Why documented wishes go unfound at the bedside — registries, EHR banners, wallet cards, and FHIR-based interoperability for point-of-care retrieval
Advance care planning research has spent three decades optimizing how directives get completed — counseling scripts, easy-read forms, digital tools. Far less attention has gone to what happens after signature: whether the document is physically or electronically present at the single moment it is needed, usually in an emergency department, ambulance, or ICU, from a patient who cannot speak for themselves and may carry no identifying documentation at all.
The SUPPORT study (Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments), published in JAMA in 1995, remains the foundational demonstration of this gap: even with structured efforts to elicit and document patient preferences for seriously ill hospitalized patients, physicians frequently did not know their patients' resuscitation preferences, and treatment often did not match stated wishes. Three decades of subsequent research — spanning paper advance directives, POLST forms, and now digital registries — has repeatedly reproduced the same underlying pattern: completion rates for advance directives, health care proxy forms, and POLST orders have all risen over time, but the rate at which those documents are actually retrieved and acted upon at the point of decision has lagged far behind.
Meta-analyses estimate that roughly one in three U.S. adults has completed some form of advance directive, with completion more common among older adults, those with chronic illness, and those who have had prior contact with palliative care or hospice services. Yet survey and chart-review studies of emergency departments and ICUs consistently find that the directive is absent from the record, unknown to the treating team, or discovered only after major interventions have already been initiated — often hours or days into an admission, well past the window in which it could have changed the first, most consequential decisions (intubation, CPR, ICU admission).
Three structural features of emergency and acute care make advance directive retrieval especially fragile, precisely when it matters most:
Decisional incapacity at presentation — the patient who most needs their advance directive honored is frequently the one least able to state that it exists. Altered mental status, intubation, cardiac arrest, or severe trauma remove the most reliable retrieval channel (asking the patient directly) at exactly the moment retrieval is time-critical.
No persistent identifier linking the person to the document — unlike allergies or blood type, which are typically captured at every registration, an advance directive has no single, universally-checked field. A patient may have completed a directive with a hospice agency, a primary care clinic, or an attorney, none of which are queried by a treating hospital's registration workflow by default.
Time pressure inverts the usual chart-review process — in routine outpatient care, a clinician has minutes to review a chart before a decision is made. In cardiac arrest or respiratory failure, decisions about resuscitation must be made in seconds. A retrieval process requiring a phone call to an outside registry, a fax request to a clinic, or a search through unstructured scanned documents cannot function on that timescale — by the time the document is found, the intervention it was meant to prevent has often already occurred.
The core finding driving every stage of this simulation: the bottleneck in advance care planning is no longer primarily about getting patients to complete a directive — it is about making an already-completed directive discoverable, structured, and computable at an arbitrary point of care, under time pressure, by clinicians who have never met the patient before.
When a directive is not found, default practice is resuscitative: absent clear documentation, emergency clinicians are trained — appropriately, from a legal and ethical standpoint — to provide full treatment, including CPR and intubation, even for patients who had explicitly declined these interventions in a document sitting unread in a different health system's file room. This produces several well-documented downstream harms: treatment inconsistent with expressed preferences, family distress when a previously-discussed plan is not honored, potentially non-beneficial invasive interventions in patients with terminal illness, and, at a systems level, higher utilization of ICU beds, mechanical ventilation, and other resource-intensive care that patients had already declined in writing.
The remaining four stages of this simulator walk through the layered set of technical and policy interventions that have been built, in roughly chronological and architectural order, to close this gap: centralized state registries, EHR-embedded flagging, patient-carried point-of-care solutions, and finally the FHIR-based interoperability standards intended to unify all of the above into a single, portable, computable record.
Beginning in the early 2000s, a number of U.S. states established official electronic registries where residents (or their clinicians) could upload advance directives, health care power of attorney forms, or POLST orders for retrieval by authorized providers. The registry model solves the storage problem — a single, queryable repository — but its real-world effectiveness has been consistently limited by low voluntary enrollment, fragmented state-by-state design, and the fact that a registry only helps if the treating clinician thinks to query it.
A typical state registry follows a common architecture: a resident (or their attorney, physician, or hospice agency) scans and uploads a signed advance directive, health care proxy designation, or POLST form to a state-run or state-contracted electronic database. The individual is issued a wallet card or registry ID number, and authorized health care providers can query the registry — usually by patient name and date of birth, sometimes by a registry ID — to retrieve the scanned document during an encounter.
