Optimizing scarce blood, ventilators, surgical teams and antibiotics across competing patient needs in a mass-casualty zone
Every mass-casualty response begins the same way: a finite, countable cache of life-sustaining resources — blood units, mechanical ventilators, surgical team-hours, antibiotic doses — must be measured against a patient population whose needs almost always exceed it. Disaster medicine differs from routine hospital care in exactly this respect: the question is no longer "what does this patient need?" but "how do we do the most good with what we have?"
Four properties define a scarce medical resource in a mass-casualty event:
• Discreteness: you cannot allocate half a ventilator or 60% of a surgical team — units are indivisible, which is exactly the structure of a combinatorial optimization problem rather than a continuous one. • Non-substitutability: a ventilator cannot be replaced by more blood; each resource addresses a distinct physiological failure mode (airway/breathing, circulation, hemostasis, infection control). • Replenishment lag: blood can be resupplied in hours by air-bridge, but a trained surgical team cannot be manufactured — the resupply curve differs by resource type and drives different allocation urgency. • Perishability: platelets last 5–7 days, packed red cells ~42 days, antibiotic potency degrades with heat exposure common in field conditions — inventory is a moving target, not a fixed number.
Disaster logisticians typically track resources in a live tally board updated every shift: units on hand, units committed, units in the resupply pipeline, and units expired/wasted.
Mass-casualty incidents rarely produce a single wave of patients. Empirically, casualty presentation follows a right-skewed surge curve: a first wave of self-transporting "worried well" and walking wounded arrives within minutes (often before EMS), a second wave of moderate-to-severe injuries arrives over 1–3 hours via organized transport, and a smaller tail of critically injured survivors trickles in over the following 24–72 hours as search-and-rescue continues.
This matters for allocation because the resource pool a patient is measured against is not static — it is whatever remains after everyone who arrived earlier has already been triaged. A patient with an identical injury profile may receive a ventilator at hour 1 and be denied one at hour 6, purely as a function of arrival order interacting with supply depletion. This is precisely why disaster medicine ethics committees insist that allocation be governed by explicit, pre-agreed rules rather than ad hoc bedside judgment.
The 2010 Haiti earthquake response treated over 300,000 injuries with a national blood supply that, before the quake, held enough stock for roughly three days of ordinary demand — illustrating how quickly baseline scarcity becomes acute scarcity.
As casualties accumulate, the gap between resources requested and resources available is the single number disaster medical coordinators watch most closely. When that gap crosses zero, care shifts from "conventional" to "crisis standards" — a formally recognized transition with its own legal, ethical and operational rules.
The US National Academy of Medicine (formerly Institute of Medicine) defines a three-tier continuum for disaster resource stress:
• Conventional care: usual resources are sufficient; care is provided under normal standards. • Contingency care: resources are stretched (e.g., one nurse covering more beds, using recovery-room space as overflow ICU) but functionally equivalent care is still delivered to every patient. • Crisis Standards of Care (CSC): resources are insufficient even after contingency measures — a fundamental shift occurs from patient-centered goals ("do everything possible for this patient") to population-centered goals ("do the most good for the most patients with what remains").
The transition into CSC is not something an individual clinician declares at the bedside. It requires formal activation — usually by a state or regional health authority — because it changes the legal standard against which care is judged and, in most US states, triggers liability protections for clinicians operating under an approved CSC plan.
A useful operational metric is the resource gap ratio for each resource type r:
gap(r) = requested(r) − available(r)
When gap(r) ≤ 0 for all r, the site remains in conventional or contingency mode. When gap(r) > 0 for a critical resource — most commonly ventilators or O-negative blood — triage officers must ration that specific resource even while other resource types remain adequate. This is why real disaster plans allocate resource-by-resource rather than treating "medical care" as one undifferentiated pool: a patient may receive antibiotics and IV fluids freely while being denied a ventilator slot, because those two resource gaps opened at different rates.
Hospitals track this in near-real time via incident command "resource status boards," refreshed every 15–60 minutes during an active surge, feeding directly into the triage committee's next allocation round.
Once a facility formally enters crisis standards, ad hoc clinical judgment is replaced by a structured scoring protocol, so that two patients with the same clinical picture receive the same priority regardless of which clinician happens to examine them. The most widely adopted instrument for this purpose is the Sequential Organ Failure Assessment (SOFA) score, repurposed from routine ICU prognostics into a rationing tool.
