Conducting a code-status conversation — from clinical prognosis to a documented, portable, revisitable decision
Before a code-status conversation ever reaches the bedside, the clinician has homework to do: synthesizing the patient's underlying disease trajectory, functional reserve, and physiologic frailty into a defensible estimate of whether cardiopulmonary resuscitation could plausibly restore a meaningful quality of life. Skipping this step turns the conversation into a reflexive administrative checkbox — "full code or DNR?" — asked identically of a fit 45-year-old post-appendectomy patient and a 91-year-old with metastatic pancreatic cancer and multi-organ failure. Good code-status practice starts with prognosis, not paperwork.
The single most common failure mode in code-status practice is treating it as an intake-form checkbox disconnected from clinical reality. Every hospitalized patient is technically asked "if your heart stops, do you want us to attempt resuscitation?" — but the honest, useful version of this question requires the clinician to already have an opinion, grounded in medicine, about what CPR would actually accomplish for this specific patient.
Key variables that shift the clinical-benefit estimate:
• Underlying diagnosis and trajectory: a reversible acute process (e.g., a drug overdose, a treatable arrhythmia) carries a fundamentally different resuscitation calculus than an irreversible terminal illness (metastatic cancer, end-stage heart failure, advanced dementia) • Functional status: Eastern Cooperative Oncology Group (ECOG) performance status and frailty indices predict both arrest survival and post-arrest quality of life; patients who are bed-bound or fully dependent before arrest rarely regain independent function even if resuscitation is "successful" • Physiologic reserve: multi-organ dysfunction, sepsis, and pre-arrest hypotension are strongly associated with failure to achieve return of spontaneous circulation (ROSC), and even when ROSC occurs, in-hospital re-arrest is common • Etiology of anticipated arrest: a primary cardiac arrhythmia in an otherwise stable patient responds far better to defibrillation than a hypoxic or metabolic arrest secondary to progressive organ failure
This is not about withholding CPR from anyone who might benefit — it is about arriving at the conversation with a real clinical opinion, so that patients and families are choosing between options a physician actually believes are viable, rather than facing an abstract yes/no with no medical anchor.
There are two opposite ways to get this stage wrong.
The first is "checkbox medicine": a rushed admission-note question, asked without context, without exploring values, and without conveying any clinical judgment — "so, do you want to be full code?" This produces defaults that reflect fear or misunderstanding rather than genuine preference, and it is the pattern most responsible for both unwanted resuscitation attempts on dying patients and, less commonly, premature DNR orders placed on patients who might actually have benefited from CPR.
The second failure is unilateral non-offering or overriding: some clinicians, having concluded CPR is futile, either fail to raise the topic at all or write a DNR order without adequately involving the patient or surrogate. While professional societies (American Medical Association, American College of Physicians) recognize that physicians are not ethically obligated to offer or perform interventions with no realistic prospect of achieving the patient's goals, the determination of "futility" is contested, institution-specific, and should almost never be exercised without transparent communication, an explanation of clinical reasoning, and — except in rare emergencies — an opportunity for the patient or family to seek another opinion.
The defensible middle path is a clinical assessment that is honest about probabilities, transparent about uncertainty, and always paired with a values-first conversation before any order is written.
A useful pre-conversation assessment answers four questions:
1. What is the primary diagnosis and its expected trajectory — reversible, chronic-stable, or progressively terminal? 2. What is the patient's current functional and cognitive baseline, and how does that compare to pre-illness baseline? 3. If cardiac or respiratory arrest occurred today, what would most likely cause it — and how reversible is that specific mechanism? 4. What does the published literature suggest for survival-to-discharge and neurologically intact survival in a population most similar to this patient?
This assessment does not predetermine the outcome of the conversation — patients and legally authorized surrogates retain the right to choose full resuscitation even against a physician's recommendation in most circumstances — but it ensures the physician enters the discussion as a clinical guide rather than an administrator collecting a signature.
