The handoff nobody owns
Two federal regulatory systems touch the same person at the same moment, and neither is designed to receive the other’s output. On one side, the Centers for Medicare & Medicaid Services (CMS) requires hospitals to run a discharge planning process as a Condition of Participation — the baseline requirement a hospital must meet to stay in Medicare. On the other, the U.S. Department of Housing and Urban Development (HUD) requires Continuums of Care (CoCs) — the regional planning bodies that coordinate homeless assistance — to operate a coordinated entry system: a single, standardized front door for crisis housing assistance, structured around four core elements HUD names as Access, Assessment, Prioritization, and Referral.
The hospital’s obligation ends at the discharge plan. The CoC’s obligation begins at the coordinated entry access point. The transition between them — the actual movement of a housing-unstable inpatient from a bed to an assessment — is not owned by either rule set. That is the boarding-to-shelter loop, and it persists not because hospitals or CoCs are noncompliant but because each can be fully compliant while the handoff fails.
What CMS actually requires — and what it does not
The discharge planning Condition of Participation at 42 CFR 482.43 is a process requirement, not an outcome requirement. CMS’s interpretive guidance for the hospital survey process describes it as a four-stage sequence: screening all inpatients to identify those at risk of adverse health consequences without discharge planning; evaluating post-discharge needs for patients identified (or who request an evaluation, or whose physician requests one); developing a discharge plan if the evaluation indicates one; and initiating implementation of the plan before discharge. The guidance is explicit that the regulation applies to inpatients, not outpatients, and that hospitals must have written policies and procedures covering each standard.
The survey instrument CMS uses to assess compliance — the Hospital Discharge Planning Worksheet attached to Survey and Certification Letter 15-12 — shows what surveyors actually check. Among the record-level items: whether the evaluation assessed the patient’s ability to perform activities of daily living; whether it assessed the patient’s and support persons’ ability to provide self-care; whether it assessed whether specialized medical equipment is available and whether home modifications can be made safely; and, critically for this argument, whether the evaluation included an assessment of available community-based services when the patient or family cannot meet all care needs (item 4.10). The worksheet also asks whether the evaluation assessed the patient’s insurance coverage and how that coverage might or might not provide for necessary post-hospital services (item 4.13).
Read those items carefully. The hospital must assess whether community-based services exist. It must assess whether insurance will cover what the patient needs. It is not required by the retrieved CMS guidance to secure a shelter bed, to enroll the patient in a coordinated entry system, or to confirm that the community-based service it identified will actually accept the patient. The regulation governs the quality of the hospital’s planning process. It does not govern the availability of the thing being planned for.
This is not a loophole in the pejorative sense. It is a jurisdictional boundary. CMS regulates hospitals. It does not regulate homeless services systems, and it has no authority to order a CoC to accept a referral.
What HUD actually requires — and what it does not
HUD’s Coordinated Entry Management and Data Guide describes coordinated entry as “larger than a single grant or a program” and as “a key component of a comprehensive crisis response” — a way of structuring a CoC’s system of care so that it “fits together intentionally and efficiently.” The guide lays out three governance functions: a policy oversight entity that establishes and oversees policies for the four core elements (Access, Assessment, Prioritization, and Referral); a management entity that runs day-to-day operations, including standardized screening and assessment and training; and an evaluation entity that plans an annual coordinated entry evaluation and feeds findings back to the policy oversight entity. The guide states these responsibilities can be executed separately by different entities or combined.
Note what the governance structure is built to do: manage a queue. Coordinated entry is a prioritization system. It takes people who present at an access point, assesses them with a standardized tool, ranks them by vulnerability or need according to local policy, and refers them to available housing resources in that order. It is designed to allocate scarcity fairly. It is not designed as a hospital discharge destination, and the retrieved HUD guide does not establish hospitals as mandatory coordinated entry partners.