SOFA assigns 0–4 points to each of six organ systems based on objective, chart-derivable measurements:
• Respiratory: PaO₂/FiO₂ ratio • Coagulation: platelet count • Liver: bilirubin level • Cardiovascular: mean arterial pressure / vasopressor requirement • CNS: Glasgow Coma Scale • Renal: creatinine or urine output
Summing across systems gives a 0–24 total. Under conventional care, SOFA is used purely for prognosis — tracking whether a patient is improving or worsening. Under crisis standards, many US state ventilator-allocation guidelines (New York 2015, updated 2023; several other states' pandemic plans) repurpose SOFA into an exclusion and priority filter: patients above a defined threshold (or with a second organ failure emerging on reassessment) are deprioritized for a scarce ventilator in favor of patients more likely to survive the acute episode, then re-evaluated on a fixed interval (commonly 48–120 hours) so that no allocation decision is treated as permanent without review.
Critically, most guidelines explicitly exclude discriminating factors such as disability status, age alone, or "quality of life" judgments from the score — restricting prioritization to short-term survivability of the acute illness, precisely to guard against embedding bias into an ostensibly objective number.
During the 2020 COVID-19 surge, the Italian Society of Anesthesia, Analgesia, Resuscitation and Intensive Care (SIAARTI) issued guidance on 16 March 2020 explicitly recommending a shift from strict first-come-first-served ICU admission toward prioritizing "the greatest chance of therapeutic success" — a move that ignited international debate over whether age and comorbidity criteria amounted to discrimination against the elderly and disabled.
For a disaster zone allocating several distinct resource types simultaneously, a composite priority score P for patient i is typically built from three components:
P(i) = w₁·(1 − SOFA_norm(i)) + w₂·survivalProb(i) + w₃·resourceUrgency(i)
where SOFA_norm is the SOFA score scaled to 0–1 (lower organ failure → higher priority), survivalProb is a short-term prognosis estimate (from SOFA-based mortality tables or clinical judgment), and resourceUrgency captures how quickly the patient will deteriorate without the specific resource in question. Weights w₁, w₂, w₃ are fixed in advance by the crisis standards committee — not chosen at the bedside — and published so the process is auditable.
Patients are then rank-ordered by P(i) within each resource queue. This produces a strict priority list, and allocation proceeds by walking down that list until the resource pool is exhausted — the ranking step is what makes the subsequent optimization step (Stage 4) tractable and defensible under after-the-fact review.
With patients scored and ranked, the allocation problem becomes formally equivalent to the classic 0/1 knapsack problem from combinatorial optimization: choose a subset of patients to receive resources, subject to a hard capacity constraint, so as to maximize total expected benefit. In practice, disaster triage almost never solves this exactly — it uses a fast greedy approximation instead, and understanding why reveals a core tension in crisis ethics.
Formally: given n patients, each requiring resource weight wᵢ (e.g., one ventilator) and offering expected value vᵢ (e.g., expected life-years or survival probability gained), and a resource capacity C (total units available), choose a selection vector x ∈ {0,1}ⁿ to:
maximize Σ vᵢ·xᵢ subject to Σ wᵢ·xᵢ ≤ C
This is the 0/1 knapsack problem — provably NP-hard, meaning an exact optimal solution over hundreds of patients and multiple resource types is computationally impractical to find within the minutes a real triage decision demands. Disaster protocols therefore substitute a greedy heuristic: rank patients by value-to-weight ratio vᵢ/wᵢ (in practice, this is exactly the composite priority score from Stage 3, since most disaster resources have near-uniform weight — one ventilator, one unit of blood) and allocate down the ranked list until capacity is exhausted.
The greedy approach is provably suboptimal in the worst case, but for near-uniform resource weights (which is the common disaster case — one patient needs one ventilator, not a fractional share) it is very close to optimal and, crucially, it is transparent and auditable: any observer can recompute the same ranking from the same inputs, which matters enormously for legal and ethical defensibility after the fact.
