Finding the right moment to start advance care planning — from wellness visits to diagnosis, care transitions, and disease progression
The single biggest structural problem in advance care planning is timing: most conversations happen far too late, often during a health crisis when capacity is already compromised and clinicians are focused on acute decisions. The most reliable countermeasure is to decouple the ACP conversation from illness altogether — anchoring it instead to something that already recurs on a predictable schedule for almost every older adult: the Medicare Annual Wellness Visit (AWV) and other age-based primary care milestones.
Advance care planning (ACP) is only meaningful if the person doing the planning still has decisional capacity — the ability to understand their situation, weigh options, and communicate a choice. Waiting for a health crisis to prompt the conversation is precisely the wrong strategy, because crises are exactly when capacity is most likely to be impaired (sedation, delirium, critical illness, stroke, advanced dementia) and when family members are asked to guess at preferences under extreme time pressure and emotional strain.
The Medicare Annual Wellness Visit, established under the Affordable Care Act and available to all Medicare Part B beneficiaries once every 12 months, is attractive as an ACP anchor precisely because it is scheduled, recurring, and completely independent of whether the patient is currently sick. It already includes a Health Risk Assessment, a review of functional status, cognitive screening, and an updated list of current providers and medications — all of which are directly relevant inputs to an ACP conversation (Who should speak for you? What matters most to you if your health changes?).
CMS formally recognized ACP as an optional elective element of the AWV in 2016, and simultaneously created two dedicated Current Procedural Terminology (CPT) billing codes — 99497 for the first 30 minutes of a face-to-face advance care planning discussion (explaining and helping complete standard forms, with or without completing those forms) and add-on code 99498 for each additional 30 minutes. Because these are separately identifiable services, a clinician can document and bill for ACP as its own encounter, rather than trying to fit it, unpaid, into the margins of a problem-focused visit.
The financial design of the ACP codes matters for how reliably the trigger fires in practice. When 99497 (and 99498, if needed) is furnished on the same day as the AWV, Medicare waives both the Part B deductible and the 20% coinsurance — the visit is effectively free to the beneficiary, matching the "no-cost preventive service" framing of the AWV itself. This was a deliberate policy lever: remove the copay barrier so that neither the clinician nor the patient has a financial disincentive to have the conversation.
Outside that specific pairing, ACP can still be billed as a stand-alone service in any care setting — office visit, hospital, skilled nursing facility, even via telehealth — but then ordinary Part B cost-sharing applies (20% coinsurance plus any unmet deductible), which introduces friction for both scheduling and patient acceptance. This asymmetry is one reason health systems that want to systematize ACP explicitly train front-desk and scheduling staff to route eligible patients into ACP-plus-AWV combined visits rather than separate encounters.
Despite the incentive design, AWV uptake itself has historically hovered around 30–35% of eligible beneficiaries in a given year, and only a minority of those visits include a documented ACP component — meaning the "easiest" systematic trigger in the entire ACP timing framework is still reaching only a small fraction of the population it was designed for. Billing-code availability is necessary but nowhere close to sufficient.
Turning 65 and Medicare eligibility is the most visible age-based trigger, but it is not the only one, and restricting ACP thinking to Medicare-age patients misses an entire population. Turning 18 is itself a meaningful trigger — legal adulthood means parents no longer have automatic authority to make medical decisions for a child, so every young adult benefits from designating a health care proxy, particularly before leaving for college or military service. Other informal but useful anchors include retirement planning appointments, first grandchild, updating a will or other estate documents (often done alongside a health care power of attorney by the same attorney), and routine periodic health examinations in primary care independent of Medicare status.
Structured patient-facing tools make these routine-visit conversations more efficient: PREPARE for Your Care (an online, plain-language decision aid) and the Five Wishes document are both designed to be completed by the patient in advance of a visit, so that the clinician's time is spent confirming and documenting rather than starting from a blank page. Trials pairing a decision aid like PREPARE with a routine primary-care visit have shown two- to three-fold increases in documented ACP completion compared with usual care, because the patient arrives having already done the reflective work.
The persistent barrier is time: the average primary-care visit runs 15–20 minutes and is usually consumed by acute complaints and chronic-disease management, leaving little room for an unhurried values conversation unless it is explicitly scheduled as its own visit — which is exactly what the AWV-plus-99497 pairing is designed to enable.
