Hospitals across the New York region spent the morning of September 11, 2001 preparing for a casualty load that mostly did not arrive. Emergency departments near the World Trade Center took a rush of ambulatory patients in the first two or three hours and then saw the flow stop, because the collapses produced deaths rather than survivors needing surgical care. The planning material in that outcome is uncomfortable and it has held up better than almost anything else written about medical surge, since the supply of beds held up that day while the information about where patients were going did not.

What the hospitals did in the first two hours

Part one of this series dealt with the movement of the daytime population out of Lower Manhattan, and this part picks up at the receiving end, in the emergency departments, operating rooms and blood centers that spent the day getting ready. The documented record of the hospital mobilization is thinner than most people assume, and it comes from three kinds of source that should not be mixed without saying which is which: peer reviewed accounts written by clinicians at individual institutions, hospital communications about their own performance, and federal summaries assembled later from state administrative data. Where those sources describe activity in the first hours, they describe existing plans being activated rather than capability being invented, which is worth stating because the popular account of that morning tends toward improvisation.

Cushman, Pachter and Beaton, writing in the Journal of Trauma in January 2003 (volume 54, pages 147 to 155) about two affiliated hospitals in Lower Manhattan, record that both institutions had established disaster plans and that both implemented them. One of the two was a state designated regional trauma center about 2.5 miles from the World Trade Center, and the other was an affiliate community hospital five city blocks away. The same paper notes that New York City had 18 state designated regional trauma centers receiving major trauma at the time, which is the number that makes the distribution question in this article a real question rather than a theoretical one, because a city with that many receiving facilities has somewhere to send patients if anybody is directing them.

NewYork-Presbyterian, in its own published account of a Weill Cornell review of the hospital’s disaster preparedness, states that its emergency departments treated 175 patients, including 22 burn patients at the Hearst Burn Center at Weill Cornell, and that the NewYork-Presbyterian Healthcare System treated 600 victims in the first 48 hours. That is an institutional statement about the institution’s own performance and it should be cited as such, though the burn center figure is useful precisely because burn beds are by far the scarcest category of surge capacity in any region and 22 patients is a long way below what a plan for a high rise fire in a complex of that size would assume.

On the blood side, the queues are documented rather than anecdotal. A 2002 Government Accountability Office testimony on the blood supply and emergency preparedness, issued as GAO-02-1095T, records that the surge of donors stressed the collection system, with shortages of collection supplies, of phlebotomists and of storage capacity, and with long waiting lines developing because there were not enough staff to draw blood from the people who had turned up. What I could not verify is any consolidated regional count of beds cleared, elective procedures cancelled or staff recalled across the New York metropolitan area, and I am not going to estimate one. The closest thing to a measure of the geographic footprint is in a 2004 GAO testimony on health effects in the aftermath of the World Trade Center disaster, GAO-04-1068T, which reports attack related emergency room visits recorded at 103 hospitals in New York State, and a picture of how many institutions were involved is not the same thing as a picture of how much capacity they held open.

The patient numbers, and which study counted what

Four published figures circulate for the number of people treated after the attack in New York, and they differ by definition rather than by error. The 2004 GAO testimony cited above gives 6,232 emergency room visits and 477 hospitalizations related to the attack at 103 hospitals in New York State for the period September 11 through September 28, 2001, along with 1,018 emergency room visits and 84 hospitalizations at nearby New Jersey hospitals for the same period. GAO adds a qualification that anyone quoting the number needs to carry with it, which is that those counts exclude people treated at more distant hospitals in New York State, New Jersey and Connecticut, and also exclude people treated at triage stations.

The Centers for Disease Control and Prevention published a rapid assessment of injuries among survivors in MMWR on January 11, 2002 (volume 51, number 1, pages 1 to 5), based on a New York City Department of Health field investigation that reviewed emergency department and inpatient records at a sample of hospitals nearest the site. That assessment counted 790 injured survivors treated within 48 hours, reported that about half of them received care within seven hours of the attack, and found that 18 percent were hospitalized. It also establishes the timing that matters most for planning, which is that injured people began arriving within minutes of the attack and that arrivals peaked two to three hours later.

