Missouri’s child welfare system runs on FACES — the Family and Children Electronic System. Every removal, every visit, every safety assessment, every permanency decision eventually has to live somewhere, and that somewhere is FACES. When the state, the legislature, or a federal reviewer wants to know how the system performed, they open the database and read what it says.
That is the entire model of oversight in use today: review the record, evaluate the record, report on the record.
It is worth asking a plain question before we go any further. A record of a case is not the same thing as the case. Reviewing what FACES says happened tells you what FACES says happened. It does not, by itself, tell you what happened.
That distinction sounds obvious stated this way. It gets lost constantly in practice, because the record is the only thing anyone in the review chain ever actually touches.
What Administrative Case Records Are Designed to Do
FACES was not built as an oversight tool. It was built as a case management tool — a way for caseworkers to document contacts, track required timelines, generate the paperwork that keeps a case moving through the system’s own compliance requirements. It exists to support the day-to-day operation of casework: did the worker make the visit, was the safety plan updated, did the required form get filed by the deadline.
Oversight was layered on top of that afterward. The state’s compliance reporting, the federal reviews, the dashboards — all of it draws from a data infrastructure whose first job was never independent verification. It was workflow documentation for the people doing the work, repurposed later as the evidentiary basis for judging whether the work was done well.
That repurposing is not inherently illegitimate. But it means every downstream review is built on top of a system that was never designed to answer the question being asked of it.
Why Documentation Matters
None of this is an argument against documentation. Documentation is the entire reason a case review is possible at all. Without a record, there is no case history, no continuity between workers, no way for a court, a supervisor, or an auditor to reconstruct what has already been tried.
The problem is not that documentation exists. The problem is what people do with it once it does — specifically, the quiet assumption that a well-documented record is a verified record. Those are not the same claim. A record can be complete, timely, internally consistent, and still not correspond to what happened in a child’s home, a foster placement, or a courtroom. Consistency is a property of the paperwork. Accuracy is a property of the world the paperwork claims to describe.
Absence of Documentation Is Not Absence of an Event
This is the distinction that gets collapsed most often, and it cuts in both directions.
If a required visit isn’t logged, two very different things could be true: the visit didn’t happen, or the visit happened and wasn’t logged. If a safety concern doesn’t appear in the case notes, it could mean no concern existed, or it could mean the concern existed and never made it into the record — because the worker ran out of time, because the intake process didn’t have a field for it, because the person who could have reported it didn’t know how, or didn’t trust that reporting it would matter.
A case-review methodology that treats the record as a complete account of reality will systematically read “not documented” as “did not occur.” That is a methodological choice, not a neutral fact-finding step, and it has a direction: it biases every finding toward looking better than the underlying reality, because omissions vanish instead of registering as unknowns.
I want to be specific about what this looks like in practice, because Missouri’s own public reporting shows the mechanism. HB1414 quarterly reports used to publish a raw county-by-county victimization compliance table. In the quarter with the worst documented outcomes on record — several counties showing severe violation rates — that raw table was the last one published. Every report since has shown improving, then perfect, compliance, with no raw table available to check the arithmetic against. The state’s own July 2026 report discloses “identified data discrepancies” between validated data and the public dashboard, in the same document reporting 100 percent compliance. I am not asserting that the underlying performance didn’t improve. I am saying that once the raw data disappears, “no violations documented” and “no violations occurred” become indistinguishable from the outside — and the report gives no way to tell which one you’re looking at.
How Downstream Reviews Inherit the Limitations of Source Data
Every review that follows the primary record inherits whatever is wrong with that record, and it usually inherits it invisibly.
A federal Child and Family Services Review draws its sample from case records in FACES. A state dashboard aggregates figures pulled from FACES. A legislative report summarizes the dashboard. A public hearing gets briefed on the report. By the time information reaches an oversight body, it has passed through several layers, each one trusting the layer beneath it, and none of them independently re-examining the original case.
This is not a hypothetical concern in Missouri right now. The definitions used for a core permanency metric are not consistent across the relevant regulations and the public dashboard — meaning the same word can be counting different things depending on which document you’re reading. A placement code has been categorized one way in one official report and a different way in another. These are not exotic edge cases; they are the kind of definitional and classification inconsistency that, once it exists anywhere upstream, propagates into every summary built on top of it, and gets harder to see the further downstream you go.
A reviewer working three layers removed from the case file has no way to know that the number in front of them was built on a shifting definition. They will treat it as solid ground, because from where they’re standing, it looks like solid ground.
