratic, and impedes states from meeting the needs of families….” The group recommended replacing the clearinghouse with an expert panel convened by the U.S. Department of Health and Human Services that would consider proposals from the states for programs they want to provide using Title IV-E funds. States would be required to submit “evidence to support the efficacy of the interventions, and data that demonstrates the why states believe that proposed interventions would reduce child welfare involvement or improve outcomes of child welfare-involved children and families.”

Congress should also change the definition of the “prevention” services that can receive federal reimbursement under Title IV-E. Instead of continuing to pay for programs that belong to the field of mental health and substance abuse or are generally funded by Medicaid or private insurance, Title IV-E funds should be used for the critical service provided either directly or under contract by child welfare agencies–case management. Agencies should not have to go through the contortions that DC’s Child and Family Services Agency went through to adapt a three-session substance abuse cessation program into a case management model only to receive reimbursement from Title IV-E. This is the main service that they provide directly and it should be funded. Congress could also add a provision for states to be reimbursed to provide services mental health or drug treatment for the small number of parents who do not have public or private coverage for such services–in other words, where a payer of last resort is needed.

In drafting and passing FFPSA, Congress hoped to make new funds available to programs that would ameliorate the problems that were putting children at risk of being removed from their homes. In its effort to ensure that effective programs were funded, Congress instead created a process that awards ratings to programs based on the mindless and mechanical application of standards that mean little–and that fails to unlock the hoped-for source of funding for foster care prevention. The law must be changed in order to fulfill the intent of its framers.

Notes

See Title VII of Public Law 115–123. ↩︎
There is something odd about the concept of services to prevent foster care. Prevention usually refers to a social problem like child maltreatment or teen pregnancy, not a government policy, which is itself a response to the problem of child maltreatment. It is like talking about preventing hospitalization rather than preventing illness. It would be more natural to think of services to prevent the underlying problem that result in foster care, which is usually child abuse or neglect, but can also be children’s behavioral health problems that are so severe that parents feel compelled to relinquish their care to the state. But this is the term used by FFPSA and I will use it here. ↩︎
A note about foster care candidacy is in order here. In general, children are placed in foster care because they are abused or neglected or are at imminent risk of abuse or neglect. However , there is another group of children entering foster care who have drawn increasing attention. These are children whose parents voluntarily relinquish them because they are unable to care for them at home or obtain needed services–usually care for severe behavioral issues. ↩︎
Federal research has estimated that as many as five percent of all children entering foster care between 2017 and 2019 may have entered care primarily to receive behavioral health
or disability services, not because of maltreatment. ↩︎
The Handbook has already been revised once, but the language on this specific issue has not been revised. ↩︎
See pages 83-84 for more on statistical significance. This author did not find a statement of the required p-level. ↩︎
Also problematic is the failure to include domestic violence services in the clearinghouse. This was always puzzling. Child welfare social workers and academics often speak of the “Big Three” factors that result in child welfare involvement– drug abuse, mental illness and domestic violence. So the omission of domestic violence was strange–especially because domestic violence programs are greatly underfunded and not chargeable to an another program like Medicaid. ↩︎

Appendix

Parents as Teachers Contrasts Cited by Clearinghouse

Two small favorable contrasts from the matched comparison group study, which was not carried out in a usual care or practice setting and two contrasts showing no effect on child welfare administrative reports;
One contrast showing no effect on out-of-home placement;
Three favorable contrasts, two contrasts showing no effect and one unfavorable contrast on child social functioning;
Two small favorable contrasts and ten contrasts showing no effect on child cognitive functions and abilities;
Three contrasts showing no effect on child physical development and health;
One contrast showing no effect on adult parenting practices;
Eight contrasts showing no effect and one showing an unfavorable effect on family functioning;
Nine contrasts showing no effect and one showing an unfavorable effect on adult economic and housing stability.

Functional Family Therapy (FFT) Contrasts Cited by Clearinghouse

Two favorable contrasts, 23 contrasts showing no effect, and one unfavorable contrast on Child Behavioral and Emotional Functioning–the core goal of the program. One of the two favorable contrasts is on “Strengths and Needs Assessment: Child Behavioral/Emotional Needs” immediately after completing the program. But another study shows no impact on the same outcome.
Two favorable contrasts, two unfavorable contrasts, and 16 contrasts showing no effect on chid delinquent behavior;
Nine contrasts showing no effect on positive parenting practices;
Two favorable contrasts and 13 contrasts showing no effect on family functioning.
Ten contrasts showing a favorable impact compared to only eight showing no effect on children’s substance abuse. But all of these effects came from one study of alcohol-abusing youths who resided in a shelter after running away–a very specific population that may not be generalizable to children who are living at home.

Motivational Interviewing (MI) Contrasts Cited by Clearinghouse

Thirteen contrasts showing no effects on child substance use;
Five contrasts showing no effect on caregiver mental health;
Sixteen contrasts showing a favorable effect, 91 showing no effect, and two showing an unfavorable effect on parent/caregiver substance use. Among the favorable contrasts, one study shows a favorable effect on the number of drinks per week while there is no effect on the number of drinks per day but the Clearinghouse did not apparently take account of such contradictions.
Seven contrasts showing no effect on parent/caregiver criminal behavior;
One contrast showing no effect on family functioning;
Ten contrasts showing no effect on parent-caregiver physical health;
One contrast showing no effect on economic and housing stability.

Healthy Families America (HFA) Contrasts Cited by Clearinghouse

Forty-three contrasts showing no effect on child safety as measured by child welfare administrative reports, medical indicators of maltreatment risk, or “maltreatment risk assessment” measures.
Five favorable contrasts, 38 contrasts showing no impact, and one showing a negative impact on child safety based on parental self-reports of maltreatment. Clearly these self-reports are less valid than more objective measures, since parents clearly know what answers are expected after having been through the program. A parent’s negative answer to the question of whether she ever used physical abuse in the past year (one of the actual indicators used) cannot be trusted to be accurate.
Seven contrasts showing no effect on child safety as measured by “medical indicators of maltreatment risk,” a strange heading for a group of questions from the Adolescent-Adult Parenting Inventory including “Inappropriate Expectations, Lack of Empathy, and “Belief in Corporal Punishment.”
Eleven contrasts showing no effect on “Maltreatment Risk Assessment” measures such as hospitalizations and injuries needing medical care.
Five favorable contrasts and two showing no effect on parent reports of child behavioral and emotional functioning, also a self-reported item;
Two favorable contrasts, six showing no effect, and one unfavorable contrast on child cognitive functions and abilities.
One favorable contrast on “child delinquent behavior;” The measure was actually “child skips school often,” self-reported by children in first or second grade, according to the study.
One favorable contrast and two showing no effect on child educational achievement and attainment. This was a positive contrast on the percentage of children retained in first grade. However, the percentage of children who performed above or below grad level did not change.
Three favorable contrasts and 24 showing no effect on positive parenting practices. The three favorable contrasts came from observations of “Positive Parenting” from researchers’ observations of parents during a puzzle task, a “delay of gratification” task, and a cleanup task. However observations of harsh parenting during those same tasks did not show a statistically significant change.
Three contrasts showing a favorable impact and 16 contrasts showing no effect on parent/caregiver mental or emotional health;
Fifteen contrasts showing no effect on parent/caregiver substance use;
Three contrasts showing a favorable effect and 28 contrasts showing no effect on family functioning. Twenty-two of the contrasts were measures of intimate partner violence (IPV) or family violencefrom one study only and not corroborated. The positive impacts were on three specific measures based on the perpetrator, the type of violence and the age of the child. Only three of those 30 contrasts showed a positive impact and no effect sizes were provided.
Five contrasts showing no effect on economic and housing stability.

Child Maltreatment 2024: Did abuse and neglect really decline?

Maltreatment Reports, Victims and Fatalities All Down in 2024, reported The Imprint upon the release of the new federal report, on child maltreatment data for Federal Fiscal Year 2024. In my last post, I showed how the child fatality numbers from that report are too flawed to allow a conclusion that child deaths decreased. Five states account for over 100 percent of the decline in child maltreatment fatalities reported by the states for FFY 2024. All of these states have acknowledged that the declines in fatalities that they reported for FFY 2024 reflected changes and errors in how child maltreatment deaths were counted, not real trends in child maltreatment fatalities. But could it be that the number of reports and child maltreatment victims are more accurate than the fatality counts, which are based on smaller and more unstable numbers? The evidence suggests that these declines are also too heavily influenced by state policy and practice changes to be taken as true indicators of actual maltreatment trends.

The annual Child Maltreatment reports, produced by the Children’s Bureau of the US Department of Health and Human Services, are based on data that states submit to the National Child Abuse and Neglect (NCANDS) data system. The new report, Child Maltreatment 2024 (CM2024), provides data for Federal Fiscal Year (FFY) 2024, which ended on September 30, 2025. Exhibit S-2 summarizes the findings of the newest report. Child welfare systems operate somewhat like a funnel, at each stage selecting a fraction of the cases or children to proceed to the next stage.. Child welfare agencies received 4.365 million “referrals” alleging maltreatment and “screened in” 47.1 percent of them as “reports,” of which 2,058,720 reports received an investigation or alternative response. The investigations resulted in a total of 532,228 children determined to be victims of child abuse and neglect.¹

In discussing each stage of the child protective services process, CM2024 follows the regular practice in these reports of showing the changes in key indicators over time for the five years leading up to and including FFY 2024. But 2020-2024 is an unfortunate period to feature, given the onset of the COVID-19 pandemic in 2020. A more thoughtful approach would have been to start from 2019 or an earlier year before the onset of the pandemic. That approach would allow the observation of trends before they were interrupted by the pandemic. To facilitate such observation, I have provided charts of the changes in key indicators between 2015 and 2024, using the five-year trend data provided in CM 2019 and CM 2024.

Referrals

NCANDS uses the term “referrals” to mean reports to child welfare agencies alleging maltreatment. According to CM 2024, child welfare agencies received an estimated total of 4,365,000 referrals, or 59.3 referrals per 1,000 children, through their child abuse hotlines or central registries in that year. The number is an estimate because the states do not provide data on the total number of referrals, only on the number that were screened in and (usually) the number that were screened out. But not all states are able to provide the number of screened-out referrals; there were five jurisdictions that were not able to do this in FFY 2023 and FFY 2024, and more in earlier years.² To estimate the total number of referrals nationally, the Children’s Bureau imputes numbers for the nonresponding states based on the national referral rate for the responding states.³

A state’s referral rate depends on what referrals it actually counts. The state-by-state tables for 2024 document large differences in referral rates, from 19.2 per 1,000 children in Hawaii to 169.1 per 1,000 in Vermont–also the top and bottom states in 2023. Vermont’s Department of Children and Families reports that the state is unique because its juvenile justice system is integrated with its child welfare system; The hotline handles calls about juvenile delinquency, truancy and other non-maltreatment concerns, helping to account for its high number of referrals.⁴ Connecticut reports that none of the calls that are assigned to alternative response are included in NCANDS, resulting in a lower number of rererrals than many states report. Hawaii does not submit commentary so reasons for its very low referral rate are unknown.

As shown in Chart 1, the total referral rate had been increasing until the onset of the pandemic in 2020, when it fell sharply; it then rose only slightly in 2021. It increased for the next two years, almost reaching its pre-pandemic level in 2023. In FFY 2024, the estimated number of referrals decreased slightly from 4.380 million to 4.365 million, and the referral rate decreased from 59.4 to 59.3 per 1,000 children.

Source: Child Maltreatment 2019 and Child Maltreatment 2024, Exhibit S-1.

The very small decrease in the referral rate between 2023 and 2024 could have been caused by changes in policy, practice, and the types of messaging coming fom child welfare agencies, state officials and advocates around mandatory reporting of child abuse and neglect. State legislators, agency heads and advocates have recently worked to discourage calls to their child abuse hotlines based on the prevailing belief that many of these calls do not concern maltreatment but rather poverty and are often influenced by bias against parents of color. Agencies have been revising their mandatory reporter training⁵ to discourage reporting of concerns that are allegedly about poverty rather than child abuse or neglect. Implicit bias training is often included based on the belief that mandatory reporters are influenced by racial, cultural and class bias. In some states, “warmlines” are being set up to accept calls that are allegedly related to poverty rather than child maltreatment.⁶ Some states have banned anonymous reports to child abuse hotlines on the grounds that such reports are frivolous and expose children and families to unnecessary trauma–an assertion that has been challenged by based on research showing that anonymous reports save children’s lives. Such bans passed by the legislatures in Texas and Arkansas took effect shortly before the beginning of FFY 2024. In Texas, the number of referrals fell from 249,283 in 2023 to 233,112 in 2024. In Arkansas, it fell from 63,732 to 57,371. (A ban on anonymous reporting has been signed by the Governor of New York State and will take effect next summer; we may see its effect in the following fiscal year.)

The changes in policy and practice related to mandatory reporting are related to a broader change in the ideological climate around child welfare. It is reflected in the 2018 Family First Prevention Services Act and similar legislation in many states. Among the tenets of this ideology are that child welfare is a racist family policing system, neglect allegations and findings reflect poverty rather than maltreatment, and that child welfare intervention does more harm than good for children. The prevalence of these beliefs among legislators and agency leaders has led to policy changes that affect not only reporting but the subsequent stages of screening and investigation.

Screening

Once a state agency receives a referral, it will be “screened in” or “screened out” by agency staff. In general, referrals are screened out if they are deemed not to contain an allegation of child abuse or neglect, contain too little information on which to act, are more appropriately assigned to another agency, or for some other reason do not fall under the mandate of the child welfare agency. In NCANDS, a referral becomes a “report” once it is screened in. “Reports” are assigned for an investigation or “alternative response.”

The Children’s Bureau, as it does every year, chooses to show screened-in referrals divided by the child population, calling the result the National Screened-In Referrals Rate. This is an ambiguous indicator that reflects both the number of referrals and the proportion of those that are screened in. It is more useful to look at the percentage of referrals that an agency decides to screen in, which is an indicator of its willingness to intervene with families to protect children. Unfortunately, the the lack of screened-out numbers from some states makes this number an estimate because the total number of referrals in those states is unknown.⁷ Nevertheless, a clear pattern emerges when the percentage of referrals that were screened in between 2015 and 2024 is plotted in Chart 2. That percentage has decreased every year since 2015, especially after 2017, with the decline continuing during and after the pandemic. Between 2017 and 2024, the percentage of referrals that were screened in dropped from 57.6 to 47.1.

Source: Author’s Calculations from Child Matlreatment 2015, Child Maltreatment 2016, Child Maltreatment 2017, Child Maltreatment 2018, Child Maltreatment 2019, and Child Maltreatment 2024. These numbers are based on screened-out referral rates for 44 states in FFY 2015, 45 states in 2016, 2017, 2018, and 2019, 47 states in 2020 and 2022, and 46 states in 2021 and 2023. The states that did not report screened-out referrals in all ten years include New Jersey, New York, Pennsylvania, and North Dakota.