Cushman and colleagues counted 911 patients received by their two affiliated hospitals. That figure is larger than the CDC total for a sample of nearby hospitals over 48 hours, which sounds contradictory until you look at what each team was counting and how. The trauma center in the Journal of Trauma paper maintained a concurrent log kept by the surgical triage officer covering the first five hours, the first day and the first week, verified against the trauma registry by September 18, while its affiliate hand counted its patients and verified the count by chart review. The CDC investigation counted injured survivors within a 48 hour window from a records review conducted afterward, and the GAO figures come from state administrative data over an 18 day window and include people who were not survivors from the towers at all, such as responders and residents presenting with respiratory complaints. None of those totals can be subtracted from another, and a briefing slide that sets them side by side without stating the counting rule for each one is misleading its audience.

A fourth number belongs to a different population entirely. Prehospital and Disaster Medicine published a study of injuries and illnesses treated at the World Trade Center site itself between September 14 and November 20, 2001 (volume 20, issue 3, 2005), which recorded 9,349 patient visits, peaking in the second week, with 2,716 visits (29 percent) for traumatic injuries excluding eye injuries and more than a fifth for respiratory problems. Those are site treatment visits by people working on the debris field over ten weeks, and part three of this series deals with that population. Put the site figure next to a hospital figure in the same paragraph of a briefing and you will produce a false impression of the acute casualty load.

The injury mix, which was escape and dust rather than collapse

The clinical picture in the published accounts is consistent across independent sources, and it is not the picture that a mass casualty plan for a building collapse anticipates. Cushman and colleagues report that 776 of their 911 patients, or 85 percent, were walking wounded with mild inhalation and eye irritant injuries, that 135 patients (15 percent) were admitted, and that 18 of those admitted patients, 13 percent of admissions, went to surgery. The CDC rapid assessment describes the same weighting from a different sample, with most of the injured treated in the first hours presenting with inhalation or ocular injuries and 18 percent of the 790 hospitalized.

Those injuries were sustained getting out rather than in the collapses. People descending stairwells and crossing the plaza and the surrounding streets sustained the kinds of injury that crowds in a hurry sustain, and then the dust cloud produced a very large number of respiratory and ocular presentations that required irrigation, observation and reassurance far more often than they required an operating room. The surgical demand that plans are built around, meaning crush injury, penetrating trauma from structural failure and major burns, arrived in small numbers, and the 22 burn patients described in the NewYork-Presbyterian account is the clearest single illustration of that.

The planning consequence is specific enough to act on. If 13 percent of admissions in the best documented surgical series from that day went to the operating room, then the operating room is not the binding constraint in an event of this shape, and the constraints are irrigation supplies, nebulizers, oxygen, eye kits, staff to register and track large numbers of ambulatory patients, and physical space to hold people who are frightened and dusty but not injured in any way that will show on imaging. Hospitals that prepared for the surgical scenario and received the ambulatory one found their scarce resources idle while their registration and triage areas were the pressure point, which is a mismatch worth exercising deliberately rather than discovering on the day.

One caution on the injury data. The studies above cover the acute period, and the respiratory illness that followed among responders, cleanup workers and residents is a separate body of literature with its own methods and its own federal program documentation. Anyone reasoning from the acute injury mix to long term health effects is crossing between two evidence bases, and the place to look for the second one is the World Trade Center Health Program administered by the National Institute for Occupational Safety and Health within the Centers for Disease Control and Prevention, which publishes its own eligibility criteria and list of covered conditions.

The reason so few patients came

The hospitals were not spared a patient load by good fortune or by efficient field triage. The people who would have filled those beds were dead, and the event killed at a rate that left very few survivors requiring definitive care. Occupants below the impact floors largely escaped, and the great majority of those above them did not, which is why the collapses produced so few rescued casualties. That is the whole explanation for the shape of the hospital data, and it should be stated once, in plain words, whenever this case is taught, because every planning inference in the rest of this article depends on it and because presenting an empty emergency department as a success would be a serious misreading.

The count of people killed at the World Trade Center site is maintained by the New York City Office of Chief Medical Examiner and has been revised over the years as remains were identified and as deaths attributed to exposure were added, so anyone publishing a figure should take the current one from that office and date it rather than repeating a number from an older article. I am deliberately not putting a death toll in this piece from memory, and neither should a briefing that has not checked the current figure.