A conversation that happens on a Tuesday afternoon during a routine wellness visit, while the patient is healthy and thinking clearly, is categorically more valuable than the same conversation attempted in an ICU family meeting. Systematizing the wellness-visit trigger is the cheapest, least emotionally fraught way to raise population-level ACP completion — yet it remains the most underused, because it depends entirely on the AWV itself being scheduled and attended.
Once a patient receives a diagnosis of cancer, heart failure, chronic kidney disease, COPD, or a neurocognitive disorder such as dementia, the abstract question "what matters to you if your health changes" becomes concrete. Evidence consistently shows that the diagnosis moment is both the most natural and most under-used opportunity to start ACP — clinicians frequently defer the conversation until prognosis is unambiguous, by which point the patient may be too sick, too rushed, or too impaired to participate fully.
Oncology has historically been the specialty most associated with structured serious-illness communication tools, largely because a cancer diagnosis creates an unambiguous, dateable moment when everyone in the room recognizes that the stakes have changed. The Serious Illness Conversation Guide, developed at Ariadne Labs, is now used across hundreds of health systems specifically to standardize this moment: a short, scripted set of open-ended questions (What is your understanding of your illness? What are your most important goals if your health worsens? What abilities are so central to your life that you cannot imagine living without them?) that a trained clinician can complete in roughly 20 minutes.
Despite these tools existing, empirical studies of real-world oncology practice — most notably work by Jennifer Mack and colleagues in adolescent and adult cancer cohorts — repeatedly find that documented end-of-life goals-of-care discussions happen at a median of less than one month before death, well after the diagnosis window has closed and typically after multiple lines of therapy have already been tried. This "prognostic paralysis," where clinicians delay difficult conversations to avoid taking away hope or because prognosis feels too uncertain to discuss, is one of the best-documented failure modes in all of oncology communication research.
The counter-evidence is equally well known: the landmark 2010 Temel et al. trial in metastatic non-small-cell lung cancer showed that patients randomized to early, integrated palliative care (which routinely includes ACP-style conversations) starting at diagnosis had better quality of life, less depression, and — despite receiving less aggressive end-of-life care — modestly longer survival than those receiving oncologic care alone. Early conversation is not only ethically preferable; it appears to be clinically neutral-to-beneficial rather than harmful.
Non-cancer serious illnesses pose a harder timing problem because their trajectories are far less predictable than cancer's more classic decline curve. Heart failure guidelines from the American College of Cardiology, American Heart Association, and Heart Failure Society of America (ACC/AHA/HFSA) now explicitly recommend that ACP discussions begin at initial diagnosis and be revisited at every hospitalization and at major changes in New York Heart Association (NYHA) functional class — because a patient can move from NYHA class II (mild symptoms) to class IV (symptoms at rest) over months, with each hospitalization carrying meaningfully elevated in-hospital mortality risk.
Chronic obstructive pulmonary disease follows a similarly unpredictable, stepwise-decline pattern staged by GOLD (Global Initiative for Chronic Obstructive Lung Disease) criteria, and, like heart failure, is notorious for underuse of ACP relative to cancer — patients with COPD are far less likely to have documented goals-of-care conversations than cancer patients with comparable estimated prognosis, in part because there is no single moment as dramatic as a cancer diagnosis to anchor the conversation.
A widely used complementary screening tool is the "surprise question" — would I be surprised if this patient died in the next 12 months? — which clinicians can apply at any encounter, not only at initial diagnosis, to flag patients across a heterogeneous set of non-cancer illnesses who would benefit from an ACP conversation regardless of exactly which diagnosis they carry. It is deliberately low-specificity and high-sensitivity: better to over-trigger a conversation than to miss a patient who is closer to the end of life than their chart suggests.
Dementia diagnoses create the single most time-urgent version of the diagnosis trigger, because the disease itself progressively erodes the exact capacity that ACP depends on. Alzheimer's Association clinical practice recommendations now explicitly call for ACP discussion "at the time of diagnosis" — ideally while the patient is still in mild cognitive impairment or early-stage dementia and can meaningfully participate in naming a health care proxy, articulating values around interventions like feeding tubes and hospitalization, and completing a durable power of attorney for health care.
In practice this window is frequently missed: many patients are not formally diagnosed until moderate-stage disease, by which point decisional capacity for complex, hypothetical future scenarios is already compromised, and any subsequent ACP effort shifts from patient-led planning to family/surrogate-led decision-making — often without the benefit of ever having heard the patient's own stated preferences. Absent a documented proxy, the eventual fallback is state default-surrogate hierarchy laws or, in the worst case, court-appointed guardianship, both of which are slower, more adversarial, and less reflective of individual values than a conversation held at diagnosis would have been.
