The Bodies the Count Forgets: How States Bury Child Abuse Deaths That CAPTA Has Required Them to Disclose Since 2003

Every year, the federal government publishes a number for how many American children were killed by abuse or neglect. For federal fiscal year 2024, that number was 1,773. Nearly every serious researcher in the field — including the federal commission Congress created to study the question — believes the real figure is two to three times higher, somewhere between 3,000 and 5,000 children. The gap is not an accident of statistics. It is the product of a reporting system in which twenty-four states submit death data only for children their child protection agencies already had a file on, in which the largest single loophole was written by HHS itself in a policy manual, and in which the disclosure mandate Congress attached to federal grant money more than two decades ago has never once cost a state a dollar. The children most likely to disappear from the count are precisely the ones the system had already touched — the ones whose deaths would document a specific agency's specific failure on a specific date.

The Statute, and the Sentence That Gutted It

The legal obligation is not ambiguous. Section 106(b)(2)(B)(x) of the Child Abuse Prevention and Treatment Act — codified at 42 U.S.C. § 5106a(b)(2)(B)(x) — conditions a state's eligibility for CAPTA grant funds on the state having "provisions which allow for public disclosure of the findings or information about the case of child abuse or neglect which has resulted in a child fatality or near fatality." The requirement entered federal law through the 1996 CAPTA amendments and was carried into the state-plan eligibility conditions by the Keeping Children and Families Safe Act of 2003. Every state governor has certified compliance with it, in writing, every year since.

The word doing the destructive work is "allow." Congress did not write "shall publish," "shall release," or "shall disclose within thirty days." It wrote a requirement that states have provisions which allow disclosure — a mandate satisfied by the existence of a permission, not the exercise of one. HHS then narrowed it further. The federal interpretation lives in the Child Welfare Policy Manual at Section 2.1A.4, Question 8, revised in September 2012, which tells states they need not release the case record and may satisfy the requirement by releasing "the full investigation; a summary of the investigation; or a statement." A statement. A state agency can meet a federal grant condition triggered by a dead child by issuing a paragraph confirming that a child died.

The near-fatality half of the mandate is weaker still. Federal and state law define a "near fatality" as an act that, as certified by a physician, places a child in serious or critical condition. Not a nurse, not a hospital record, not a caseworker's observation — a physician's certification. As Texas officials have themselves conceded, there is no agreed-upon clinical meaning for "serious" or "critical condition" among hospitals or medical professionals. A child beaten into an intensive care unit generates a federal disclosure obligation only if a doctor happens to write the right two words. The result is a category that states can define into near-nonexistence.

A 2024 study of all fifty states and the District of Columbia — Keeping the Public in the Dark, produced for the Lives Cut Short project at the American Enterprise Institute — found that twelve states allow but do not require release of fatality and near-fatality information, and only seventeen have laws or policies requiring the release of any information without someone first filing a request. Only eleven states — Arizona, Arkansas, Colorado, Florida, Nevada, Oregon, Pennsylvania, Rhode Island, Tennessee, Texas and Wisconsin — post public notification of any child fatality, near fatality, or egregious incident at all. In the other thirty-nine, a child can be killed by a parent whose file sits open on a caseworker's desk, and nothing whatsoever appears in public unless a reporter knows to ask.

Half the Bodies: How the National Count Gets Made

The federal count is assembled by the National Child Abuse and Neglect Data System, or NCANDS, run by the Children's Bureau within the Administration for Children and Families. Submission to NCANDS is voluntary. Massachusetts simply did not submit fatality data for federal fiscal year 2023 — Child Maltreatment 2023 was compiled from 49 states, the District of Columbia and Puerto Rico — and Massachusetts had also failed to report FFY 2019 fatalities. No consequence followed either omission.

For those states that do report, the definitions do not match. Some count only deaths substantiated as maltreatment under state civil law; some count only deaths where the death certificate says homicide or child maltreatment; some exclude accidental deaths made possible by egregious neglect, such as an infant who drowns in a bathtub while a supervising adult is unconscious from fentanyl. NCANDS does not even collect a date of death, which makes it impossible to reconcile the federal file against a coroner's records or a state vital-statistics database.

