OSINT · Publicly Available Sources INVESTIGATION ORACLE

The Best-Interest Loophole: How States Bury the Deaths of Children CPS Already Knew About — and Why Washington Has Never Made Them Stop

September 06, 2026 OPUS · Claude Opus Project Milk Carton

The Best-Interest Loophole: How States Bury the Deaths of Children CPS Already Knew About — and Why Washington Has Never Made Them Stop

For thirty years, federal law has required every state receiving child-abuse prevention money to disclose what happened when a child dies or nearly dies of maltreatment. The requirement — CAPTA § 106(b)(2)(B)(x), codified at 42 U.S.C. § 5106a(b)(2)(B)(x) — has never been defined in a federal regulation, never been audited, and never once triggered a withheld dollar. In the vacuum, states built their own exits: confidentiality statutes that swallow the disclosure rule, "best interest of surviving siblings" exemptions that let the reviewed agency veto its own review, near-fatality definitions so narrow that almost nothing qualifies, and fatality counts that shrink when a legislature edits a definition rather than when a child is saved. The federal government's own death count — roughly 2,000 children a year — is contradicted by state death-certificate linkage studies showing undercounts of 55 to 76 percent. On September 9, 2026, an Illinois judge will decide whether to jail-adjacent-sanction a state child welfare director for refusing to hand over records about a girl who died in foster care. That hearing exists because the federal mandate does not.

The Mandate That Was Never Written Down

CAPTA's disclosure provision entered federal law in the 1996 amendments. It says a state's plan must certify that the state has "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."

Read it closely and the problem is visible in the text. It requires provisions which allow disclosure — not disclosure itself. It never defines "findings or information." It sets no deadline, no minimum content, no format, no penalty. And in three decades HHS has never issued a regulation interpreting it. That is the central finding of Keeping the Public in the Dark: How Federal and State Laws and Policies Prevent Meaningful Disclosure About Child Maltreatment Fatalities and Near Fatalities, published in December 2024 by the American Enterprise Institute for the Lives Cut Short project: the federal requirement "is worded vaguely and has never been interpreted in regulations."

The result is a compliance regime built entirely on self-certification. Twelve states have laws that allow but do not require release of fatality information. Only sixteen states have laws or policies requiring the release of any information without someone first filing a request. Only eleven — Arizona, Arkansas, Colorado, Florida, Nevada, Oregon, Pennsylvania, Rhode Island, Tennessee, Texas and Wisconsin — proactively post notifications of any child fatality, near fatality, or "egregious incident" at all.

This is not new information to the federal government. In July 2011, GAO published Child Maltreatment: Strengthening National Data on Child Fatalities Could Aid in Prevention (GAO-11-599). It found that "more children have likely died from maltreatment than are counted" in the National Child Abuse and Neglect Data System; that nearly half of states report only data originating inside their own child welfare agencies, structurally excluding every child who died without a prior CPS case; that HHS collects information on the circumstances of these deaths — including perpetrators' prior maltreatment history — and simply does not publish it; and that NCANDS collects no near-fatality data whatsoever. Fifteen years later, NCANDS still collects no near-fatality data.

The Definitional Trapdoor

Federal law defines "near fatality" at 42 U.S.C. § 5106g as "an act that, as certified by a physician, places the child in serious or critical condition." Every word of that definition is a filter.

Certified by a physician means a child who is nearly killed but whose treating doctor never signs a certification is not a near fatality. Texas's own Department of Family and Protective Services has conceded the deeper problem: there is no agreed-upon clinical meaning of "serious or critical condition" among hospitals, so DFPS grafted on its own gloss — a near fatality is an act of abuse or neglect after which the child, "without imminent medical intervention, would likely have died." Nevada (NRS ch. 432B) widened the certifying universe to include registered nurses and other licensed providers. Twenty states embed "near fatality" in their abuse definitions; three use "near death" instead. The categories are not comparable across state lines, and nobody at the federal level harmonizes them.

Downstream, the category nearly vanishes. Among the roughly eighteen states that publish child maltreatment fatality information without a request, only a small handful publish anything about near fatalities — and the Lives Cut Short analysis concludes the discrepancies are severe enough that "these numbers cannot be relied upon to tell us how many children actually suffer life-threatening injuries due to abuse or neglect." For egregious incidents that fall short of the near-fatality threshold — the beatings, the starvations, the confinements that don't put a child in the ICU — "there is almost no information at all."