The framing of a code-status conversation determines its outcome almost as much as the medical facts within it. "Do you want us to do everything?" is a leading, binary, guilt-laden question that few people can answer honestly — of course a loving family wants "everything" done for someone they love. Structured communication approaches instead open with the patient's values and goals, explain what CPR actually involves in concrete terms, and only then connect those values to a recommendation. This single reframing — values first, intervention second — is one of the most consistently replicated findings in palliative-care communication research.
Structured serious-illness communication frameworks — most notably the Serious Illness Conversation Guide developed at Ariadne Labs (Bernacki & Block, and colleagues, building on work from the VitalTalk collaborative) — share a common architecture: establish prognosis understanding, explore the patient's goals and fears, identify what abilities are so central to their life that their loss would be unacceptable, and only then discuss specific interventions like CPR in light of those values.
A typical opening sequence:
1. "What is your understanding of where things stand with your illness?" — surfaces prognostic awareness gaps before any decision is discussed 2. "As you think about the future, what is most important to you?" — elicits values (time with family, avoiding suffering, remaining mentally present, a specific milestone) 3. "What are you hoping for? What are you worried about?" — surfaces both hope and fear explicitly, rather than assuming 4. Only after this groundwork: "Given what you've told me matters most, I want to talk with you about a specific medical decision — what happens if your heart or breathing were to stop."
This sequencing matters because CPR preference is not really a standalone medical fact to be elicited — it is a downstream consequence of what a person values. A patient who says "I want to be there for my daughter's wedding next spring, even if it means some suffering" is answering the code-status question before it is even asked; the clinician's job is to make that connection explicit.
Most patients and families have never seen real CPR performed and have no concrete mental model of the intervention they are being asked to accept or decline. Concrete, plain-language description is essential before any decision-point is offered:
• Chest compressions: forceful, rhythmic compression of the chest wall (approximately 2 inches deep, 100–120 per minute) — commonly causing rib fractures, especially in older or osteoporotic patients • Defibrillation: electrical shock delivered if a "shockable" rhythm (ventricular fibrillation or pulseless ventricular tachycardia) is present — many arrests, especially in patients dying of progressive illness, present with non-shockable rhythms (asystole, pulseless electrical activity) where defibrillation is not applicable • Airway management: bag-mask ventilation, followed frequently by endotracheal intubation and mechanical ventilation • Medications: epinephrine, amiodarone, and other agents delivered through IV or intraosseous access during the resuscitation attempt • What "success" means: restoring a spontaneous heartbeat (return of spontaneous circulation, ROSC) is only the first hurdle — many patients who achieve ROSC do not survive to hospital discharge, and a meaningful fraction of survivors sustain neurological injury from the period of inadequate brain perfusion during the arrest
Describing these mechanics — without either sanitizing or sensationalizing them — gives patients and families the concrete information needed to weigh the intervention against their own values, rather than reacting to an abstraction.
Framing research consistently shows that clinicians who ask "what matters most to you" before discussing CPR mechanics produce documentation that better reflects patients' actual values, whereas clinicians who lead with "do you want everything done" more often produce default full-code orders regardless of prognosis — largely because the question itself supplies no information to reason with.
Popular media has taught most patients an inaccurate script for what happens during cardiac arrest — a script this simulator addresses directly in Stage 3. But the framing stage is where the clinician should first signal, gently, that the coming conversation may challenge assumptions: "A lot of what people picture when they think about CPR comes from television, and it's often quite different from what really happens in the hospital. Would it be okay if I shared some real numbers with you?"
This kind of "ask permission before informing" technique (a core VitalTalk skill, sometimes abbreviated as an "ask-tell-ask" structure) respects patient autonomy over the pace of difficult information, reduces the likelihood of a defensive reaction, and sets up the misconception-correction that follows as a collaborative act rather than a paternalistic lecture.