Some states with larger decreases in the number of screened-in referrals attributed them in their commentaries to more consistent screening processes with the implementation of centralized intake, more structured processes for determining if a report meets criteria for acceptance, or new screening tools such as a Structured Decision-Making model. Nebraska reports that in July 2023 it dropped a policy that was implemented in June 2019 that required the acceptance of all reports made by medical professionals about a child aged five or under. The state reports a modest decline in the number of screened-in referrals in the following year. Considering changes in the prevailing child welfare climate mentioned above, it is likely that other states made policy or practice changes designed to reduce the number of reports that were screened in.

National Child Disposition Rate

An informative rate reported by the Children’s Bureau is the “national child disposition rate,” which is the number of children subject to an investigation or alternative response. This rate is a good measure of the “footprint” of CPS–what proportion of children it touches. This rate ranges from 12.0 per 1,000 children in Hawaii (the lowest again) to a high of 90.4 per 1,000 in West Virginia–nearly one out of every ten children. The West Virginia rate, which has fallen significantly from the high of 148.8 in FFY 2019k presumably reflects the state’s substance abuse (particularly opioid) epidemic.

The National Child Disposition Rate rose from 2015 to 2019, falling drastically with the pandemic in 2020 and again 2021. In 2022 it was back to the 2020 level of 40.1 per 1,000 children, but then began declining. The pandemic may have masked what would have been an annual decline in this rate every year between 2018 and 2024. The proportion of children touched by CPS has dropped significantly since its high of 47.8 per 1,000 in 2018 to 40.6 in 2024. Given the lack of a decrease in referrals during that time period except for the very small decrease in 2024, this pattern must reflect the drop in the proportion of reports that were screened in over the period.

“Victimization”

Once a referral is assigned for investigation or alternative response, the next phase in the funnel is the determination of whether abuse or neglect has occurred, or in NCANDS parlance, whether the child is a “victim.” A “victim” is defined as “a child for whom the state determined at least one allegation of maltreatment was substantiated or indicated⁸; and a disposition of substantiated or indicated was assigned for a child in a report.” “Victims” include children who died of abuse or neglect if the maltreatment was verified. Those children who receive an “alternative response”⁹ instead of an investigation are not counted as victims because alternative response does not result in substantiation or “indication” of a report. The Children’s Bureau somewhat tautologically treats “alternative response” as one of the possible dispositions for children assigned to an investigation or alternative response.

As I explain every year (with apologies to my faithful readers), the number of “victims” reported by states does not represent the true number of children who experienced abuse or neglect, which is unknown. Many cases of child maltreatment go unreported. Children assigned to alternative response are not found to be victims unless their case is reassigned to the investigation track. And finally, substantiation may not be an accurate reflection of whether maltreatment occurred. Making a determination of whether maltreatment occurred is difficult. Adults and children do not always tell the truth, the youngest children are nonverbal or not sufficiently articulate to answer the relevant questions. So it is not surprising that research suggests that substantiation decisions are inaccurate¹⁰ and a report to the hotline predicts future maltreatment reports and developmental outcomes almost as well as a substantiated report.¹¹

The vast difference in state “victimization rates” illustrates how these rates may reflect agency policy and practices as well as underlying rates of abuse and neglect. These rates range from a low of 1.2 per 1,000 children in New Jersey (even lower than Hawaii and even lower than last year’s rate of 1.5) to a high of 15.4 in Maine. It is unlikely that Maine has more than ten times more child abuse and neglect victims than New Jersey. “”Victimization rates” reflect what happens at earlier stages in the funnel of child welfare–reports, screening, and the use of alternative response to divert a child from the investigative track. They also reflect policies including different definitions of abuse or neglect and levels of evidence required to confirm maltreatment. They also state idiosyncracies in state systems: referrals concerning less serious neglect that are routed to General Protective Services in Pennsylvania do not appear at all in NCANDS. In Calendar Year 2024, the Pennsylvania Department of Human Services received 176,496 General Protective Services reports and only 41,070 Child Protective Services reports, so the overwhelming majority of referrals in Pennsylvania are not included in NCANDS In view of the deceptiveness of these terms, I have put the terms “victims” and “victimization rates,” when not qualified, in quotation marks in this post.

But what about changes over time? The fact that “victimization rates” reflect different policies and practices does not mean that trends over time don’t reflect real changes in the incidence of abuse and neglect. The national “victimization rate” declined in 2024 as it has done every year since 2018, even after the pandemic. It was down to 7.2 per 1,000 children in FFY 2024, compared to 9.2 per 1,000 children in 2018. Couldn’t this be good news about a decline in child abuse and neglect?

Victimization rates depend on screening decisions, and investigators’ decisions about whether a screened-in report will be substantiated as abuse or neglect. And these things can change over time. As mentioned above, recent changes have reflected the increasing prevalence of an ideology professing that child welfare involvement does more harm than good to children and families, and that agencies should scale back child removals into foster care and perhaps even in-home services. One way to do this is to make it more difficult to substantiate maltreatment, as New York and Texas (two states on the opposite ends of the political spectrum) have recently done. In Texas, the Legislature changed the definition of neglect to require both “blatant disregard” for the consequences of a parent’s action or inaction and either a “resulting harm or immediate danger.” New York changed the level of evidence required to substantiate an allegation of abuse or neglect from “some credible evidence” to “a fair preponderance of the evidence.” Both of these states mention these changes in explaining their declines in “victimization” counts.

Surprisingly, plotting the number of children who received an investigation or alternative response by disposition (see Table 1) casts some doubt on opinion of states like New York and Texas that the decreased number of victims they found reflects state policies making it more difficult to substantiate maltreatment. The proportion of children judged to be victims of abuse or neglect dropped from 16.8 percent in 2018 to 15.8 percent in 2024. Yet the percentage of children who received a disposition of “unsubstantiated” hardly changed during the period, nor did the proportion of children receiving an alternative response. The only category that increased enough to account for the decrease in substantiations was a category called “no alleged maltreatment,” which increased from 10.4 to 11.8 percent of the children. According to CM 2024¹² that category would include children who are not the subject of an allegation but who are investigated because they live in a state that provides a response to all children in a family, even those who are not the subject of an allegation. It is not clear why the percentage of children in this category would have increased so much since 2018.

Table 1 Children who Received an Investigation or Alternative Response by Disposition (Percentage), FFY 2018-2024

FFY	Victims (Substantiated/Indicated)	Alternative Response	Unsubstan-tiated	No alleged maltreatment	Other
2018	16.8	14.0	56.3	10.4	2.3
2019	16.7	13.8	56.5	10.6	2.4
2020	17.6	13.2	56.4	10.4	2.4
2021	17.8	13.3	56.0	10.3	2.6
2022	16.1	13.9	56.3	11.1	2.5
2023	15.8	13.7	56.3	11.8	2.4
2024	15.8	13.9	56.2	11.8	2.2

Source: Child Maltreatment 2024, Child Maltreatment 2022.

The number and rate of child maltreatment referrals registered by states saw a small decline in Federal Fiscal Year 2024. This decline may reflect changed messaging from states regarding mandatory reporting of child maltreatment as much or more than a real reduction in concerns regarding children being maltreated. States report a declining number of child maltreatment victims as well, but it is not possible to say with any confidence that the number of victims and their proportion of the population decreased in FFY 2024 or in the past five or ten years. There are simply too many factors influencing these numbers and how they change over time, including recent changes in policy and practice by state child welfare agencies designed to reduce their intervention with families. Adding to the confusion is the existence of unexplained changes, such as the reduction in the percentage of children who received an investigation or alternative response with “no alleged maltreatment.” Sadly, Child Maltreatment 2024, like its predecessors, tells us little about the true state of maltreatment in America.

Note: This blog was edited on February 17, 2024 to incorporate information about the disposition category of “no alleged maltreatment” and also to add a mention of Pennslvania’s General Protective Services cases not being included in CM 2024. It was modified again on February 19, 2024 to add information provided by the Vermont Department of Children and Families about why it receives such a high number of referrals.

Notes

Of those victims, 296,738 received some type of services, which are not discussed in this blog. ↩︎
I use the term “states” to mean “jurisdictions,” including the District of Columbia and Puerto Rico. Four states were unable to report complete data for both FFY 2023 and FFY 2024: New Jersey, New York, North Dakota and Pennsylvania. West Virginia did not report complete data for FFY 2023 so its data could also not be used for a year-to-year comparison. ↩︎
See CM 2024, page 11. ↩︎
Email from Commissioner’s Office, Vermont Department of Children and Families, February 19, 2026. ↩︎
See also DCYF Updates Mandatory Reporter Training (Washington), Family Experiences Shape Efforts to Reform Missouri’s Overburdened Foster Care System. ↩︎
See A Phone lifeline for families in need in DC receives support from Doris Duke Foundation, Warmline offers alternative to calling protective services (Colorado). ↩︎
See notes to Chart 2. ↩︎
Substantated is defined as “supported or founded by state law or policy.” “Indicated” is a less commonly used term meaning a “disposition that concludes maltreatment could not be substantiated understate law or policy, but there is a reason to suspect that at least one child may have been maltreated or is at risk of maltreatment.” ↩︎
An “alternative response” includes an assessment and referral to appropriate services if the parent agrees to participate. There is no determination on whether abuse or neglect occurred and no child removal unless the case is transferred to the investigative track. ↩︎
Theodore Cross and Cecilia Casanueva, “Caseworker Judgments and Substantiation,” Child Maltreatment, 14, 1 (2009): 38-52; Desmond K. Runyan et al, “Describing Maltreatment: Do child protective services reports and research definitions agree?” Child Abuse and Neglect 29 (2005): 461-477; Brett Drake, “Unraveling ‘Unsubstantiated,’” Child Maltreatment, August 1996; and Amy M. Smith Slep and Richard E. Heyman, “Creating and Field-Testing Child Maltreatment Definitions: Improving the Reliability of Substantiation Determinations,” Child Maltreatment, 11, 3 (August 2006): 217-236. ↩︎
Brett Drake, Melissa Jonson-Reid, Ineke Wy and Silke Chung, “Substantiation and Recidivism,” Child Maltreatment 8,4 (2003): 248-260; Jon M. Hussey et al., “Defining maltreatment according to substantiation: Distinction without a difference?” Child Abuse and Neglect 29 (2005): 479-492; Patricia L. Kohl, Melissa Jonson-Reid, and Brett Drake, “Time to Leave Substantiation Behind: Findings from a National Probability Study,” Child Maltreatment, 14 (2009), 17-26; Jeffrey Leiter, Kristen A. Myers, and Matthew T. Zingraff, “Substantiated and unsubstantiated cases of child maltreatment: do their consequences differ?” Social Work Research 18 (1994): 67-82; and Diana J. English et al, “Causes and Consequences of the Substantiation Decision in Washington State Child Protective Services,” Children and Youth Services Review, 24, 11 (2002): 817-851 ↩︎
See page 23. ↩︎

Child Maltreatment 2024: Clear Evidence that Federal Trend Data Cannot be Trusted

by Marie Cohen

This post was originally published on Lives Cut Short website.

The Children’s Bureau’s eagerly awaited Child Maltreatment report has been published and the child welfare newsletter The Imprint lost no time in announcing that Maltreatment Reports, Victims and Fatalities All Down in 2024. Others are sure to follow with commentaries celebrating a 10 percent drop in the number of child maltreatment fatalities. This would be great news if it were true, but making any conclusion based on these data is not warranted. The state-by-state data show that the reductions reflect changing policies and practices, as well as reporting problems, rather than actual reductions in child maltreatment deaths.

On January 8, 2025, the Administration on Children and Families (ACF) of the U.S. Department of Health and Human Services published Child Maltreatment 2024 (CM2024), its latest compendium of data on states’ responses to reports of child abuse and neglect. These annual publications, produced by the Children’s Bureau of ACF, are based on data that states submit to the National Child Abuse and Neglect (NCANDS) data system. CM2024 provides data for FFY 2024, which ended on September 30, 2024. Based on reports from all 50 states, the District of Columbia, and Puerto Rico, CM 2024 reports that the number of child maltreatment fatalities dropped from 1,979 or 2.68 per 1,000 children in Federal Fiscal Year (FFY) 2023 to 1,773 or 2.41 per 1,000 children in FFY 2024. Of course,we should celebrate if these numbers are believable. But that is not the case.

An analysis of last year’s Child Maltreatment report explained how the numbers in the annual federal reports cannot be taken as a good estimate of child maltreatment deaths for any given year. It is widely accepted that these numbers greatly understate the actual number of deaths due to child abuse or neglect. Well-functioning child death review teams identify many more child maltreatment fatalities than are reported to NCANDS. Based on only publicly available sources, Lives Cut Short has counted more child maltreatment deaths than those reported to NCANDS in more than half of states in either 2022 or 2023. The undercounts in NCANDS are due to a variety of state policies and practices including restrictive definitions of child maltreatment fatalities, failure to consult all available sources, restrictions on which reports are investigated, and constraints on finding maltreatment in fatality investigations.

Of course it is theoretically possible that while NCANDS underestimates child maltreatment deaths, the trends it finds over time are still valid. But state-by-state data clearly disprove that thesis. The 50 states, the District of Columbia and Puerto Rico reported 206 fewer child maltreatment deaths in FFY 2024 than in FFY 2023. Five states–Texas, Maryland, North Carolina, Illinois and Virginia–together account for a decrease of 210 fatalities–more than 100 percent of the total decrease for the year overall. Explanations from all of these states indicate that the drops they reported reflect factors other than the actual number of child maltreatment deaths.

Child Maltreatment Fatalities, FFY 2023 and 2024
State	FFY 2023	FFY 2024	Total Decline, 2023-2024
Illinois	83	66	17
Maryland	83	46	37
North Carolina	107	46	61
Texas	187	107	80
Virginia	55	40	15
TOTAL (5 states)	515	305	210
United States	1,979	1,773	206

Source: US Department of Health and Human Services, Child Maltreatment 2024, January 2026.

The number of child maltreatment deaths in Texas that were reported in Child Maltreatment 2024 fell from 187 in FFY 2023 to 107 in FFY 2024, an enormous drop. In its own annual report on child fatalities and near fatalities for FY 2024, the Texas Department of Family and Protective Services (DFPS) reports a smaller but still substantial decrease from 164 child maltreatment fatalities in state fiscal year (SFY) 2023 to 99 in SFY 2024.¹ The number of investigated child fatalities fell from 690 to 587–a decline of 587, or 14.9 percent. Out of the investigated families, the number of fatalities where abuse and neglect was confirmed fell from 164 or 23.7 percent of fatalities investigated to 99, or 16.9 percent of fatalities.

In its report, DFPS suggests that the decline in both investigations and maltreatment findings reflects a change in the definition of child neglect adopted by the Legislature that took effect in FY 2022. This change required the agency to find that a parent exhibited “blatant disregard” for the consequences of an act or failure to the act that results in harm or immediate danger to a child. DFPS believes this law affected investigations and dispositions in both FY 2023 and FY 2024. (NOTE: This does not explain the drop in findings of abuse, however.)