The live rescues from the debris are heavily mythologized and the numbers quoted for them vary, so I checked what I could and I am leaving the rest out. What is well established across multiple accounts, including the Washington Post’s 2021 retrospective and her own memoir, is that Genelle Guzman-McMillan, a Port Authority employee who had been on the 64th floor of the North Tower, was trapped for about 27 hours and was pulled out on September 12, and that she was the last person recovered alive from the rubble. The total number of people rescued alive from the debris field is given differently in different places, I could not verify it against a source I would defend, and so it does not appear here.

Volume is the easy half of surge

Medical surge is normally planned as an inventory problem, expressed as beds, blood, ventilators, operating rooms and the staff to run them. The doctrine side of that is covered separately on this site in the piece on hospital surge capacity, which sets out the staff, stuff and space framing alongside the structure of a healthcare coalition. That same piece is where the conventional, contingency and crisis standards of care continuum is explained, so I am not going to restate any of it here. What the 2001 event in New York tested was the other half of the problem, which is whether a region can route patients deliberately, whether hospitals can learn what is happening to them in the first hour, and whether anyone can authoritatively say that the wave is over.

The routing question failed first, and it failed in the way it usually fails, because patients who can walk do not wait for an ambulance and nobody assigns them a destination. The CDC data showing arrivals beginning within minutes and peaking two to three hours later describes a self distributed flow, not a managed one, and the two hospitals in the Journal of Trauma series received 911 people between them in a city with 18 designated trauma centers. The mechanism here is worth being precise about, because a plan that assumes an incident commander will distribute casualties across a region is assuming a control point that did not exist in the path that most of these patients took.

The information question failed next. A paper in Prehospital and Disaster Medicine on the Centers for Disease Control and Prevention’s emergency phase response in the first 24 hours records that CDC staff assisted with assessing hospital capacity among other tasks, which tells you that capacity information was being assembled during the event by people who had flown in for the purpose. The inference I draw from that is mine rather than the paper’s, and it is that there was no standing system from which the regional capacity picture could simply be read, which is the same gap that hospital bed reporting systems were built to close in the years afterward.

The third failure is the one nobody writes an after action item about. Resources were committed and held for a demand that did not come, and the Journal of Trauma account covers the day of the attack and the week that followed, which is the window over which those two hospitals were still tracking the response. Holding surge posture is not free, since cancelled elective procedures are cancelled care for patients who needed it, recalled staff who stand in an empty trauma bay for twelve hours are fatigued staff on day two and three when the site actually needed them, and beds held empty in a system that ordinarily runs near capacity are beds unavailable to the ordinary emergencies that continued to happen across the region that week.

The blood figures and where they come from

Blood is the part of this story that gets quoted most often and sourced least often, and the figures are actually well documented, so there is no excuse for the round numbers that circulate. The 2002 GAO testimony cited earlier, GAO-02-1095T, estimates that roughly 572,000 additional units were collected in response to September 11, that about 364,000 of those units, or about two thirds, entered the nation’s blood inventory, and that approximately 208,000 units, about one third, expired and were discarded. The same testimony reports that monthly collections rose nearly 40 percent over collections earlier in 2001 in the weeks immediately following the attacks, and that the proportion of units expiring and being discarded in October and November 2001 was five times the proportion for an average two month period earlier that year.

Two further findings in that testimony deserve to be read out at any planning meeting where somebody proposes a blood drive as a community response. As much as 20 percent of some blood banks’ donations was collected improperly and had to be discarded, primarily because donors had not completed the donor questionnaire correctly, which is what happens when collection staff are overwhelmed by volume. The collection system itself was stressed for supplies, for phlebotomists and for storage, so the surge of goodwill consumed the capacity of the organizations it was meant to help.

The demand side figure is the one to handle carefully. A 2002 news item in the Canadian Medical Association Journal reporting on the post-attack donation slump states that fewer than 260 units were actually needed to treat victims of the attacks in New York and Washington, and that same item notes that donated blood is discarded if it remains unused after 42 days. I could not trace that 260 figure to its primary source, so it belongs in a briefing as a published report rather than as an established measurement, and the honest version of the sentence is that the number of units transfused to victims was very small in relation to the hundreds of thousands collected, with the precise transfusion figure attributed to whoever you are quoting. The 42 day shelf life for red cells is the mechanism underneath the whole problem, because a unit donated in the emotional week after an event cannot be used for a shortage two months later.