The measured size of the resulting gap is consistent across two decades of independent work. GAO-11-599, published in July 2011, found that child welfare officials in 28 states told federal auditors their own state's official number was probably or possibly an undercount. Peer-reviewed research examining three states found state child welfare records missed between 55 and 76 percent of maltreatment fatalities. Death certificates are no better as a backstop: more than 60 percent of child abuse deaths may never be coded as abuse on the certificate itself. The Commission to Eliminate Child Abuse and Neglect Fatalities — created by the Protect Our Kids Act of 2012 (P.L. 112-275) and chaired by Dr. David Sanders, executive vice president of Casey Family Programs — concluded in its March 2016 final report Within Our Reach that against a NCANDS figure of 1,670 for 2015, the true toll was likely 3,000 to 5,000.

The most damning contemporary evidence is that amateurs with a search engine can beat the government. Lives Cut Short — a joint project of AEI and the UNC-Chapel Hill School of Social Work, involving AEI senior fellow Naomi Schaefer Riley and UNC's Emily Putnam-Hornstein, faculty co-director of the Children's Data Network — builds its database from media reports, publicly released state fatality reviews, and medical examiner records requested from selected jurisdictions. Using only those incomplete public sources, the project counted more child maltreatment deaths than NCANDS received from ten states for FFY 2022 and twelve states for FFY 2023. As of 2026 it has compiled over 4,000 incidents from 2022 onward, many of which appear in no state or local statistic.

Then there is the trend line, which is where the undercount becomes something closer to a falsification. NCANDS reported approximately 2,000 maltreatment deaths for FFY 2023 at a rate of 2.73 per 100,000 children. One year later it reported 1,773. An eleven-percent single-year drop in child homicide and fatal neglect is not a public health achievement anyone announced, funded, or can explain. It is a reporting artifact — and HHS published it as a trend.

The Twenty-Four-State Rule: Only the Children We Already Knew

Buried in the data-quality notes of the federal report is the single most consequential fact in this entire system: nearly half of states — 24 — report fatality information to NCANDS only for children who had a previous CPS case. GAO documented the same practice in 2011 and it has not been fixed in fifteen years.

Read plainly, that is a rule guaranteeing the number is wrong. Every child killed by abuse who was never previously reported to a hotline vanishes from those states' submissions entirely.

But it also produces a second, stranger distortion that cuts the other way and matters more for accountability. The federal report's own prior-contact figures show that in 15.2 percent of confirmed fatalities, CPS was involved with the family or the child at the time of the death, and in 37.8 percent, CPS had been involved with the child or the perpetrator in the past. Those are the cases that carry institutional liability — the ones where a screening decision, a closed investigation, a reunification order, or an unmade home visit can be traced to a named worker, a named supervisor, and a dated document. Those are exactly the cases where a state's disclosure obligation bites hardest, and exactly the cases that a "we may release a statement" standard is designed to absorb.

The Commission's most-cited finding drives the point home: a call to a child protection hotline — regardless of whether it was screened in, investigated, or substantiated — is the single best predictor of a later child abuse or neglect fatality. The screening decision is the moment of maximum lethal risk and minimum public visibility. Of roughly 10.9 million referrals in a representative year, about 60.9 percent were screened in and only 12.8 percent of all referrals ended in substantiation. Everything filtered out at the front door is invisible at the back.

The Money: Ten Billion Dollars, No Strings on the Count

The financial architecture explains why nothing changes.

The disclosure mandate is attached to the smallest pot. CAPTA Title I State Grants have run in the range of roughly $90 to $105 million annually in recent years — the Community-Based Child Abuse Prevention program under CAPTA Title II received $70.7 million in FY2024, and Children's Justice Act grants distribute about $17 million a year to states and territories out of the Crime Victims Fund, at $50,000 per state plus a per-child-population increment. HHS's own Office of the Assistant Secretary for Planning and Evaluation has published a brief stating flatly that the CAPTA State Grant program "represents less than one percent of child welfare spending" while imposing extensive requirements on states.