Nowhere is the gap between public perception and medical reality larger than for CPR outcomes. Decades of televised medical drama have taught viewers that cardiac arrest is usually survivable and usually leaves the patient neurologically intact — an impression contradicted by nearly every published outcomes study. Correcting this misconception, without shaming or overwhelming the patient, is one of the most consequential communication tasks in the entire code-status discussion, because unrealistic survival expectations are among the strongest predictors of a default full-code order that does not reflect what the patient would actually choose if accurately informed.
The most cited empirical demonstration of media-driven misinformation about CPR is Diem SJ, Lantos JD, Tulsky JA, "Cardiopulmonary Resuscitation on Television — Miracles and Misinformation," New England Journal of Medicine, 1996. The authors reviewed CPR portrayals across three popular medical dramas of the era (ER, Chicago Hope, Rescue 911) and found that approximately 75% of depicted resuscitation attempts were for trauma, drowning, or other conditions poorly representative of the actual causes of cardiac arrest in real patients (who far more often arrest from chronic cardiac disease, sepsis, or progressive illness). Of the patients shown undergoing CPR on screen, roughly 67% survived immediately and about 65% of that group appeared to survive to hospital discharge, with most depicted as returning to their prior baseline function.
This stood in stark contrast to the actual published survival literature available at the time, which showed survival-to-discharge rates after in-hospital cardiac arrest in the range of roughly 15%, and dramatically lower rates for patients with pre-existing terminal or severely debilitating conditions. Subsequent replications of the "TV-CPR" analysis in the following decades have found the pattern persists, and follow-up survey research has repeatedly found that patients who report medical dramas as a primary information source about CPR are more likely to overestimate survival odds and more likely to request full-code status without being able to accurately describe what the intervention involves.
Contemporary in-hospital cardiac arrest outcome data (drawn from large registries such as the American Heart Association's Get With The Guidelines–Resuscitation database) paint a more sobering — though not hopeless — picture than either TV or a purely nihilistic stance would suggest:
• Overall survival to hospital discharge after in-hospital cardiac arrest: historically cited in the 15–20% range in older literature, with more recent registry-based estimates trending somewhat higher (approximately 20–25%) reflecting improvements in resuscitation systems and early defibrillation — the number varies significantly by cohort, rhythm, and era of study, so ranges rather than a single figure are most honest to quote • Shockable rhythms (ventricular fibrillation/pulseless VT) carry meaningfully better survival than non-shockable rhythms (asystole/PEA), which are more common in patients dying of progressive medical illness • Patients with metastatic cancer: multiple studies report survival to hospital discharge after in-hospital CPR in the low single digits, with several cohorts reporting 0% survival to discharge among patients with widely metastatic disease undergoing CPR • Advanced frailty, multi-organ failure, and pre-arrest sepsis or hypotension are each independently associated with survival rates well below the general in-hospital average • Even among survivors, a meaningful fraction sustain some degree of new neurological impairment, and post-arrest hospital courses frequently include prolonged ICU stays, mechanical ventilation, and further invasive procedures
Communicating these numbers is not about frightening patients out of a preference for resuscitation — some patients, fully informed, still rationally choose full code, and that choice deserves respect. It is about ensuring the choice being made is actually informed.
Key Insight: The single most important sentence a clinician can say in this stage is often not a statistic — it is the clarification that "DNR does not mean we stop treating you." A do-not-resuscitate order applies narrowly to the specific event of cardiac or respiratory arrest; it does not restrict antibiotics, dialysis, chemotherapy, surgery, ICU-level monitoring, pain control, or any other treatment unless those are separately and explicitly declined. Surveys of patients and even some clinical staff repeatedly find this distinction poorly understood, and the resulting fear — that agreeing to DNR means "giving up" on all care — is one of the most common reasons patients decline a status that otherwise matches their values.