Real allocation runs one knapsack per resource type in parallel, since a patient needing a ventilator and a patient needing a unit of blood are not competing for the same slot. A single patient may appear on multiple resource queues simultaneously (e.g., a hemorrhaging, respiratory-failure patient needs both blood and a ventilator) — in which case allocation software or the triage officer must check that all required resources are jointly available, otherwise a partial allocation (blood but no ventilator) may still leave the patient unable to survive, wasting the blood that was allocated.
Patients who rank below the cutoff on a resource queue are recorded as denied for that resource and that round — not denied care altogether. They continue to receive whatever contingency-level care remains available (oxygen by mask instead of a ventilator, pressure dressings instead of surgery) and are automatically re-queued for the next allocation round as capacity frees up (a treated patient dies or recovers, resupply arrives). No CSC framework treats a denial as final without scheduled reassessment — this is one of the most important, and most frequently misunderstood, safeguards in the model.
The same patient population, run through different allocation rules, produces measurably different outcome sets — not just in who survives, but in how the process itself is perceived as fair. No allocation rule is ethically "free": each one optimizes for a different value at the deliberate expense of another, which is why every credible disaster-medicine framework insists the rule be chosen and published before a crisis, not improvised during one.
When Hurricane Katrina flooded New Orleans in August 2005, Memorial Medical Center lost power, air conditioning and running water, and evacuation became a slow, resource-constrained triage problem in itself — who gets carried down dark stairwells first, who waits, who is deprioritized for evacuation. Over the five-day ordeal (chronicled in Sheri Fink's Pulitzer Prize-winning investigation, later the book "Five Days at Memorial"), some of the sickest patients — including several tagged as unlikely to survive evacuation — were given lethal doses of morphine and midazolam by physician Dr. Anna Pou, who was later indicted on charges of second-degree murder (a grand jury ultimately declined to indict).
The case became the defining cautionary tale in American disaster-medicine ethics precisely because no formal, pre-agreed crisis-standards protocol existed at the facility — decisions were made bedside, under extreme physical and psychological duress, by individual clinicians who had received no advance guidance and no legal protection. It directly motivated the IOM's 2009 push to formalize Crisis Standards of Care frameworks nationally, with explicit emphasis on removing individual bedside clinicians from the burden of ad hoc life-and-death rationing by assigning that responsibility instead to a separate triage officer or committee operating under a published protocol.
"The whole idea of triage inherently, fundamentally is discriminatory... the question is whether we do it consciously, transparently, and by rule — or unconsciously, in the dark, by whoever happens to be standing there." — a framing repeated across post-Katrina disaster ethics literature to argue for pre-committed protocols over bedside improvisation.
The Sphere Handbook — the most widely used humanitarian minimum-standards reference, now in its fourth edition (2018), produced by a coalition including WHO-aligned bodies — frames resource allocation around a "protection principle" that sits in tension with pure utilitarian optimization: aid must be provided impartially, based on need alone, without discrimination on the basis of age, gender, disability, ethnicity, religion or political affiliation, even when a purely survival-maximizing calculation might favor one demographic group.
In practice this means real humanitarian and disaster protocols rarely implement a "pure" utilitarian knapsack solver. Instead they apply utilitarian ranking within demographic-blind clinical criteria (SOFA-style scores using only physiological measurements) precisely so the optimization cannot be gamed or perceived as targeting a protected characteristic — the WHO/Sphere equity constraint acts as a side-constraint on the knapsack problem, not a value term inside it.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Utilitarian (maximize survivors) | Rank by SOFA / survival probability; treat highest-benefit patients first | Greedy knapsack by value/weight ratio | Maximizes total lives or life-years saved; used in NY/Italy CSC ventilator guidelines |
| Egalitarian (first-come-first-served / lottery) | Treat in arrival order, or randomize among eligible patients | Queue order or random draw, ignores prognosis | Perceived as maximally fair and non-discriminatory; used in routine (non-crisis) triage and organ-allocation lotteries |
| Prioritarian (sickest-first) | Treat worst-off patients regardless of survival odds | Rank by severity of need, not benefit | Honors traditional medical ethic of treating the neediest; can waste scarce resources on low-probability saves |
| Life-cycle / youngest-first ("fair innings") | Favor patients who have had less life-years to date | Rank inversely by age or life-years already lived | Reflects intergenerational equity intuitions; explicitly excluded from most US CSC protocols as age discrimination |