GHOST VISITS: THE NATIONAL INVESTIGATION — ARTICLE 8 When Visits Fail

Child Fatalities and the Human Cost of Non-Compliance

Jul 11, 2026

A Note on This Article

This is the heaviest article in this series. Articles 1 through 7 documented statutes, budgets, compliance rates, and legislative maneuvers. This article documents what those abstractions cost. It names one child, because her name is already part of the public record of criminal proceedings and state review, and because the system that failed her counted her, for five years, as a case in compliance.

The names of the adults charged in the Felmlee case appear here as they appear in the public charging record. Criminal charges are allegations; every defendant is presumed innocent unless and until proven guilty.

One case discussed in Section III involves the death of a child by apparent suicide. If you or someone you know is struggling, call or text 988 — the Suicide & Crisis Lifeline — any time, free and confidential.

Articles 1 through 7 of this Ghost Visits series documented a federal mandate requiring monthly caseworker visits for at least 95% of children in foster care — visits that were supposed to be well-planned, focused on case planning, and designed to ensure the child’s safety, permanency, and well-being. We documented the $9.5 billion in federal Title IV-E and related funding that flows through the system with almost no accountability for the actual verification that children are safe once removed. We documented state after state failing to meet even the most basic quantitative thresholds for visitation, with up to 140,000 children per month receiving no safety check at all under realistic assumptions. We documented the impossible caseloads far exceeding evidence-based standards and the catastrophic turnover that destroyed case continuity and made compliance structurally impossible. We documented the falsified records, copy-paste notes, misclassified phone calls, and drive-by checks that revealed the compliance data itself could not be trusted — because in multiple documented cases, the visits that were counted had never happened. We documented the twenty-year enforcement failure in which the graduated financial penalty for missing visits was never applied to any state — not even Connecticut, whose externally audited 57.4% compliance rate sat squarely in the maximum penalty tier. And we documented the formal repeal of that penalty in P.L. 118-258 — making the unenforced mandate now an unenforceable one, with no federal financial consequence for any state that fails to visit 95% of its foster children monthly.

This article documents what happens at the bottom of that ledger. When the visit does not happen — or happens only on paper — children die.

Mackenzi Felmlee died on May 11, 2024, at 18, after five years in Illinois state care. Ten caseworkers. A final ‘contact’ that was a phone call with the foster mother — the person now charged with her murder — two days before her death. Notes filed five days after she died. Zero abuse findings in five years. She weighed 90 pounds.

In Connecticut, a child under active DCF investigation — who had said they felt unsafe and asked to be placed in foster care — died by apparent suicide within one hour of a DCF visit. The visit happened. It still failed.

Child welfare fatality reviews across the country consistently identify inadequate caseworker visitation as a contributing factor in child deaths. The pattern is documented, repeated, and — after P.L. 118-258 (Article 7) — now entirely without federal consequence.

These two cases were chosen because they bracket the failure. One shows what happens when the visit is missing. The other shows that even a visit that occurs can fail a child when the system behind it is broken. Together they are the human cost of every number in Articles 1 through 7 — the cost that every prior article in this series has been adding up to, case by case, statute by statute, and now, child by child.

Executive Summary: The Human Cost of Every Number

The monthly caseworker visit exists for one reason: it is the moment when a trained professional walks into a home, looks at a child, and determines whether that child is safe. Article 1 documented the mandate. Article 2 documented the $9.5 billion that funds it. Articles 3 through 5 documented its failure — no state passing CFSR Round 4, impossible caseloads, and visit records falsified outright. Articles 6 and 7 documented the death of its enforcement: a penalty never used, then repealed.

This article documents what happens at the bottom of that ledger. When the visit does not happen — or happens only on paper — children die.

Section I — Why the Visit Exists: The One Moment of Truth

Strip away the statutes and the funding streams, and the monthly caseworker visit is a simple thing: an adult whose job is child safety stands in the same room as the child and looks. Weight. Bruises. Fear. The condition of the home. Whether the child, asked in private, says something is wrong.

Article 1 documented why Congress made that moment a condition of every state plan (42 U.S.C. § 622(b)(17)) and set a 95% monthly floor: because in-person observation is the only reliable way to detect many forms of abuse and neglect. A camera angle hides a bruise. A case file does not lose weight. A phone call cannot step inside the house.

Every substitute for the face-to-face visit — the drive-by checks of Article 5, the doorstep conversations, the calls logged as contacts — removes one of those checks. The two cases that follow show what is on the other side of that removal.

Section II — Mackenzi Felmlee: 18 Years Old, 90 Pounds, 10 Caseworkers

Mackenzi Felmlee died on May 11, 2024, at a hospital in the Metro East area of Illinois, after being found unresponsive in her Fairview Heights foster home. She was 18 years old. She had been in state care since 2019 — five years under the supervision of the Illinois Department of Children and Family Services.

The facts of her case read like a checklist of every failure documented in this investigation. What follows is not a list of isolated administrative errors. It is the record of a system that counted her as compliant for five years while the conditions that would end her life were developing in plain sight — or, more accurately, out of sight, because the visits designed to see them were not meaningfully occurring.