Some states operate a specific registry for POLST (Physician/Provider Orders for Life-Sustaining Treatment) forms, which are actionable medical orders signed by a clinician, distinct from a traditional advance directive (a legal document expressing wishes, not a standing order). Because POLST orders are meant to be followed immediately by EMS and hospital staff without further interpretation, several of the earliest and most actively used electronic registries were built specifically around POLST — for example, Oregon's POLST registry, one of the first in the nation, and New York's eMOLST (electronic Medical Orders for Life-Sustaining Treatment) system, which requires clinician verification and directly informs EMS protocols in participating regions.
Other states maintain a broader advance directive registry open to any legally valid document, generally with lower barriers to enrollment but correspondingly less clinical urgency attached to the content — an advance directive states preferences; it does not, by itself, instruct EMS to withhold CPR the way a signed POLST order does.
Even in states with a mature registry program, voluntary enrollment consistently falls far short of advance directive completion rates. Several compounding factors explain the gap:
Awareness — many patients who complete a directive with an attorney, a hospice intake coordinator, or during a hospital admission are never told that a state registry exists, let alone how to submit their document to it.
Friction of the upload step — registry enrollment is typically an additional, separate action after the directive itself is signed: scanning, mailing, or manually uploading a document to a state web portal is a task that competes with the emotional and administrative burden already involved in completing the underlying directive, and it is frequently skipped.
No enrollment requirement — registry participation is voluntary in essentially every state program; there is no mechanism analogous to organ donor registration at the DMV that captures the majority of the population by default.
Provider-side unfamiliarity — even where a registry is well-populated, treating clinicians in a true emergency often do not know the registry exists, do not have quick credentialed access to query it, or are unaware their state operates one at all — particularly for out-of-state patients presenting far from home.
The practical result is that registries function well for the subset of engaged, typically older and chronically-ill patients whose care teams are already registry-aware — commonly patients enrolled through hospice or palliative care programs — while doing little for the much larger population of adults who complete a directive once, informally, and never interact with the registry system again.
Because each state built its registry independently — different legal document types accepted, different verification requirements, different technology vendors, different query interfaces for providers — there is no mechanism for a Massachusetts hospital to query an Oregon POLST registry, or for EMS crews to check a registry from a neighboring state during interstate travel, a common scenario for retirees, seasonal residents ("snowbirds"), and patients transferred between regional health systems.
This fragmentation mirrors a broader pattern in U.S. health IT: strong point solutions built at the state or institutional level, with no federal standard requiring them to interoperate with one another. A patient who dutifully registers their POLST in their home state gains essentially no benefit if they suffer a cardiac event while traveling, because the treating EMS crew and hospital in another jurisdiction have no way to know the registry — or the record inside it — exists. Stage 3 examines the next layer of the solution stack: rather than relying on an external registry a clinician must remember to query, embed the flag directly inside the electronic health record the clinician is already looking at.
Rather than requiring a clinician to remember an external registry exists and separately query it, the EHR integration approach embeds advance directive and POLST status directly into the patient's chart — typically as a persistent, high-visibility banner at the top of the record, visible on every screen, alongside allergies and code status. This dramatically improves retrieval within a patient's home health system, but exposes a second-order problem: interoperability across the dozens of competing EHR vendors and thousands of independently-operated health systems that make up U.S. health care.
Human-factors research on clinical alerting consistently shows that information buried several clicks deep in a chart is far less likely to influence real-time decisions than information placed persistently in the clinician's direct line of sight. Leading EHR platforms now support a dedicated, color-coded banner — separate from the general problem list — specifically for code status and advance directive presence, often paired with a direct link to the scanned document or structured order.
Well-designed implementations distinguish between several related but distinct pieces of information that are easy to conflate: (1) whether an advance directive exists at all, (2) whether a POLST/MOLST order exists specifying actionable resuscitation and treatment preferences, (3) the current code status order actually active in this admission (which may have been updated by the current care team and can legitimately differ from an older directive), and (4) who the legal health care proxy or surrogate decision-maker is. Conflating these — for example, treating "has an advance directive on file" as equivalent to "do not resuscitate" — is a documented source of dangerous misapplication of a patient's actual wishes, and best-practice banner design keeps them visually and structurally distinct.
The EHR banner solves retrieval beautifully within a single health system's instance of a single vendor's software — the same database, the same patient identifier, the same interface every clinician in that system already uses. It does essentially nothing for a patient who is unconscious in an emergency department that has never seen them before, running a different EHR vendor, with no shared patient identifier and no live connection to the home system's database.
This is the same fragmentation problem seen with state registries in Stage 2, recurring one layer down the stack: instead of fifty uncoordinated state registries, the barrier is now thousands of uncoordinated EHR instances across independent hospitals, health systems, and outpatient practices, frequently running different vendor platforms that historically stored advance directive information as an unstructured scanned image (a PDF or TIFF of the paper document) rather than as discrete, machine-readable data — meaning even where a data-sharing connection exists, the receiving system may only be able to display a static scan, not query specific fields like "does this patient decline mechanical ventilation."