The same CMAJ item records that Canadian Blood Services collected about 37,000 units in the eight days after September 11, roughly twice its normal volume, and that a spokesperson stated none of it was wasted, which is a useful reminder that outcomes differed between collection systems and that the American figures should not be generalized to every blood service. The item also describes a task force responding with a call for multi-year donor awareness campaigns encouraging regular periodic donation instead of one-time surges, and I did not verify that body’s formal name, so I am describing its recommendation rather than naming it. This site has a separate piece on unsolicited donations and spontaneous volunteers that covers the same dynamic in physical goods.

The same shape in later incidents

The New York case is unusual in its cause and ordinary in its pattern, and the pattern is documented in the disaster research literature well before 2001. Erik Auf der Heide, in a review published in Annals of Emergency Medicine in January 2006 (volume 47, pages 34 to 49), assembled the empirical findings that contradict common disaster planning assumptions, among them the finding that casualties are typically transported to the closest hospital rather than distributed across a region, and that the least seriously injured tend to arrive first, a pattern the Disaster Research Center work he cites characterized as a kind of reverse triage. Any plan that assumes field triage will meter the flow into hospitals is arguing with several decades of observation.

The clearest modern example is the shooting at the Route 91 Harvest festival in Las Vegas on October 1, 2017. News reporting carried by KING 5, which I could not date precisely from the copy available to me, gives Sunrise Hospital, a Level II trauma center, as having treated 199 patients in a few hours, and University Medical Center, Nevada’s only Level I trauma center, as having taken 104, with a rumor circulating that UMC had closed or gone on diversion that was not true, and with UMC’s trauma director quoted as recalling that he had room and had offered to take patients from Sunrise. Sunrise’s own officials, speaking to FOX5 Las Vegas in an article published on October 2, 2025, gave different totals for their hospital, including 125 gunshot victims, more than 250 people treated, 14 people deceased on arrival and two who died later, 58 surgeries in the first 24 hours and 87 in total. Those two accounts disagree, they disagree partly because they count different things over different windows and partly because one is contemporaneous reporting and the other is an institutional recollection eight years later, and the right way to use them is to give both sets of figures and name both sources.

A review titled “A day like no other: the Las Vegas mass shooting lessons from America’s largest mass casualty event”, available in PubMed Central, describes the mechanism in the same terms: patients self transported to the nearest facilities, one hospital reported a line of private vehicles more than a quarter mile long waiting to reach its emergency department, and patients arrived without prehospital care and without advance notice. That review also describes what Southern Nevada built afterward, a Hospital Area Command collaboration arrangement under which hospitals can request fire personnel and paramedics through 911 to help with extrication from private vehicles, triage, vascular access and airway management as self transported patients arrive at the door.

Smaller events produce the same concentration. McGory and colleagues, writing in The American Surgeon in 2004 about the Santa Monica farmers market crash of July 16, 2003, record 73 people injured and 10 killed, eight of them at the scene, with 26 of the victims treated at two facilities, 15 at UCLA and 11 at Santa Monica. A jurisdiction reading its own mass casualty plan should ask what fraction of its casualties in a plausible event would be delivered by people who have never heard of the plan, because in every one of these cases that fraction was large and in none of them was it anticipated in the numbers that actually occurred.

What a coalition or an EOC is actually for

Regional medical coordination functions are usually sold on their ability to find beds, and finding beds is the part of the job that matters least in the first two hours, because in the first two hours the patients are already going wherever they are going. The function that would have changed outcomes in the cases above is knowing where patients are actually arriving, in something close to real time, and having a route by which that picture reaches every hospital in the region along with an instruction about what to do next. That is an information function before it is a logistics function, and it is the reason a coalition that cannot reach all of its member hospitals on one call within fifteen minutes has not yet built the notification capability it exists to provide.

The harder half is the stand-down, and it is harder for institutional reasons rather than technical ones. No hospital administrator has ever been criticized for having been too ready, while the person who releases the recalled surgical teams and reopens the elective schedule owns the consequences if a second wave arrives twenty minutes later, so the rational individual choice is always to hold, and the accumulated cost of everyone holding is paid quietly by the patients whose procedures were cancelled and by the staff who are exhausted before the long work begins. Getting this right requires the authority to be named in advance, exercised, and backed by whoever signs the hospital’s disaster plan, because an authority that appears for the first time during an incident will not be believed by the people it is instructing to stop.