A child who survives is a child whose case file stays sealed. That is the incentive the definition creates.

The Numbers Game

The federal count is Child Maltreatment, an annual NCANDS report now in its 35th edition. For FFY 2023 it published a national estimate of 2,000 maltreatment deaths, a rate of 2.73 per 100,000 children, with infants under one accounting for 44 percent of all fatalities at a rate of 24.11 per 100,000. The prior year's figure of 1,990 was quietly revised upward to 2,050 in the same volume.

Then Child Maltreatment 2024 reported 1,773 deaths from 52 reporting jurisdictions and was announced as roughly a ten percent decline. It was not a decline in dead children. The Lives Cut Short analysis of the state-level tables found that five states — Texas, Maryland, North Carolina, Illinois and Virginia — together account for a decrease of 210 fatalities, more than 100 percent of the entire national year-over-year drop. Every one of those states offered an explanation that had nothing to do with fewer children dying:

Texas narrowed its statutory definition of neglect via HB 567 in 2021 — neglect being the largest share of maltreatment deaths nationally — and DFPS acknowledged in its own fatality report that this drove the decline. In 2023 the department separately changed when it opens a fatality investigation at all, now requiring "explicit concern" for abuse or neglect at the time of death. That single policy change produced 30 percent fewer child fatalities assigned for investigation in its first year. Deaths that are never investigated are deaths that are never counted.

Maryland told reporters it will report 46 fatalities — after years of reporting high numbers it now characterizes as over-reporting. But the state's own data showed the required Form 1080 was missing for 68 children who died of suspected abuse or neglect in 2023 and 2024. The count went down while the paperwork went missing.

North Carolina fell from 107 fatalities in FFY 2023 to 46 in FFY 2024 — a 57 percent collapse. NCDHHS attributed it to an "outdated database platform that became increasingly unstable and subject to error," replaced by a new Child Welfare Fatality Information System that launched for state users in 2025.

Against these self-reported figures sit the linkage studies. A study of fatal child maltreatment across three states found state child welfare records undercount maltreatment deaths by 55 to 76 percent. A capture-recapture study estimated that more than 60 percent of child abuse deaths are never coded as such on the death certificate. The consensus expert estimate is that NCANDS undercounts true maltreatment deaths by a factor of two to three. The independent CANDID database assembled by AEI and the University of North Carolina has documented more than 4,000 child maltreatment deaths from 2022 through 2026 from public sources alone — and the researchers report that many of those cases were never counted in state or local statistics.

Between a third and a half of these deaths involve families already known to CPS through prior maltreatment reports. Separate analysis puts the share of fatality victims who had a completed risk assessment or were enrolled in family preservation services before dying at 30 to 40 percent. Those are precisely the cases the disclosure mandate was written to expose — and precisely the cases that trigger the most aggressive confidentiality claims, because those are the files that show what the agency knew.

The Money: A Mandate Stapled to Less Than One Percent

The reason nothing happens is arithmetic.

State child welfare agencies spent more than $34.3 billion on child welfare activities in state fiscal year 2022. Total CAPTA funding for FY2026 is $212 million, of which the CAPTA State Grant program — the pot to which the disclosure mandate is attached — is $105 million. HHS's own Office of the Assistant Secretary for Planning and Evaluation published a July 2026 brief with the finding in the title: The CAPTA State Grant Program Represents Less Than One Percent of Child Welfare Spending but Places Many Requirements on States to Drive System Design.

Since FY2018, Congress has further directed $60 million of the annual CAPTA appropriation toward plans of safe care for substance-exposed infants — meaning a majority of the money is already earmarked before the disclosure mandate gets near it. The open-ended money that actually runs the system is Title IV-E foster care, which carries no comparable public-disclosure condition at all.

So the federal government's entire enforcement lever over child death transparency is a grant worth roughly 0.3 percent of what states spend. A Child Welfare League of America vice president put the consequence plainly: "we just don't have a huge stick right now because the funding for CAPTA is so low."

