OSINT · Publicly Available Sources INVESTIGATION ORACLE

The 95 Percent Illusion: How States Manufacture Foster Care Visit Compliance — and How Congress Quietly Deleted the Only Penalty

September 12, 2026 OPUS · Claude Opus Project Milk Carton

The 95 Percent Illusion: How States Manufacture Foster Care Visit Compliance — and How Congress Quietly Deleted the Only Penalty

I've completed the research. Here is the full investigation.

The 95 Percent Illusion: How States Manufacture Foster Care Visit Compliance — and How Congress Quietly Deleted the Only Penalty

Federal law requires that a caseworker lay eyes on nearly every child in American foster care, in person, every single month. States report to Washington each December that they are doing it, and almost all of them say they are: the national median for fiscal year 2024 was 95.5 percent, comfortably at the statutory bar. No federal agency audits the case notes behind that number. No independent party confirms that the children were seen. And on January 4, 2025, President Biden signed Public Law 118-258, which struck the financial penalty for missing the standard out of the U.S. Code entirely — meaning that as of today, the monthly visit remains a legal requirement with no consequence attached to failing it. What follows is an account of how a safety mandate became a self-graded number, what that number is worth in federal dollars, and what auditors, prosecutors, and child fatality reviewers find when anyone actually checks the underlying records.

The Standard That Counts Itself

The mandate lives at 42 U.S.C. § 624(f), enacted in its modern form by the Child and Family Services Improvement and Innovation Act (P.L. 112-34) in 2011. Each state must ensure that monthly caseworker visits to children in foster care total "not less than 90 percent (or, in the case of fiscal year 2015 or thereafter, 95 percent) of the total number of such visits that would occur during the fiscal year if each such child were so visited once every month while in such care." A companion provision requires that at least 50 percent of those visits occur in the child's residence. The visit requirement is also a condition of the state plan under § 422(b)(17).

The critical design flaw is in the arithmetic. As the Children's Bureau's own data documentation states, "the number of caseworker visits is the aggregate number of monthly caseworker visits made to children in the reporting population. The number of visit months is the aggregate number of calendar months for all children spent in care." It is a pooled ratio, not a per-child guarantee. A child who is never visited for an entire year contributes twelve unvisited months to the denominator — and is fully offset by twelve other children visited once each. A state can hit 95 percent statewide while an identifiable subset of children is invisible for months at a time, and the federal measure will never surface them.

The reporting is self-certification. States submit the numerator and denominator annually each December. There is no federal requirement that a state verify the entries, no requirement that it sample case notes, no requirement that anyone outside the agency confirm a single visit happened. The Children's Bureau receives two integers per state per year.

Pulling the full FY2024 dataset from ACF's Child Welfare Outcomes data site produces a picture that is both reassuring and revealing. Across 52 reporting jurisdictions, the median was 95.5 percent and 33 met or exceeded the standard — meaning 19 jurisdictions reported failing it, and nothing happened. (Applying the same method to FY2021 reproduces exactly the figure ACF published in its Report to Congress — a median of 96 percent with 37 states meeting the standard — confirming the extraction.) The reported range for FY2024 runs from Puerto Rico at 60.1 percent and Alaska at 61.6 percent up through Rhode Island at 98.8 percent, Nebraska at 98.6, and Indiana, Michigan, and Washington all at 98.5.

What Congress Did on January 4, 2025

Until last year, § 624(f) contained a second half. Paragraphs (1)(B) and (2) authorized HHS to reduce a noncompliant state's rate of federal financial participation under Title IV-B, subpart 1 — the Stephanie Tubbs Jones Child Welfare Services program — by one to five percentage points for each failed performance measure, from the baseline 75 percent federal share.

Section 112(d) of P.L. 118-258, the Supporting America's Children and Families Act, struck both paragraphs. The U.S. Code amendment note is unambiguous: "Pub. L. 118–258 struck out par. (1)(A) designation before 'Each State' and struck out pars. (1)(B) and (2) which related to reduction in allotment of funds to a State for failure to comply with child visitation requirements." (138 Stat. 2963.) What remains is a single sentence instructing states to take "such steps as are necessary." The obligation survives; the price of ignoring it is now zero.