DFPS also reported that two other factors affected the fatality numbers. Starting in 2022, the agency changed its screening policy so that reports involving a child fatality but include no explicit concern for abuse and neglect are not investigated if the reporter “or first responders” had no concern for abuse or neglect. According to DFPS, this contributed to the decrease in investigations and ultimately in fatalities reported both in FY 2023 and in FY 2024. DFPS also reported that in FY 2024 it imposed an additional level of review for certain investigations that found neglect (like those involving drowning or unsafe sleep) to ensure that the investigator had followed the new neglect definition incorporating “blatant disregard.” This change almost certainly contributed to the reduced number of maltreatment fatalities found.

North Carolina reported a large drop in child maltreatment fatalities from 107 in FFY 2023 to 46 in FFY 2024. In response to a question about the cause for this large decrease, a spokesperson for the North Carolina Department of Health and Human Services (NCHDHHS) provided the following explanation, which makes clear that the reported decline in child maltreatment fatalities in FFY 2024 is not valid and should not have been included in CM 2024:

NCDHHS is aware of probable discrepancies in the state’s historic child fatality data, particularly in 2023-2025. Prior to 2025, the state was using an outdated database platform that became increasingly unstable and subject to error. Recognizing this, NCDHHS made significant investments in designing and building a new, comprehensive Child Welfare Fatality Information System that launched for state users in 2025. While data from the old database was transferred to the new system, these numbers likely include inconsistencies and should not be used to analyze trends in child fatalities.

Maryland reported 46 child maltreatment deaths in FFY 2024, compared to 83 in FFY 2023.The large drop in Maryland’s child maltreatment fatalities was predictable in advance and is as irrelevant to actual child deaths as those in Texas and North Carolina. In 2025, Maryland’s child maltreatment fatality reports to the Children’s Bureau made the news when a reporter discovered the state had the second highest number in the country. Department of Human Services (DHS) officials soon realized that the agency had for years been reporting the number of death cases investigated for abuse and neglect, whether or not they were confirmed as due to maltreatment. Maryland reported this issue in its commentary to CM2024, stating that it had submitted correct information for 2024. However, Maryland’s entries for previous years have not been corrected, resulting in the misleading appearance of a large reduction in child maltreatment fatalities between FFY 2023 and FFY 2024.

Illinois reported a smaller but still sizable decline from 83 child maltreatment fatalities in FFY 2023 (itself a drop from 110 in FFY 2022) to 66 in FFY 2024. In its commentary, DCFS reported that the decrease in confirmed maltreatment fatalities in both years stems from the introduction of an administrative review process for sleep-related deaths. For any sleep-related death allegation, a senior administrator reviews the investigation “to ensure consistent evaluation of whether the sleep-related death included evidence of blatant disregard,” referring to the parent’s disregard of the danger to the child of unsafe sleeping arrangements. This is the same concept that is included in Texas’ new definition of child neglect. DCFS reports that this new review process has continued to result in fewer sleep-related deaths being “indicated” as due to maltreatment.

Virginia reported a significant decrease in child maltreatment fatalities from 55 in FFY 2023 to 40 in FFY 2024. In its commentary, the Virginia Department of Social Services suggests that this decline was due to a decrease in fatality investigations in FFY 2024, which in turn stemmed from “more local agencies having overdue investigations that were not closed timely or in accordance with VDSS guidance.”

The explanations provided by Texas and Illinois have something in common. In both states, the decline in confirmed child maltreatment fatalities was the intended result of policies–like a changed definition of neglect in Texas and the imposition of administrative reviews to enforce such a restrictive definition in both states–that were designed to find fewer child maltreatment fatalities. Those policies and practices are only part of an overall effort designed to reduce the involvement of child welfare agencies in the lives of families in those two states and in many other states around the country. So it is not surprising that the number of reports accepted for investigation and the number of children found to be victims of abuse or neglect also declined in FFY 2024.

In conclusion, five states account for over 100 percent of the decline in child maltreatment fatalities reported in Child Maltreatment 2024. All of these states have acknowledged that the decline in fatalities that they reported for FFY 2024 was caused by the fatality determination and reporting process itself, not real trends in child maltreatment deaths. The bottom line is that the numbers contained in the Children’s Bureau’s Child Maltreatment reports are not reliable for determining national or state-level trends in child maltreatment deaths, or in child maltreatment in general. Responsible reporters and researchers should know by now not to rely on these data, but they continue to do so, confusing the public and depriving policymakers of the information they need to make good decisions.

It is not clear why the SFY numbers are so different from the FFY numbers, as the two years overlap by 11 months, DFPS has not yet responded to a request for an explanation. ↩︎

How did peer reviewers let this one get by? A flawed study of foster care removals and child maltreatment mortality

by Marie Cohen

This blog has often focused on the uncritical use of flawed research to support desired conclusions. A good example of this has now become available. The authors of a new study conclude that decreasing rates of entry into foster care are not associated with changes in mortality from child abuse and neglect. Advocates of continuing reductions in foster care or its total abolition have lost no time in sharing the results, undaunted by what should have been obvious to any knowledgeable reader. The study is too flawed by the nature of its data on child maltreatment fatalities to make any conclusions about the presence or absence of a connection between foster care reductions and changes to child maltreatment mortality.

The number of children entering foster care in the United States has fallen from 252,198 in Federal Fiscal Year (FFY) 2019 to 170,955 in FFY 2024. As computed by Frank Edwards, Kelley Fong, and Robert Apel in their new paper, foster care entry rates have decreased from 3.49 to 2.47 entries per 1000 children between 2018 and 2023, a decrease of about 29 percent. This large drop in foster care entries was in large part the result of a growing bipartisan consensus on the federal level and in red and blue states around the need to reduce or entirely terminate the use of foster care to protect children from continuing abuse or neglect.

Some child advocates and media outlets around the country have argued that the reductions in the use of foster care have gone too far. The Seattle Times Editorial Board recently told readers to “Stop dodging and face facts: More kids are dying on DCYF’s watch,” citing a 35% drop in dependency petitions to take kids into foster care since 2021 along with a 67 percent increase in child deaths and near-deaths between 2020 and 2024. A series from Texas Public Radio called When Home is the Danger suggests that a 40 percent reduction in removals of children to foster care, at the same time as in-home services were being curtailed, has resulted in more children dying as a consequence of abuse or neglect. Child advocates across the country have expressed similar fears, but there has been no good data to support or refute this claim.

Edwards, Fong and Apel say their new study shows these fears are baseless. They report that they “found no evidence of a negative association between foster care entries and child maltreatment fatalities at the population level.” They came to this conclusion by estimating the association between child maltreatment mortality rates and foster care entry rates at the state level, using state administrative data from Federal Fiscal Years 2010 to 2023. Even after controlling for variations in population disadvantage, other state features, and national temporal effects they found no association between foster care entry rates and child maltreatment mortality rates.

Advocates of continuing reductions in, or outright elimination of, foster care have been circulating the article widely, using it to counter any claim that excessive foster care reductions may be leading to child fatalities. But a cursory review shows that the study is fatally flawed and would not have been published had the peer reviewers been familiar with the sources of the data used. Four of the leading academic researchers in the field have already posted comments pointing to issues with the data and the model. These comments are summarized and explained below, along with examples from my research on child maltreatment fatality data.

Flawed Data

The Edwards study uses child maltreatment fatality data from the National Child Abuse and Neglect Data System (NCANDS) to represent child maltreatment deaths. These numbers appear in the annual Child Maltreatment reports released by the Children’s Bureau. In a report called A Jumble of Standards: How State and Federal Authorities Have Underestimated Child Maltreatment Fatalities, I summarized the results of my review of ten years of federal Child Maltreatment reports, including all the state commentaries provided in each report. My work points to multiple reasons that these numbers reported by the states should never be used in such a study.

Screening Policies

The number of child maltreatment fatalities a state will find depends first on which deaths it accepts for investigation and which ones it screens out. And when these change, the number of fatalities found will probably change. For example, some states or counties historically screened out cases where there were no other children in the home,¹ because investigations were not considered necessary to protect the other children in the home. West Virginia began investigating cases with no surviving children in the home in FFY 2016. Alaska changed its policy in 2023 so that it will now investigate cases where there are no surviving children in the home, which will affect future counts. We do not know how many states may have made this or similar changes over the span covered by the Edwards paper.

Texas changed its screening policy as of September 2022 so that “reports that involve a child fatality but include no explicit concern for abuse and neglect” are not investigated until the reporter or first responders are interviewed to find out if they had any concerns for abuse or neglect. Only if they had such concerns is the report investigated. DFPS reported that as a result, the number of child fatalities it investigated decreased from 997 in FY2022 to 690 in FY2023 (a 31 percent decrease) due to this new screening policy. And the number of child maltreatment fatalities found fell from 182 to 164.

Changed screening policies can also lead to an increase in reported child maltreatment fatalities. In 2022, Maryland implemented a policy that requires local agencies to screen in sleep-related fatalities, which often involve parental substance abuse. In its commentary, the agency indicated that this was the reason it reported an increase in child deaths from 68 in FFY 2022 to 83 in FFY 2023.

Definitions

NCANDS defines a child maltreatment fatality as “the death of a child as a result of abuse or neglect because either: (a) an injury resulting from child abuse or neglect was the cause of death; or (b) abuse and/or neglect were contributing factors to the cause of death.” So a state’s definition of a maltreatment fatality depends on its definition of child abuse and neglect. And if that definition changes over time, that will affect the trend in child maltreatment fatalities reported. For example, Texas changed its definition of child neglect in 2021 to require “blatant disregard” for the consequences of an act or failure to act. Texas reported a drop of in the number of child maltreatment fatalities from 255 in 2020 to 205 in 2021 to 176 in 2022. The state has attributed that decline in reported neglect deaths to the 2021 law.

A similar decrease in reported child fatalities occurred in Illinois, where “blatant disregard” was already included in the definition of neglect but the requirement was enforced through a new administrative review process for sleep-related deaths. A senior administrator reviews the investigation to ensure that death included evidence of “blatant disregard.” In its 2023 commentary, Illinois links this new policy with a decrease of 24.6 percent in reported child fatalities in FFY 2023.

Data Sources

Experts widely agree that NCANDS underestimates the total number of child maltreatment fatalities, as Edwards et al. acknowledge. One reason is that many states do not use all of the available sources for possible child maltreatment fatalities, which include reports to child protective services, state vital statistics departments, state child death review teams, law enforcement agencies, and medical examiners or coroners. Some states have recently been broadening the sources of data they use, and such changes may have resulted in an increase in reported fatalities. For example, North Carolina originally reported only deaths that the chief medical examiner classified as “homicide” by a parent or caregiver as child maltreatment fatalities. (This does not jibe with the NCANDS definition of child maltreatment fatality, but there is no consequence applied to states who do not use it.) After the agency began to include child deaths that were substantiated as maltreatment by CPS, as well as increasing their collaboration with vital statistics and criminal justice agencies to identify maltreatment deaths, North Carolina’s reported child fatalities increased from 64 in FFY2018 to 111 in FFY2019 (a 73 percent increase).

Historically, Mississippi also reported only deaths that the chief medical examiner had classified as homicide by a parent or caregiver but in 2007 began including fatalities substantiated by CPS as due to maltreatment. In FFY 2014, Mississippi developed a Special Investigations Unit to investigate all reports of child fatality that meet criteria for investigation. Mississippi reported that the having dedicated, specialized investigators resulted in a higher count of maltreatment fatalities in that year and thereafter.² The state also reported that public awareness campaigns to prevent deaths caused by unsafe sleep and children left in hot cars led to an increase in reporting of such fatalities, also contributing to an increase in reported deaths. The number of child maltreatment fatalities reported by Mississippi increased from 17 in FFY 2010 to 76 in FFY 2023.

Reporting Errors

In 2025, a reporter at the Baltimore Banner noticed that Maryland was reporting an alarming rate of child fatalities. With a reported 83 children dying of abuse or neglect in FFY 2023, Maryland had one of the worst child maltreatment fatality rates in the nation. Taken by surprise, state Department of Human Services officials soon realized that they had been reporting incorrect numbers for the past five years. Instead of reporting the deaths for which maltreatment was confirmed, they had been reporting all child deaths that had been investigated for maltreatment. The department now believes that there were 173 child maltreatment fatalities, not the 285 reported, between FFY’s 2020 and 2023. We do not know how many other states have been reporting erroneous numbers, given the little attention that these reports usually receive.

Time Lags

A key flaw of Edwards et al.’s analysis is the assumption that child maltreatment fatalities reported for a specific year reflect the deaths that occurred in that year. That is far from being the case. As the Children’s Bureau explained in Child Maltreatment 2023, “The child fatality count in this report reflects the federal fiscal year (FFY) in which the deaths are determined as due to maltreatment. The year in which a determination is made may be different from the year in which the child died.”³ There is similar language in reports from the previous years. States have explained in their commentaries that the actual dates the children died may have occurred as much as seven years before the year they were reported. As Sarah Font of Washington University and Emily Putnam-Hornstein of the University of North Carolina explain in their comment (posted below the article), “the Edwards et al. analysis largely models whether state foster care entry rates predict past [child maltreatment fatalities] — a nonsensical question with no bearing on whether foster care is a protective intervention. Indeed, if states respond to rising [child maltreatment fatalities] by placing more high-risk children in foster care, one would expect higher past [child maltreatment fatality] rates to be associated with higher current foster care entry rates.”

Even if the fatalities did occur in the same year as they were reported, it would not make sense to expect an increase in deaths only in the same year as a decrease in foster care, Brett Drake of Washington University raises this concern in his comment. Any maltreatment fatality resulting from a failure to remove a child will not necessarily occur in the same year. A foster care removal may save a child’s life in a subsequent year. Many of the most horrific child fatality cases involve child fatalities that could have been prevented by removal in prior years. In some of these cases, the child suffered through years of abuse or neglect before dying. But as Font and Putnam-Hornstein point out, the Edwards model attempts to correlate removal rates with past child maltreatment fatalities. And there is no way that makes sense.

Other flaws in the model

Font and Putnam-Hornstein, as well as Richard Barth of the University of Maryland, draw attention to another problem. Children under age five account for 82 percent of reported child maltreatment fatalities but less than half of foster care entries. But proportional decreases in foster care entries were largest among older children: there was a 49 percent reduction in entries of 17-year-olds versus an 18 percent reduction in infant entries. It might have been more sensible to focus on younger children only, though fixing this problem would not address all the other problems with the analysis.

Barth suggests another problem with the model. Placing a child in foster care may prevent that child dying from other causes, not just maltreatment. And indeed, we have learned that maltreatment is correlated with deaths from all causes, not just from child maltreatment. Research shows that children who have been reported for maltreatment are more likely to die of accidents in their first five years, sudden infant deaths and medical causes as infants, and suicide as teenagers. It is also obvious from my personal experience reviewing child fatalities in the District of Columbia that children who die of homicide as teenagers and young adults are likely to have a family history of child abuse and neglect reports. Child maltreatment mortality is only one type of child mortality we should care about.