Effective misconception-correction follows an "ask-tell-ask" pattern rather than a lecture:
• Ask what the patient already believes: "Can you tell me what you picture happening if your heart were to stop?" This surfaces the specific misconception (often drawn from television or a prior family experience) rather than assuming a generic knowledge gap • Tell — briefly, in plain numbers, without jargon: "In real hospitals, for someone with your kind of illness, CPR restarts the heart in only a small number of cases, and even then many people don't survive to leave the hospital. That's quite different from what's often shown on TV, where it usually works." • Ask again to check understanding and invite reaction: "What are your thoughts, hearing that?" This step is frequently skipped but is where the actual values-based decision-making happens — the clinician learns whether the new information changes the patient's preference, and whether further emotional support is needed before continuing
This approach preserves patient autonomy — the goal is never to argue a patient into a particular code status, but to ensure the decision, whatever it is, rests on an accurate model of what the intervention can and cannot do.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| TV medical dramas (composite) | Any depicted arrest | Diem, Lantos & Tulsky, NEJM 1996 — content analysis of ER, Chicago Hope, Rescue 911 | ~67% immediate survival, ~65% to discharge shown |
| General in-hospital arrest | Mixed inpatient population | Registry data (e.g. AHA Get With The Guidelines–Resuscitation) | ~15–25% survival to discharge |
| Advanced frailty / multi-organ failure | Pre-arrest hypotension, sepsis, dependence | Reduced physiologic reserve limits ROSC and post-ROSC survival | ~5–10% survival to discharge |
| Metastatic cancer | Widely disseminated malignancy | Multiple cohort studies, several reporting 0% survival to discharge | 0–5% survival to discharge |
A code-status conversation that is never translated into an unambiguous, correctly scoped, and portable medical order has accomplished nothing durable. Vague documentation ("patient wishes to be DNR, discussed with family") or orders that fail to travel with the patient across care transitions are a leading cause of goals-discordant resuscitation — patients receiving CPR they had explicitly declined, or, less commonly, being denied resuscitation that was never actually refused. Rigorous documentation practice closes this gap.
"DNR" is not a single monolithic status — it describes a family of possible order configurations that must be specified precisely to avoid dangerous ambiguity at the bedside:
• Full code: attempt chest compressions, defibrillation, intubation, and all standard resuscitation medications in the event of cardiac or respiratory arrest • DNR/DNI (do not resuscitate / do not intubate): no chest compressions or defibrillation attempted, and no endotracheal intubation performed — but note DNR and DNI are conceptually separable orders and can, in principle, be specified independently (e.g., a patient may accept non-invasive ventilation or intubation for a reversible respiratory event while still declining chest compressions) • Partial or "slow" codes: orders that specify some but not all resuscitative measures (e.g., defibrillation permitted but chest compressions declined, or medications but no intubation). These are ethically controversial — many institutions and ethicists discourage truly "partial" resuscitation protocols as internally inconsistent or difficult to execute safely under the time pressure of an actual arrest — and any partial order must be documented with unusual precision about exactly which interventions are authorized • DNR-comfort care (DNR-CC, sometimes DNR-CC Arrest): the most restrictive designation, typically indicating that beyond declining resuscitation, the patient's overall goal is comfort-focused care — new interventions aimed at prolonging life (e.g., new antibiotics for a new infection, transfer to ICU, dialysis initiation) are also generally declined, though this always should be spelled out explicitly rather than assumed, since DNR-CC is a specific and stronger designation than DNR alone
Ambiguous documentation — a free-text progress note stating "family agrees with DNR" without specifying DNI status, without specifying whether comfort-focused care is also intended, and without a corresponding order actually entered into the physician order system — is a well-recognized source of resuscitation errors during actual emergencies, when a responding team has seconds, not minutes, to determine what is authorized.
Physician (or Practitioner) Orders for Life-Sustaining Treatment (POLST), also implemented under names such as Medical Orders for Life-Sustaining Treatment (MOLST) or Physician Orders for Scope of Treatment (POST) depending on the state, is a standardized, brightly colored, single-page medical order form designed specifically to solve the portability problem that plagues advance directives and hospital-only DNR orders.