Three questions will tell you where your own region stands, and none of them takes longer than a week to answer. Somebody, by title, has to be able to tell hospitals in your region that patient movement is complete, and you need to know who that person is and whether the decision travels to a hospital outside your county. Patient distribution information also travels some path in the first hour, so find out whether an emergency department charge nurse learns the regional picture from a coordination center, from a news broadcast or from the ambulances arriving at the ramp. Your blood supplier’s actual position on unsolicited donation during an incident is in most regions a request to book appointments over the following weeks rather than to queue on the day, which is the opposite of what the public assumes and of what a well meaning public information officer will say if nobody has briefed them.

The command and control story of September 11 in New York, including the two command posts and the location of the city’s emergency operations center, belongs to the Communications section of this site, which is running its own series on that record, and I have deliberately kept it out of this piece. What belongs here is narrower, which is that the medical side of a regional response needs a coordination function with a name, a roster, a communications path that is tested against the hospitals rather than against itself, and the standing authority to tell a hospital to stop preparing. Part three of this series moves to the debris field and the ten months of work that followed, where the health consequences that the hospitals eventually did see began.

What to do at your agency

• Ask your healthcare coalition coordinator, by name and in writing, who holds the authority to tell hospitals in your region that patient movement from an incident is complete, and file the answer with the date it was given, since an unanswered question here is itself the finding.
• Have your emergency manager read the patient distribution section of your mass casualty plan this month and confirm whether it states how a hospital learns, within the first hour, how many patients are coming and from where, and write that path into the plan in one paragraph if it is missing.
• Telephone your regional blood supplier’s disaster contact and ask what they want said publicly in the first six hours of a mass casualty event, then give that answer to your public information officer as a pre-scripted message before you need it.
• Add one item to the agenda of a meeting that already exists, such as your EMS medical advisory committee or your coalition’s quarterly meeting, asking each hospital representative to state how many patients arrived by private vehicle in their last real mass casualty activation.
• Test the hospital notification path you believe you have by sending an unannounced message through it during a business day and recording how many facilities acknowledge within fifteen minutes, then report the number to the same meeting.
• Ask your EMS agency director for the current bed availability and diversion reporting system’s actual refresh interval, and find out who checks it during an incident, because a system nobody reads during the event is a system that exists only on paper.
• Write a stand-down criterion into your existing mass casualty annex specifying who declares it, what they say, and which facilities get told, and rehearse that step at the end of your next exercise instead of ending the exercise when the last patient moves.

Takeaways

• Hospitals in the New York region activated existing disaster plans on September 11, 2001, and the best documented surgical account, by Cushman, Pachter and Beaton in the Journal of Trauma in January 2003, records 911 patients received by two affiliated Lower Manhattan hospitals, 85 percent of them walking wounded with inhalation and eye irritant injuries.
• Published patient totals differ by definition rather than by error, running from 790 injured survivors treated within 48 hours at a sample of nearby hospitals in the CDC rapid assessment published in MMWR on January 11, 2002, to 6,232 attack related emergency room visits at 103 New York State hospitals between September 11 and September 28 in the 2004 GAO testimony.
• Arrivals began within minutes of the attack and peaked two to three hours later according to that CDC assessment, which describes a self distributed flow rather than a managed one.
• Few patients arrived because the event produced deaths rather than survivors needing hospital care, and the last person recovered alive from the rubble was Genelle Guzman-McMillan, rescued on September 12 after about 27 hours.
• Roughly 572,000 additional units of blood were collected in response to the attacks, of which about 364,000 entered the national inventory and approximately 208,000 expired and were discarded, according to the 2002 GAO testimony GAO-02-1095T.
• Disaster research summarized by Erik Auf der Heide in Annals of Emergency Medicine in January 2006 finds that casualties are typically taken to the closest hospital and that the least injured arrive first, which is the same pattern seen at Sunrise Hospital in Las Vegas after October 1, 2017.
• Standing down is institutionally harder than standing up, because nobody is criticized for having been ready, and the cost of holding is paid in cancelled procedures, fatigued staff and beds unavailable for ordinary emergencies.
• A regional medical coordination function earns its place by knowing where patients are arriving and telling hospitals when to stop preparing, which requires a named authority, a tested notification path and a stand-down step that appears in exercises.