It is unfortunate that so many child welfare leaders and commentators are so quick to grab hold of any research that supports their point of view without critically examining the methodology and considering whether the findings justify the claims. I hope that the peer reviewers for this study were as unfamiliar with state child maltreatment data as the study authors themselves; otherwise their motives in approving the study are suspect. As Font and Putnam Hornstein put it, “[t]he conclusions of Edwards et al. rest on a mis-specified model and deeply-flawed data; deficiencies unaddressed by the authors’ numerous sensitivity analyses.” These findings have no relevance to the discussion of whether the decline in foster care removals is putting children at risk of dying.

Notes

Some jurisdictions still do this, including the District of Columbia (as I learned from requesting child fatality information) and some counties in Ohio, as described in Child Maltreatment 2022. ↩︎
Child Maltreatment 2014, Child Maltreatment 2015, and Child Maltreatment 2016. ↩︎
Like many statements in the CM reports, this is not exactly true. A few states (including California) do attempt to capture only the deaths that occurred in the reporting year. This involves reporting on only deaths that occurred in the reporting year and then amending the numbers in future years. See A Jumble of Standards. In their eAppendix 1, Edwards, Fong and Appel note that they include the most recent revisions into their estimates. But as I describe in A Jumble of Standards, it appears that only a few states make such revisions. ↩︎

A Home for Every Child: an inadequate framework for assessing state child welfare programs

On December 19, 2025, the Administration for Children and Families (ACF) released CFSR Technical Bulletin No. 14, the first step in its plan to replace the the existing process by which it reviews state child welfare performance with one that is less burdensome and more useful. The replacement is long overdue. But ACF’s attempt to reorganize all outcomes under one slogan–A Home for Every Child–is highly problematic. It disregards the need for larger homes for sibling groups and residential treatment for some young people. And it devalues and disincentivizes the central function of child welfare–protecting children from harm.

In an article in The Imprint called Why We Are Putting the PIP on a PIP, Alex Adams, the Assistant Secretary for ACF, explains the shortcomings of the Child and Family Services Reviews (CFSR’s), which are the periodic federal reviews of state child welfare systems to assess their performance and ensure they are in compliance with federal requirements. In 25 years, no state has been in substantial compliance with all the outcomes and systemic factors that are evaluated, and every state has been placed on a Performance Improvement Plan (PIP). Meanwhile, the administrative burden is heavy. Adams rightly states that “this is busywork–expensive, repetitive, punitive busywork–and it has not improved child or family outcomes.” I don’t think many child welfare stakeholders would disagree. It is time for the CFSR’s to be put out of their misery.

But what will replace the CFSR’s? As a first step to a solution, ACF has announced a pilot PIP process that states can opt to follow instead of the standard PIP procedure. As described in the Technical Bulletin, “this Administration is creating a pilot opportunity for states….to center their program improvement efforts around the A Home for Every Child goal.” Interestingly, the goal is not defined in the Technical Bulletin;¹ the definition is found in several other places including a press release stating that “A Home for Every Child sets an ambitious goal of achieving a foster home-to-child ratio greater than 1:1 in every state. The initiative will focus on both sides of the equation — increasing the availability of safe homes through diligent recruitment, prioritizing kin, and improving retention of existing caregivers while reducing entries into foster care through effective prevention and faster pathways to permanency.” This sounds reasonable at first, but there are problems with centering program improvement planning around this goal–problems with the concept itself and with the implications of using it as an organizing principle in the assessment and improvement of child welfare performance.

A Flawed Concept

A Home for Every Child as a goal of child welfare seems to make sense on its face. To take a child into foster care when there is no place to put that child seems like folly, and a state that cannot equalize the supply and demand for foster care might need to rethink its rules about when to remove a child. But ACFs definition of that goal–one licensed home per foster child– makes no sense. Many foster homes can accommodate more than one child, and we need more larger homes to welcome sibling groups. And some young people in foster care need residential treatment before they can flourish in a foster home. For these reasons, child welfare administrators tend to use the concept of “beds” rather than “homes” when counting available placements for foster youth. What matters is not how many homes there are, but how many children can be accommodated in a placement that meets their needs. The two issues of multi-child homes and placements that are not foster homes are discussed below.

Foster homes with more than one child

As mentioned above, many if not most foster homes can accommodate more than one child. Many children come into care as part of sibling groups, and child welfare agencies hope to place as many of them together as possible. Despite their efforts, siblings are often separated in foster care for lack of homes that can accommodate them. We need more programs like Together California, a new community of 12 homes each built to house up to six siblings with full-time, professionally trained foster parents. Similar programs exist in Illinois, Florida and other states. Such homes could also be offered to relatives who are do not have room to take in larger sibling groups. The requirement of one home per child encourages states to focus on assembling as many foster homes as possible, rather than developing programs to keep sibling groups together.

Placements that are not foster homes

Secondly, A Home for Every Child should not include just foster homes. Most experts acknowledge that some children and youth need residential treatment before they can thrive in a foster home. Secretary Adams wrote in The Imprint that when the number of foster homes exceeds the number of children needing them, “[c]hildren avoid sleeping in offices or cycling through short-term rentals.” But most of the children sleeping in hotels, offices, short-term rentals, and other inappropriate placements have complex behavioral health needs and have been through many foster homes already. Media reports from California, Colorado, Hawaii, Illinois, Kentucky, Maryland, Massachusetts, Michigan, Missouri, New Mexico, North Carolina, Ohio, Oregon, Pennsylvania, Rhode Island , South Carolina, Tennessee, Texas, Washington, West Virginia, and Wisconsin document the preponderance of high-needs children among those in unlicensed, inappropriate placements.² In Adams’ own state of Idaho, the child welfare agency under his leadership reported that short-term rentals were until recently used to “temporarily house children with complex needs.” Many of these high-needs young people need residential treatment, maybe for six months or a year, before they can live in a foster home. Otherwise, they will simply be bounced from home to home until they end up in another inappropriate placement or worse.

Unfortunately Adams gives short shrift to the need for residential treatment. He writes in The Imprint that “I saw firsthand the power of this ratio [of licensed homes to children] when I ran Idaho’s child welfare system. In just one year, Idaho increased from 74 to 104 foster homes per 100 children—eliminating emergency short-term placements, expanding kinship care, augmenting prevention services, improving licensing timeliness, reducing congregate care, lowering costs, and restoring public trust.” But the press release announcing the end of Idaho foster youth being placed in short-term rentals also credited the opening of a new 16-bed assessment center which “immediately reduced the department’s reliance on temporary housing.” The center is “designed to serve as a homelike setting for foster youth until an ideal placement in a foster home or residential treatment facility [my italics] is located.” Another such center is planned. So reducing the child-to-home ratio may have helped in closing the short-term rentals, but it certainly was not the only factor that allowed these rentals to be closed.

Moreover, Idaho is still struggling to find treatment for the youth with the most complex needs, the Idaho Press reported in October 2024, shortly before the last short-term rental closed. The article concluded that “Many of the youth require residential care for mental health, substance abuse disorder or other behavioral health issues, and there still isn’t enough of this treatment available in the state.” As of December 30, 2025, DHW reports having 173 children in “congregate care,”³ of which 32 were out of state.⁴ If Idaho is like most states, it is the lack of high-quality, trauma-informed residential treatment facilities, rather than foster homes, that is the main cause of the stays in inappropriate placements. The Family First Prevention Services Act defined a new option for providing such care–a Qualified Residential Treatment Program (QRTP). But based on a recent listing of residential facilities, Idaho appears to have only two QRTP’s. Having a goal like “A Home for Every Child” provides no incentive to develop the needed therapeutic placements for youth with complex needs.⁵

Equalizing the number of licensed homes with the number of children in foster care is an arbitrary goal that does not encourage the development of needed resources, like larger foster homes and residential treatment programs. But replacing it with a one-to-one ratio of available placement slots to youth is not a good idea either. A well-functioning system needs a sizable excess capacity, because existing vacancies may not be in the right geographical area or in the right type of placement for a child who comes into care. Moreover, many existing homes and residential treatment programs are probably substandard and should be closed.⁶ A Home for Every Child encourages states to maximize the number of homes, regardless of quality, capacity, or the need for residential treatment.

What about safety?

The ratio of homes to children can be improved in two ways, as ACF has stated: increasing the number of foster homes or decreasing the number of children in foster care. But the latter means taking fewer children into foster care. An easy way to do this is to redefine “safety” to make it harder to find a child unsafe. It appears that states have been doing this already, driving foster care numbers down to their lowest level in years and raising concerns about children being left in unsafe conditions in states as diverse as Washington, Texas, Indiana and New Jersey. As Naomi Schaefer Riley puts it, “foster care numbers are close to record lows and there are plenty of reasons to think that states are pushing them lower and sacrificing child safety in the process. If the federal government provides them with an excuse to double down on their plans to reduce the number of kids in care, many states will jump at the chance.”

The Technical Bulletin states that A Home for Every Child “provides a unifying construct that reflects the full continuum of good practice in child welfare—from prevention and family strengthening to permanency and post-permanency support—offering a cohesive framework for aligning state efforts with what works.” “From prevention to permanency” conveniently leaves out the screening and investigation of child maltreatment reports. With A Home for Every Child as the metric, the incentive is clearly to reduce entries into foster care regardless of trends in child abuse and neglect.

Some language in the Technical Bulletin suggests that other factors can be considered outside of the child-to-home ratio. States participating in the pilot must also include in their PIP’s “wraparound measures related to safety, permanency, and child and family well-being, such as adverse placement scores.” But the next sentence is “Wraparound measures should be chosen such that they will show performance improvement in areas that are expected to lead to improvements in the ratio of homes to child.” But, again, the only way child protective services can do that is reducing foster care removals. Clearly, A Home for Every Child does not center child safety.

The leadership of ACF is to be commended for seeking to reduce the burden imposed by the Child and Family Services Reviews. But the attempt to build a new quality review around a simplistic slogan like A Home for Every Child is not adequate for such a complex and serious problem as child abuse and neglect. It ignores the needs for larger foster homes to keep siblings together and for high-quality residential placements. It also creates an incentive to reduce foster care entries regardless of safety. One hopes that the current pilot will be succeeded by a more comprehensive approach that keeps safety in the forefront and recognizes the diverse needs of foster youth.

Note: After this post was published, Dr. Sarah Font of Washington University in St. Louis reminded me of another problem with A Home for Every Child. Many states rely heavily on unlicensed relatives to care for children who need placements. These states can greatly increase their count of licensed homes simply by licensing these relatives, without actually increasing the number of available placements.

Notes

It does appear in the technical bulletin as an adverbial phrase in the following sentence: “In furthering efforts to achieve the 1:1 ratio of licensed foster homes to children in foster care, states will not be required to address every item or practice area identified in the CFSR Final Report.” ↩︎
In New York, these young people are often sent to residential programs that are not equipped to provide the level of care that they need. ↩︎
“Congregate care” is a term used, often pejoratively, to describe placements that are not foster homes. ↩︎
Email from A.J. AJ McWhorter, Public Information Officer, Office of the Director |Idaho Department of Health and Welfare, December 30, 2025. ↩︎
The Technical Bulletin does say that “The 1:1 ratio does not preclude states from pursuing placements outside of foster homes or from placing multiple children, for example, siblings, in one home but rather is intended to serve as an indicator spanning the full child welfare spectrum.” But the meaning of “intended to serve as an indicator spanning the full child welfare spectrum” is unclear. Placements outside of foster homes, or of more than one children in one home, do not count toward the goal of A Home for Every Child as defined by ACF. ↩︎
This statement is based on my own personal experience in the District of Columbia and Maryland as well as personal accounts from youth around the country and media reports of abuse and neglect in foster care. ↩︎

Charitable contributions to prevent abuse and neglect and protect children when it happens

by Marie Cohen

It’s the time of year when many commentators are recommending charities to which their readers can contribute. So I thought it might be useful to recommend some nonprofits whose work may prevent child abuse and neglect or that are working protect children against maltreatment when it happens. Finding such nonprofits is not easy. A charity called “Prevent Child Abuse America” devotes much of its funding to a program that has not been shown to prevent child maltreatment of any kind, while my very favorite child protection nonprofit does not even have child abuse or neglect in its name.

Protecting homeschooled kids

My favorite child protection nonprofit is the Coalition for Responsible Home Education (CRHE), a small but mighty nonprofit that advocates for legislation to protect homeschooled children, researches homeschooling outcomes and policy and develops resources for homeschooling parents, children, and those who want to help them. There have been many recent reports of children who were withdrawn from school after repeated calls by teachers to child abuse hotlines and then endured years of horrific abuse, often ending in death. In 2025, the Connecticut Office of the Child Advocate (OCA) reported that nearly a quarter of children withdrawn to be homeschooled were in families that had at least one report of abuse or neglect that was accepted for investigation by child protective services. But around the country, any attempt to regulate homeschools is met with massive resistance from well-funded organizations representing homeschoolers, spearheaded by the Home School Legal Defense Association (HSLADA) as well as state homeschooling organizations. Standing against this behemoth is CRHE, with only two paid staff. CRHE spent less than $122,000 in 2024 compared to the $16.7 million spent by HSLADA. CRHE needs more resources to achieve its mission of making homeschool safe for all children.

State Child Advocacy Groups

One might expect every state to have a group that advocates for policies to make child welfare systems more protective of abused and neglected kids. But that is not the case. Some groups with names that include “child advocacy” or “child advocates” have actually bought into the prevailing “families first” ideology, acting as though CPS investigations are more traumatic than abuse or toxic neglect. Abused and neglected children in a few states, including Arizona, Maryland, Massachusetts, Minnesota, Pennsylvania, and Washington, are lucky enough to have an advocacy group that puts children first. New Jersey and New Mexico have foster parent support organizations that also advocate for improved child protection policies. I have been shocked to find no such a group in large states like California, New York, Texas and Florida, Illinois, or in states like Connecticut, where ideology has clearly trumped concerns about child safety and wellbeing. But please let me know if I have missed any organizations that are worthy of support.

Family planning

Prevention of child maltreatment before it starts is all the rage nowadays, and of course it would be fabulous if we know how to do it. Unfortunately there is not much evidence that any programs work to prevent child maltreatment. But we do know very well how to prevent pregnancy and birth As I have written more than once, teen pregnancy, larger numbers of children, and closer child spacing are associated with child abuse and neglect. Access to family planning is not what it should be. Longer-acting methods such as IUD’s and implants, may require a repeat clinic visit, be unavailable or not covered by insurance. I support Upstream, which works to integrate family planning into primary care so that when family planning clinics close (as has been sadly happening around the country), access to contraception is not diminished. When contraception is not used or fails, abortion is the last resort, and many experts have expressed fear that the restriction of access to abortion would increase abuse and neglect and entries into foster care. I support two abortion funds, and an entire list of funds can be found here.