The guide is also candid about data. It states that “the data collection needs for coordinated entry are significantly more complex than for a single project or even a collection of projects,” and that new data points must be combined with existing Homeless Management Information System (HMIS) data — the local database CoCs use to record client-level information — to understand whether the system is functioning as planned. That is an admission that coordinated entry’s measurement depends on joining datasets. The question is which datasets, and whether the hospital’s record is among them.
The feedback loop that measures the wrong boundary
Here is the mechanism that keeps the loop in place. Each system measures its own side of the handoff, and each measurement is satisfied by compliance on its own side.
On the hospital side, the CMS survey worksheet asks whether the hospital tracks readmissions as part of its review of the discharge planning process (item 3.2) and whether its assessment of readmissions includes an evaluation of whether readmissions were potentially due to problems in discharge planning or implementation (item 3.3). This is a readmission feedback loop. It tells the hospital whether its discharge process is associated with patients coming back. It does not tell the hospital whether a housing-unstable patient ever reached a coordinated entry access point, because that event is not a readmission and does not appear in the claims or medical record data the hospital reviews.
On the CoC side, the HUD guide requires an annual coordinated entry evaluation covering access, assessment, prioritization, and referral, with findings routed back to the policy oversight entity. That evaluation can tell the CoC whether its queue is moving, whether assessment is consistent, and whether referral outcomes match prioritization. It does not tell the CoC how many people were discharged from hospitals into housing instability, because hospitals are not a data source in the HMIS pipeline unless a local agreement puts them there.
The result is a structural blind spot. A patient can be screened as high-risk, evaluated, given a discharge plan that lists community-based services, and discharged — fully compliant with 42 CFR 482.43 as surveyed. The same person can later present at a coordinated entry access point, be assessed, and enter the queue — fully consistent with the CoC’s coordinated entry policies. Both institutions can report success. The transition between them was never measured by either.
Why a shared identifier is the load-bearing failure
The hospital’s discharge plan lives in the medical record. The coordinated entry assessment lives in HMIS or the CoC’s designated data system. These are separate systems with separate consent regimes, separate data dictionaries, and no shared person-level identifier. HUD’s guide devotes an entire chapter to data privacy and security, including uses and disclosures, consent requirements, and comparable databases — a signal that cross-system data sharing is governed and constrained, not incidental.
This is the load-bearing failure. Without a shared identifier, you cannot join the two datasets to ask the only question that matters: for patients flagged as housing-unstable at discharge, what fraction completed a coordinated entry assessment, and how long did it take? The hospital cannot answer because it does not hold coordinated entry data. The CoC cannot answer because it does not hold discharge data. The surveyor cannot answer because the CMS worksheet does not ask. The annual coordinated entry evaluation cannot answer because hospitals are outside its scope.
This is not a research gap in the sense of “more study needed.” It is a measurement architecture gap. The data required to close the loop are generated by both systems and joined by neither.
What a compliant-but-failed handoff looks like
Consider a hypothetical but structurally ordinary case, clearly labeled as illustrative rather than observed. An inpatient with no fixed address is screened at admission as needing discharge planning. The evaluation documents that the patient cannot meet care needs in the current environment and that no support person can provide them. Item 4.10 of the survey worksheet is satisfied: the discharge planner documents available community-based services, including a local shelter and a coordinated entry access point. Item 4.13 is satisfied: the planner notes that the patient’s coverage does not include housing. A discharge plan is written and implementation begins before discharge. The hospital is compliant.
But the discharge plan lists the access point’s phone number. The patient is discharged to a temporary location. The coordinated entry access point has no record that a hospital referral is coming, because no referral was transmitted — the plan documented the service, it did not connect to it. The patient may present, or may not. If the patient does present, the assessment proceeds on the CoC’s timeline, not the hospital’s. If the patient does not present, no one in either system is notified, because no one owns the transition.
This is the loop. It is not a story about bad actors. It is a story about two rule sets that each stop at their own property line.
