On my first day as a pediatric resident at Cedars-Sinai Medical Center in Los Angeles, I was assigned to the neonatal intensive care unit. I was in awe of it. Three (and later four) large rooms were lined with clear plastic incubators, and inside each, a baby. Many of them were extremely premature.

Some were small enough to fit in the palm of my hand and weighed barely a pound. Their little chests heaved with each heartbeat. Their skin was paper-thin. Ventilators breathed for them; intravenous lines traveled to their bodies to nourish them; sensitive electronics tracked every subtle change in their vital signs.

NICUs are the triumph of modern medicine’s investment in technology, pharmacy and know-how. They exist to finish nature’s work because 500,000 times a year — more than anywhere else in the industrialized world — an American baby is born prematurely. The most precarious are born at the margin of life: somewhere between 23 and 26 weeks of gestation, or what’s called the limit of viability.

That limit has changed dramatically over the past half century. In the 1960s, when the first NICUs opened, premature infants had a 95 percent chance of dying. Today, they have a 95 percent chance of survival. This has, in the words of one neonatologist, Dr. Nicholas Nelson, changed our perception of the premature baby as “a patient to be cared for, rather than an object to be pitied.”

Now we face a difficult choice, one not unlike that facing physicians who take care of adults near the end of their life: whom to fight for and whom to let go. The decision says volumes about how we have come to regard the tiniest, frailest of patients.

Saving lives this young is not benign. Survivors of extreme prematurity have frequent, and often severe, complications during their time in the NICU. In the worst cases, these children will suffer lifelong disabilities: cerebral palsy; severe visual impairment that thick glasses and eye surgery can only partly correct; scarred lungs that will leave them reliant on oxygen tanks; intellectual and behavioral problems that put them well behind their peers.

Partly because prospects for these children can be so dark, the American Academy of Pediatrics suggests not resuscitating babies born before 23 weeks, while babies born after 26 weeks are usually resuscitated.

Between 23 and 26 weeks, the risks remain high but survival improves with each week. This range is treated by the pediatrics organization as a gray zone, and doctors and parents must make the hard decision about whether to try to resuscitate a baby without firm guidance.

The circumstances are rarely ideal. Neonatologists walk into a patient’s room, day or night, amid the intense activity of obstetricians and nurses trying to manage labor. It’s an emotional, tense and uncertain time not conducive to detailed discussion or reflection. Unsurprisingly, then, these life-or-death decisions are made inconsistently.

In a 2005 study, researchers at McGill University in Montreal surveyed 165 pediatric and obstetric residents in four Quebec medical centers about resuscitating babies born between 23 and 26 weeks. Some residents, the researchers found, worked at hospitals with an aggressive culture of resuscitation. Other facilities embraced far less aggressive approaches: Even at 26 weeks, when a premature baby’s chance of survival is over 70 percent, residents at these medical centers indicated that they would attempt resuscitation only about half the time. Pediatricians in the United States also have highly variable approaches to resuscitating premature babies, studies suggest.

Parents hold to a far more consistent ethos. A 2001 study by researchers at McMaster University in Ontario showed that a significant majority believed that attempts should be made to save all infants, irrespective of condition or weight at birth. Just 6 percent of health professionals said the same. Older studies in the United States have suggested that