High Quality Early Care and Education

Free high-quality early care and education (ECE) for low-income children, starting in infancy, protects children through multiple pathways, as I have written in the past. Participation in an ECE program with staff trained in detection of abuse and neglect ensures that more adults will be seeing the child and able to report on any warning signs of maltreatment. Taking young children away from home for the day provides respite to the parent, gives them time to engage in services, and may reduce their stress, which contributes to child maltreatment. Attending quality ECE all day improves child safety by reducing the amount of time the children spend with the parents. Quality ECE programs that involve the parents can also improve child safety by teaching parents about child development, appropriate expectations, and good disciplinary practices. They may also connect parents with needed resources in the community and help them feel less isolated. I support a Washington, DC program called Educare, which provides full-day, year-round educational childcare to children in poverty from the age of six months to five years of age. Educare DC is part of a national network of 25 schools that “serve as centers of excellence, where children flourish, families lead, and communities unite to shape systems that ensure every child has access to high-quality early education.”

Helping foster youth

Let us not forget about children who are already in foster care, where they often are neglected if not actually abused. Moreover, adults who have been in foster care are more likely to abuse and neglect their children, so supporting foster youth may help break that cycle. One of my favorite District of Columbia programs is called Family and Youth Initiative, which connects caring adults with teens who are in foster care or have aged out. Adults may choose to mentor a teen or young adult, become a weekend host, or become an adoptive parent. Monthly fun events brings adult and youth together to make the connections that may ultimately lead to deeper involvement. Perhaps there is a program like this in your area that needs support.

Prevent Child Abuse, America?

Having the words “Prevent Child Abuse” in its name does not mean that an organization actually prevents child abuse. Prevent Child Abuse America describes itself as “the nation’s oldest and largest organization committed to preventing child abuse and neglect before it happens.” PCAA runs a program called Healthy Families America (HFA), which has become the largest home visiting program in the country. Billed as a child abuse and neglect prevention program, HFA has never proven its efficacy at achieving that goal. PCAA, then called The National Committee to Prevent Child Abuse, reports that it launched HFA after it “learned of the success” of a home visiting program in Hawaii.” But as I wrote in a post called The Power of Wishful Thinking, that report of success was greatly exaggerated. It was based on a pilot study with no control group and a short follow-up period. When a randomized controlled trial was finally completed and the results released in 1999, it found no impact on abuse or neglect. Early HFA evaluation results, published in the same issue, also failed to find effects on abuse and neglect in three randomized trials. But the HFA juggernaut was already in motion. PCAA does not say what percentage of its budget goes to HFA, but twenty-three members of the 45-person staff work on Healthy Families America. PCAA had total expenditures of $15.6 million in 2024. According to its 2023 IRS Form 990, its CEO earned $339,770 and the 13 members of the executive suite tegether earned over $2 million.

It is not easy to find organizations whose work actually prevents child abuse and neglect or protects children when it happens. Because withdrawal from school is such a common recourse by abusers to isolate their victims from protective adults, my most highly recommended charity is a homeschoolers’ advocacy organization called the Coalition for Responsible Home Education. Because the antecedents of child abuse and neglect are so diverse and reach so far back in time, I recommend supporting organizations that encourage family planning and. provide high-quality early care and education. And I also recommend giving to the woefully small group of organizations that truly advocate for children who are abused or neglected. Unfortunately, the name of an organization is not a good indicator of its purpose or impact. An organization with “Prevent Child Abuse” in its name spends much of what it takes in on a program that has no documented effect on child maltreatment.

Child Welfare Monitor’s Recommended Nonprofits

Abortion funds list
Center for Children’s Justice (Pennsylvania)
Center for the Rights of Abused Children (Arizona, also works in other states)
Coalition to Protect Maryland’s Children
Coalition for Responsible Home Education
Educare early childhood schools
Family and Youth Initiative connecting DC foster youth with caring adults
Friends of Children (Massachusetts)
Keeping Kids Safe (Washington State)
Miriam’s Heart Child Advocacy Network (New Jersey)
New Mexico Child First Network
Safe Passage for Children (Minnesota)
Upstream (contraceptive care)

This post was edited on Monday, December 22, to add friends of the Children, which advocates for abused and neglected children in Massachusetts.

Child maltreatment deaths raise questions about Michigan’s funding priorities

by Sarah Font (Washington University in St. Louis) and Emily Putnam-Hornstein (University of North Carolina at Chapel Hill)

I am honored to publish this post by two of the leading academic researchers in child welfare. They are also the Principal Investigators of the Lives Cut Short project, which documents child abuse and neglect fatalities around the country.

Armani Evans. Zemar King. Leviathan Froust. These are just three of Wayne County’s children who have been killed by their caregivers in recent years. Wayne County is the home of Michigan’s largest city, Detroit.

As part of the Lives Cut Short project, which aims to document child abuse and neglect fatalities nationwide, we requested and reviewed the Wayne County Medical Examiner records for child deaths since 2022. At least 52 children died due to abuse or neglect in the last 3.5 years, accounting for more than 1 in 10 of all child deaths in the county. Nearly two-thirds of child maltreatment deaths involved children ages 3 years and under.

At least nine children under the age of 3 died of illicit drug poisonings – involving fentanyl, heroin, and methamphetamine.

Equally disturbing, more than half of the child maltreatment deaths – 27 – involved intentional injury rather than negligence: children who were shaken, stabbed, beaten, and smothered. Many young children’s deaths received no media attention – all that is known is that they were killed by homicide, with the injury description merely stating “found beaten.”

The 52 children who died of maltreatment in Wayne County are likely the tip of the iceberg – these deaths are challenging to identify due to limitations in the death investigation process, minimal release of information, and other factors.

What would prevent children from dying at the hands of caregivers and family members?

Wayne County recently announced an expanded partnership with RxKids to provide thousands in no-strings-attached cash to all new and expectant mothers in 6 cities within the county. The county’s $7.5 million investment adds to a statewide investment of $250 million in RxKids for 2025-2026 alone. The governor’s FY2026 budget recommendation further includes $27 million to provide “economic and concrete supports” with the goal of reducing or avoiding involvement with Child Protective Services.

The leaders of RxKids imply on their website and other materials that their cash transfers can produce a large decline in child maltreatment and reduce the need for CPS intervention. Fortunately, a rigorous evaluation of the program was conducted in Flint.

The punchline? No impact.

Such findings should come as little surprise when we take seriously the threats that children face. Neither drug addiction nor extreme violence seems likely to be ameliorated with short-term monthly checks. And many children died after CPS ignored clear warning signs. A wrongful death lawsuit filed on behalf of murdered Detroit toddler Chayce Allen reveals that relatives asked CPS to intervene on at least 13 occasions.

The likely reason so many kids are left to die in horrifying circumstances is that Michigan has a severe shortage of child protection caseworkers. Statewide vacancy rates are 20% and the problem is worse in Wayne County, which has 46 fewer caseworkers than intended, leading to high caseloads and turnover. High caseloads were one of the systemic problems that the state was expected to address as part of the Dwayne B. settlement – a case filed nearly two decades ago. Michigan seeks to exit court supervision as soon as this summer, despite their continued failure to adequately staff their system.

Before massive expansions of cash assistance – much of which is going to families who are not impoverished – perhaps the state should fulfill its existing obligations to kids.

Note: some deaths handled by the Wayne County Medical Examiner may stem from incidents occurring in surrounding counties (from which children were brought to and then died at a Wayne County hospital). Our data do not provide the location of the maltreatment incident.

Home visiting: more hype than hope

Image: homvee.acf.gov

by Marie Cohen

Home visiting has been a highly popular intervention for the prevention of child abuse and neglect and for addressing a much broader set of social problems as well. But the research has never supported the efficacy of home visiting programs as a whole for improving child and family outcomes. The latest study of four popular home visiting programs found that all these programs have negligible impacts after five to seven years. But there was no hint of this message in the government’s press release or the report itself. The bipartisan belief in home visiting is prevening a needed examination of home visiting’s impact and the level of resources devoted to it.

A Brief History of US Home Visiting

While home visiting has existed since Elizabethan times in England, its history in the U.S. began in the late nineteenth century with charities seeking to address urban poverty by changing the behavior of poor families. While it is now considered to be the solution to a number of different social problems mostly related to poverty, modern home visiting was conceived as a way to prevent child abuse and neglect. Publication of Henry Kempe’s The Battered Child in 1968 brought about the recognition of child maltreatment as a national problem. To address child abuse, Kempe called for universal prevention through a network of home health visitors. Inspired by Kempe, Hawaii’s Healthy Start Project (HSP) began in 1975. In 1977, David Olds began testing his Nurse Home Visiting program in Elmira, New York. The first Parents as Teachers program was created in 1991. In 1992, the National Committee to Prevent Child Abuse (now Prevent Child Abuse America) rolled out Healthy Families America (HFA).

In 1993, the Future of Children, an influential academic journal produced by Princeton University and the Brookings Institution until 2021, devoted an issue to home visiting. In the summary article, the authors cautioned that the research so far was limited and had mixed results, but opined that the results were “promising enough” to recommend the expansion of existing programs and the continuation of evaluation efforts. Home visiting programs burgeoned in the wake of that issue, with funding from federal, state, and foundation sources.

In 1999, The Future of Children released its second issue on home visiting, containing evaluations of six demonstration programs. The results were sobering. In their analysis of all six studies, Deanna Gomby and colleagues concluded that “[I]n most of the studies described, programs struggled to enroll, engage and retain families. When program benefits were demonstrated, they usually accrued only to a subset of the families originally enrolled in the programs, they rarely occurred for all of a program’s goals, and the benefits were often quite modest in magnitude.” The one exception was the Nurse Home Visiting Program, (now Nurse-Family Partnership), which differed from the other programs in being delivered by nurses rather than paraprofessionals, and which produced some sizable impacts on child abuse and neglect and second births to mothers.

But the home visiting juggernaut was already in motion. Programs continued to grow, funded by multiple sources, and most of the growth was not in the most promising (and expensive) Nurse-Family Partnership. The National Center to Prevent Child Abuse, renamed Prevent Child Abuse America in 1999, made HFA its signature program despite the lack of evidence that it prevents child abuse. According to the National Home Visiting Resource Center, “evidence-based home visiting was implemented in all 50 states, the District of Columbia, 5 territories, 32 Indigenous communities, and 65 percent of U.S. counties” by 2024. These programs reached over 284,000 families through more than three million home visits in that year, and another 70,000 families were reached by 11 “emerging models.” Of the more than three million home visits provided, approximately 14 percent were provided virtually, down from nearly 23 percent in 2024, as services gradually returned to in-person after the pandemic. Today, there are multiple home visiting programs with different target groups, curricula, goals, and personnel. In addition to the 24 models recognized by the federal government, there are an unknown number of “emerging models” which have not yet earned the label of “evidence-based.”

Undaunted by the scant evidence of success, Congress established, with bipartisan support, the Maternal, Infant and Early Childhood Home Visiting Program (MIECHV) as part of the Affordable Care Act of 2010. The goals of the program are to improve the health of mothers and children, prepare children for success in school, improve families’ economic well-being, connect families to other resources in the community, prevent crime and domestic violence and prevent child injuries, abuse and neglect. The funds can be used to implement any one of 24 models that meet the eligibility criteria established by HHS’ Administration for Children and Families (ACF). An evaluation of this program was required by the legislation.

Home visiting programs also became a popular intervention in child welfare with the growing emphasis on keeping children with their families. This began in 1994 with the Title IV-E waivers and continued with the Family First Prevention Services Act (FFPSA), signed by President Trump in 2018. FFPSA allowed states to use Title IV-E funds, formerly reserved for foster care and subsidized adoptions, to support children and families and prevent foster care placements through in-home parent skill-based programs, as well as mental health, drug treatment and kinship navigator services. Programs had to be approved by the new Title IV-E Prevention Services Clearinghouse as “evidence-based” before they could be included in a state’s Family First Prevention Services Plan. According to a research brief from Chapin-Hall at the University of Chicago, at least one home visiting model was included in the Family First Prevention Plans in 28 states as of April 2023. Most commonly included were Parents as Teachers (28) and Healthy Families America (23), with Nurse Family Partnership in third place with 15 programs, and SafeCare in fourth with seven programs.

Home visiting also became popular in growing efforts by child welfare agencies to invest in preventing child abuse and neglect before it occurs, or at least before a family comes to the attention of child protective services. A small source of federal funds, the Community Based Child Abuse Prevention Program (CBCAP), was established by Congress in 1996 to fund such prevention programs and is commonly used to fund home visiting as well.

The bipartisan enthusiasm for home visiting has been unwavering. Created under Obama, MIECHV has been supported by every succeeding administration. Total federal funding on this program is slated to increase from $500 million in 2023 to $800 million in 2027.Earlier this year, the Senate even passed (unanimously) a bipartisan resolution designating April 21 through April 25, 2025 as National Home Visiting Week. Even the Trump Administration has heartily endorsed the home visiting. Yet, the much-vaunted evidence for the value of home visiting really consists of a series of modest impacts affecting different outcomes, often based on less reliable indicators like self-reports, and dwarfed by a sea of findings of no effect. Even the one program (Nurse-Family Partnerhip), that had the most promising early resultsm has no stood up to recent replications–though additional trials with the population that seems to benefit most may be warranted.

Home visiting program evaluations

There have been multiple studies of home visiting programs, including both randomized controlled trials (RCT’s) and comparison group studies, and together these studies have generated hundreds of papers. Therefore, Child Welfare Monitor (CWM) drew from a summary of research on Nurse-Family Partnership from the Arnold Ventures Social Programs that Work website; the evidence assembled on the website of the Title IV-E Prevention Services Clearinghouse for Healthy Families America, Parents as Teachers, and SafeCare; and the Home Visiting Evidence of Effectiveness (HomVEE) Review conducted by the Administration on Children and Families of HHS for Early Head Start Home-Based Option. CWM consulted the original studies as needed, focusing on RCT‘s because randomization is the best way to rule out selection bias as the explanation for any differences between the intervention group and the control group. Otherwise, one cannot know whether the group that participated in the program differed in significant but unmeasured ways from the members of the comparison group. Appendix I includes more details about the program evaluations. Appendix II focuses on the challenges in measuring child abuse and neglect and what the research suggests.

Nurse-Family Partnership

Nurse -Family Partnership (NFP) connects first-time mothers and their babies with a specially trained nurse, who works with the mother and child from early in the pregnancy through the child’s second birthday. It differs from other models in using registered nurses to deliver the visits, making it more expensive and dependent on a scarcer group of providers. Nurse Family Partnership (NFP) has been the subject of RCT‘s in Elmira, NY (launched in 1988); Memphis, TN (launched in 1990), Denver, CO (1994) and in a larger statewide trial in South Carolina that started in 2016. It has also been tried internationally in British Columbia, the Netherlands, the United Kingdom and Germany; though the differences between populations and systems make these results less applicable to the United States, they can be seen as suggestive. The participants in the demonstrations were all first-time mothers but other criteria for inclusion varied by study. Among the effects that were replicated in two or more of the studies identified by the Arnold Ventures’ Social Programs That Work website were reductions in medical treatment for injuries and accidents in Elmira and Memphis), reductions in subsequent births to mothers in their late teens and early twenties (in Elmira, Denver, and Memphis), and an improvement in cognitive or academic outcomes for the children of mothers with lower psychological resources, like intelligence, mental health, and self-confidence (in Memphis and Denver). There were few significant impacts on children’s emotional and behavioral outcomes and mothers’ life trajectories in terms of employment, income and crime, and those impacts were not repeated in more than one study.