The money that actually runs the system carries no fatality-data condition at all. Federal Title IV-E foster care, prevention and permanency support was estimated at $10.1 billion for FY2026, following $9.5 billion in FY2023 — $5.1 billion reimbursing states for foster care, $4.3 billion for adoption and guardianship subsidies, $172 million for placement prevention. Title IV-E is a mandatory, open-ended, permanent entitlement: it expands with what states spend and claim, and it is not conditioned on whether a state counts its dead children honestly or discloses a single fatality file.

So the leverage is 1 percent, the entitlement is 99 percent, and the 1 percent has never been pulled. In 2010, the last time Congress substantially renewed CAPTA, a Senate committee — already aware states were underreporting deaths — directed HHS to develop policies that would force transparency. HHS did not. The GAO's 2011 recommendations were not implemented. There is no record of HHS withholding CAPTA funds from any state over a missing, incomplete, or falsified fatality count.

The comparison to the Children's Bureau's other enforcement tool is instructive. In roughly twenty-five years of Child and Family Services Reviews — the federal audit of every state child welfare system, which carries statutory financial penalties under 45 C.F.R. § 1355.36 — no state has ever passed. Penalties are routinely rolled into a Program Improvement Plan and suspended, then renegotiated and suspended again. Funds have actually been withheld once: Idaho, $935,712, in November 2022 — and only because Idaho conceded outright that it had failed to complete its improvement plan. A system that has assessed one real penalty in a quarter century against universal failure is not an enforcement regime. It is a paperwork ritual.

The Closed Room

Where states do review deaths seriously, they do it behind a legal wall built specifically to keep findings out of court and out of print.

Virginia's statutes governing its state and local child fatality review teams (Va. Code §§ 32.1-283.1 and 32.1-283.2) require every team member and every person presenting information to execute a sworn confidentiality statement, and provide that team information and records "shall not be subject to subpoena or discovery or be admissible in any criminal or civil proceeding." South Carolina closes its Child Fatality Review Team meetings and exempts them from the state Freedom of Information Act when individual deaths are discussed. Across states, the portions of meetings where actual cases are examined are closed, and minutes and materials generated in closed session are sealed from public inspection. Findings may be published only "in statistical or other form which shall not identify individuals."

The structure is defensible in the abstract — candor requires protection — and indefensible in practice, because the confidentiality shield extends to the agency's own conduct. A review board can determine that a caseworker ignored three reports and closed a case in violation of policy, and that determination is sealed, unsubpoenaable, and inadmissible. The family gets nothing. The public gets an aggregate. The next family gets the same caseworker.

What It Costs to Pry One File Loose

The record of what happens when someone tries to obtain these records is a record of attrition by legal expense.

Kentucky. The Louisville Courier Journal and the Lexington Herald-Leader fought the Cabinet for Health and Family Services for seven years before the state conceded that its files on child deaths and near-fatalities are public records. The Kentucky Court of Appeals ordered disclosure; the Cabinet ultimately paid the newspapers nearly $700,000 in penalties and legal costs.

Texas. The Austin American-Statesman sued the Department of Family and Protective Services and settled out of court for roughly 1,000 pages of day-care death records plus $5,000 toward its fees. Texas Public Radio, pursuing fatality documents after a 2022 survey found two-thirds of child welfare judges believed cases were arriving in court "more severe," was quoted a state cost estimate of more than $30,000. Texas has since deemed its more robust fatality review audits confidential; the last such report made public dates back nearly a decade. A Pulitzer Center StoryReach U.S. Fellowship investigation independently documented more than 1,200 Texas children who died from abuse and neglect between 2018 and 2023.

Illinois. Reporting by Capitol News Illinois and the Illinois Answers Project established that more than 1,200 deaths and more than 3,000 serious injuries have met the statutory criteria for incident-specific public reports since July 2018 — and DCFS simply did not produce them, arguing that inspector general reviews and internal death review teams sufficed, though those cover fewer deaths, exclude hundreds of injuries annually, and are not public records. Cook County Public Guardian Charles Golbert asked the state auditor general and the DCFS inspector general to investigate. In June a judge found DCFS had "willfully and intentionally violated" the Illinois FOIA, ordered production within five days, and imposed costs, fees, and a $5,000 civil penalty. On August 11, 2026, a judge ordered the DCFS director to personally answer for repeated failures to release public documents in a child death case.