Because dementia care so often intersects with functional decline, this diagnosis-triggered conversation is also where POLST (Physician/Portable Orders for Life-Sustaining Treatment) documents begin to become relevant — a distinction worth making early: advance directives describe general future wishes and name a proxy for any capable adult, while POLST is a medical order set, signed by a clinician, intended for patients whose health status already suggests they could die within a year, and travels with the patient across care settings including EMS response.
Every hospital admission, hospital discharge, and nursing-facility placement is a legally recognized checkpoint for advance care planning in the United States. The 1990 Patient Self-Determination Act requires Medicare- and Medicaid-participating hospitals, skilled nursing facilities, home health agencies, and hospices to ask every adult patient about existing advance directives at admission and to provide written information about their rights — but being asked a screening question and having a genuine conversation are very different things, and this is where systematic ACP timing frequently breaks down into box-checking.
Hospitalization is a forcing function: whatever else is happening in a patient's outpatient life, admission clerking under the Patient Self-Determination Act requires the intake team to ask whether the patient has an advance directive and to document the answer in the chart. In well-run systems this question is the opening of a genuine conversation, ideally involving the primary team, a palliative care consult, or a trained ACP facilitator; in many systems, however, it is reduced to a single checkbox in the admission electronic health record template, satisfying the letter of the federal requirement while accomplishing none of its substantive intent.
Discharge is an equally important, and often better-timed, second opportunity — particularly after a stay involving the intensive care unit, a major surgery, or a new diagnosis made during the admission itself. The patient is now more concretely aware of their vulnerability than they were on the day they walked into the emergency department, and discharge planning already involves a structured conversation about follow-up care, medications, and functional needs, into which an ACP or goals-of-care check-in fits naturally.
The 30-day post-discharge period carries the highest documented risk of adverse events, readmission, and death for many hospitalized populations, especially older adults with multiple comorbidities — which is exactly why transition-of-care ACP triggers are timed at discharge rather than left to the next scheduled outpatient visit, which may not occur until well after this critical window has passed.
Admission to a skilled nursing facility (SNF) or long-term care setting is one of the most reliably systematized transition triggers in American health care, because it is embedded directly into regulatory and clinical documentation requirements. The Minimum Data Set (MDS) — the federally mandated resident assessment instrument completed at SNF admission and at regular intervals thereafter — includes specific items on advance directive status, and POLST paradigm programs (active in most US states under varying names — POLST, MOLST, MOST, POST) are frequently initiated or reviewed at exactly this point of care transition.
Because POLST is a signed medical order rather than a general statement of preference, it travels with the patient and is immediately actionable by paramedics, hospital staff, and nursing facility personnel — which is precisely the population for whom SNF-admission ACP matters most: patients whose overall health trajectory already suggests a realistic possibility of dying within the coming year, for whom a document limited to naming a future proxy is no longer sufficient on its own.
Some state Medicaid and nursing-home quality-reporting frameworks now include advance directive completion or POLST completion as a facility-level quality indicator, giving nursing homes a direct institutional incentive — beyond the federal PSDA floor — to make this transition trigger fire reliably rather than perfunctorily.
Not every meaningful transition is a change of physical location. A patient who begins needing help with activities of daily living (ADLs — bathing, dressing, transferring, toileting) or instrumental activities of daily living (IADLs — managing medications, finances, transportation) has crossed a threshold that is, on its own, a strong and diagnosis-independent predictor of mortality and hospitalization risk over the following year, comparable in magnitude to many single-organ prognosis models.
The frailty phenotype described by Fried and colleagues — unintentional weight loss, self-reported exhaustion, low grip strength, slow gait speed, and low physical activity — offers a structured, measurable way to detect this decline before it manifests as a hospitalization, and geriatric assessment programs increasingly embed an ACP trigger directly into frailty screening: a patient who meets three or more Fried criteria, or who has a new fall, or whose gait speed drops below a threshold on a timed walk test, is flagged for an ACP conversation regardless of whatever specific diagnoses happen to be on their problem list.
This illness-agnostic framing matters because a large share of the oldest and most vulnerable patients — those with multimorbidity rather than one single dominant terminal diagnosis — will never experience a single dramatic "new diagnosis" moment of the kind described in Stage 2. For them, functional decline is the trigger, and it can be captured systematically only if primary care and geriatric programs are actively screening for it rather than waiting for it to present as a crisis.