A recent scaled-up NFP replication in South Carolina was disappointing, producing no significant positive effects on any of the three primary outcomes studied: the rate of adverse birth outcomes, mothers’ rate of subsequent births within 21 months, and child health, measured as “a composite of mortality or health care utilization associated with major injury or concern for abuse or neglect.” Nor did the researchers find any significant improvements for a prespecified subsample of “socially vulnerable families” that were similar to the families for which earlier studies found positive impacts. There were also no significant findings for the secondary outcomes, which related to healthcare utilization. These results were sobering, and the researchers suggest that the rapid scaling up of the program and the broader population served may have contributed to the weaker results. David Olds, the founder of NFP, suggests¹ that the program’s effectiveness was affected by the relative inexperience of the nurses (due to the rapid implementation), the use of impersonal recruitment methods (unlike Elmira and Memphis, where nurses personally recruited mothers in clinic waiting rooms), and the relatively more advantaged clientele. The program included any pregnant woman who qualified for Medicaid in South Carolina, which funded 46 percent of births in that state in 2023 and includes women up to 200 percent of the poverty level. So it was a less disadvantaged group than was included in the other US demonstrations, and earlier studies suggested that the more disadvantaged benefited more from this program.² Thus, further study of NFP with a highly disadvantaged population should be considered.

Healthy Families America (HFA)

Healthy Families America, an initiative of the national organization, Prevent Child Abuse America, is a flexible program that allows local communities to choose their eligibility criteria, parenting materials, and staff. Services last for a minimum of three years and up to five years. Based on three RCTs of Healthy Families America, the Title IV-E Prevention Services Clearinghouse counted 18 favorable “contrasts” (their term for comparisons between the intervention and control groups) compared to four unfavorable contrasts and 211 contrasts displaying no effect. Among the 18 favorable contrasts, 11 were on self-reports of maltreatment or child behavioral and emotional functioning. Of the other impacts, all were from only one RCT. Many outcome categories showed no favorable impacts. These included child safety measured by administrative reports; child safety based on injuries, Emergency Room use or hospitalizations; child permanency based on out-of-home placements; child social functioning; child physical development and health; parent/caregiver substance abuse; and economic and housing stability. One of the four RCT‘s (Healthy Families Oregon) showed no positive impacts at all.

Parents as Teachers

Parents as Teachers is delivered by “parent educators” who work with families from pregnancy through kindergarten. The Clearinghouse identified one American RCT of this program, one RCT from Switzerland, and one matched comparison group study. The one US RCT, which evaluated two separate demonstrations in California, found that PAT had “little effect on parenting knowledge, attitudes or behaviors as measured in these demonstrations. Nor were significant effects noted on child health or health care.” The demonstration did find small positive effects (a two-month gain at the age of three) on “self-help development” in one of the two sites but no significant effects on cognitive development at either site. In the Swiss study, the clearinghouse found one significant effect on one of two measures of the child’s expressive language.

SafeCare

SafeCare is a brief home visiting model that is delivered in 18 one-hour sessions. SafeCare is designed for parents and caregivers of children birth through five who are either at-risk for or have a history of child neglect and/or physical abuse. It was included in the Title IV-E Prevention Plans of seven states in April 2023. The contrasts reported by the Clearinghouse include only one favorable effect–on foster care placement–based on a matched comparison group study and not an RCT and one unfavorable effect (on child welfare reports), along with 19 findings of no effect.

Early Head Start-Home-Based Option

Early Head Start Home-Based Option provides weekly visits to pregnant women, infants, and toddlers until the child is three years old. The goal is to promote school readiness of young children by enhancing their cognitive, social and emotional development. The federal reviewers of Early Head Start’s Home-Based Option used nine publications based on a large federal RCT of the program at 17 sites as well as the early results of the federal study of MCHIEV programs, which is described below. There were no favorable findings on maternal or child health; child maltreatment; or delinquency, family violence and crime. There were a few favorable impacts on child development and school readiness; positive parenting practices; and family economic self-sufficiency scattered among multiple findings of no effect. These effects were not consistent across age groups or outcomes where one would expect some alignment (like reading to children vs. reading at bedtime). These impacts appeared to be small, although the lack of standardized effect sizes complicates interpretation. In the final report on Early Head Start, the authors stated that at “the end of the program, when children were three, impacts were modest in size and Early Head Start children continued to perform below national norms on cognitive and language assessments.” By the time the children reached fifth grade, all but one favorable impact earlier reported was gone.

The MIHOPE Study

The legislation establishing the Maternal, Infant and Early Childhood Home Visiting Program (MIECHV) required an evaluation of the program in its early years. The study, named the Mother and Infant Home Visiting Program Evaluation (MIHOPE), included 88 local programs in 12 states and was carried out by an evaluation nonprofit called MDRC. All of the programs were based on one of the four models most commonly chosen by states in their initial MIECHV plans–Early Head Start Home-based Option, Healthy Families America, Nurse-Family Partnership, and Parents as Teachers. A total of 4,229 families enrolled in the demonstrations between 2012 and 2015. Strangely, even though the sample sizes were large enough to estimate the effects of each program model, the researchers opted to report findings for all four models taken as a whole, a decision that has been criticized by experts and funders because of the significant programmatic differences between the models. Early findings released in 2019 from when the children were 15 months old were disappointing. There was little difference between the experimental and control groups. About a third of the 63 outcomes measured were statistically significant and though most comparisons did favor the home-visited groups, the effect sizes were extremely small–too small to be of any practical significance. The authors reported that for most outcomes the effects were slightly smaller than the average effects found in past studies of the models (which were already modest).

On September 11, 2025, the Administration on Children and Families released the long-term results of the MIECHV program evaluation conductd five to seven years after enrollment, when children were in kindergarten or first grade. Summing up their findings in a press release, ACF asserted that “MIECHV-funded home visiting significantly improved maternal and family wellbeing for participants five-to-seven years after enrolling in services…MIHOPE found statistically significant and positive effects of home visiting” for the five categories of maternal and family well-being outcomes. For the three categories of child outcomes, the researchers found “some evidence of positive effects,” but only one was statistically significant and positive.

Unfortunately, the researchers did not distinguish between a statistically significant effect and an effect which is large enough to be meaningful. A more sober analysis by the Coalition for Evidence-Based Policy shows that home visiting produced “negligible outcomes” for each of the eight expected impacts. The review points out that the average effect size across the six statistically significant or near-significant categories of effects was 0.03 – the equivalent of moving the average child or mother from the 50th to the 51st percentile. The study did not report model-specific effect sizes, but the reviewers noted that they are likely to be small or modest given that few reached statistical significance. It is also worth mentioning that many of the outcomes came from maternal answers to a caregiver survey, introducing the possibility of bias..

It is also concerning that the evaluation team at MDRC used a different analysis plan to assess impacts for the kindergarten study than it did for the earlier results. The new analysis plan was dated January, 2025 (but not released until May 2025). The new methodology combined all the 66 outcomes studied into eight new “research questions” some of the outcomes were included under more than one research question. The researchers chose to focus not on the significance of each individual impact, as was done in the earlier report. Instead, they decided to use a method called “omnibus testing” to compute an overall significance level for each research question.³ By choosing this method, they were able to find significant results (at the 0.10 level) for five of the eight research questions, where looking at each outcome would have shown only eight significant outcomes out of 86, or less than 10 percent of the outcomes. Since the new analysis plan was dated January 2025, it seems likely that it was developed after an analysis of the data (which was collected in 2021 and 2022) under the old plan yielded unsatisfactory results.⁴ The Imprint has published a more detailed critique of the MIHOPE study by Sarah Font and Emily Putnam-Hornstein called The Odd Bipartisan Effort to Oversell the Evidence for Home Visiting.

Why have home visiting programs been so unsuccessful at changing outcomes for most children and families?

As Deanna Gomby stated back in 1990, “home visiting programs have struggled to engage and retain families.” Research has documented low levels of enrollment and engagement of families at risk of maltreatment in voluntary services in general and home visiting in particular. According to the MIHOPE implementation report, 17 percent of the home visiting group never even received one visit, as compared to 12 to 22 percent in previous studies. All the models expected families to participate at least until the child’s second birthday, with services available for two or three years longer in three of the programs. Yet, only 46 percent of families were still participating in home visits 12 months after their first visit, consistent with previous research. On average, families who received at least one home visit went on to participate for an average of eight months. While participating, families received fewer visits than expected by the models in which they were enrolled. In the first 12 months, less than 60 percent of families received at least half of the visits prescribed by their model, a result consistent with prior research. Part of the problem might be that many people who need the kind of help that home visiting is designed to provide do not want to let a stranger into their home to scrutinize their parenting and family functioning. Child Welfare Monitor has heard in the District of Columbia and elsewhere that there is an oversupply of home visiting slots, with too few people wanting to participate.

Misleading Congress, the Media and the Public

Ever since the initial excitement about home visiting, there have been high hopes for this service delivery method and unwillingness among policymakers on both sides of the aisle to discard their hopes. At the same time, the federal government along with many advocacy groups, has endorsed a vision of “evidence-based practices” that asks only for a minimal number of statistically significant impacts, with no concern about the size of the impacts or the nature of the evidence–whether it is self-reported, self-contradictory, or unsupported by more than one study. As a result, studies that show only a few modest impacts that may be statistically significant but not meaningful in size or corroborated by other studies can be cited as evidence of program success.

Unfortunately, program evaluations are technical enough that readers who are not schooled in the intricacies of research methods are often forced to rely on the researchers’ interpretation of their findings. The usually well-informed Congressional Research Service has stated that “A large body of research suggests that some home visiting models or services can benefit children and their parents.” Less surprisingly, the press is easily misled. Due to lack of time or expertise in the intricacies of social science research, reporters often simply report what is in the press release announcing new research results. For example, the Imprint, a widely circulated outlet for child welfare content, repeated in its podcast the government’s misleading proclamation about good news from the MIHOPE study.

—–

No matter how painful the process, legislators, agency officials and advocates should remove their blinders about home visiting. It is time to phase out the MIECHV program. State and local governments should begin scaling down their home visiting programs and funneling the money to other uses that are currently underfunded. In this time of budget scarcity, it is time to stop throwing good money after bad. We need new ideas and meaningful evaluations that can bring about the implementation of programs that actually work. If money is being wasted on home visiting services that are not making a difference, or not even being used, surely there are better uses of these scarce funds.

Appendix I

In the absence of time to review the hundreds of publications on the Nurse Family Partnership Program, I used the excellent evidence summary on the Arnold Ventures Social Programs that Work website. For the studies of HFA, Parents as Teachers, and SafeCare, I relied mostly on the compilation of study results provided by the Title IV-E Prevention Services Clearinghouse. For Early Head Start, which was not included in the Clearinghouse, I used the Home Visiting Evidence of Effectiveness review conducted by the Office of Policy Research and Evaluation (OPRE) of the US Department of Health and Human Services.

Nurse Family Partnership

Each of the four U.S. RCT’s of NFP had a different population and eligibility criteria for participants. Elmira is in a small, semirural county in New York State which had the highest rate of child abuse and neglect in the state at the time of the study. In 1980, the community was rated the lowest Standard Metropolitan Statistical Area in the United States for economic conditions. Program participants In Elmira were either teens, unmarried, or low-income, and all were White. They were actively recruited by nurses at the prenatal clinic, private obstetricians’ offices, Planned Parenthood, schools, and other health and human services agencies. In Memphis, the program admitted pregnant women with no previous live births who did not have a chronic illness that might affect the fetus and who had at least two of the following risk factors: being unmarried, having less than 12 years of education, and being unemployed. Ninety-two percent of the women enrolled were Black, 98 percent were unmarried, 64 percent were eighteen years old or younger and 85 percent came from households with incomes at or above the poverty level. In Denver, women were recruited at any point in their pregnancy if they had no previous live births and either qualified for Medicaid or had no health insurance. Almost half of the two groups were Hispanic, another 35 percent were Caucasian, and 16-17 percent were Black. The South Carolina demonstration recruited pregnant women who were 15 years or older and eligible for Medicaid, which meant their incomes were less than two percent of the poverty level. The participants were mainly non-Hispanic Black (54.9 percent) and non-Hispanic White (35.0). In addition to enrolling a different population, each study looked at a different set of outcomes, and sometimes at different times as well, making it hard to compare the results. The findings of the demonstrations are summarized below.