Ten Caseworkers in Five Years

Mackenzi was assigned 10 different caseworkers during her time in state care. One was on her case for only three weeks. The constant turnover — the caseworker crisis of Article 4 — meant no single person developed the relationship, trust, or institutional knowledge needed to detect the gradual deterioration of her care. Each new caseworker started from a case file that, as Article 5’s copy-paste evidence suggests, may not have reflected reality. When a child has ten caseworkers in five years, there is no continuity of observation, no accumulated institutional memory of what her baseline looked like, and no one person who has watched her over time and can say: something is wrong here. The system that removed her from her family for her protection then failed to provide her with even the most basic continuity of professional oversight.

A Phone Call, Not a Face-to-Face Visit

The last caseworker ‘contact’ before Mackenzi’s death was a phone call with the foster mother — not a face-to-face visit with Mackenzi herself. The call occurred two days before she died. A phone call with the foster mother — the person now charged with Mackenzi’s murder — cannot substitute for in-person observation of the child. A caseworker on the phone cannot see bruises. Cannot assess the child’s weight. Cannot observe fear, pain, or deterioration. Cannot step inside the home and notice the conditions that might explain why an 18-year-old who had been in care for five years weighed 90 pounds. The final documented interaction between the child welfare system and Mackenzi Felmlee was a conversation with the person later charged with killing her. That is not oversight. It is its opposite.

Delayed Documentation, Inaccurate Reporting

The caseworker did not submit notes from the phone call until five days after Mackenzi’s death. Lutheran Child and Family Services of Illinois initially reported the contact as an in-person visit. It was not. The documentation system failed at every level — the visit was not face-to-face, the notes were filed late, and the initial reporting was inaccurate. Article 5 documented ghost visits as a systemic phenomenon; this is what one looks like attached to a name. The records that were supposed to document her safety were created after she was already dead. The system that was supposed to be watching her did not even record its own failure until after the outcome was irreversible.

Zero Findings in Five Years

Illinois DCFS reported receiving zero abuse or neglect cases involving Mackenzi while she was in foster care. Five years. Ten caseworkers. Zero findings. And then she was found dead at 90 pounds with bruising across her body. The system designed to detect abuse detected nothing — because the visits designed to detect abuse were not meaningfully occurring. Zero findings over five years is not evidence that nothing was wrong. It is evidence that the detection mechanism was not functioning. When the only person charged with looking at a child in state custody does not look — or looks only through a phone call with the person later charged with her murder — the absence of findings tells us nothing about the child’s safety and everything about the system’s failure to perform its most basic function.

The Physical Evidence

When Mackenzi was found, she weighed 90 pounds. Her neck, shoulders, legs, and face were bruised. She was severely dehydrated. She died from a blood clot that reached her lungs. Her foster mother, Shemeka Williams, and Williams’ mother, Cornelia Reid, face first-degree murder and child abuse charges. The physical evidence of prolonged abuse and neglect was overwhelming — and none of it was detected by the child welfare system charged with her protection. The bruises, the weight loss, the dehydration, the conditions that would produce a fatal blood clot — all of it developed while ten caseworkers, over five years, recorded her as a case in compliance. The system that removed her from her family for her protection then counted her as safe while she was being killed.

Section III — Connecticut: The Visit That Happened — and Still Failed

In what may be the most devastating case in this investigation, a child under active Connecticut DCF investigation died by apparent suicide within one hour of DCF staff visiting the family. The details, reported by the Connecticut Office of the Child Advocate:

The child had explicitly stated they felt unsafe.

The child had specifically requested to be removed from the home and placed in foster care.

DCF decided to leave the child with the parent, indicating that coming into care ‘was not an option.’

The child died within one hour of the visit.

The family had a long history with DCF, did not have stable housing, and none of the children in the household were enrolled in school.

This case is devastating precisely because a visit DID occur. The visit happened, the child spoke, the child asked for help, and the system said no. It demonstrates that even when the visit mandate is technically met, systemic failures in training, judgment, resource availability, and institutional culture can render the visit meaningless. The visit is the floor of child protection — necessary, and not sufficient.

The Connecticut OCA’s subsequent investigation found the 57.4% visit compliance rate documented in Articles 3 and 6 and characterized DCF’s deficits as ‘increasingly alarming,’ noting that ‘ongoing deficits’ in case practice ‘leave children at risk.’ For cases where a safety plan was in place, reviewers found that visitation was not conducted as required in 35% of cases — the children the system had already flagged as at risk were missing the visits designed to watch over them.

Section IV — The Pattern: Visit Failures Precede Tragedies

These are not isolated cases. Child welfare fatality reviews across the country consistently identify inadequate caseworker visitation as a contributing factor in child deaths. The pattern repeats:

Caseworker visits are infrequent, brief, or non-existent.

Documentation does not reflect the child’s actual condition.

Warning signs are missed because no one is looking.

The child’s condition deteriorates over weeks or months.

The child dies or is seriously injured.

A review finds that ‘systemic failures’ contributed to the outcome.

A performance improvement plan is developed. The cycle continues.