Advance care planning is not a one-time legal task to be completed and filed away; it is an iterative process that needs to be revisited as illness progresses and as patients' values genuinely change in response to their evolving health. Recurrent hospitalizations, an ICU admission, or any new life-limiting complication — dialysis initiation, a new metastasis, ventilator dependence — are the clearest signals that an existing ACP document may no longer reflect the patient's current situation or current preferences, and that the conversation needs to happen again, not for the first time but as an update.
A single hospitalization is common and, by itself, only a modest signal of overall trajectory. Two or more hospitalizations within a six-month window is a materially different signal: it indicates that whatever chronic disease is driving admissions is not being adequately controlled by outpatient management alone, and it is used explicitly as an escalation trigger in condition-specific protocols — most notably in heart failure programs following HFSA guidance, where recurrent HF hospitalizations prompt not just medication optimization but a parallel palliative-care or ACP referral.
General-purpose readmission-risk models originally built to predict 30-day readmission — the LACE index (Length of stay, Acuity of admission, Comorbidities, Emergency department visits) and the HOSPITAL score among others — are increasingly repurposed by health systems as ACP triggers rather than purely as operational/financial tools: a patient flagged as high readmission risk is, by the same underlying logic, a patient whose overall illness trajectory warrants a proactive goals-of-care conversation before the next crisis rather than during it.
The key conceptual shift at this stage is from "has this patient ever had an ACP conversation" to "is this patient's most recent ACP conversation still accurate" — a distinction that most first-generation ACP quality metrics, which simply track whether any advance directive exists in the chart, fail to capture.
Intensive care unit admission is the highest-stakes environment in which ACP timing operates, because decisions about mechanical ventilation, vasopressors, renal replacement therapy, and resuscitation status often must be made within hours, frequently from a surrogate decision-maker rather than the patient directly. Structured communication frameworks — the VitalTalk curriculum and the Ariadne Labs Serious Illness Conversation Guide among the most widely disseminated — are used to run family meetings that explicitly connect the patient's previously stated values (if documented) to the specific decisions now on the table, and to negotiate "time-limited trials" of a given intervention with a pre-agreed point of reassessment rather than an open-ended commitment.
Multiple randomized and quasi-experimental studies of structured ICU family communication interventions, including work led by J. Randall Curtis and colleagues, have found that proactive, structured goals-of-care communication in the ICU reduces ICU and hospital length of stay and reduces family psychological distress (anxiety, depression, post-traumatic stress symptoms), without increasing mortality — directly rebutting the common clinician fear that raising goals-of-care conversations "gives up" on patients or hastens death.
A foundational and frequently overlooked task at ICU admission, independent of any broader goals-of-care discussion, is simply confirming who the legally authorized surrogate decision-maker is and documenting it clearly — a large share of ICU communication breakdowns trace back not to disagreement about the patient's wishes but to ambiguity about who has the authority to speak for them.
One of the most robust and counter-intuitive findings in the ACP and quality-of-life literature is "response shift": patients' own stated preferences about acceptable health states and acceptable trade-offs systematically change as they actually experience declining health, usually in the direction of accepting states they would have rejected as intolerable when healthy — sometimes described as the disability paradox. An advance directive completed five years earlier, while accurate at the time, may no longer represent what the same person would say today, precisely because the person has adapted to circumstances they could not have accurately imagined in advance.
This is the strongest argument for treating every new life-limiting complication as its own trigger for revisiting ACP rather than treating a single completed document as durable indefinitely: initiation of chronic dialysis, a new distant metastasis, a feeding-tube decision, or a period of ventilator dependence each represent a fundamentally new health state that the patient has not previously been asked to weigh in on, and each is common enough across serious illness trajectories that health systems can build explicit re-trigger rules around them.
At the far end of this trajectory sits formal hospice eligibility — a clinical judgment, certified by two physicians, that a patient has a life expectancy of six months or less if the disease follows its usual course. Reaching this threshold functions as a final, unambiguous trigger tier: by this point ACP is no longer optional forward planning but urgent, concrete decision-making about the remaining months of life, and any prior document should be actively reconciled with the patient's current, present-tense wishes rather than simply carried forward unchanged.
Every trigger described so far — wellness visits, new diagnosis, care transitions, disease progression — depends on some individual clinician noticing it and acting on it in real time, which is precisely why so many opportunities are missed even in health systems that know, in principle, exactly when ACP conversations should happen. The final and most consequential piece of ACP timing strategy is not identifying more triggers, but building infrastructure — electronic health record decision support, dedicated non-physician facilitators, and quality measurement — that makes the triggers fire reliably without depending on any one person's memory.
