This post appears as an editorial in the July 2020 special COVID-19 issue of The American Journal of Bioethics
by Sheri Fink, M.D., Ph.D.
The anthrax mailings following the September 11, 2001 attacks in the United States led to fears that victims of bioterrorism could overwhelm hospitals. The federal government convened experts to define how medical treatments should best be allocated across a population affected by a mass casualty disaster, a concept at first referred to as “altered standards of care,” later changed to the more palatable “crisis standards of care.” This work informed triage plans developed in the wake of the SARS outbreak in 2003, a novel respiratory pathogen that stressed critical care resources in advanced hospital systems, including Toronto’s.
Although the H1N1 influenza pandemic in 2009 turned out to be relatively mild, fears of a more severe pandemic spurred additional protocols for rationing ventilators, critical care services and hospital admission, building on the previous guidelines. According to their own descriptions, the plans aim to save as many lives or life years as possible–prioritizing people who are expected to survive their current illness or live the longest after being treated. Accompanying ethical analyses assert that their application will avoid the potential biases and moral distress expected to result from providers making ad hoc choices between patients at the bedside.
The emergence of Covid-19 has renewed interest in these plans. However, controversies and unresolved ethical, legal, conceptual and practical aspects of the guidelines, many of them explored in this issue, have arguably limited their use. This article reviews how the plans evolved, how they met the reality of a burgeoning outbreak, and questions that remain to be answered. Is a transparent, thoughtful, but ethically problematic triage protocol, backed by little evidence that it will achieve its stated goals, even as it relieves clinicians of the burden of making difficult choices without guidance, better than no protocol at all?
By 2016, at least 18 states from New York to California and numerous hospitals, including the 152 medical centers of the Veterans Health Administration, had drawn up guidelines for who would, and would not, get access to potentially lifesaving resources in a severe pandemic and from whom they could be taken away without patient or proxy consent to be reassigned to others. Some of these efforts were supported by federal hospital preparedness funding.
The intellectual ancestry of the plans included work by Dr. Frederick M. Burkle Jr., a former Vietnam War physician who laid out ideas for how to handle the victims of a large-scale bioterrorist event. After the SARS outbreak stressed Toronto hospitals in 2003, some of his ideas were proposed by Canadian doctors, and they made their way into many American plans after the H1N1 pandemic in 2009. “I have said to my wife, ‘I think I developed a monster here,’” Dr. Burkle told me several years ago, commenting on the inflexibility of many early versions of the plans. He had emphasized the importance of reassessing the level of resources sometimes on a daily or hourly basis in an effort to minimize the need to deny care.
What stood out to me when I began writing about these plans in 2009, as a reporter who was trained as a physician and served as a relief worker in armed conflicts and disasters, was that little effort had been made to ensure that the general public who could be subject to them—as well as the medical staff members who would be implementing them—knew that the protocols existed. Likewise, little input from the broader public or vulnerable groups had been solicited in crafting them.
The plans had been hashed out by professionals who, by dint of being professionals, could be expected to share many relevant values and implicit biases. The protocols themselves often gave lip service to the ethical duty to solicit public input. A typical article stated that the authors of a draft protocol would hold focus groups with “older adults, people with disabilities, and communities of color.” But when I would follow up, asking researchers whether they had engaged with the public, and what the results had been, I was often told that they had run out of funding. I was even assured that the presence of ethicists in the planning process ensured that the interests of vulnerable groups were “well accounted for.”
Perhaps unsurprisingly, guidelines developed in this era typically excluded people with certain chronic or disabling conditions from treatment in a severe pandemic or called for lowering their priority, and sometimes subjected people who lived in congregate settings such as nursing homes, mental institutions, prisons and facilities for the “handicapped” to rationing rules earlier in a pandemic than patients in the general population.
The plans typically did and still do call for doctors to withdraw treatments, including ventilators, from patients without their consent if they fail to improve after various time trials (or in some cases if they arrive at the hospital with their own ventilators that they use every day and depend on to live, such as people with end stage lung disease from cystic fibrosis). While provisions to withdraw care regardless of consent are a nearly universal feature of the triage plans, there is little or no evidence that the implementation of these measures—potentially illegal and certainly distressing to providers and families—would actually promote the plans’ utilitarian goals.
For example, if a disease caused by a new pathogen requires longer than anticipated periods of ventilator support, rigid implementation could paradoxically worsen survival outcomes, because patients would not be given enough time to improve. Commonly used systems to score disease severity, such as the Sequential Organ Failure Assessment (SOFA), may lack the predictive value ascribed to them in this context. And questions about the triage of children and neonates are particularly problematic.
Few outcomes studies exist, and those that do are troubling. Researchers at Mercy Regional Medical Center in Durango, Colorado, found that had a triage protocol been used, 67 percent of H1N1 flu patients who survived would have been considered for exclusion from the ICU or for triage to “expectant” care with extubation and withdrawal from ICU support (Blumhardt 2010).