National interoperability efforts — health information exchanges (HIEs), vendor-agnostic networks such as Carequality and CommonWell Health Alliance, and the federally-created Trusted Exchange Framework and Common Agreement (TEFCA) — have made progress connecting EHR instances for general clinical data (labs, medication lists, discharge summaries) across vendors. Advance directive and POLST data, however, has historically lagged behind other data classes in these exchange pipelines, partly because it was rarely captured in a structured format standardized enough to exchange automatically in the first place.
Federal policy has increasingly pushed EHR vendors and health systems toward exchange rather than isolation. The 21st Century Cures Act (2016) and its subsequent ONC information-blocking regulations prohibit health care providers and health IT developers from unreasonably interfering with the access, exchange, or use of electronic health information — creating regulatory pressure that, over time, is expected to extend more forcefully to advance care planning data as it becomes more consistently structured.
The practical effect for advance directive accessibility is indirect but important: as the broader U.S. health IT ecosystem is pushed toward structured, standards-based exchange for all clinical data categories, advance directive and POLST information is increasingly being pulled into that same structured pipeline rather than remaining a special-case scanned document — setting up the standards work described in Stage 5. In the meantime, Stage 4 examines a category of solutions that sidesteps institutional interoperability entirely by putting retrieval directly in the patient's own hands.
A parallel line of solutions bypasses institutional data infrastructure entirely by putting a retrieval mechanism directly on or with the patient: wallet cards, medical alert bracelets and necklaces, smartphone apps, and QR codes linking to a hosted copy of the directive. These solutions are attractive precisely because they do not depend on any registry or EHR being queried correctly — but they introduce a different point of failure: the item must be present, current, and discoverable by first responders at the exact moment of the emergency.
The simplest and longest-standing point-of-care solution is a physical wallet card or a medical alert bracelet/necklace (e.g., through established medical ID jewelry programs) indicating that the person has a DNR order, POLST form, or advance directive, sometimes with a phone number or registry ID for further verification. EMS and emergency personnel are broadly trained to check for medical alert jewelry as part of a rapid physical assessment of an unresponsive patient — a workflow that predates electronic health records entirely and remains one of the fastest retrieval paths that exists, because it requires no database query, no login, no network connectivity, and no interoperability between systems at all.
The well-documented weakness is reliability of presence: a card can be left in a coat pocket at home; a bracelet can be removed for a shower and not replaced; the item may not be updated if the person's wishes or code status change; and regional variation in EMS protocols means the significance of a given card or bracelet is not always uniformly recognized or immediately actionable without a corroborating physician order (in many jurisdictions, a wallet card alone is not legally sufficient for EMS to withhold resuscitation — a properly executed POLST form or state-specific EMS DNR order is required).
A newer generation of solutions layers a QR code onto a wallet card, bracelet, or phone lock screen that, when scanned, links directly to a hosted digital copy of the patient's advance directive, POLST form, and emergency contact/proxy information — often maintained through a dedicated advance care planning platform. Notable examples in this space have included MyDirectives (a cloud-based advance directive platform designed for interoperable, EMS- and clinician-accessible retrieval), MIDEO (Medical Interventions: Definitions and Outcomes, a video-based advance directive format intended to be more comprehensible and less ambiguous than text-only documents), and Vynca, which focuses on structured electronic POLST completion and registry integration for health systems.
These platforms attempt to solve several of the weaknesses of a pure physical card: content can be updated any time without reissuing a physical item, a video or structured-question format can reduce ambiguity relative to free-text legal language, and a hosted link can, in principle, be queried by any responder with a smartphone and camera regardless of which EHR or state registry the patient's home clinicians use — making this approach one of the few genuinely vendor-agnostic and state-agnostic retrieval paths available today.
The corresponding limitations are adoption and awareness rather than technology: a QR-linked directive is only useful if (a) the patient consistently carries or wears the code, (b) the responding clinician or EMS provider knows to look for and scan it, and (c) the hosting platform remains active, accessible without excessive authentication friction in an emergency, and legally recognized as sufficient documentation in that jurisdiction.
None of the patient-carried approaches solve the underlying systemic fragmentation described in Stages 2 and 3 — they route around it for the specific case where the item is physically present and recognized. Because of this, most health-system and EMS medical-direction guidance treats wallet cards, bracelets, and QR-linked apps as an important complementary layer, not a substitute for a registry entry or EHR flag: the safest architecture is redundant — the same information present in a state or regional registry, flagged prominently in the patient's EHR, and also carried physically or digitally by the patient, so that at least one retrieval path succeeds regardless of where or how the emergency unfolds.