Where the redesign levers already exist
The argument that follows is a recommendation, not a finding from the retrieved sources. But it rests on requirements that are already in the rules, which means it does not require new legislation to begin.
First, point the CMS reassessment requirement at the handoff. The discharge planning Condition of Participation includes a reassessment standard: the plan must be reassessed based on changes in patient condition, changes in available support, and changes in post-hospital care requirements. The survey worksheet asks whether the hospital’s policy includes a process for ongoing reassessment (item 2.8) and whether the plan was updated when significant changes occurred (item 4.17). A hospital could define “confirmed connection to a coordinated entry access point” as a post-discharge care requirement for patients flagged as housing-unstable, making the handoff part of what reassessment tracks. That is a local policy choice within the existing standard, not a new federal mandate.
Second, point the HUD annual evaluation at the same outcome. The HUD guide requires the evaluation entity to assess the coordinated entry process and identify policy and process improvements. A CoC could add a measure of hospital-origin referrals — how many people entering coordinated entry were discharged from a hospital within a defined window — and report it to the policy oversight entity. That is a local evaluation design choice within the existing requirement.
Third, define a shared transition outcome and assign an owner. The outcome should be specific and measurable: for patients flagged as homeless or housing-unstable at hospital discharge, a confirmed coordinated entry assessment within a defined number of days. The number is a local policy decision, not a research finding. The owner is the harder question. Neither CMS nor HUD assigns it. A CoC and its largest hospital systems could assign it by agreement — a named staff role on each side responsible for the handoff — without waiting for federal direction.
Fourth, build the data-use agreement before building the dashboard. HUD’s guide treats data sharing as a governed activity requiring defined disclosure protocols and consent. Any joint measurement of the handoff requires a data-use agreement that specifies what is shared, under what consent, and for what purpose. The agreement is the prerequisite; the metric is downstream of it.
What would need to be true for the loop to close
Three conditions, stated plainly. A common flag at hospital screening that identifies housing instability in a way that can be transmitted outside the medical record under a defined consent. A coordinated entry assessment that can be triggered before discharge rather than after, so the queue position is established while the patient is still in a bed and the discharge plan can reflect a real referral rather than a listed service. And a shared outcome measure that both the hospital’s quality review and the CoC’s annual evaluation report against, so that each institution’s compliance is conditioned on the transition succeeding, not merely on its own process running.
None of these requires new statutory authority. All of them require someone to own a boundary that the current rules leave unowned. Until that happens, the boarding-to-shelter loop will continue to produce the same result: two compliant systems, one failed handoff, and no dataset that can see it.
Questions readers ask
Does CMS require hospitals to enroll patients in coordinated entry or HMIS? The retrieved CMS guidance does not say this. The discharge planning Condition of Participation requires assessment of community-based services and insurance coverage, not enrollment in any specific housing system.
Does HUD require hospitals to participate in coordinated entry? The retrieved HUD guide describes CoC participation expectations for coordinated entry partners but does not establish a hospital mandate. Hospitals are not mandatory coordinated entry partners under the retrieved guidance.
Is there a federal dataset that tracks hospital-to-shelter transitions? Not in the sources retrieved for this analysis. The hospital’s readmission tracking and the CoC’s annual coordinated entry evaluation measure different things on different sides of the handoff.
Can a hospital be cited for discharging a patient to homelessness? The retrieved CMS survey worksheet assesses whether the discharge planning process was followed — screening, evaluation, plan development, reassessment — not whether a shelter bed was available. The regulatory hook is process compliance, not placement outcome.
Sources
CMS, State Operations Manual Transmittal 87, Revised Appendix A, Interpretive Guidelines for 42 CFR 482.43, Discharge Planning (July 19, 2013). cms.gov
CMS, Survey and Certification Letter 15-12, Attachment 3, Hospital Discharge Planning Worksheet. cms.gov
HUD, Coordinated Entry Management and Data Guide. hudexchange.info