The most direct way to remove reliance on individual clinician memory is to have the electronic health record itself detect when a trigger condition is met and surface a best-practice advisory (BPA) — a pop-up alert or a flagged item on a problem list — prompting the clinical team to have or document an ACP conversation. Trigger logic can be built directly from the categories already described: a new problem-list entry for metastatic cancer, a diagnosis code for dementia, an NYHA class III/IV designation, a frailty index above a threshold, or an admission count exceeding two in six months can all be encoded as structured rules that fire without anyone having to remember them.
A particularly influential trial by Courtright and colleagues (JAMA Internal Medicine) used a default-based nudge embedded in the EHR — rather than requiring an opt-in action, the palliative-care or ACP referral was pre-selected as the default order, requiring the clinician to actively decline rather than actively choose it. This behavioral-economics framing of "make the desired action the path of least resistance" produced significantly higher rates of palliative-care and ACP engagement than an equivalent opt-in prompt, illustrating that even small differences in how a trigger is presented in the workflow materially change whether it results in action.
More recent systems layer machine-learning mortality-prediction models on top of standard EHR data (labs, vitals, admission patterns, diagnosis codes) to estimate a patient's probability of death within 6–12 months and use that score, rather than any single diagnosis or admission count, as the trigger for outreach — in effect automating the "surprise question" at scale across an entire patient panel rather than relying on any one clinician's individual, encounter-by-encounter judgment.
Even a perfectly designed EHR alert only creates an opportunity; someone still has to conduct a skilled, unhurried conversation, and physicians' time is the scarcest resource in the entire system. Many of the most successful systematized ACP programs instead train non-physician facilitators — nurses, social workers, chaplains, or dedicated ACP navigators — to conduct structured conversations using a standardized curriculum, freeing physician time for the encounters where their specific medical expertise is required (interpreting prognosis, explaining specific interventions) while the facilitator handles the values-elicitation and documentation work.
The best-known example is Respecting Choices, originally developed in La Crosse, Wisconsin by the Gundersen Health System, which trains community facilitators through a tiered curriculum (First Steps for healthy adults, Next Steps for patients with chronic progressive illness, Last Steps for patients nearing end of life) matched to exactly the trajectory-based staging used throughout this framework. La Crosse became a widely cited case study in the ACP literature because community-wide advance directive completion reached roughly 90%+ of adults who died in the region — a figure dramatically higher than typical US population rates — attributed largely to the systematic, facilitator-driven, community-embedded nature of the program rather than to any single clinical intervention.
Health systems adopting this model typically embed navigators at multiple trigger points simultaneously — during AWVs, at oncology intake, at SNF admission, and post-discharge follow-up calls — so that whichever trigger fires first for a given patient, a trained person, not an overburdened physician, is available to actually carry out the conversation.
Systems cannot manage what they do not measure, and ACP has historically suffered from weak, inconsistent measurement — many health systems can report whether any advance directive exists somewhere in a patient's chart, but far fewer can report whether that directive is current, whether it reflects a documented conversation (rather than a form handed over unread), or whether a "missed opportunity" occurred — a case where a recognized trigger was present but no conversation took place.
National Quality Forum (NQF)-endorsed measures and related payer quality programs increasingly track ACP documentation as a discrete, reportable metric, giving health systems a financial and reputational incentive — beyond intrinsic clinical motivation — to build and maintain the infrastructure described above. Internally, the more sophisticated systems maintain a longitudinal registry that merges all four trigger categories (wellness, diagnosis, transition, progression) into a single view per patient, flagging anyone who is overdue for review under any category, and routing that flag to whichever combination of EHR alert and human navigator the system has established as its workflow.
Closing the loop also means auditing failure, not just success: retrospective chart-review studies that specifically quantify missed-opportunity encounters — visits or admissions where a clear trigger criterion was met but no ACP conversation was documented — consistently find that the majority of triggered encounters, in systems without systematic prompts, still result in no conversation at all. That gap, more than any single missing billing code or undertrained clinician, is the actual target of "systematizing" ACP timing: not inventing new triggers, but making sure the triggers that already clearly exist are never silently ignored.
The Gundersen/La Crosse Respecting Choices experience is frequently cited alongside data showing that communities with systematized, facilitator-driven ACP achieve not only higher documented advance-directive completion but also lower end-of-life hospital utilization and costs — evidence that reliable timing infrastructure, not any single conversation in isolation, is what ultimately converts ACP from a well-intentioned ideal into consistently goal-concordant care.