This redundancy principle motivates Stage 5: rather than treating registries, EHR banners, and patient-carried tools as separate, competing point solutions, the current direction in health informatics is to standardize the underlying data itself — so that whichever system captures it, any other authorized system can retrieve, interpret, and act on the same structured record.
The long-term fix being pursued by health informatics standards bodies, federal policy, and national advance care planning coalitions is not another registry or another app — it is a shared data standard that lets any of the previous four layers interoperate: HL7 FHIR-based structured advance directive data, national POLST and registry harmonization efforts, and information-blocking policy that treats advance care planning data as a first-class category of health information entitled to the same interoperable exchange as medications and lab results.
HL7 FHIR (Fast Healthcare Interoperability Resources) is the modern data-exchange standard underlying most current U.S. health IT interoperability work, including the APIs mandated by federal information-blocking rules for patient and provider access to health records. Rather than exchanging a scanned PDF of a signed form — as most legacy registry and EHR integrations have historically done — a FHIR-based approach represents the advance directive as structured, discrete data resources: for example, a Consent resource capturing the scope and provisions of the directive, DocumentReference resources linking to the source legal document, and CarePlan or ServiceRequest-style resources capturing specific actionable orders such as a POLST's resuscitation and treatment-intensity preferences.
HL7's Advance Directive Interoperability (ADI) implementation guide work specifically targets this problem: defining a standardized, computable way to represent advance directive and POLST content so that any FHIR-conformant system — regardless of vendor, regardless of state, regardless of whether it is a hospital EHR, an EMS electronic patient care report system, or a national registry — can query, retrieve, and correctly interpret the same underlying record, including which specific interventions are declined or requested, not merely that "a document exists somewhere."
Structured data also enables something scanned documents cannot: automated, real-time clinical decision support. A FHIR-conformant system can, in principle, surface a discrete alert — "patient has a POLST order declining mechanical ventilation" — directly inside an order-entry workflow, rather than requiring a clinician to locate, open, and manually read a multi-page scanned legal document during a resuscitation.
National POLST — the nonprofit coalition that maintains the POLST paradigm across adopting states — has worked for years to harmonize form content, legal standing, and clinical training across the patchwork of state programs described in Stage 2, reducing (though not eliminating) the state-by-state fragmentation that limits cross-jurisdiction recognition of POLST orders. Harmonized form design also makes it easier to build a single national structured-data schema on top of state programs that otherwise evolved independently.
At the exchange-infrastructure level, the Trusted Exchange Framework and Common Agreement (TEFCA), established under the 21st Century Cures Act and operationalized by ONC with an initial cohort of Qualified Health Information Networks (QHINs) beginning in 2023, is intended to create a single "network of networks" so that a participating provider can query for a patient's available health information — in principle including advance care planning data — across the entire connected national ecosystem, rather than needing bilateral connections to every possible source system individually. As advance directive data becomes more consistently structured under FHIR-based standards, it is positioned to travel through exactly this kind of national query infrastructure rather than remaining locked in a single state registry or EHR instance.
Policy levers reinforce the technical ones: information-blocking regulations increasingly discourage health systems and vendors from treating any category of patient health data, including advance directives, as proprietary or siloed by default, and quality/accreditation frameworks are beginning to incorporate advance care planning documentation and accessibility as measurable performance domains — creating institutional incentive, not just technical capability, to keep this data exchange-ready.
Layering the four preceding stages under a common FHIR-based standard, with national exchange infrastructure connecting them, produces a retrieval workflow qualitatively different from any single layer alone: a patient registers a directive once, through whatever channel is most convenient for them — a hospice intake team, a primary care visit, a state registry portal, or a smartphone app — and that structured record becomes queryable by any authorized clinician or EMS provider at any connected point of care, automatically surfaced as a banner or decision-support alert the moment the patient is registered in an unfamiliar system, without requiring anyone to remember which registry, which app, or which state to check.
None of the earlier layers become obsolete under this model — a wallet card or QR code remains valuable as an immediate, network-independent fallback for a first responder in the field before any database query is even possible, and state POLST registries remain the authoritative source feeding the structured national layer. The goal of systems-level standardization is not to replace these tools but to stop treating them as competing silos and instead make them consistent, redundant views onto the same underlying, computable, patient-controlled record — closing the loop that Stage 1 opened: an advance directive that is not just completed, but reliably found.
The trajectory across all five stages is a shift from "where is the document" to "what does the structured data say" — moving advance care planning from a records-retrieval problem (find the right scanned PDF, in the right registry, in the right state) to a data-interoperability problem (query a standardized FHIR resource through a nationally connected network) that mirrors the same evolution already underway for medication lists, problem lists, and lab results across U.S. health IT.