Child safety based on child welfare administrative reports: There was no statistically significant difference among confirmed child maltreatment reports between the experimental and control groups in Elmira. But there is some evidence that the Elmira program reduced child maltreatment among the participants who were most at risk, those who were teenagers, poor and unmarried. Among this group, 19 percent (or a total of eight) of the poor, unmarried teens had a verified maltreatment report compared to only one of their nurse-visited counterparts. But this effect was statistically significant at the p-0.7 level, not the standard level of 0.05. And there were no treatment-control differences in verified maltreatment reports during the two years after the program ended. However, the Elmira study did find a large impact on verified CPS reports when the children were 15; nurse-visited children had received on average of 0.29 verified CPS reports compared to 0.54 in the comparison group. This result is difficult to explain and one wonders if it was due to chance.This outcome was not examined in Memphis or Denver. The Netherlands study also found a large, statistically significant impact on child welfare administrative reports, where the researchers found that 11 percent of the visited children had a CPS report compared to 19 percent of the control group children during the first three years of their lives.
Child safety based on health care for injuries and ingestions: Nurse visited children in Elmira had fewer emergency room visits for injuries and ingestions in their second year of life than the control group. (This was not measured during the first year, when they were less mobile and able to get into trouble. No effect size was provided). When they were between 25 and 50 months old, they had 40 percent fewer mentions for injuries and ingestions in their medical records and 45 percent fewer mentions of child behavioral or parental coping problems. In Memphis the researchers found that in the first two years of life, nurse-visited children had an average of 0.43 health encounters for injuries or ingestions compared to 0.56 for the control group, or 23 percent fewer encounters. They also spent an average of 0.04 days hospitalized for injuries and ingestions, compared to 0.18 days for the control group– a 78% decrease. A more recent evaluation of a large scale implementation of NFP in South Carolina found no difference between the experimental and control groups on child health, measured as “a composite of mortality or health care utilization associated with major injury or concern for abuse or neglect.” In the UK Study, there were no differences in the rates of emergency hospital visits for the experimental and control groups.
Child wellbeing based on behavioral and emotional functioning: This outcome was not included in Elmira at two and four years. But there were large favorable effects on self-reported arrests and convictions for the Elmira children at age 15-19–a finding that was not reported anywhere else and was not matched by effects on other outcomes like high school graduation, teen pregnancy, engagement in work or school at 19, or self-reported substance use or welfare receipt, all of which could have been expected to covary with the arrests and convictions. So it is not clear whether these results occurred by chance. In Memphis there were no effects found on children’s reported behavioral problems at two years, nor were there any results on youth behavioral functioning when they got older. There were significant favorable effects on child emotional functioning in Denver at two years and four years. By ages six to nine in Denver, behavioral and emotional effects were consistently favorable but did not reach statistical significance at standard levels, perhaps because the sample size was not large enough.
Child wellbeing based on cognitive functions and abilities: In Elmira there were no statistically significant cognitive effects on children in the first two years, although the researchers observed “improved intellectual functioning of nine to 11 points on the developmental tests for children from the highest risk families. Although only marginally significant statistically, the researchers observed that it is of clinical importance. Treatment effects in this range are consistent with those obtained for children of this age enrolled in intensive early childhood intervention programs aimed specifically at enhancing cognitive development.” In Memphis, there were no effects at two years on children’s mental development but there were substantial statistically significant effects on academic performance at age 12 for the children whose mothers were in the lower half of the sample on intelligence, mental health and self-confidence. In Denver, there were favorable effects on the cognitive development of children born to mothers with low psychological resources in the two-year follow-up. This group also did better cognitively at ages six to nine but the findings only occasionally reached statistical significance and may be due to attrition differences between the intervention and control groups. But at age 18 there were “sizable, significant” effects on two of three cognitive outcomes for this subgroup in Memphis. Cognitive effects were not studied for Elmira or Denver 18-year-olds.
Maternal life course: When the children were aged 15, the Elmira study found that nurse-visited mothers had 19 percent fewer births than control mothers, an average of 1.3 births compared to 1.6. In Memphis, the mothers had 16 percent fewer births in the first six years of the program. They had caught up by the time the children were 12, but the increase in birth spacing is still a significant favorable outcome. In Denver, home-visited women had fewer subsequent pregnancies (29 percent vs. 41 percent) and births (12 percent vs. 19 percent) by their children’s second birthdays. There was no impact on the rate of second pregnancies after two years in the South Carolina, British Columbia, and UK studies. Where reported, there were no effects on adverse birth outcomes, maternal employment, likelihood of partnership or marriage with the child’s father, substance abuse, psychological distress or foster care placements (mentioned only in Memphis).

Healthy Families America

The contrasts presented by the Title IV-E. Clearinghouse were based on four RCT’s that were rated highly for design and execution by clearinghouse staff. The results of each RCT are based on multiple research papers published for each major study. Reviewing the Clearinghouse’s tabulation of the data, and sometimes comparing it to the actual publications to which it referred, raised several questions about the overall effectiveness of the program:

Child safety measured by child welfare administrative reports: There were no favorable or unfavorable outcomes, as compared with 43 contrasts showing no statistically significant effect.
Child safety, based on maternal self-reports about whether they maltreated their children: There were five favorable contrasts, 38 contrasts with no effect, and one unfavorable contrast. It is hard to be confident about the validity of self-reports of maltreatment, as one could easily imagine the program participants having learned more about what to report, and under-reporting behaviors (such as spanking) that they had been taught were undesirable. The large number of contrasts with no effect is worth noting.
Child safety based on injuries needing medical care, hospitalizations, and emergency use: There were no favorable or unfavorable impacts and 11 contrasts showing no effect.
Child permanency based on out-of-home placements: There were six contrasts showing no effect, and none showing a positive or negative effect.
Child well-being: Behavioral and Emotional Functioning: Five contrasts showed a positive effect, two with no effect, and none with a negative effect. All of the five positive effects were reported by Healthy Families Alaska and were fairly large. But all of these were based on the caregiver’s report of the child’s behavior, and self-reports are not sufficient on their own for making conclusions about impact. Moreover, these outcomes and measures were not replicated in any other study.
Child well-being: social functioning: The Clearinghouse reports no favorable or unfavorable effects and and two contrasts showing no effect.
Child well-being: cognitive functions and abilities: There were two favorable impacts, one unfavorable impact, and 6 contrasts showing no effect. The two favorable impacts came from Alaska and were not found in any other evaluations.
Child well-being: physical development and health: The Clearinghouse reported no favorable or unfavorable impacts and six contrasts with no impact.
Child well-being: delinquent behavior. There was one favorable effect in the one contrast available, which was “child skips school often.” A look at the publication containing this result, which was a report on the RCT of Healthy Families New York (HFNY) seven years after random assignment, showed that fewer children self-reported skipping school, but this result was not supported by reports from their mothers.
Child well-being: educational attainment: The Clearinghouse reported one favorable impact and two findings of no impact. All three findings came from one publication from the HFNY RCT. The researchers found that children in the HFNY group were about half as likely to be retained in first grade (3.54 percent) than children in the control group (7.10 percent), based on official school data. However, there were no impacts found for the other two educational attainment outcomes used by the Clearinghouse–performing above or below grade level in reading or math. Moreover, this contrast was not available from any other study.
Adult well-being: positive parenting practices: There were three favorable impacts and 24 findings of no impact. All of the favorable impacts were from another report on HFNY that was based on observations of how the mothers interacted with their children as they completed three tasks–a puzzle solving task, a delay of gratification task, and a cleanup task. I was not able to judge the size of the effects; all were statistically significant at the 0.05 level. However, there were no significant effects on observed presence of harsh parenting during the same tasks. Moreover, this outcome was not included in the evaluation of any other program.
Adult well-being: parent/caregiver mental/emotional health. The Clearinghouse found three favorable impacts and 16 contrasts showing no impacts from a total of three RCT’s.
Adult well-being: Parent/Caregiver Substance abuse: There were no favorable or unfavorable effects, and 15 instances where no statistically significant effect was found.
Adult well-being: family functioning: There were three favorable impacts, one unfavorable impact, and 28 instances of no impact. The three favorable impacts stemmed from three different contrasts related to Intimate Partner Violence (IPV)–overall maternal IPV victimization rate (child age 1-3), maternal IPV victimization rate: physical assault (child aged 1 to 3), and maternal IPV perpetration rate: physical assault (child age 1 to 3). The size of the effect was not provided and there were nine other maternal IPV contrasts when the child was aged 1 to 3 that showed no effect. There were ~~were~~ no impacts on IPV when the child was older.
Adult well-being: economic and housing stability. There were no favorable impacts, five contrasts showing no impact, and one showing an unfavorable impact.

Parents as Teachers

The results presented by the Title IV-E Clearinghouse are based on two RCT’s and one study based on a matched comparison group. Even when counting all these programs, the results are not impressive.

For child safety based on administrative reports, the Clearinghouse noted two contrasts with a favorable effect and two with no effect. The effect size and implied percentile effect calculated by the Clearinghouse were very small. Moreover, these results were based on a matched comparison group rather than an RCT, casting doubt on the validity of the results.
Child permanency (out-of-home placement): The clearinghouse cited no favorable or unfavorable findings and one finding of no effect.
Child well-being: social functioning. The original article cited by the Clearinghouse, based on an RCT in two California sites, reported that PAT children in one of the sites benefited significantly, advancing by about two months of the control group in self-help development but did not report significant results for the other site or for social development at either site.
Child well-being: cognitive functions and abilities: Based on the American and Swiss RCT’s, The Clearinghouse reported two favorable findings and 10 findings of no effect. But one of the findings was actually of no effect for the PAT-only group; it was the “PAT plus case management group” that experienced an impact.
Child well-being: Physical development and health: The clearinghouse reported no favorable or unfavorable effects and three findings of no effect from one RCT.
Adult well-being: positive parenting practices: The Clearinghouse reported no favorable or unfavorable effects and one finding of no effect from an RCT.
Adult well-being: family functioning: The Clearinghouse reported no favorable effects, 8 findings of no effect, and one unfavorable effect, all from one RCT.
Adult well-being: economic and housing stability. The Clearinghouse reported no favorable effects, one unfavorable effect, and nine findings of no effect, all from one RCT.

SafeCare

SafeCare is a brief home visiting model that is delivered in 18 one-hour sessions. SafeCare is designed for parents and caregivers of children birth through five who are either at-risk for or have a history of child neglect and/or physical abuse.It was included in the Title IV-E Prevention Plans of seven states in April 2023. The contrasts reported by the Clearinghouse include only one favorable effect–on foster care placement–based on a matched comparison group study and not an RCT and one unfavorable effect (on child welfare reports), along with 19 findings of no effect.

Early Head Start Home-based Option

The Early Head Start Home-based option serves low-income women and families with children under three years old. They receive a minimum of weekly 90-minute home visits and two group socialization activities per month. The findings discussed here are based on the HHS Office of Policy Research and Evaluation (OPRE) review of the research on home visiting. OPRE reports that it reviewed 23 “manuscripts” and identified nine of those manuscripts that were based on “impact studies rated high or moderate quality.” By focusing on “manuscripts” instead of studies, OPRE obscured the fact that seven of these manuscripts were actually based on the same study–a large federal demonstration of EHS programs in 17 sites conducted between 1996 and 2002. Of the nine manuscripts, five were based on the full study and two were based on results from one Utah site only. The other two studies reviewed were based on results of the MIHOPE study of four home-visiting models when the children were 15 months old. All but two of the manuscripts were rated high by the OPRE staff in quality for methodology. The manuscripts based on the Utah study and the grade five follow-up for the national study were rated “moderate” in quality because of high attrition. In the nine publications reviewed, there were no favorable findings on maternal or child health; child maltreatment; or delinquency, family violence and crime. There were a few favorable impacts on child development and school readiness; positive parenting practices; and family economic self-sufficiency. These effects were not consistent across age groups or similar outcomes (like reading to children vs. reading at bedtime.) These impacts appeared to be small, although the lack of standardized effect sizes makes the importance of the effects hard to estimate.

For child development and school readiness, the reviewers reported five favorable findings from the 17-site study and the Utah study. All the other 66 contrasts related to child development and school readiness in the two studies showed no effect. By the time the children in the main study reached fifth grade, no effects remained.
For positive parenting practices, the reviewers reported 10 favorable findings from the 17-site study and the Utah study. In total, there were 64 findings of no effect in this area. By the time children reached fifth grade, one favorable impact (which was not noted for the three or five year-olds) was observed.
For family economic self-sufficiency, the reviewers reported 16 favorable findings, one unfavorable finding and 88 findings of no effect from 3 publications in a total of two studies. No economic effects remained by the time the children were in fifth grade.

Appendix II: Home visiting and child maltreatment

Analyzing the effect of any program on child maltreatment poses unique difficulties because it is such a difficult outcome to measure. Obviously, the evaluators cannot see what goes on in a household after the visitor has gone home. Evaluators have used three types of measures to estimate the effects of home visiting programs on child maltreatment–verified child protective services (CPS) reports, health care encounters for injuries or ingestions (or simply emergency room visits), and self-reports of abusive or neglectful behaviors through surveys like the Conflict Tactics Scale.

The most obvious measure of abuse and neglect is official Child Protective Services (CPS) data, but there are several problems with CPS data as a measure of maltreatment. The number of maltreatment reports that are confirmed (substantiated) by CPS is most frequently used, but it is known to be an understatement. Many cases go unreported, and reported cases are often not substantiated. Another problem is that verified abuse is a relatively rare event in a population and a study may not have enough participants to detect it. Finally, h visitors are mandatory reporters and their presence in the home introduces surveillance bias; these families are under more surveillance than families in the control group and may receive more reports for that reason.

Olds and his colleagues did not find statistically significant differences in substantiated CPS reports for the whole program group during the two year period that families participated in the Elmira demonstration or in the subsequent two years. But they found some evidence that the Elmira program did reduce child maltreatment among the participants who were most at risk–those who were teenagers, poor and unmarried. About 19 percent (or a total of eight) of the ~~the~~ poor, unmarried teens had a verified maltreatment report compared to four percent (or one) of their nurse-visited counterparts. But this effect was statistically significant at the p-0.07 level, not the standard level of 0.05. And there were no treatment-control differences in verified maltreatment reports for this subgroup or the whole treatment group during the two years after the program ended. The researchers speculated that this may be due to increased surveillance on the nurse-visited group, because the nurses connected them to other providers before the programs ended.

However, a surprising finding emerged when the children were 15 years old. By that age, nurse-visited children had received on average of 0.29 verified CPS reports compared to 0.54 in the comparison group–a large and highly statistically significant difference. The investigators hypothesized that as young first-time parents mature and develop, small positive changes that [occur while they are in the program] can build and multiply over time, yielding larger effects in later years.” The mechanism by which the Elmira program had such delayed effects is hard to understand. Perhaps it occurred by chance. But in any case, a replication would be necessary to give it credence, and this outcome was not measured in Memphis or Denver.

As an alternative to CPS data, some researchers have used data on health care encounters for children’s injuries or ingestions. Many of these encounters may reflect abuse or neglect but they also would include cases that are not due to either abuse or neglect and would leave out many instances of maltreatment as well. But it is certainly a good indicator of safe parenting. In the four-year followup of the NFP Elmira group, when the children were 25 to 50 months old, the researchers found that nurse-visited children had 40 percent fewer injuries and ingestions (according to notations in their medical records) and ~~and~~ 45 percent fewer notations of or child behavioral or parental coping problems. Nurse-visited children also made 35 percent fewer visits to the emergency room. In the NFP Memphis trial, the evaluators found that nurse-visited children had an average of 0.43 health encounters for injuries or ingestions compared to 0.56 for the control group, or 23 percent fewer encounters in the first two years of their lives. They also spent an average of 0.04 days hospitalized for injuries and ingestions, compared to 0.18 days for the control group. But a more recent evaluation of a large scale implementation of NFP in South Carolina, described above, found no difference found between the experimental and control groups in its composite measure of child mortality and major injury related to abuse or neglect.

Other studies have used parent self-report measures such as the Conflict Tactics Scale. This measure is less valid than the other two because many parents are reluctant to report abusing or neglecting their children. A few studies found positive effects on such measures but without any corroboration from more objective measures.

Notes

Conversation between Marie Cohen and David Olds, October 22, 2025
↩︎
Similar disappointing results from a study in the United Kingdom may have been influenced by a control group that received an average of 16 home visits from a public health nurse and 11 visits from a midwife through the child’s second birthday, as well as targeting a lower-risk population than most of the other studies. A study in British Columbia found no reductions in its primary outcome – child injuries by age two years – or in subsequent maternal pregnancies by the child’s second birthday. The authors speculate that British Columbia’s more comprehensive health and social services may explain the lack of effects. ↩︎
The authors organized the outcomes into five “research questions” focusing on maternal outcomes and three research questions focusing on “child outcomes.” The “maternal outcomes” included “outcomes that could be improved through direct interaction between parents and home visitors;” maternal mental and behavioral health; parent-child interactions; conflict, violence, aggression and maltreatment;” and families’ economic circumstances. The child outcomes included “children’s social-emotional functioning in the home context; children’s social-emotional functioning at school; and children’s cognitive, language and early math skills. ↩︎
Conversation with Emily Putnam-Hornstein, who made me aware of the revised 2025 research plan. ↩︎

This post was edited on November 10, 2025 to add a sentence and links about enrollment and engagement in home visiting and a link to an article about the MIHOPE report.