Key distinctions from a standard advance directive:
• An advance directive (living will, healthcare proxy designation) documents a patient's wishes and appoints a decision-maker, but it is not itself a medical order — clinicians and emergency responders cannot act directly on a living will's language without an intervening clinical interpretation • A POLST/MOLST form is a standing physician order, signed by a physician, nurse practitioner, or physician assistant, that emergency medical services, nursing facilities, and receiving hospitals can act on immediately and directly, exactly as written, without requiring a new conversation or new order at each transition • POLST/MOLST is intended for patients with serious illness or frailty who are likely to face a decision about life-sustaining treatment in the near term — it complements rather than replaces an advance directive, and is not recommended for generally healthy patients • The form typically documents not only resuscitation status but also broader scope-of-treatment preferences: hospitalization/transfer preferences, use of antibiotics, and artificial nutrition, giving a fuller picture of the patient's overall goals than a code-status order alone
Because the form is designed to be immediately recognizable (most US programs use a distinctive bright pink or similarly high-visibility paper) and to travel with the patient — attached to the medical record, given to family to carry, and honored across hospital, nursing facility, ambulance, and home settings — it directly addresses the historically common failure of DNR orders that existed only in one hospital's chart and were unknown to paramedics or a receiving facility during a subsequent crisis.
A code status is only as good as its ability to be found and honored at the moment it matters — often in a different building, by different clinicians, under time pressure. Practices that reduce transfer failures include:
• Placing the POLST/MOLST (or local equivalent) in a standardized, predictable physical and electronic location — commonly the refrigerator door at home (a well-established EMS convention in many regions) and the front of the chart or a flagged banner in the electronic health record in facilities • Re-verifying code status explicitly at every transition of care: hospital admission, transfer between units (especially to or from an ICU), transfer to a skilled nursing or rehabilitation facility, and hospital discharge — a step frequently omitted under time pressure, but responsible for a large share of goals-discordant resuscitation events • Ensuring the electronic health record's active order (not just a scanned form or a note) reflects current status, since responding code teams generally act on the active order set, not on narrative documentation • Communicating the order, and the reasoning behind it, to the full care team and to family members who might otherwise call emergency services reflexively during a crisis at home
None of this replaces the conversation itself — documentation is the last step, not a substitute for the values-based discussion that should precede it — but a well-run code-status program treats documentation and portability with the same seriousness as the conversation, because an order that cannot be found or is misread at the bedside fails the patient just as completely as no conversation at all.
Perhaps the most important — and most frequently violated — principle of code-status practice is that it is never a one-time decision permanently fixed at a single admission. Clinical trajectories change, patients' values and priorities evolve as illness progresses, and family understanding deepens or fractures over time. Treating a DNR order as "settled" after one conversation ignores all of this, while treating every conversation as starting from zero exhausts patients and families unnecessarily. Skilled longitudinal practice reassesses at the right moments and actively manages disagreement when it arises.
The landmark SUPPORT study (Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments), published in JAMA in 1995, remains one of the most consequential pieces of evidence in the history of end-of-life communication research. In a multi-site cohort of seriously ill hospitalized patients, the study found substantial and troubling gaps: physicians frequently misunderstood or were entirely unaware of their patients' actual resuscitation preferences, DNR orders when written were often entered very late in a hospitalization (frequently only in the final days of life), and many patients experienced substantial pain and aggressive intervention inconsistent with preferences they had, in fact, expressed when actually asked.
A structured nurse-facilitated communication intervention tested in the same study, intended to improve communication between patients, families, and physicians about prognosis and preferences, failed to significantly change these outcomes — a sobering finding that reshaped the field's understanding that better communication requires more than a single structured conversation or a single extra conversational touchpoint; it requires an ongoing, systematized practice of revisiting goals as illness evolves, embedded in how care is delivered rather than treated as an isolated task.