Maryland. The Baltimore Banner found that the Department of Human Services failed to complete the required Form 1080 reports for about one-third of suspected abuse or neglect deaths in 2023 and 2024. The agency first denied the records request as too "burdensome," then conceded the documents did not exist. Spokesperson Ben Shnider attributed the gap to recordkeeping predating the current administration; the state is building a $210,000 database fix. Those caseworker reports are the raw material of fatality review — without them, Maryland's review process was examining patterns in a file that was one-third empty.

District of Columbia. The Child and Family Services Agency refused to release the names of children, parents and caregivers in fatality cases despite D.C. Code language requiring it, redacted parental criminal history, substance abuse, mental illness and domestic violence as "unrelated to the child fatality," and took more than six months to respond to a single request.

Ohio. InvestigateTV's "Failed and Forgotten" series found that in 2017 nearly 100 Ohio children died in unsafe sleep environments while the state counted 20, and that five of 34 children whose deaths coroners ruled homicides involving abuse or neglect were left out of the state maltreatment tally entirely.

The civil settlements, meanwhile, establish that these deaths are worth enormous sums once liability is proven — Oregon approved an $80 million settlement in 2026 for two children its agency failed to protect; Washington paid $17 million over the death of two-year-old Jose Fernandez Armas, killed days after DCYF closed an investigation the agency admitted was incomplete, and $14 million over four-year-old Kennan Ford; a $60 million claim was filed against Arizona's Department of Child Safety after the death of eleven-year-old Chaska Davis; California's Tulare County paid $32 million in J.G. v. County of Tulare. Every one of those figures is a market valuation of information the state fought to keep sealed.

Who Is Supposed to Be Watching

CAPTA requires citizen review panels, and requires each panel to evaluate whether the state agency is fulfilling its child protection duties, including a review of fatalities and near fatalities. The panels have no subpoena power, no enforcement authority, and no mechanism compelling a state to act on a recommendation. Academic assessment of them has described them as "misunderstood, neglected, and underutilized."

The Children's Bureau reviews CAPTA state plans and accepts the governor's certification. It has never treated a fatality-disclosure failure as a plan violation. The GAO recommended fixes in 2011 that remain unimplemented. The Senate directed action in 2010 that never came. The CECANF delivered a national strategy in 2016 whose central recommendation — that states examine all child abuse and neglect fatalities from the prior five years — has not been adopted federally.

The incentive structure is coherent from every seat. A state agency that reports more deaths generates litigation exposure, legislative hearings, and a director's resignation; a state that reports fewer generates a press release about declining maltreatment. HHS gains nothing from a finding that would embarrass an agency it must partner with and force it to defend a fund cutoff in front of a congressional delegation. Nobody in the loop is paid to produce an accurate number, and the one entity with standing to demand it — the family of a dead child — is the one the confidentiality statutes were drafted to exclude.

What Would Actually Fix It

The remedies are unglamorous and entirely within Congress's existing authority.

Change one verb. Amend § 5106a(b)(2)(B)(x) from "provisions which allow for public disclosure" to a mandate: affirmative release, without request, within a fixed number of days, of a specified minimum record set — the fatality investigation, all prior reports and dispositions concerning the child and the perpetrator, and every agency action taken or declined. Repeal the CWPM 2.1A.4 #8 "or a statement" option.

Make NCANDS submission mandatory and uniform. Require a single federal definition of a maltreatment fatality, require a date of death, prohibit the practice of reporting only children previously known to CPS, and require every state to reconcile its submission against state vital records, medical examiner findings, and child death review team determinations — publishing the discrepancy.