It is worth being precise about what was deleted, because the penalty was never the deterrent it sounded like. ACF's own program page — still marked "Current as of: June 27, 2024," and still describing the penalty as if it exists — explains the mechanics: "States may continue to receive their full federal grant award but must provide a higher state match to receive the full funding." The sanction never took a dollar away from a state. It required the state to put up slightly more of its own money to draw the same federal grant, out of a national Title IV-B subpart 1 appropriation of $268.7 million in FY2024 — roughly $800 per child in foster care per year, spread across 50 states. A five-point match shift on a mid-sized state's slice of that pool is a rounding error in a child welfare budget. I could locate no public ACF record of the penalty ever being applied to any state, and ACF does not publish penalty determinations — an opacity that is itself part of the story.

Meanwhile, the carrot survived and grew. The Monthly Caseworker Visit formula grant, funded at $20 million in FY2024, rises to $26 million beginning in FY2026 under the same law that killed the penalty. States receive that money on the basis of their foster care population, not their visit performance. P.L. 118-258 also newly authorizes virtual monthly visits for youth in care aged 18 and older, with informed consent.

The Money: A Visit Is a Billable Event

The visit standard is nominally enforced through a $268 million grant. The visit record is attached to something far larger.

A February 10, 2026 brief by ACF and the HHS Office of the Assistant Secretary for Planning and Evaluation, Administrative Costs Drive Foster Care Claims and Are Increasing Per Child as Caseloads Decline, lays out the structure. From FY2019 to FY2023, administrative costs — not payments to foster parents — were the largest driver of Title IV-E foster care spending. In-placement administration alone was "the largest reported expenditure by over one billion dollars each year"; in FY2023, in-placement administration total computable claims came to "nearly five billion dollars," while foster care maintenance payments to the adults actually raising the children came to "just over two billion dollars." Within in-placement administration, case planning and management was generally the single largest reimbursement category.

Title IV-E administrative costs draw a flat 50 percent federal match in every state. States allocate those costs using Random Moment Time Studies — statistically sampled snapshots of what child welfare staff are doing, extrapolated to a statewide cost pool and applied against the IV-E-eligible share of placement days. Documented caseworker activity is therefore not merely a compliance artifact. It is an input to a multi-billion-dollar federal claim.

The trend line is striking. Average monthly children served through in-placement case planning and management fell from 197,245 in FY2019 to 138,163 in FY2023 — a 30 percent decline — while total administrative expenditures stayed flat. Average monthly administrative cost per child rose 40 percent, from $2,945 to $4,147. For in-placement administration alone, states averaged $2,576 per child per month in FY2023, with a range across states from $263 to $10,041. Combined administrative spending per child ranged from $420 to $12,659 a month. Fewer children, same money, and a forty-fold spread between states in what it costs to administer a case — with no federal verification of the underlying activity records that justify the claim.

Layer private contractors on top. Illinois contracts case management to nonprofits like Lutheran Child and Family Services. Florida runs its entire system through community-based care lead agencies paid on per-child-per-month rates, which subcontract case management to organizations such as Camelot Community Care, Gulf Coast Jewish Family and Community Services, and One Hope United. Texas contracts regions to Single Source Continuum Contractors. Pennsylvania's own reporting refers to "public and private children and youth agencies." For these entities, the documented visit is simultaneously the safety record, the contract performance measure, and the billing record. The same keystroke satisfies all three, and nobody checks any of them against the world.

Pennsylvania and the Workbooks

Pennsylvania reported 73.8 percent for FY2024 — the lowest figure among the fifty states, and roughly 11,300 children in care on an average monthly basis. Its explanation, published in its own Annual Progress and Services Report to the Children's Bureau, is the most candid public description of how these numbers are produced.