Compare the countervailing incentive. A single disclosed fatality file is discovery material in a wrongful death case. Los Angeles County agreed in 2025 to a $20 million settlement in the death of Noah Cuatro, where the family alleged child welfare authorities failed to act on documented warning signs and never followed through on protective steps already in the court record. A $60 million claim was filed against the Arizona Department of Child Safety in August 2025 over the death of 11-year-old Chaska Davis. New Mexico's CYFD settled a wrongful death case for $4.9 million. In January 2026 the family of five-year-old Ethan Belcher sued two Michigan CPS workers, alleging they failed to investigate abuse reports, never ensured the mother's boyfriend completed required services, and never petitioned the court for removal as Michigan law requires.

A state facing that exposure is being asked to publish the evidence against itself in exchange for a share of $105 million. The math answers itself.

The Suppression Machinery, by Name

New York. State law requires OCFS to produce a report after a child known to CPS dies, examining whether caseworkers followed protocol. The commissioner may withhold it under a single exemption: that release is not in the "best interest" of surviving siblings or other children in the household. In response to an open records request, OCFS disclosed it had suppressed 725 child death reviews since 2010. Roughly a quarter of 2022 reports were withheld on these grounds — including most of the cases that had drawn press coverage. The mechanism is the tell: OCFS asks the county social services agency for a written recommendation on release — the same county whose employees' conduct the report critiques. The reviewed party gets a vote on publication, and the entire deliberation is itself secret, so no one outside can assess whether a single suppression was justified. Eight-year-old Thomas Valva died in 2020 after school staff made repeated hotline calls over sixteen months describing that he and his brother were starved, beaten, and made to sleep in a garage; nearly five years on, the full official account of what the system did remains unpublished. State Comptroller Thomas DiNapoli's January 2023 audit found OCFS identifies deficiencies case by case but fails to convert them into recommendations addressing systemic statewide problems; a September 2024 follow-up found one of three recommendations fully implemented.

Illinois. Under state law there is a presumption that public disclosure serves the public interest in child death investigations. In March 2025, Capitol News Illinois and the Illinois Answers Project reported that DCFS had failed to produce the legally required incident-specific reports in a staggering share of qualifying cases: more than 1,200 deaths and more than 3,000 serious injuries met the reporting criteria since July 2018. When a records request was filed for the agency's most recent report, DCFS answered that no reports exist. No report was ever produced for eight-year-old Navin Jones, found dead at his family's Peoria home in 2022. Cook County Public Guardian Charles Golbert called for an investigation. The agency's position was that it had been operating under an "understanding" that what it already did satisfied the statute.

The sequel is live. Capitol News Illinois sued for the personnel records of two caseworkers and a supervisor responsible for Mackenzi Felmlee, an 18-year-old who died May 11, 2024 — found struggling to breathe at the bottom of the basement stairs of a Fairview Heights foster home, killed by a pulmonary embolism. On June 10, 2026, Judge Jack Davis II found DCFS had "willfully and intentionally violated" the Freedom of Information Act, ordered production within five days, and imposed costs, attorney fees, and a $5,000 civil penalty. DCFS did not comply and did not appear. Davis has ordered DCFS Director Heidi Mueller to appear personally on September 9, 2026 to show cause why the agency should not remain in contempt, with coercive penalties of up to $1,000 per day of continued noncompliance on the table.

That is the actual enforcement architecture for CAPTA's disclosure mandate in 2026: a county judge and a nonprofit newsroom's legal fund.

Citizen Review Panels: Oversight With No Lever

CAPTA requires states to operate citizen review panels that evaluate whether agencies are discharging their child protection duties, publish annual public reports, and — critically — receive a written state response within six months describing whether and how recommendations will be adopted. On paper it is a closed accountability loop.

In practice the loop has no clasp. There is no consequence for a late response, a non-response, or a response that says no. New Jersey's Child Fatality and Near Fatality Review Board issued recommendations on June 30, 2004; the state response was due December 30, 2004 and did not arrive until August 2005 — eight months late, with nothing following. Panel members are appointed by the agency they review, funded through the agency's CAPTA allocation, and given access to the very case files whose public release the agency controls. Where panels succeed — Arizona reports that few of its panel's recommendations have gone unadopted — it is because the agency chose cooperation, not because the panel could compel it.

The Accountability Gap

The entity responsible for enforcing all of this is the Children's Bureau within HHS's Administration for Children and Families, now led by Associate Commissioner Ryan Hanlon. Its enforcement posture was documented in the 2019 Boston Globe–ProPublica investigation that surveyed all fifty states against CAPTA's 27 mandated provisions and found that not one state complied with all of them. When presented with the findings, the Children's Bureau did not dispute them. Its stated position was that it does not question states' annual reports claiming compliance — that the assurance letter a state files is the compliance determination.

That is the whole audit. A state signs a page; the money moves. Child welfare expert Michael Petit's assessment at the time was that "every single state" is "vulnerable to successful class action litigation for being in violation of federal law."

The vacuum has been filled by private actors. Casey Family Programs, the nation's largest operating foundation focused on child welfare, holds $2.68 billion in assets (EIN 91-0793881), employs roughly 379 people, works in all fifty states plus D.C., Puerto Rico, the U.S. Virgin Islands and tribal nations, and runs a "Fatality Data Consistency" initiative. AEI and the University of North Carolina built CANDID because the federal database does not do the job. The most reliable public accounting of American child maltreatment deaths is now produced by a think tank and a university from news clippings — not by the agency that collects the data and writes the checks.

Meanwhile the 2016 federal blue-ribbon answer sits on a shelf. The Commission to Eliminate Child Abuse and Neglect Fatalities released Within Our Reach: A National Strategy to Eliminate Child Abuse and Neglect Fatalities on March 17, 2016 — 168 pages, ten recommendations, built on three pillars: leadership and accountability, decisions grounded in better data and research, and multidisciplinary support for families. A decade on, NCANDS still collects no near-fatality data, the disclosure mandate still has no regulation, and no state has lost a dollar.

Why It Matters, and What Would Actually Fix It

The children this concerns are not statistical abstractions. They are, disproportionately, infants — 44 percent of maltreatment fatalities are under one year old — and they are, in a third to half of cases, children whose names were already in a state database, whose families had already been assessed, whose risk had already been scored. The public disclosure requirement exists for exactly one reason: a system that reviews itself in secret does not learn. Every suppressed report is a pattern that the next caseworker, the next legislator, the next mandated reporter never gets to see. That is why Indiana's Rep. Julie McGuire could not get basic answers about how DCS responded to reports about Kinsleigh Welty, found starving in a closet in 2024, or Zara Arnold, who died in May 2025 after her mother's documented pleas about the father's history of violence.

Indiana shows the fix is not exotic. House Bill 1257 passed unanimously and took effect July 1, 2026. It requires a public summary of how DCS staff responded in every fatality and near-fatality case, mandates more detail in the annual fatality report, and requires DCS to answer media inquiries within ten days with the child's age, the child's sex, and what the agency actually did. One state legislature, one session, unanimous.

What would close the gap nationally:

Write the regulation. HHS has had thirty years to define "findings or information" under § 5106a(b)(2)(B)(x). A rule specifying a minimum content floor — prior reports received, dispositions, services offered, dates of contact, supervisory decisions — and a maximum release deadline would end the "we had an understanding" defense in a paragraph.

Kill the reviewed-party veto. No agency should be able to suppress a review of its own conduct, and no county should get a vote on whether the report criticizing its caseworkers is published. New York's "best interest of surviving siblings" exemption should be replaced with targeted redaction of identifying information, the model already used in Pennsylvania, Florida, Oregon, Washington, and Colorado.

Move the mandate to Title IV-E. A disclosure condition attached to 0.3 percent of child welfare spending is not a condition. Attach it to the entitlement that actually funds the system, and non-disclosure acquires a price.

Collect near-fatality and multi-source fatality data federally. NCANDS should require linkage to vital records, child death review teams, and the National Violent Death Reporting System — GAO recommended exactly this in 2011 — and should collect near fatalities under a clinically operable definition, not one that depends on whether a physician happened to sign a form.

Publish the denominator. States should be required to report how many deaths they declined to investigate, and why. Texas's 30 percent drop in assigned fatality investigations after a 2023 policy change would have been visible instantly.

Until then, the operative federal standard remains what a state writes on its assurance letter, and the operative enforcement mechanism remains a reporter with a FOIA request and a judge willing to count the days.