Attach the condition to the money that matters. Fatality-count integrity should be a condition of Title IV-E administrative claiming, not of a grant worth less than one percent of child welfare spending. A state that cannot say how many children in its care died should not be drawing federal administrative reimbursement while it figures that out.

Split the shield. Preserve confidentiality for the identities of surviving siblings, reporters and clinical informants. Strip it from the agency's own conduct — its decisions, its policy compliance, its timelines. There is no legitimate privacy interest of a government agency in the record of how it handled a child who is now dead.

Fix the near-fatality definition. Replace physician certification of "serious or critical condition" with an objective clinical trigger — ICU admission, intubation, or an injury-severity threshold — so a category that currently depends on a doctor's word choice becomes a category that depends on what happened to the child.

Until then, the arithmetic stands as it is. Somewhere between 1,200 and 3,200 American children a year are killed by abuse or neglect and do not appear in the federal number. A substantial share of them were already known to the agency charged with protecting them. Their deaths are reviewed in closed rooms, their findings sealed against subpoena, their names withheld from the public record — and the federal grant condition written to prevent exactly this has, in twenty-three years, cost no state anything at all.

---

Sources: - 42 U.S. Code § 5106a — Grants to States for child abuse or neglect prevention and treatment programs (Cornell LII) - Child Welfare Policy Manual, Section 2.1A.4 Q&A (ACF/HHS) - Child Maltreatment 2024 (HHS Children's Bureau) - Child Maltreatment 2023 (HHS Children's Bureau) - GAO-11-599, Child Maltreatment: Strengthening National Data on Child Fatalities Could Aid in Prevention - Within Our Reach: A National Strategy to Eliminate Child Abuse and Neglect Fatalities (CECANF) - Keeping the Public in the Dark: How Federal and State Laws and Policies Prevent Meaningful Disclosure (AEI) - Keeping the public in the dark about child fatalities and near fatalities (Child Welfare Monitor) - A Jumble of Standards: How State and Federal Authorities Have Underestimated Child Maltreatment Fatalities (AEI) - Child Maltreatment 2023: How the Government Continues to Undercount Maltreatment Fatalities (AEI) - Child Maltreatment 2024: Clear Evidence that Federal Trend Data Cannot be Trusted (Lives Cut Short) - Lives Cut Short — About Us / Data (AEI & UNC-Chapel Hill) - Near-fatalities and egregious incidents: Information is scarce and inconsistent (Lives Cut Short) - District violates federal law requiring release of child fatality information (Lives Cut Short) - Act 33's Promise Failed: Pennsylvania Faces an Urgent Reckoning (Lives Cut Short) - The CAPTA State Grant Program Represents Less Than One Percent of Child Welfare Spending (HHS ASPE) - Child Welfare: Purposes, Federal Programs, and Funding (CRS IF10590) - 45 C.F.R. § 1355.36 — CFSR penalties (Cornell LII) - Child and Family Services Reviews: Idaho Final Report (ACF) - Failed and Forgotten (InvestigateTV) - Illinois' child welfare agency failed to produce critical reports after child deaths (Capitol News Illinois) - Judge orders DCFS director to answer for repeated failures to release public documents in child death (WGLT) - Maryland DHS failed to document many deaths from suspected child abuse or neglect (The Baltimore Banner) - Texas Keeps Child Abuse and Neglect Deaths out of Sight (Pulitzer Center) - Statesman wins access to child death records (Austin American-Statesman) - Child abuse records must be disclosed to newspapers, Ky. appellate court rules (Reporters Committee for Freedom of the Press) - High-profile child deaths spark push for welfare agency transparency (Stateline) - Va. Code § 32.1-283.1 — State Child Fatality Review Team; confidentiality - Va. Code § 32.1-283.2 — Local and regional child fatality review teams; confidentiality; immunity - $80 Million Settlement Highlights the Cost of Child Welfare Agency Failures (Ross Law PDX) - $60 Million Claim Filed Against Arizona Department of Child Safety (Gallagher & Kennedy) - Citizen review panels in child protection: Misunderstood, neglected, and underutilized (Children and Youth Services Review) - Child Fatality Review (Pediatrics, AAP)