Applying a disaster protocol to 255 ICU patients treated under normal conditions at Derriford Hospital in Plymouth, researchers found that, of the 46 percent of patients who would have been denied or withdrawn from intensive care treatment under the protocol, 61 percent actually survived to hospital discharge. In the subset of those who would have been deemed too sick for ICU on admission, 69 percent survived. “The proposed triage tool failed adequately to prioritise patients who would benefit from intensive care,” the authors concluded. “A new model of triage needs to be developed.”
Another British study focused on H1N1 patients admitted to an ICU in June and July 2009. Had the triage guidelines been followed, five of the eight patients would have been considered for withdrawal of care at 48 h based on their SOFA scores. All survived. The authors concluded: “SOFA score based triage could lead to withdrawal of life support in critically ill patients who could survive with an acceptably low length of stay in the intensive care unit.” In this cohort, none of the patients who needed mechanical ventilation would have received it for the length of time required to survive had the triage protocol been used. “New patients fulfilling the same high priority admission criteria would have replaced those palliated, and potentially had the same outcome after 2 days’ treatment in critical care.”
The few public engagement exercises held in those early days raised issues that had been overlooked by the experts. In Seattle, members of the public at large were concerned that using survival statistics to determine access to resources might be “inherently discriminatory,” a project report from 2009 said, “because of institutional racism in the health care system; if some groups (e.g., African Americans and immigrants) do not receive the same quality of care, then their rates of recovery and other survivability measures would be biased.” That critique has particular resonance today, given the disproportionate share of Covid-19 deaths occurring among racial minorities.
Pandemic triage plans also assume that patients who are not prioritized for potentially life-saving treatments will be palliated. Little acknowledged, however, is the limited capacity that many medical systems have to provide that palliative care. I saw the importance of that when reporting from an Ebola treatment unit in West Africa in 2014. There was no need to allocate effective treatments for the disease because none existed. But medical workers still experienced moral distress related to their limited ability to provide supportive care. Wearing Tyvek suits in the heat, they had to triage their own time in the infectious hot zone providing fluids, pain medications and comfort. Creative thinking contributed to a solution, as it often does: field hospitals mobilized Ebola survivors, who now had immunity to the disease, expanding resources and extending the ability to provide this care.
As a witness to several deliberative democracy exercises concerning pandemic triage held by researchers in Maryland beginning in 2012, I noted another phenomenon that has rarely been documented or discussed. Participants who were asked to decide between ventilator allocation strategies frequently attempted to escape the conundrum by proposing innovative ways to avoid rationing altogether, demonstrating a quintessentially American can-do attitude and a refusal to accept limits. What about rapidly producing more ventilators? Could more than one patient be treated with a single ventilator? One participant, a nursing student, said that on her ward she wouldn’t give up on anyone. The leaders of the sessions were frustrated. They considered the responses naïve. I recall thinking: this is exactly what might save us.
Some people involved with drafting triage plans have said that they were motivated by what happened at Memorial Medical Center after Hurricane Katrina, a community hospital that was surrounded by water, lost power, and, like every other hospital in New Orleans, took days to fully evacuate in 2005. After the storm and a lengthy investigation, a doctor and two nurses were arrested and accused of second-degree murder in the deaths of multiple patients. That concerned providers around the country, given that the three were well respected, maintained their innocence, and had kept working in abysmal conditions.
But although doctors made triage decisions at Memorial–namely concerning whom to prioritize for rescue by helicopter, designating many of the sickest and most dependent for last—it is important to note that the arrests were made for allegedly hastening deaths, not for making tough triage calls. I spent years investigating the incident, culminating in an article and the book Five Days at Memorial. Nearly two dozen patients were injected with morphine, midazolam or a combination of the two and died in a short time period. Two other doctors who participated told me that their goal was death, not palliation.
The drugs were injected into at least one patient who, according to his providers, had been awake, alert, had fed himself breakfast that morning and had expressed the desire to be rescued (“Don’t let them leave me behind.”). But he was paraplegic and morbidly obese, and so his rescue posed logistical challenges, according to staff members I interviewed who had participated in discussions about what to do with him. At the time most injections took place, rescue helicopters were finally focused on the hospital. But there was no reassessment of the resources, which might have allowed a change in the conclusion that these patients could not be saved.
Over the years, work on pandemic triage plans has continued. More efforts have been made to engage the public. Aspects such as categorical exclusion criteria—forbidding people with certain diseases or above a certain age to get treatment—have been challenged. Newer guidelines are often more nuanced and flexible to context. Some assign a priority score to each patient based on a combination of factors, using random selection to decide between people when survival chances cannot be meaningfully distinguished. (Some authors have raised concerns about whether these combination protocols contain the same ethical pitfalls, only making them less obvious by burying them in complex algorithms.)