Article 6 documented what the PIP cycle looks like from the enforcement side: process substituted for consequence, twenty years running. This is what it looks like from the child’s side. The review happens after. The plan is written after. The visit — the one intervention that operates before — is the step the system keeps skipping.

Section V — The Series Converges: Every Failure, One Outcome

Each article in this series documented one structural failure. The cases in this article are what those failures look like when they land on a single child at the same time.

This convergence is the central finding of the series so far: the failures are not independent. Caseworker turnover (Article 4) produces the rushed or skipped visit (Article 3), which the documentation system records as compliance anyway (Article 5), which no federal actor ever penalized (Article 6), and which Congress has now formally excused (Article 7). A child in the system experiences all of it as one thing: nobody came.

As of FY2026, following P.L. 118-258, there is no federal financial consequence for any of it. The 328,947 children in foster care are protected by a standard that Article 7 showed is now, in law as in practice, a suggestion. The cases in this article are what the stakes of that suggestion are.

Section VI — Recommendations

PMC editorial position, following from the documented record of children dying while counted as cases in compliance:

Count the Deaths Against the Visits

Every child fatality and near-fatality review should be required to document, as a standard field, the child’s caseworker visitation history: dates, whether each visit was face-to-face, whether the child was seen alone, and whether documentation was contemporaneous. The Felmlee case shows every one of those fields failing; no review should be able to omit them. If the system cannot produce a complete, accurate visitation history for a child who has died in its care, that failure itself should be documented as a contributing factor in the review.

No Phone Calls Logged as Contact for Children at Risk

A telephone call with a caregiver must never satisfy — or be recordable as — a required contact for a child in care or under a safety plan. The final ‘contact’ of Mackenzi Felmlee’s life was a phone call with the person now charged with killing her. That is not oversight; it is its opposite. The in-person requirement exists for a reason: a caseworker on the phone cannot see bruises, cannot assess weight, cannot observe fear or pain, and cannot step inside the home. Logging a phone call as a contact for a child at risk is not a documentation error. It is a policy failure that should be prohibited in statute or regulation.

The Child’s Request Must Trigger Independent Review

When a child explicitly states they feel unsafe or asks to be removed, that statement should trigger mandatory review outside the assigned caseworker’s chain — before the case is closed or the child is left in place. In Connecticut, the child said the words. The system’s answer — ‘not an option’ — was reviewed only after the child was dead. A child’s direct statement of fear or request for placement change is not a routine case note. It is a safety signal that should activate an independent review process with authority to override the assigned caseworker’s decision if necessary.

Continuity as a Safety Metric

Caseworker continuity should be tracked and reported as a child-level safety metric — ten caseworkers in five years is a detectable, reportable failure condition, not an administrative footnote (Article 4). When a child has ten caseworkers in five years, that pattern itself should trigger mandatory supervisory review and, where appropriate, reassignment of the case to a more stable team or external oversight. High turnover is not just an operational problem. It is a safety problem that should be measured, reported, and acted upon at the child level.

Restore the Consequence

Everything above is procedure, and Articles 6 and 7 documented what happens to procedure without consequence. The penalty architecture recommended in Article 7, Section VIII — mandatory, automatic, keyed to independently verified data — is the enforcement floor under every recommendation in this series. Children died under the unenforced mandate. The unenforceable one offers them less. Without a real consequence for systemic failure to visit children in care, every other reform remains aspirational. The children who died while counted as compliant deserve more than another round of recommendations without enforcement.

The children in America’s foster care system deserved a federal government that was willing to impose a cost on states that failed to look at them. For twenty years, that cost was never imposed. Then it was formally removed.

Every child deserves to be seen.

Every visit deserves to be real.

The human cost of counting children as compliant while they die is the ledger that every prior article in this series has been adding up to.

Project Milk Carton is a 501(c)(3) nonprofit dedicated to U.S. child welfare transparency, CPS and foster care accountability, human trafficking prevention, and OSINT-driven investigations.

Data current as of July 2026. Sources include Capitol News Illinois, NPR Illinois, KSDK, NBC Connecticut, CT Mirror, WTNH, WFSB, Connecticut Office of the Child Advocate, and PMC CivicOps analysis from Articles 1–7. Full methodology and replication files available at projectmilkcarton.org. Criminal charges described are allegations; defendants are presumed innocent unless and until proven guilty.

Share this investigation. Demand that every fatality review count the visits. Demand that phone calls never count as contacts for children at risk. Demand that a child’s request for safety trigger immediate independent review. Demand that Congress restore a real enforcement mechanism with independent verification. The children cannot wait.

If you enjoyed this work and feel encouraged by independent investigations like this, please consider becoming a paid subscriber to The Constitutional Republic on Substack or donate at We Built the Tools. Help Us Keep Them Running.

Project Milk Carton Inc. Your gift keeps our site alive, spotlighting high-risk children!

Your support helps us continue:

investigative reporting
public records research
educational transparency projects
community outreach
and future investigations like this series

Most importantly, please share this article.

The Constitutional Republic

Discussion about this post

Ready for more?