The Tiniest Patients: Rethinking How We Decide

When a baby is born at 22 or 23 weeks of pregnancy — half the normal gestational period — doctors and parents face one of the most agonizing decisions in all of medicine. Should they fight to keep the baby alive, knowing survival is uncertain and disability common? Or should they focus on comfort and let nature take its course?

Over the past three decades, a broad consensus has emerged that these decisions should be made collaboratively between clinicians and families — a process called shared decision-making, or SDM. Professional societies, bioethics commissions, and parent advocates have all endorsed this approach. And yet, despite the consensus, outcomes for babies born at the edge of viability vary wildly from hospital to hospital, from country to country. Something isn't working.

In a paper just accepted by Seminars in Fetal and Neonatal Medicine, I argue that the problem isn't the principle of shared decision-making — it's how we've been putting it into practice.

A debate that has fundamentally shifted

Not long ago, the central question for a baby born at 22 weeks was simply: should we even try? The prognosis seemed so grim that many hospitals didn't offer treatment at all. A 2015 review of policies from 23 countries found that none recommended routine active care for babies born at 22 or 23 weeks.

That ground has shifted dramatically. In 2022, the New England Journal of Medicine presented a clinical case involving a baby at 22 weeks gestation — and asked readers to vote on whether aggressive resuscitation should be mandatory or optional. The debate was no longer about whether treatment was futile. It was about whether treatment was morally obligatory. Of nearly 1,900 respondents, 71% preferred a policy of individualized, selective resuscitation over a universal mandate.

The question has changed. The answers, however, remain deeply contested.

The flaw at the heart of shared decision-making

SDM frameworks assume that good decisions follow from good information. First, establish the facts. Then, with a clear picture of the baby's prognosis, parents and clinicians can together decide whether treatment should be offered, required, or withheld.

But the facts, it turns out, are anything but clear.

Consider the most basic question: what is the survival rate for a baby born at 22 weeks? The answer depends enormously on who's counting and how. Studies routinely pool together babies who received aggressive NICU care with babies who received no treatment at all. One major NICHD study reported an 11% survival rate for babies at 22 weeks — but among babies who actually received active NICU treatment, the survival rate was 30%. Only 36% of babies in that study received active treatment.

The numbers that matter

Reported survival at 22 weeks: ~11% (all births) vs. ~30% (among those who received active NICU treatment). Among mothers given antenatal steroids, survival climbed to ~39%. The difference isn't biology — it's treatment.

Reporting survival rates that include babies who were never treated is a bit like reporting cancer survival rates without accounting for whether patients received chemotherapy. The data become self-fulfilling prophecies: pessimistic statistics discourage treatment, which produces worse outcomes, which reinforce the pessimism.

The picture gets murkier when we turn to neurodevelopmental outcomes. Many studies use a composite variable of "death or survival with severe disability" — treating the two as ethically equivalent. But that equivalence is itself a value judgment, and one that many parents would strongly reject. The nature and severity of disability matters enormously to families, and their assessments of what constitutes an acceptable outcome consistently differ from those of health professionals.

Gestational age is not enough

Most hospital policies — and most clinical decisions at the bedside — hinge on a single number: the baby's estimated gestational age. Yet gestational age is only one of many factors that shape a preterm baby's prognosis. Sex, birth weight, whether the mother received antenatal steroids, the presence of infection, the quality of prenatal care, the experience of the clinical team, and even the parents' socioeconomic status all make a real difference.

A female infant at 23 weeks whose mother received steroids has a substantially better prognosis than a male infant at 25 weeks whose mother did not. And yet, in many centers, the older baby would be aggressively treated while the younger one would not. This is not individualized medicine — it's policy by proxy.

"There is no other area of medicine in which decisions about life-sustaining treatment are based on a single integer."

In every other domain of medicine, a 30% chance of survival would be considered reason enough to at least offer treatment and have a full conversation. Neonatology is the only field with professional guidelines that sometimes recommend not offering such treatment at all.

What parents actually want

One of the foundational assumptions of the current SDM framework — articulated by early advocates like William Silverman and Helen Harrison — was that parents, if fully informed of the risks of disability, would generally prefer to forgo aggressive treatment. The thinking was that overtreatment was being driven by physician enthusiasm rather than parental preference.

Research has repeatedly shown this assumption to be wrong. Well-informed parents frequently want more treatment than medical staff think is appropriate. Their choices don't reflect ignorance of the data — they reflect different values, particularly about what constitutes a life worth living with a disability.

The result is that the most common ethical conflicts in NICUs today are not about undertreated babies. They are about parents who want continued treatment when doctors believe the treatment is only prolonging dying.

A better model: iterative, not one-and-done

The deepest problem with current SDM frameworks is that they treat the decision as a single event — a conversation before birth or at the delivery room door, after which a course is set. But prognosis evolves rapidly. Information that is unavailable at birth — how the baby is responding to treatment, whether complications are emerging, what the trajectory looks like — becomes available over days and weeks.

We actually get better at predicting outcomes the longer we watch a baby respond (or not respond) to care. A policy of initial resuscitation with ongoing review, with the explicit understanding that withdrawal of treatment is always an option if the picture worsens, may allow for better-informed decisions than forcing a binary choice in the delivery room.

This isn't merely a clinical recommendation — it's a different philosophy of how these conversations should unfold. Better SDM would start by acknowledging the emotional devastation of premature birth before presenting statistics. It would create space for parents to express fear, grief, hope, and confusion. It would be explicit that no single conversation settles everything, and that families will be supported through repeated decision points as the situation develops.

"Patients and surrogates encounter repeated decision points in the care of the seriously ill patient. Past decisions influence present decisions."— Feudtner, Schall & Hill, Pediatrics 2018

Looking ahead

The boundaries of viability are not fixed. Treatment of babies at 22 weeks was once considered experimental; survival with good outcomes is now documented at specialized centers. Emerging technologies — including research into artificial placenta systems — may push those boundaries further still.

Each advance brings new ethical terrain. Technology tells us what is possible. Ethics asks us what we should actually do — and for whom, and why, and at what cost to the baby, the family, and the caregivers who bear the weight of these decisions every day.

What will not change is this: the decisions are made by human beings under conditions of profound uncertainty, and they carry consequences that last a lifetime. The least we can do is make sure the process of deciding is as honest, as humane, and as genuinely shared as we can make it.

For more, see "An Emerging Consensus About Treatment of the Tiniest Babies" forthcoming in Seminars in Fetal and Neonatal Medicine (DOI: 10.1016/j.siny.2026.101712).