Who Goes First: Triage, the Constitution, and the Calculation Nobody Wants to Make

The Hook: A Law the Court Demanded – and the Same Court Struck Down

On 16 December 2021, the First Senate of the German Federal Constitutional Court did something rare in the history of the Basic Law: it condemned the legislature not for regulating too much, but for regulating nothing at all. Nine people with disabilities had filed a constitutional complaint, arguing that the state left them unprotected in the event of pandemic-related triage. The court agreed. The legislature had violated Article 3(3) sentence 2 of the Basic Law – "No person shall be disfavoured because of disability" – by failing to put safeguards in place. It was ordered to act "without delay."

The legislature acted. At the end of 2022, § 5c of the Infection Protection Act entered into force: a short, densely written provision specifying the sole criterion by which life-sustaining intensive care capacity may be allocated, which characteristics are forbidden, how many physicians must be involved, and what must be documented. By general assessment it was Europe's first triage statute.

On 23 September 2025, the same First Senate declared that law void – by six votes to two. The decision was published on 4 November 2025 (press release no. 99/2025). This time the complainants were fourteen physicians, backed by the German Medical Association and the Marburger Bund. For non-lawyers the reasoning is at first disappointingly dry: the federation lacks legislative competence. Article 74(1) no. 19 of the Basic Law permits measures against communicable diseases; the triage rules, however, were "pure pandemic aftermath law" – they attached to a scarcity resulting from a pandemic but did not serve to combat one. In addition, the mandatory allocation criteria infringed physicians' therapeutic freedom under Article 12(1) of the Basic Law, without constitutional justification.

The result is a situation that would be hard to invent: the 2021 duty of protection still stands. The law meant to discharge it is gone. Competence now lies with the sixteen federal states, which have so far done nothing. What actually governs is a medical societies' guideline with no force of law.

This article is not only about that episode. It is about the structure underneath it – a decision situation studied in ethics for fifty years, in which every conceivable rule violates a moral intuition we want to keep. And it ends at an empirical finding that undercuts the entire normative debate: the criterion on which nearly all parties agreed – probability of survival – can barely be measured reliably with the instruments available. Build a just rule, insert a poorly calibrated predictor, and you have not built a just rule.

Part 1: What Triage Is – and What It Is Not

The term and its origin

"Triage" comes from the French trier, to sort. The term stems from the military medicine of the Napoleonic wars; Dominique Jean Larrey, surgeon of the Grande Armée, organized care of the wounded by urgency of treatment rather than by rank – a revolutionary idea at the time. Classical military triage sorts into three groups: those who will survive without treatment; those who will survive with it; those who will die even with it. The middle group is treated first.

One distinction matters above all, and it is almost always lost in public debate. Ordinary clinical triage – in every emergency department, every day – is a sequencing decision under time pressure: who is seen first? It produces waiting, not forgoing. The triage at issue here is an allocation decision under absolute scarcity: more people need a ventilator or an ECMO machine than there are machines, and whoever gets nothing will probably die. That is not a problem of prioritization but of selection.

The four features that make the case hard

What distinguishes allocation triage from almost every other ethical problem is four structural properties occurring at once:

1. Rivalry. The resource is not divisible. Half a ventilator saves no one by halves.
2. Complete exhaustion of alternatives. There is no option that harms nobody. Doing nothing is a decision with the same consequence as a decision made.
3. Irreversibility. The outcome is death, not disadvantage.
4. Decision under ignorance. Each individual patient's prognosis is a probability statement, and one with large uncertainty.

Point 4 is often omitted in philosophical discussion because it spoils the thought experiments. In practice it is the hardest of the four.

Ex ante and ex post

Two cases that look morally entirely different and pose medically the same question:

Ex ante triage: Two patients arrive simultaneously; there is one bed. You choose.

Ex post triage: A patient has been ventilated for four days and his prognosis has deteriorated dramatically. A new patient arrives with a markedly better chance of survival. May the first be taken off the machine?

Almost everyone feels a difference here. Almost every consequentialist analysis says there is none: if the resource is scarce and saving life is the goal, time of arrival must not matter – otherwise first come, first served has been reintroduced through the back door. This conflict is not theory. It was the core of the German legislative debate of 2022, and it is the reason physicians sued. More on that in Part 7.