Why Reviewing FACES Against FACES Cannot Independently Validate FACES
This is the part that I think matters most, and it’s the part that’s easiest to miss because it sounds like a tautology rather than a methodological failure.
If the question is “does FACES accurately reflect what happened in this case,” you cannot answer that question by pulling more data out of FACES. Checking one field in the database against another field in the same database can catch internal inconsistency — two entries that contradict each other — but it cannot catch a case where the entire record is wrong in a consistent way. If a visit wasn’t made and wasn’t logged, every field related to that visit will agree with every other field. The record will be perfectly self-consistent and completely inaccurate.
Missouri’s own current case review transition illustrates the stakes of getting this wrong. The state’s Response and Evaluation Team discontinued full federal-style case reviews as of June 2026, with a state-specific replacement methodology not expected for six to nine months — leaving no fully described interim review process in the gap. And under existing state regulation, only providers — the case management contractors themselves — are authorized to submit anomalous-case exclusions or alternative metrics to that Response and Evaluation Team. Families, lived-experience advocates, and the public have no formal channel into that process at all. The people with the most reason to notice a record that doesn’t match reality are structurally the people least able to flag it.
An internal audit checking the system’s data against itself is not oversight. It’s quality control. Those are different functions, and mistaking one for the other is how a system ends up certifying its own accuracy.
Triangulation: Court Records, Providers, Families, and Service Data
Independent verification requires an independent source — something that was not generated by the same system being reviewed, using the same incentives, filled in by the same staff, under the same time pressure.
That means:
- Court records, which are created by a separate institution with its own record-keeping obligations and its own incentive structure, not the case management system’s.
- Provider documentation — from residential facilities, therapists, medical providers — generated for reasons unrelated to CD’s own compliance reporting.
- Family and youth accounts, including the people the case is actually about, who experienced the events the record claims to describe and have no stake in the record matching their experience.
- Service records from the actual entities delivering services, which can confirm or contradict a case note claiming a service was provided or a referral was made.
None of these sources is perfect on its own. Court records can be incomplete. Family accounts can be shaped by memory, trauma, or distrust of the system asking the questions. Provider documentation has its own gaps. That is exactly the point of triangulation — no single source is trusted as the ground truth. Where independent sources converge, confidence in the record increases. Where they diverge, that divergence is the finding. It tells you exactly where the system’s account of itself needs scrutiny.
Why This Isn’t an Accusation
I want to be precise about what this argument is and is not, because it is easy to hear “the record may not match reality” as an accusation that caseworkers are falsifying documentation. That is not the claim.
Caseworkers operate inside caseloads, timelines, and documentation requirements that were not designed with slack for the unexpected. A record can fail to capture reality for reasons that have nothing to do with dishonesty: a visit that ran short because of a crisis elsewhere, a note written from memory at the end of a long day, a system that has no field for the thing that actually mattered most in the room. None of that is malice. Most of it is exactly what you’d predict from asking overloaded people to produce a complete written account of every consequential thing they do, in real time, using a database designed for compliance tracking rather than narrative accuracy.
The argument here is structural, not personal. It is a claim about what any documentation system — built by anyone, staffed by anyone, operated in good faith by everyone involved — can and cannot verify about itself from the inside.
How Independent Systems Review Could Work
If reviewing the record against itself cannot answer the accuracy question, independent review has to look different. At minimum, that means:
- Sampling that pulls from outside the primary database — cases selected and cross-checked using court dockets or provider referral logs, not just FACES case IDs.
- A formal channel for families and lived-experience advocates to flag discrepancies between the record and their experience, with actual standing to have that discrepancy investigated — something Missouri’s current regulatory structure does not provide.
- Published raw data, not just aggregated summaries — the county-level tables that disappeared from the HB1414 quarterly reports are exactly the layer that makes independent recalculation possible; once only the summary is public, verification stops.
- Reviewers who are organizationally separate from the agency whose performance is being measured, with access to the same court and provider records the agency itself can see.
- A defined, funded interim review process during any transition period — like the current gap between the discontinued CFSR model and Missouri’s still-unbuilt replacement — so that “we’re changing methodology” never functions, even temporarily, as “no one is currently checking.”
None of this requires assuming bad faith anywhere in the system. It requires accepting that no system can be its own auditor, and building the review function outside the boundary of the thing being reviewed.
The question underneath all of this is the one I keep coming back to:
Who independently verifies that the system’s record of what happened matches what actually happened?
Right now, in Missouri, I don’t think anyone does.
Leave a Reply