The Office of Children, Youth and Families acknowledged failing Measure #1 for FFY 2024, then wrote that "PA's failure to meet the statutory performance standard, in part, is a symptom of the new AFCARS data submission and not solely based on non-compliance." The remedy it describes is this: "Twice per AFCARS reporting period, OCYF has been sending out manually generated workbooks that show all of the youth in their population who do not have at least one visit every full month in care. This effort intends to ensure that if visits were completed in those months, workers can make the necessary corrections in the county case management system so that the correct data is available and sent via the AFCARS submission."

The following year's report states the result: "Data measures related to caseworker visitation for FFY 2025 showed a 13.5% improvement in Measure #1."

Pennsylvania frames this as correcting the under-recording of real visits, and that may be exactly what it is. But read as a control system, it is the whole problem in four sentences. The state generates a list of children with no evidence of a visit, hands it back to the agency whose funding and performance rating depend on the answer, and invites workers to enter records after the fact for months that have already closed. There is no step in that loop at which anyone asks the child, the foster parent, or any third party whether the visit occurred. The state cannot distinguish a correction from a fabrication. Neither can ACF. And the mechanism moved a statewide federal safety measure by 13.5 points.

What Happens When Someone Actually Checks

The gap between the self-reported number and the audited record is not subtle.

Texas reported 98.2 percent for FY2024. In its Round 4 Child and Family Services Review, federal reviewers examined 100 cases through record review plus interviews with children, parents, caregivers, and workers. Item 14, caseworker visits with child, was rated a strength in 86 percent — below the 90 percent threshold, an Area Needing Improvement. Item 15, visits with parents, came in at 51 percent. Texas was found not in substantial conformity with Well-Being Outcome 1. The state's own internal case reviews over prior years rated Item 14 between 79 and 90 percent. Two measurements of the same activity: one tallied from database entries, one verified against human beings, separated by 12 to 19 points.

Connecticut reported 95.4 percent and passed. On April 30, 2026, the state's Office of the Child Advocate sent DCF Commissioner Susan Hamilton a letter of findings describing "alarming" and "ongoing deficits" in case practice, prompted by notification that a child with an open DCF investigation died by apparent suicide within one hour of a caseworker visit — a visit during which, per DCF's own records, the child said they did not feel safe and asked to come into foster care. The caseworker left the child with the parent. The letter's data covers investigations and in-home cases rather than foster care, but the documentation picture is the point. DCF's own quality-assurance reviews scored contact with children at 55.8 percent for July–December 2024 and 57.9 percent for January–June 2025; all four foundational components — contact with children, contact with adults and collaterals, risk and safety assessment, supervision — were rated a strength together in just 22.7 percent and 25.4 percent of cases. In in-home reviews for May–October 2025, only 55.8 percent of cases had all required documented face-to-face visits with the children, and 8.7 percent had none at all. Quality of visits with children was rated adequate in 37.4 percent. All four quality measures together: 7.4 percent. OCA wrote that it was "particularly concerned about the fact that there are cases with no documented face-to-face visits with the child."

New York's State Comptroller, reviewing OCFS foster care oversight, found that of 150 children's records, 33 lacked evidence of the two required contacts within the first 30 days of placement, 14 were not visited in the foster home in that window, and for 30 children a total of 162 contacts were entered outside the 30-day documentation limit. In one case, 427 days elapsed between a July 2015 contact and the September 2016 date the record was finally entered. A contact entered fourteen months late is indistinguishable, in the aggregate federal count, from one entered the same day.

California reported 93.3 percent for FY2024 and 83.2 percent in-residence — one of the few large states to publicly fall short. A California State Auditor review of Los Angeles County DCFS found 40 percent of sampled assessments inaccurate and 10 percent completed without the social worker ever visiting the home.

Utah reported 98.1 percent and 99.8 percent in-residence — the highest in-home figure in the country. A January 2026 legislative audit found DCFS workers failed to make face-to-face contact with children within required deadlines in more than 3,200 cases.

Arizona's Auditor General, in Report 25-109 (September 2025), found the Department did not always complete, timely complete, or properly document key investigative activities including contacts with alleged child victims.

The Prosecutions, and What They Cost

Falsification is not hypothetical. It is prosecuted, rarely, and punished lightly.