Part 2: The Map – Eight Principles in Four Families

The cleanest systematic survey of the field dates from 2009. Govind Persad, Alan Wertheimer, and Ezekiel J. Emanuel published a paper in The Lancet arranging eight allocation principles into four families – and refuting each one, at least as a sole principle.

The authors do not conclude that one must choose a single principle, but that each captures a morally relevant aspect and none captures all of them. Their own proposal, the complete lives system, combines five: younger people who have not yet lived a complete life are favoured, modulated by prognosis, saving the most lives, lottery, and instrumental value. The age curve is not linear but bell-shaped: infants are prioritized below adolescents, because more has been invested in a fifteen-year-old than in an infant and more is destroyed by that loss.

The proposal has a remarkable reception history: it became a political target ("death panels") and remained professionally influential. Above all it shows what happens when allocation ethics is spelled out honestly – the results almost always feel wrong, including when the arguments are sound.

The pandemic version: four values, six recommendations

In March 2020, Emanuel and colleagues published an operational version for COVID-19 in the New England Journal of Medicine. They name four fundamental values – maximize benefits, treat people equally, promote and reward instrumental value, give priority to the worst off – and derive six recommendations. Four matter for our purposes:

• Maximizing benefits takes priority. Justified by the exceptional nature of the situation, not by general utilitarianism.
• Health care workers are given priority – not because they are worth more, but because they act as multipliers.
• With equal prognosis: lottery, not first come, first served. The reasoning matters: first come, first served appears neutral but systematically favours those who live closer, are better informed, and are more mobile.
• No distinction between COVID patients and other patients. Scarcity is scarcity, not a question of diagnosis.

And: the recommendations explicitly provide that ongoing treatments be reassessed after defined intervals – that is, ex post triage.

Part 3: Should the Numbers Count? Taurek, Parfit, and the Aggregation Dispute

Behind every rescue rule lies an assumption so self-evident that it is easily missed: that it is better to save six people than one.

In 1977, John Taurek denied precisely that in Philosophy & Public Affairs. His essay "Should the Numbers Count?" constructs the case: one drug; six people on one island each need a sixth of it; one person on another island needs the whole dose. Taurek's argument is not that the six count for less. It is that there exists no subject for whom the sum of the suffering is worse. Each individual's loss is total for that individual; "sixfold death" is not an experience anyone has. To add the numbers is to treat persons as vessels for well-being. Taurek's conclusion: flip a coin. Everyone is entitled to an equal chance.

Derek Parfit replied in 1978 with "Innumerate Ethics." His counter: if each person has an equal claim to a chance, it does not follow that the number may be ignored – for the six together hold six claims, and a procedure that saves the one with 50 percent probability gives each of the six only half a chance when all six could have been given a full one. The debate that followed – Frances Kamm, T. M. Scanlon, Alastair Norcross – is among the longest-running in moral philosophy and remains unsettled.

For triage this is no academic side issue. It decides whether "save as many lives as possible" is a principle or a category mistake.

Why German constitutional law sits close to Taurek

On 15 February 2006 the Federal Constitutional Court declared § 14(3) of the Aviation Security Act void (1 BvR 357/05) – the power to shoot down a hijacked passenger aircraft to prevent an attack. The reasoning is central to the triage debate: the state may not make innocent people the mere object of a rescue calculation. Weighing life against life by the numbers is denied to the state, because human dignity is not amenable to quantitative offsetting.

Here lies the genuinely German peculiarity. In a legal order that expressly forbids such offsetting, a triage statute that makes "as many lives as possible" its standard is constitutionally delicate. Prevailing doctrine resolves this by a distinction: in the aircraft case the state acts by killing uninvolved persons; in triage a physician distributes a chance of rescue that he cannot give to everyone, and no one holds a claim that he in particular be the one saved. In legal terms this is the figure of a justifying conflict of duties. I am of the opinion that this distinction holds, but that it is thinner than it is usually presented – it holds exactly as long as the decision turns on the prospect of rescue and not on an evaluation of the life. Which brings us to the question of how well the prospect of rescue can be predicted at all.

