"Nurse refuses to perform CPR," read the caption on an ABC newscast in California. "911 dispatcher's pleas ignored." Several days earlier, an elderly woman at a senior living facility had gone into cardiac arrest. The dispatcher instructed an employee to perform CPR, or cardiopulmonary resuscitation. But the employee refused.

"Is there anybody there that's willing to help this lady and not let her die?" the dispatcher said. It made the local news, which elicited a national outcry and prompted a police investigation. But the woman was already dead — her heart had stopped. And according to family, the woman had wished to "die naturally and without any kind of life-prolonging intervention."

So why the controversy? It comes down to a widespread misconception of what CPR can, and can't, do. CPR can sometimes save lives, but it also has a dark side.

The discovery that chest compression could circulate blood during cardiac arrest was first reported in 1878, from experiments on cats. It wasn't until 1959 that researchers at Johns Hopkins applied the method to humans. Their excitement at its simplicity was clear: "Anyone, anywhere, can now initiate cardiac resuscitative procedures," they wrote. "All that is needed is two hands."

In the 1970s, CPR classes were developed for the public, and CPR became the default treatment for cardiac arrest. Flight attendants, coaches, and babysitters are now often required to be certified. The allure of CPR is that "death, instead of a final and irrevocable passage, becomes a process manipulable by humans," writes Stefan Timmermans, a sociologist who has studied CPR.

"This is the truest of emergencies and you give people the simplest of procedures," Timmermans told me. "It seems too good to be true," he said, and it is.

Many people learn what they know about CPR from television. In 2015, researchers found that survival after CPR on TV was 70%. In real life, people similarly believe that survival after CPR is over 75%. Those sound like good odds, and this may explain the attitude that everyone should know CPR, and that everyone who experiences cardiac arrest should receive it. Two bioethicists observed in 2017 that "CPR has acquired a reputation and aura of almost mythic proportions," such that withholding it might appear "equivalent to refusing to extend a rope to someone drowning."

But the true odds are grim. In 2010 a review of 79 studies, involving almost 150,000 patients, found that the overall rate of survival from out-of-hospital cardiac arrest had barely changed in thirty years. It was 7.6%.

Bystander-initiated CPR may increase those odds to 10%. Survival after CPR for in-hospital cardiac arrest is slightly better, but still only about 17%. The numbers get even worse with age. A study in Sweden found that survival after out-of-hospital CPR dropped from 6.7% for patients in their 70s to just 2.4% for those over 90. Chronic illness matters too. One study found that less than 2% of patients with cancer or heart, lung, or liver disease were resuscitated with CPR and survived for six months.

But this is life or death — even if the odds are grim, what's the harm in trying if some will live? The harm, as it turns out, can be considerable. Chest compressions are often physically, literally harmful. "Fractured or cracked ribs are the most common complication," wrote the original Hopkins researchers, but the procedure can also cause pulmonary hemorrhage, liver lacerations, and broken sternums. If your heart is resuscitated, you must contend with the potential injuries.

A rare but particularly awful effect of CPR is called CPR-induced consciousness: chest compressions circulate enough blood to the brain to awaken the patient during cardiac arrest, who may then experience ribs popping, needles entering their skin, a breathing tube passing through their larynx.

The traumatic nature of CPR may be why as many as half of patients who survive wish they hadn't received it, even though they lived.

It's not just a matter of life or death, if you survive, but quality of life. The injuries sustained from the resuscitation can sometimes mean a patient will never return to their previous selves. Two studies found that only 20-40% of older patients who survive CPR were able to function independently; others found somewhat better rates of recovery.

An even bigger quality of life problem is brain injury. When cardiac activity stops, the brain begins to die within minutes, while the rest of the body takes longer. Doctors are often able to restart a heart only to find that the brain has died. About 30% of survivors of in-hospital cardiac arrest will have significant neurologic disability.

Again, older patients fare worse. Only 2% of those over 85 who suffer cardiac arrest survive without significant brain damage.

CPR can be harmful not just for patients, but also for medical providers. In 2021, a study found that 60% of providers experienced moral distress from futile resuscitations, and that these experiences were associated with burnout. Another study linked intrusive memories and emotional exhaustion to difficult resuscitations. Holland Kaplan, a physician and bioethicist, told me that "the bad experiences far outnumber the good ones, unfortunately."

She has written about performing chest compressions on a frail, elderly patient and feeling his ribs crack like twigs. She found herself wishing she were "holding his hand in his last dying moments, instead of crushing his sternum." She told me that she's had nightmares about it. She described noticing his eyes, which were open, while she was performing CPR. Blood spurted out of his endotracheal tube with each compression.

"I felt like I was doing harm to him," she told me. "I felt like he deserved a more dignified death." It's no wonder that many doctors are not fond of CPR, and choose not to receive it themselves.

The true purpose of CPR is to "bridge the person to an intervention," Jason Tanguay, an emergency physician, told me. "If they can't get it, or there isn't one, then what is it accomplishing?" This is the crucial insight that doctors have and most others don't. CPR is a bridge, nothing more. Sometimes it spans the distance between life and death, if the cause can be quickly reversed, and if the patient is fairly young and relatively healthy. But for many that distance is too great. "The act of resuscitation itself cannot be expected to cure the inciting disease," the Hopkins researchers wrote in 1961.

A patient with terminal cancer who is resuscitated will still have terminal cancer. In those cases, the most humane approach may be to ease the pain of the dying process, rather than build a bridge to nowhere.

How can physicians help patients make these choices in advance? Part of it is education. Studies have found that half of patients changed their wishes when they learned the true survival rates of CPR, or after watching a video depicting the reality of CPR.

Another part is communication. According to one survey, 92% of Americans believe it's important to discuss end-of-life care, but only 32% have done so. Physicians (or patients) should initiate these conversations early, especially for those who are elderly or have chronic medical problems, so that their wishes are known in advance if they suffer a cardiac arrest.

Language matters too. Doctors often ask if patients "want everything done" if their heart stops. But that puts a burden on patients and families. "Who wants to feel like they don't want everything done for their loved one?" Kaplan says. Instead, if CPR would likely be futile, doctors could recommend "allow natural death" instead of "do not resuscitate," suggests Ellen Goodman, director of a non-profit that encourages end-of-life conversations.

"Give people something they can say yes to," she told me. Physicians have the knowledge and experience to guide patients in choosing measures they may benefit from, declining those that may harm, and aligning interventions with their wishes and values. The most important thing, instead of always taking action, is to ask.

Clayton Dalton is a writer in New Mexico, where he works as an emergency physician.