Sylvia Mbilo Mundala, an Indiana Department of Child Services family case manager from February 2021 to August 2024, documented face-to-face visits that never occurred. Investigators cracked it with evidence the agency itself never collects: a photo of a child supposedly taken at a July meeting had been copy-pasted from a May meeting — the child's braided hairstyle in a photo taken by her guardian was "distinctly different" on the claimed date — and cell phone location data placed a guardian an hour from home during a purported visit. Mundala pleaded guilty in Marion Superior Court to official misconduct, a Level 6 felony; the misdemeanor count of falsifying child abuse or neglect records was dismissed. Sentence: 360 days, with 356 suspended and four days of credit. Indiana reported 98.5 percent compliance to the federal government for FY2024.

Hester Renee Nelms of Moffat County, Colorado, was found to have fabricated details in at least 50 child abuse and neglect reports, describing the décor of a child's bedroom she had never entered and inventing illnesses parents did not have. A crew of 15 caseworkers from across Colorado deployed to re-investigate more than 80 reports. She was charged with forgery and abuse of public records; the county human services director left the job.

Ashley Steele, a Muskingum County (Ohio) Children Services caseworker, faces 11 felony counts of tampering with records, ten stemming from documenting client visits that never occurred. Haleigh Ingham of Franklin County Children Services has pleaded not guilty to felony tampering over her official activity log, in a matter connected to the 2024 death of an 8-year-old Columbus boy found in an attic.

Georgia offers the only systematic count. An investigation of state disciplinary records covering January 2013 through August 2018 found more than 800 DFCS child protective services employees disciplined — from a workforce of roughly 1,900 caseworkers — with about 14 percent of those disciplined, roughly 112 people, having falsified a record or lied about checking on a child in state custody. That is one state, six years, and only the ones who were caught.

And the case that named the problem: Rilya Wilson, a Florida foster child last seen alive in January 2001, whose disappearance was not noticed until April 2002. Caseworker Deborah Muskelly admitted she stopped conducting in-person visits, faked travel vouchers, and filed monthly progress reports describing Rilya as "happy and well cared for." Fifteen months of documented visits to a child who was not there. She pleaded guilty to official misconduct and received five years' probation.

Ghost Visits Produce Ghost Children

The monthly visit is the only routine, scheduled mechanism in the entire system designed to confirm that a child in state custody is alive and where the state believes they are. When the visit is fabricated, the state loses not just contact but proof of life — and the fabricated record actively suppresses alarm.

The scale of the downstream failure is documented. HHS-OIG audit A-07-21-06102 examined 74,353 missing-children episodes reported by 46 state agencies. It estimated that 51,115 of them — 69 percent — were not reported in accordance with federal requirements: 34,869 episodes (47 percent) were never reported to the National Center for Missing and Exploited Children, and another 16,246 (22 percent) were reported late. Nearly all the sampled states lacked any system for tracking whether missing children had been reported at all.

The Definition Keeps Getting Looser

During COVID, the Children's Bureau permitted videoconference visits to count as monthly caseworker visits — and, during the declared disaster period, to count as occurring "in the child's residence." The effect on the metric was immediate and backward: the national median for in-home visits rose from 88 percent in 2020 to 92 percent in 2021, the period when caseworkers were least likely to be inside anyone's home.

The in-residence figures states report today strain credulity in both directions. For FY2024: Utah 99.8 percent, Alabama 99.6, Florida 99.6, Pennsylvania 99.0 — Pennsylvania reporting that essentially every visit it counted happened in the child's home, in the same year it reported the nation's worst visit rate. At the other end: Hawaii 59.2 percent, Arizona 63.2, South Carolina 65.1, Alaska 70.7. These are supposedly measurements of the same thing under the same federal definition.

The direction of legislative travel is toward more flexibility, not more verification. P.L. 118-258 authorized virtual visits for youth 18 and over. California's AB 2684 (2026) would let nonminor dependents placed out of state take monthly visits virtually, with a minimum of one in-person visit per quarter — though notably it would also require the caseworker to document in the case file and the court report the location of any visit not occurring in the residence and the reason why, one of the few verification-adjacent provisions moving anywhere.