Part 4: German Legal History in Three Acts

Act I: The decision of 16 December 2021

What the court held in 2021 was not that triage is unconstitutional. It held that the state bears a duty of protection under Article 3(3) sentence 2 of the Basic Law which it may not delegate to professional societies. The court identified the risk precisely: if comorbidities and frailty enter the decision as negative prognostic factors, there is a danger that a disability will be stereotypically linked to poor prospects of success, or conflated with a comorbidity. That medical guidelines expressly prohibit prioritization by disability did not satisfy the court – because guidelines are not legal norms, and because unconscious stereotypes cannot be switched off by prohibition.

The sharpest observation in that decision is the most uncomfortable: a risk of discrimination can exist even where nobody intends to discriminate.

Act II: § 5c of the Infection Protection Act

The legislature answered with a norm an engineer can read, because it is built like a specification:

The first eight rows of this table are, taken alone, a considerable achievement: they translate a constitutional mandate into auditable procedural requirements. The last row blew up the statute.

Act III: The decision of 23 September 2025

The court declared § 5c(1)–(3) void for lack of legislative competence and extended the nullity to subsections (4)–(7) because they are inseparably bound to the allocation criteria. Two lines of reasoning:

Competence. Article 74(1) no. 19 of the Basic Law covers only measures with "a certain orientation towards containment or prevention." The triage rules, "as pure pandemic aftermath law, attach to a scarcity resulting from a pandemic but do not serve to combat the pandemic." They regulate who is treated, not how the disease is fought. Article 74(1) no. 7 (public welfare) also fails to support them.

Occupational freedom. Article 12(1) protects medical practice "free from professional directives" and thus therapeutic freedom. Binding allocation criteria infringe that freedom – and cannot be justified absent a basis of competence.

One can read this decision as a formality. I am of the opinion that this would be a mistake. At bottom the Senate says: the question of who lives under absolute scarcity is not a question of infection control but a general question of health and professional law – and there the federation has less power than it assumed. That is a statement about the architecture of federalism, not about ethics.

What governs now

From the federation: nothing. Competence lies with the states, with three conceivable routes – sixteen state statutes, coordinated legislation via the conference of health ministers, or an interstate treaty. None has been taken. The 2021 duty of protection remains unfulfilled; a renewed constitutional complaint over inaction is foreseeable the moment the situation becomes acute again. What actually governs is the professional societies' S1 guideline.

Part 5: What Physicians Actually Have in Hand

The clinical-ethical recommendations "Decisions on the allocation of intensive care resources in the context of the COVID-19 pandemic" (AWMF register number 040-013, version 3 of 14 December 2021) are issued by eight professional societies, among them DIVI, DGAI, DGIIN, and the Academy for Ethics in Medicine. Their core:

The criterion. "Prioritization of patients should therefore be guided by the criterion of clinical prospect of success" – meaning the probability of surviving the current acute illness with intensive care. Not life expectancy afterwards, not quality of life.

The prohibitions. Prioritization is "not permissible on the basis of chronological age, on the basis of social characteristics, or on the basis of particular underlying conditions or disabilities, nor on the basis of SARS-CoV-2 vaccination status."

The instruments. The Clinical Frailty Scale to capture pre-morbid frailty – relevant only insofar as it demonstrably lowers the chance of survival, and validated essentially for patients aged 65 and over. The SOFA score to quantify accompanying acute organ failures.

The procedure. A multiple-eyes principle, "with the involvement of at least two physicians experienced in intensive care," where possible with nursing representation and, where needed, clinical ethics. Consensus sought; dissent procedures agreed in advance.

Re-evaluation. The guideline expressly records that ongoing intensive therapy may be reassessed and discontinued when the therapeutic goal is unattainable, when the patient's documented wishes oppose it, or when agreed time windows elapse without improvement: "Re-evaluation should take place and be documented upon changes in the patient's state of health and upon changes in the care situation."

That last clause is the point of conflict. "Changes in the care situation" is ex post triage – precisely what § 5c(2) sentence 4 forbade.

Part 6: Where Ethics Fails on Measurement

Nearly everyone involved – legislature, professional societies, constitutional court, disability advocacy groups – settled on the same criterion: probability of survival with respect to the acute illness. That is a remarkable convergence, because it is the smallest common denominator of all positions: it values no life, counts no life-years, asks nothing about desert.

It has only one drawback. It cannot be measured well.

The SOFA score

The Sequential Organ Failure Assessment score is the world's most widely used instrument for quantifying organ failure in intensive care, and it was the central allocation parameter in numerous crisis standards of care. In 2021, Robert Raschke and colleagues examined in JAMA how well it does the thing it was used for in the pandemic: 675 mechanically ventilated COVID-19 patients from 18 intensive care units in the southwestern United States, March to August 2020, with nearly 60 percent mortality or discharge to hospice.

For orientation: an AUC of 0.5 means a coin flip. 0.59 means that for randomly drawn pairs consisting of one survivor and one decedent, the score identifies the decedent as the worse-off in 59 percent of cases. The authors conclude that the SOFA score has "inadequate discriminant accuracy" to be used for ventilator triage of COVID-19 patients. More uncomfortable still: age – a criterion expressly forbidden by the German guideline and by § 5c – discriminated significantly better (p = 0.02).

The miscalibration

A second finding from the same year strikes the fairness question directly. Deepshikha Ashana and colleagues showed in the American Journal of Respiratory and Critical Care Medicine that the SOFA score predicts not merely weakly but in a systematically biased way: it overestimates mortality among Black patients and underestimates it among white patients. In simulation this resulted in 81 percent of Black patients from the lower priority categories – and 9 percent of all Black patients – being improperly excluded from the highest priority category.

This is the decisive methodological point of this article. An allocation procedure can be perfectly non-discriminatory on paper – no age, no disability, no origin, no social status – and still discriminate systematically, because the predictor through which it reaches its decision is calibrated differently for different groups. The discrimination migrates into the statistics, and there it becomes invisible.

And the frailty scale

The Clinical Frailty Scale, named as an instrument by the German guideline, is validated essentially for patients aged 65 and over. Applied to younger people with disabilities, it captures not frailty but a functional status that, in a forty-year-old with a congenital disability, is permanent and has nothing to do with the acute illness. That was precisely the constitutional court's concern in 2021. It was not abstract; it was a validation problem.

The general lesson

From this follows a rule that reaches far beyond medicine and applies to anyone who builds prioritization logic:

The justice of a rule is bounded above by the quality of its predictor. A rule that refers to a quantity estimated at AUC 0.59 is, in its individual decisions, barely distinguishable from a lottery – yet it feels like expertise, generates documentation, confers authority, and thereby escapes the criticism that an openly declared lottery would instantly attract. A poorly calibrated score is morally worse than a coin flip, because it delivers the same outcome with false legitimacy.

Part 7: The Hardest Case – Ex Post Triage

This is where every thread converges.

The positions

The legislature in 2022 prohibited ex post triage in § 5c(2) sentence 4. The reasoning was not consequentialist but protective: people with disabilities and serious pre-existing conditions would stand under permanent pressure of justification if reassessment were continuous; whoever has a bed should not have to compete for it afresh every day. Disability organizations strongly supported the prohibition.

The medical profession sued against it. The criticism, put pointedly in the constitutional-law literature: "The prohibition of ex post triage leads in effect to 'first come, first serve.'" The statute prescribes probability of survival as the sole criterion and in the next sentence forbids applying it once treatment has begun. Fourteen physicians, backed by the German Medical Association and the Marburger Bund, saw themselves compelled to act against their own prospect-of-success logic. In 2025 they prevailed – though on competence grounds, not because the court endorsed their ethical position.

Why the intuition is so strong

The difference almost everyone feels on reading the two cases is the same one that structures the whole debate about acting and omitting. To take a running machine away from someone is to kill him in a sense in which one does not kill the person one never gave it to. In criminal law this is the distinction between commission and omission; in ethics, between intended harm and harm merely accepted as a side effect. The consequentialist rejoinder is seductively simple: both patients need the same machine, both die without it, order of arrival is morally irrelevant. Yet an ongoing treatment relationship has a normative quality a queue does not – and a patient already intubated has invested a trust that is not nothing.