The coronavirus pandemic led newer and older plans alike to be hastily dusted off, updated and considered for use, particularly after overwhelmed doctors in Italy reported withholding ventilators from older, sicker adults in favor of younger, healthier patients—a strategy as straightforward as it was, to many, chilling. In the absence of federal and, often, state level guidance, networks of experts, including bioethicists, shared best practices and the most recent protocols.
At last, public awareness of the protocols greatly expanded, and with it, scrutiny. Feedback from stakeholders came not from long-promised focus groups but instead from civil rights complaints and lawsuits.
While expert groups had spent years hashing out differences over details, such as exactly which groups of patients should be prioritized or deprioritized, the plans—which reference each other and were built upon one another—retained the same DNA and lacked an evidence base as to whether they were likely to achieve their intended utilitarian goals. Conducting that research in the context of Covid-19 seems imperative.
And although conceptually similar, the plans have notable discrepancies that could affect which patients survive depending on which side of a state border they live on, further highlighting the diversity of values held by different groups. Triage plans have as much to do with values as with medicine.
In late March this year, a colleague and I examined ten plans deemed ethically acceptable in different states. For a pregnant woman with a healthy fetus, the Maryland guidelines gave a one-point credit, while Pennsylvania offered a two-point benefit. Utah guidelines provided one or two points depending on how far into the pregnancy the mother was. The Pennsylvania guidelines also offered a one-point benefit to health workers playing a role in responding to a crisis; whether or not to prioritize health workers remains an area of ethical debate and is discussed in many articles in this issue. Maryland’s plan gave young people better odds of getting treatment when estimated survivability was the same.
Political leaders in some states, including New York and Maryland, refrained from endorsing draft guidelines that experts had worked on for years. The Office of Civil Rights at the Department of Health and Human Services rejected several common elements of the plans, including age and disability-based rationing. The office received complaints from disability advocacy groups against at least seven states for their crisis standards of care plans and took action on some of them.
For example, in an effort to incorporate the likelihood of long-term survival into triage scores, Pennsylvania’s plan explicitly mentioned patients with Alzheimer’s disease, those with cancer expected to live fewer than 10 years, and those with kidney failure. In settling its civil rights complaint, the state changed its guidelines on April 10, removing those specific diseases and stating that patients must receive individualized assessments and that nobody may be excluded based on age and disabilities.
Disability advocacy groups filed suit in Washington state, alleging that the triage algorithms could disadvantage people living with disabilities. That led to the addition of a statement that only “prognostically relevant decompensation from a person’s baseline function” should be considered.
What has struck me the most is that transparency about the possibility of rationing and the horrors of what that might look like, for patients and providers, led to a scramble to avert it. The grim prospect of losing people due to a shortage of ventilators and critical care spurred hospital officials, scientists, manufacturers and political leaders to source and build more ventilators, retrofit other types of breathing devices, construct field hospitals and even dramatically expand intensive care units within institutions, all in an effort to avoid rationing. Earlier investments mobile hospitals and stockpiling ventilators, among other preparedness steps, also paid off.
This may be the most important lesson of all: creativity, improvization and national values that reject rationing can help avoid some tragic choices. Perhaps the nurse in the Maryland triage exercise had the right idea after all.
As the pandemic continues and its realities are better understood, it may be time to consider more profound conceptual changes to the triage guidelines. For example, authors in this issue have discussed prioritizing people from disadvantaged groups for resources even if they are less likely to benefit from them. And researchers now recognize that having a conceptual framework is very different from operationalizing a triage plan within a hospital.
Meanwhile, the realities of this new pandemic have shed light on other ethical dilemmas that had not been as well anticipated, many of them also highlighted in this issue, including whether social distancing during a pandemic constitutes a right or a restriction and when it might not be permissible or legal to ban hospital visitors (after a civil rights complaint, Connecticut required hospitals to allow a support person for people with disabilities when needed). Quieter forms of rationing have occurred, involving critical care, experimental medicines, extracorporeal membrane oxygenation, and personal protective gear. In many places, triage guidelines were not widely shared. Other dilemmas loom, including how the first doses of hoped-for vaccines will be distributed.
I reported from seven hospitals during the spring 2020 surge of cases in New York City and three in Texas during the state’s summer 2020 surge. One marked difference from the typical standard of care was arguably a positive one: earlier efforts were made to have conversations about life support preferences. However palliative care experts were in short supply. End-of-life discussions were complicated by the fact that family members were not allowed to be present, and the ability to prognosticate was limited by a lack of data and shifting knowledge about the disease and the efficacy of various interventions. Some families felt pressured by providers to remove life support too early, and some providers felt pressured by families to maintain it too long.
Other ethically troubling issues included that where a patient’s ambulance went could determine his or her ability to receive certain modalities of care; that hospitals had to compete for personal protective gear; and the compromises in care that came from stretching hospital capacity. Hospital leaders struggled over whether the duty to care is limited when procedures of unknown value might put providers at an unknown level of risk. How much benefit to the patient does there need to be? What if the efficacy is unknown? What is certain is that the field of bioethics has much important work to do.