Subsequent research, including the influential Respecting Choices program in La Crosse, Wisconsin, and Detering et al.'s 2010 randomized controlled trial in the BMJ, demonstrated that structured, facilitated advance care planning conversations — when actually integrated longitudinally into care rather than performed once — improve documentation of and adherence to patient wishes, increase family satisfaction, and are associated with reduced depression and anxiety among surviving family members, without evidence of increased mortality or hastened death.
Code status should be actively reassessed, not merely inherited from a prior chart entry, at each of the following junctures:
• Any significant change in clinical trajectory — a new diagnosis, a major complication, an ICU admission, or a marked functional decline changes the calculus underlying an earlier decision and should prompt a fresh conversation, not silent carry-forward of an old order • Transitions of care — hospital admission, transfer between units, transfer to a skilled nursing or rehabilitation facility, and hospital discharge are each recognized moments where orders are prone to being lost, mistranscribed, or simply not re-verified, and are recommended checkpoints for active reassessment • Milestone time intervals in chronic serious illness — many palliative-care and oncology programs build routine goals-of-care check-ins into follow-up visits at defined intervals (e.g., every few months, or at each major treatment-response evaluation) rather than waiting for a crisis • Explicit patient or family request — a change of heart in either direction deserves immediate engagement, not deferral to the next scheduled encounter • Approaching a recognized prognostic threshold — for example, a new estimate of a life expectancy measured in weeks to months, which many frameworks (including the "surprise question": would I be surprised if this patient died in the next year?) use as a trigger for a more explicit goals-of-care conversation
Each reassessment need not be as extensive as the original conversation — often it is a brief confirmation ("Given how things have changed, I want to check in — does the plan we discussed before still reflect what you want?") — but it must actually happen, and it must be documented as a fresh confirmation, not silently assumed.
Disagreement — between the patient and family, between family members, or between family and the clinical team — is common in serious illness and, left unmanaged, tends to default toward maximal intervention, since escalating care is rarely challenged in the moment while withholding it often is. Constructive approaches include:
• Clarify whose voice governs: if the patient has decision-making capacity, their own expressed wishes take precedence over family preference, and this should be stated clearly and compassionately, even when family members disagree • When the patient lacks capacity, identify the correct legally authorized surrogate (per the patient's advance directive, healthcare proxy designation, or the jurisdiction's default surrogate hierarchy) and clarify that the surrogate's role is substituted judgment — representing what the patient would have wanted, not what the surrogate personally prefers • Separate grief and guilt from genuine disagreement about goals: family conflict often stems less from a true dispute about the patient's wishes than from unprocessed grief, guilt about past absence or conflict, or fear of being blamed for "giving up" — palliative care and chaplaincy consultation are frequently more effective here than repeated purely medical explanation • Convene a structured family meeting with all key decision-makers present rather than having fragmented conversations with individual family members, which tend to produce inconsistent understanding and mutual suspicion • Use ethics consultation when disagreement persists despite good-faith communication efforts — most hospitals maintain an ethics committee specifically for these situations, and involving them early, rather than only after conflict has escalated, tends to produce better outcomes for everyone • Recognize that time-limited trials of intervention ("let's try the ventilator for 48 hours and reassess based on how she responds") can sometimes resolve disagreement more effectively than an immediate binary decision, by giving family members a concrete, evidence-based off-ramp
Structured, longitudinally embedded goals-of-care conversations do not increase mortality or hasten death — a finding replicated across multiple studies including Temel et al.'s 2010 NEJM trial of early palliative care integration in metastatic lung cancer, which found improved quality of life, less aggressive end-of-life care, and, notably, modestly longer survival in the group receiving earlier, more structured conversations about goals and prognosis. Revisiting code status well is not merely an ethical nicety — it is measurably good medicine.