I am of the opinion that this conflict cannot be dissolved but only constructively defused – in exactly the way the professional societies propose: through time-limited therapeutic trials agreed in advance with defined success criteria. If it is established and documented at the start of treatment that reassessment will occur after, say, 72 hours against specified parameters, then discontinuation is not the retroactive repurposing of a promised resource but the honouring of a promise that was limited from the outset. The difference lies not in the outcome but in the expectation created in the patient – and creating expectations one cannot keep is the real moral error.

Part 8: Five Questions Every Allocation Rule Must Answer

This grid is abstracted from the medical debate and applies to any scarcity rule – including ones with nothing to do with medicine.

Question 2 is the one most often skipped, and question 5 the one that costs the most trust when it is not answered up front.

The Central Takeaway

The triage debate is not a special case of medical ethics but the clearest available case of a general structure: a rule that decides about people under scarcity is only as good as the predictor it decides by – and only as legitimate as the procedure that carries it. Four concrete prompts:

1. An advance directive and a healthcare power of attorney. This is the one part of this debate that lies entirely in your own hands. The German guideline names documented patient wishes as the first reason not to begin or to discontinue intensive therapy – ahead of any question of resources. Whoever specifies what they want makes their own decision instead of delegating it into an allocation situation. What matters is not the form but the specificity: "no life-prolonging measures" is too vague to have effect.
2. For any score that prioritizes: ask for the AUC and the calibration per group. This applies to credit decisions, fraud detection, incident prioritization, candidate ranking, and every model that sorts people. A score without a group-wise calibration check is a discrimination risk with documentation attached.
3. In engineering: write capacity rules before the crisis, not inside it. When a system overflows, something decides – a timeout, a queue order, a random load balancer, an exhausted on-call engineer. Every unwritten rule is an implicit rule, and implicit rules are almost always first come, first served. The same holds for support queues, emergency rollouts, and budget cuts.
4. Time-limit commitments before you make them. The most ethically robust finding of the ex post debate is a general one: a commitment declared from the outset as limited and reviewable can be withdrawn without a breach of trust; an implicitly open-ended one cannot. That holds for ventilators, for project resources, and for promises made to people.

A Question to Think About

In 2021 the Federal Constitutional Court demanded that the state make a rule, and in 2025 it held that the federation was not allowed to make it. Between those two decisions lies an insight more uncomfortable than either: the rule everyone could agree on fails not on ethics but on the measurability of its own criterion.

Which decision in your own area of responsibility is made by a number whose discriminative power you have never checked – and would you still make the decision that way if you knew that number gets the ordering right in only 59 out of 100 cases?

Cross-References in the Vault

• Behind the Veil of Ignorance: How Rawls Reinvented Justice – and Why His Thought Experiment Still Echoes in the Ethics of Artificial Intelligence – the obvious thought experiment for allocation rules: which triage rule would you choose if you did not know whether you were the first to arrive or the second?
• Meant or Merely Foreseen: The Doctrine of Double Effect and the Power of Intention – the conceptual basis of the ex post problem: the difference between intended and merely foreseen death.
• The Switch in the Machine: The Trolley Problem and the Ethics of Autonomous Vehicles – the same counting question inside a machine, except that there it is programmed in advance and here it must be decided at the bedside.
• The Billions Barely Alive: Parfit's Repugnant Conclusion and the Ethics of Future Generations – aggregation as the underlying problem: why adding well-being across persons leads into abysses.
• The Child in the Pond: How a Three-Page Essay by Peter Singer Launched a Global Movement — and Why the Question of How Far Our Duties Reach Echoes All the Way into the Boardrooms of Tech Billionaires – "save as many lives as possible" as a principle for living; triage is the case where that principle meets its hardest limit.
• The Promise on the Form: How the Crimes of Nuremberg Produced Informed Consent – and Why the Signature Still Means Less Than We Think – the prehistory of procedural ethics in medicine, and why documentation duties are not bureaucracy.
• The Child in the Road: Moral Luck and the Limits of Responsibility – what it does to a physician whose moral standing depends on the accident of bed occupancy.
• The Algorithm as Judge: Article 22 GDPR, the SCHUFA Ruling, and the Right to a Human Decision – the same pattern in law: a score decides, and the question is whether a human can object.
• The Pyramid of Risk: How the EU AI Act Tames Artificial Intelligence – and Why It Concerns the Whole World – risk-based regulation of decision systems; triage scores fall squarely into the category treated most strictly there.
• The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage – a finding that breaks on methodological scrutiny; here an allocation criterion breaks on the same scrutiny.
• The Gentle Push: Why a Pre-Ticked Box Achieves More Than Any Argument — and Whether It Is Permissible to Nudge People Toward Their Own Good – choice architecture as a question of power: whoever designs the procedure designs the outcome.
• The 38 Witnesses Who Never Were: The Bystander Effect Between Laboratory Finding and Video Evidence – helping behaviour under diffusion of responsibility; triage is the inverse case, where responsibility rests indivisibly on two named people.

Sources

• Persad, G., Wertheimer, A., & Emanuel, E. J. (2009): Principles for allocation of scarce medical interventions. The Lancet, 373(9661), 423–431. DOI: 10.1016/S0140-6736(09)60137-9
• Emanuel, E. J., Persad, G., Upshur, R., Thome, B., Parker, M., Glickman, A., Zhang, C., Boyle, C., Smith, M., & Phillips, J. P. (2020): Fair Allocation of Scarce Medical Resources in the Time of Covid-19. New England Journal of Medicine, 382(21), 2049–2055. DOI: 10.1056/NEJMsb2005114
• Taurek, J. M. (1977): Should the Numbers Count? Philosophy & Public Affairs, 6(4), 293–316.
• Parfit, D. (1978): Innumerate Ethics. Philosophy & Public Affairs, 7(4), 285–301.
• Raschke, R. A., Agarwal, S., Rangan, P., Heise, C. W., & Curry, S. C. (2021): Discriminant Accuracy of the SOFA Score for Determining the Probable Mortality of Patients With COVID-19 Pneumonia Requiring Mechanical Ventilation. JAMA, 325(14), 1469–1470. DOI: 10.1001/jama.2021.1545
• Ashana, D. C., Anesi, G. L., Liu, V. X., Escobar, G. J., Chesley, C., Eneanya, N. D., Weissman, G. E., Miller, W. D., Harhay, M. O., & Halpern, S. D. (2021): Equitably Allocating Resources during Crises: Racial Differences in Mortality Prediction Models. American Journal of Respiratory and Critical Care Medicine, 204(2), 178–186. DOI: 10.1164/rccm.202012-4383OC
• DIVI, DGINA, DGAI, DGIIN, DGNI, DGP, DGP (Palliative Medicine) & AEM (2021): Entscheidungen über die Zuteilung intensivmedizinischer Ressourcen im Kontext der COVID-19-Pandemie. Klinisch-ethische Empfehlungen. S1 guideline, AWMF register number 040-013, version 3 of 14 Dec 2021. https://register.awmf.org/de/leitlinien/detail/040-013
• Federal Constitutional Court, Order of the First Senate of 16 December 2021 – 1 BvR 1541/20 – (duty of protection in pandemic-related triage, Art. 3(3) sentence 2 Basic Law). https://www.bundesverfassungsgericht.de/
• Federal Constitutional Court, Order of the First Senate of 23 September 2025 – 1 BvR 2284/23, 1 BvR 2285/23 – (§ 5c Infection Protection Act void; press release no. 99/2025 of 4 November 2025). https://www.bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/DE/2025/bvg25-099.html
• Federal Constitutional Court, Judgment of the First Senate of 15 February 2006 – 1 BvR 357/05 – (§ 14(3) Aviation Security Act void; no quantitative offsetting of lives). https://www.bundesverfassungsgericht.de/SharedDocs/Entscheidungen/EN/2006/02/rs20060215_1bvr035705en.html
• § 5c Infektionsschutzgesetz (until declared void in 2025): Verfahren bei aufgrund einer übertragbaren Krankheit nicht ausreichend vorhandenen überlebenswichtigen intensivmedizinischen Behandlungskapazitäten. https://www.gesetze-im-internet.de/ifsg/__5c.html

← All articles