Child welfare and community norms: a troubling divergence

This summer, I was asked by a reporter to comment on a terrible case in the District of Columbia. Twenty-month old Kemy Washington died of starvation and dehydration, after her mother had had overdosed on a mix of MDMA, cocaine, ethanol and the animal sedative xylazine. An older sibling had been removed from Kemy’s mother due to her neglect and that child’s guardianship with a relative had been ratified only days before Kemy was born. Yet, Kemy was never on the radar screen of the Child and Family Services Agency until her grandmother made two calls, which were apparently screened out as not involving child abuse or neglect. When I read the more than 600 comments on the article, I was struck that over and over again, commenters asked the same question. How is it that a mother who had a previous child removed from her due to neglect could give birth to another child without triggering an investigation, close monitoring, or even removal of the child? The reporter asked me the same question and I explained that such a policy, though clearly logical to members of the community, would never be accepted by the current child welfare establishment, where it would be viewed as an unacceptable infringement on parents’ rights.

I have often remarked on situations where child welfare policy or practice departs from general community norms. Whether it is the continued screening out of calls on the same family, even if a child was previously removed; the refusal to consider policies that would trigger investigations when a new child is born to a parent who committed severe abuse or neglect; the push to “reunify” children with parents who have proved over and over again that they cannot keep them safe or even stop harming them, it seems that policymakers and practitioners of child welfare are operating from a different set of norms than the public. What would be clear to a grandparent, a neighbor, or a random layperson do not seem so evident for those who are charged with protecting our children. This was made very clear in a devastating report on child fatalities in Minnesota. As the authors put it,

Members of the public often express dismay and outrage to us over stories such as those recounted in this report. We infer from this that the professional norms currently guiding child protection and foster care are out of alignment with those of the broader community.

Examples of this divergence abound, and I am sharing just a few here.

“B.B.” was born in the State of Washington in 2022 and died of fentanyl poisoning in March, 2023. Starting in 2014, the Department of Children, Youth and Families (DCYF) had received 30 reports about B.B.’s family for the use of heroin, marijuana and alcohol in the home; lack of supervision of the children; domestic violence; an unsafe adult living with the family; an unsafe and unclean physical environment; unsecured guns in the home, “out-of-control” behaviors by B.B.’s older siblings at school to which the mother was unresponsive, seeming “out of it;” concerns about the children’s hygiene; and the mother driving under the influence of marijuana. A few days before B.B.’s death, a caseworker told B.B.’s mother that the agency was closing a voluntary services case that had been open for about half a year. But the family was already under investigation again at the time B.B. died.

There have been multiple cases around the country of child protection workers disregarding reports of hungry children eating food from the floor or the garbage, until a child finally died or escaped from the torturers. School staff in Utah reported making at least four calls in the 2022-2023 school year (confirmed by a timeline released by DCFS) to CPS alleging that Gavin Peterson was always hungry and eating food from the trash. School staff were buying his lunch but had to stop after his father and stepmother forbade it. One school staffer “said Gavin’s father and stepmother justified Gavin’s small frame and constant desire for food as side effects from his medication, but she didn’t buy it.” Why did CPS? That summer, Gavin was withdrawn from school. A year later he was dead after years of starvation and beatings.

Seven-month-old Emmanuel Haro is missing and presumed dead. Investigators believe that he was abused for an extended period of time before he was killed. But his suffering and death could have been avoided. His father had been arrested for abusing a child from a previous marriage in 2018–abuse so serious that the child is now bedridden. A simple “birth match” policy could have prevented the death of Emmanuel Haro. If birth records were linked to criminal and CPS records, Emmanuel’s birth could have triggered a mandatory investigation and monitoring because the father had been convicted of child abuse–the kind of policy that commenters in the Kemy Washington case were asking for. It is hard to think of a more common-sense idea than birth match. Yet, only five states had such a policy in 2022, according to my report on birth match for the American Enterprise Institute. And most of these programs are very limited both in terms of which parents are covered and of the state response.

Even a birth match policy would not help in cases where a parent’s violent history is known but disregarded. Four-year-old Rykelan Brown died from a beating by his father, Joshua Emmons, in May 2024, two months after he was removed from a loving foster home to be placed with Emmons. The foster parents had repeatedly reported that Rykelan came home from visits with his father bruised and saying his father hit him and he never wanted to go there again. The local Department of Social Services knew that in 2019 Emmons had beat his then-girlfriend’s three-year-old son so severely that he damaged the child’s liver, which must now be checked regularly. But the social services commissioner told an interviewer that the event occurred too long ago to be considered. Really? Even when paired with Rykelan’s bruises and reports of beatings?

As the above examples show, some things that are intuitive to ordinary people – -like that a child going to school hungry (and not because of poverty) – is a sign that something is deeply wrong at home–seemed to be missed by people engaged in child welfare practice, administration and policymaking. Much of the problem stems from a dominant ideology that preaches that abused and neglected children are almost always better off with their own families. The same viewpoint holds that what child welfare calls neglect is just poverty, as if all poor parents neglect their children, and that child welfare is a a racist system that was created to destroy Black and Brown families.¹ Social work schools have adopted and promulgated these positions and agencies have incorporated them in the training for new social workers. Deep-pocketed groups like Casey Family Programs have used their money to foster this ideology through training and technical assistance to state and local agencies. The entire child welfare community in many states has found itself endorsing policies and practices that defy common sense thinking.

We must bring child welfare policy and practice back into alignment with community norms. But that is easier said than done. The public pays little attention to child welfare until there is a tragic fatality or egregious incident that is covered in the media. But many of these cases are never known to the media and therefore to the public. And even when they are, child welfare agencies often refuse to release information about their past involvement with the family, in violation of federal law. So the press, the public and legislators cannot identify what went wrong and what would be needed to prevent future tragedies in the future.

A small but useful first step to align child welfare systems with community norms would be to make the public aware of decisions that clearly violated these norms and harmed children. The federal government should enforce the requirements of the Child Abuse Prevention and Treatment Act (CAPTA), which as interpreted in the federal Child Welfare Policy Manual, requires states to issue specific information and findings on all child maltreatment fatalities and near fatalities caused by maltreatment. That includes information about past dealings between the children’s families and the child welfare agency. Ultimately, the requirement must be expanded to cover all “egregious incidents” where maltreatment is suspected.² Increased public awareness how child welfare agencies knowingly and routinely leave children in harm’s way may help elevate child welfare into a major issue, not a backwater that gets addressed only when there is a tragedy.

Three family members are awaiting trial for murdering 11-year-old Arabella McCormick in August 2022 and torturing her sisters. A kindergarten aide in Arabella’s class told a reporter that she got a disturbing note from Arabella’s foster mother, who went on to adopt her and then allegedly participate in her murder. “In the envelope, it said, ‘Arabella is,’ and it was line items such as ‘a terrible child,’” she said. “‘She’s a liar. You can’t believe anything she says. She’s a thief. She steals everything. Don’t trust her.’ It was just one thing after another of horrible things that you would never say about a 6-year-old.” The teacher’s aide told the grand jury that she contacted child protective services (CPS) after Arabella arrived at school school in the same dirty clothes on several occasions. She also told CPS that Arabella wasn’t allowed to eat fruit, accept rewards or participate in recess with other children. “And the lady from CPS said to me on the phone — after I told her everything, she said, ‘Well, it could be worse,’” the teacher’s aide told the grand jury. Really? I don’t think most members of the public would agree.

Notes

In fact, child welfare systems initially involved White children only. Black children were originally excluded from public child welfare systems. ↩︎
Both Colorado and Wisconsin release information on cases meeting this description. ↩︎

The continued decline in foster care placements: What, if anything, are children and families getting instead?

On May 9, 2025, the US Administration for Children and Families (ACF) announced that the number of children entering foster care had continued to decrease in FFY 2023. Entries to foster care dropped from an estimated 264,000 in FFY 2018 to 175,282 in FFY 2023–a drop of 33.6 percent. But foster care is not the only service provided by child welfare agencies after an investigation or assessment finds that a child needs protection. Instead of being placed in foster care, some children and their families receive “in-home services,” which aim to ameliorate the risks to their safety without removing them from their homes. With the large drop in child removals, it is natural to ask whether home-based services are being provided to more families as fewer children are placed in foster care. Unfortunately, the data to answer to that question is not available on a national basis or for most states.

Clearly we cannot determine what is happening to the children who “would have been placed in foster care” under previous policies and practices. But at least we can ask if declining foster care placements are being offset by the opening of more cases for in-home services–which some states call Intact Family Services, Family Maintenance, or Family Preservation Services, among other terms. These are the services that are provided to children and their parents after an investigation or assessment determines that the situation does not meet the criteria for foster care placement (usually imminent danger to the child) but that the family does need services and monitoring to reduce risks to the children. In a major investigation, Texas Public Radio (TPR) found that a 40 percent drop in child removals by the state’s Department of Family and Protective Services (DFPS) over a six-year period was not accompanied by an increase in family preservation services (as they are called in Texas) but instead by a “radical curtailment” of such services. Children were left in dangerous homes with no services or monitoring.

For many years, many powerful groups like Casey Family Programs blamed the alleged lack of federal reimbursement for services to families, like drug treatment, mental health care, and parenting support, for the removal of many children who could have been helped at home. As I have pointed out, this argument was largely spurious because states were already providing these services using Medicaid and other sources. In any case, the passage in 2018 of the Family First Prevention Services Act (FFPSA) was supported by advocates as a means to alleviate this alleged problem by making funds available for these in-home services. Given the repeated use of that narrative by proponents of FFPSA, it would have been logical to mandate that states report on the numbers of children receiving such in-home services, the characteristics of these children and the services they receive. But such a mandate was not included in the Act.

States are currently required to provide child welfare data to the Children’s Bureau via two separate systems. They must submit foster care and adoption data through the Adoption and Foster Care Analysis and Reporting System (AFCARS). A separate system, the National Child Abuse and Neglect Data System (NCANDS), includes data on abuse, neglect, and child protective services. NCANDS includes numbers of children receiving “postresponse services,” or services provided as a result of needs discovered during an investigation or assessment. But these numbers are provided in a form that is neither meaningful nor comparable to the AFCARS data, and therefore does not allow the comparison of foster care and in-home numbers for each state and the nation.¹

Not only does the federal government not ask the states for meaningful data on in-home cases; most states do not provide such data on public-facing platforms. A review of reviewed publicly available data for the 12 states with the largest number of children in foster care yielded only two (California and Texas) that provided enough data on in-home services to answer the question of whether declines in foster care removals have been offset by the opening of in-home cases. The results are discussed below.

California

California has far more children in foster care than any other state, 43,095 children at the end of Federal Fiscal Year 2023, which was 12.6 percent of the national total, according to AFCARS data. Fortunately, California Child Welfare Indicators Project (CCWIP), a collaboration between University of California at Berkeley and the California Department of Social Services, provides excellent data on all child welfare services provided in California. The CCWIP dashboards include data on the number of entries into foster care and the opening of Family Maintenance cases, as in-home cases are called in California. According to CCWIP, entries into foster care declined from 26,766 in Federal Fiscal Year (FFY) 2019 to to 17,071 in FFY 2024. Rather than increasing to make up for the drop in children entering foster care, the number of children with Family Maintenance case openings as the first service component declined from 25,887 to 18,441 over the same period. The total number of children with cases opened for child welfare services dropped from 44,747 to 29,936 over the five-year period–a drop of 33.1 percent.² Rather than a shift from foster care to in-home services, there has been an reduction in the number of children reached by child welfare services.

Texas

Texas has the third largest state foster care caseload, behind California and Florida. The number of Texas children entering foster care dropped from 16,028 in 2021 to 9,623 in 2022, an incredible 40 percent, and then stayed about the same for the following two years. Confirming the conclusions of TPR, the number of children entering family preservation services dropped from 51,806 in FY 2020 to 39,655 in FY2021 and then to 26,132 in 2022–half of the 2020 total. The drops in foster care and in-home case openings were presumably related to a new Texas law that changed the definition of neglect and put restrictions on child removals. In addition, a 2020 policy change restricted eligibility for Family Preservation Services in Texas, according to the TPR investigation. TPR found that the number of families that DFPS rejected for family preservation services jumped from 300 to about 2,800 in the first year the policy took effect. Perhaps in an effort to rectify the extreme curtailment of services, the number of children with in-home case openings increased in 2023 and 2024, while the number of children entering foster care remained stable. Nevertheless, the overall trend over time was a decrease in foster care entries, and in children with new family preservation cases.³

A look at the two figures above shows a major difference between the California and Texas. In California, substantially more children entered foster care (17,071) than entered family maintenance services (13,473) in FFY 2024. But in Texas, the number of children entering family preservation services (42,855) was more than four times the number of children entering foster care (9,220) in FY 2024. This vast difference makes clear that nationally we have no idea whether more investigations result in foster care or in in-home services. And we don’t know whether the number of children with new in-home cases has increased as the number being placed in foster care has gone down. Moreover, we don’t know what kind of in-home services parents and children are getting. Are children in California’s Family Maintenance Services getting a similar package of services as those in Texas’ Family Preservation Services? We just don’t know. That is not an acceptable state of affairs, especially given the use of federal funds for these services.

In order to make the needed information available, new legislation might be required. Section 479 of the Social Security Act required the establishment of a system to collect data on foster care and adoption in the United States. AFCARS was created based on general guidelines laid out in the law and more detailed regulations promulgated by HHS. But now it is time to expand AFCARS to include data on child welfare services delivered in children’s homes when a case is opened for the provision of such services. The same sorts of data on entries, exits, caseloads, demographics, and reasons for the initiation of services as is required for foster care should be included for in-home servcies. Required variables should also include the types of services provided to each family and whether they are voluntary or mandatory with court involvement.

While data from Texas and California suggest that child welfare has been reducing its footprint since 2018, we cannot assume that this is happening nationwide without data. That is why ACF should request that Congress act to require that states submit data on in-home cases. Even with this information we will not know how effective in-home services are in protecting children andhelping their parents ameliorate the conditions that put their children at risk; there is very little research that addresses this question. More and better research on that issue is needed. But not even knowing whether children who would have once been placed in foster care are now being protected at home while their parents are helped to make them safe is simply not an acceptable situation.

Notes

The numbers provided include foster care as well as in-home services. The number of children provided is a duplicated count. And the number of children receiving each type of services is not provided. ↩︎
These totals are lower than the sum of the children with foster care entries and in-home case openings, presumably because some children entered in-home services and were placed in foster care in the same year. ↩︎
Texas does not provide the total number of children entering services of any type as does California. Adding those entering foster care to those entering in-home services would produce an inflated estimate since we do not know the number of children who may have entered both types of services during the year ↩︎