Who Is Supposed to Be Watching

The Children's Bureau receives a self-reported ratio once a year and publishes it. Its most recent Child Welfare Outcomes Report to Congress covers 2021; FY2024 data sits on a public data site with no accompanying analysis. It has no statutory authority left to penalize noncompliance and no practice of auditing the case notes that generate the number.

The Child and Family Services Review does examine records and interview families — and demonstrably catches what the self-report misses — but it runs on a multi-year cycle, samples on the order of 100 cases per state, and its financial sanction is suspended while a state works a Program Improvement Plan.

HHS-OIG has not conducted a national audit of caseworker visit documentation since OEI-04-03-00350 in 2005, which found that states largely lacked the capacity to track visit frequency reliably or verify that visits occurred. Its recent state-specific work, such as the Kentucky pandemic-era audit (A-06-22-07001), found that the state "did not always comply with State and Federal requirements related to background checks and caseworker visits" — and recommended that Kentucky "identify ways to address the challenges." Kentucky reported 89.8 percent for FY2024.

Inside agencies, the supervisor is the control. Connecticut's data shows what that control is worth: monthly supervisory conferences were documented for every month of the review period in 70.1 percent of in-home cases. The supervisor is reviewing the same unverified entry the caseworker typed, in the same system, with no independent signal.

What Would Actually Fix It

Verify at the source. Require confirmation from someone who is not the caseworker and not the caseworker's employer — a foster parent or caregiver attestation, and for age-appropriate youth, a direct child-side confirmation channel independent of the agency. Indiana's prosecutors proved a case using cell-site data and family photographs; the state's own compliance system possessed none of it.

Make the record immutable and transparent. CCWIS systems should capture and report the entry timestamp alongside the claimed visit date, and states should publish the distribution of that lag. An entry made 427 days later should not be silently interchangeable with one made same-day. Entries beyond a short window should be flagged, reviewed, and separately reported.

Replace the pooled ratio with a per-child measure. Report the number and percentage of children who had zero visits in any month of the year. Pennsylvania already generates precisely this list twice a reporting period — it just sends it to the agency to fix rather than to the public to see.

Independent audit with real sampling. A statistically valid annual sample, verified against non-agency evidence, conducted by HHS-OIG or state auditors — not by the agency being measured.

Restore a consequence that bites the right pool. The repealed penalty only obliged states to raise their own match. Tie verification instead to Title IV-E administrative claiming, the roughly $5 billion in-placement pool where the case note already functions as a billing record. An unverifiable visit should not be a reimbursable administrative activity.

Publish everything. State by state, county by county, and contractor by contractor, including every private case management agency, so that the entities paid per child per month are visible on the same board as the public agencies.

Why It Matters

Four hundred thousand-plus children pass through American foster care each year; 328,947 were in care at the close of FY2024. For most of them, the monthly caseworker visit is the state's entire physical presence in their lives — the one recurring moment when an adult with legal responsibility is supposed to enter the room where they sleep and look at them.

Rilya Wilson was documented as "happy and well cared for" for fifteen months after anyone last saw her. A Connecticut child told a caseworker they did not feel safe and asked to be removed; the caseworker left, and the child was dead within the hour — and that visit, having occurred, counts. Mackenzi Felmlee died at 18 in a Fairview Heights foster home licensed through a private DCFS contractor; her case had cycled through ten caseworkers, a detective reviewing the file alleged a caseworker had been copying and pasting home visit notes, and the contact two days before her death was logged as an in-person visit before the agency later said it had been a phone call — documented five days after she died. Illinois is still fighting in court, in September 2026, over releasing the records.

Every one of those files reported compliance. The number at the top of the page was fine. That is not an accident of the system; it is the system working as designed — a safety mandate that measures typing instead of presence, graded by the party being graded, now with the last consequence formally removed from the statute. The children who disappear into the gap between the record and the world are not statistical outliers. They are the specific, named, foreseeable cost of a metric that was never built to detect them.

Sources: