When the Facts Aren't Really Facts: Rethinking Decisions at the Edge of Newborn Life

A baby is born at 23 weeks. The room is full of people who need to make a decision in minutes. The parents are terrified. The clinicians have done this before, but never with this baby. Someone, somehow, has to decide whether to begin life-sustaining treatment.

In the United States, the standard answer to “how should this decision be made?” is shared decision making. Parents and clinicians sit down—or, more often, stand together at the bedside—and reason through the options together. The clinicians supply the medical facts. The parents supply their values. Together they reach a choice.

It sounds good. It is, in many ways, good. But in a recent paper in BMC Medical Ethics, I argue that the model has a quiet problem at its center: it assumes that the facts being shared are actually fixed. At the edge of neonatal viability, they aren’t.

The hidden assumption

Shared decision making was a response to two earlier models that bioethics had grown uncomfortable with: old-fashioned medical paternalism, where doctors decided and patients deferred; and a thin, legalistic version of informed consent that asked patients to sign forms while clinicians stayed studiously neutral. Shared decision making was supposed to be the better third option—a collaboration, drawing on different kinds of expertise.

But built into most accounts of how it should work is a quiet premise: that there is a stable set of facts (survival rates, disability outcomes, treatment burdens) that the clinician brings to the table, and that the parents’ job is to weigh those facts against their values.

At the margins of viability, this premise quietly breaks down. The facts themselves are produced by the decisions that get made. They don’t sit outside the process, waiting to be discovered. They emerge from it.

How a fact gets made

Consider two babies born at 23 weeks at two different hospitals.

At hospital A, clinicians, expecting a poor outcome at this gestational age, don’t give the mother antenatal steroids. The baby is born floppy, with weak respiratory effort. Comfort care is provided. The baby dies within hours.

At hospital B, steroids were administered. The baby is intubated, given surfactant, admitted to the NICU. Two years later, the child has moderate developmental delay but is walking and engaging with the world.

Now: what are the survival statistics for babies born at 23 weeks? The answer depends entirely on which babies you count. If you average together infants who received aggressive intervention and infants who never did, you get one number. If you look only at babies who were actively treated, you get a very different one. Both numbers are factually accurate. Neither is innocent.

This isn’t a hypothetical. The American College of Obstetricians and Gynecologists’ 2017 consensus statement on periviable birth reported survival rates of 5–6% before 23 weeks—numbers that mixed treated and untreated infants together, implicitly treating variation in human decisions as if it were variation in human biology. The neonatologist Keith Barrington pointed out that this framing presents the consequences of one set of choices as though they were the natural limits of medicine.

The data then gets used to justify the practice that produced it. Centers that intervene less aggressively cite their own outcome data to explain why they don’t intervene. The prediction shapes the action; the action shapes the outcome; the outcome confirms the prediction. Philosophers and sociologists have a name for this—a self-fulfilling prophecy—and it shows up across critical care medicine, especially in adult neurology after cardiac arrest, where pessimistic early prognoses tend to lead to withdrawal of treatment, which then appears to confirm the prognosis.

Why this matters ethically

You might say: so what? Doctors have to make decisions under uncertainty all the time. The data we have is the data we have.

But the ethical force of shared decision making depends on parents being told something true about the situation they’re in. If a clinician says “the survival rate at this gestational age is 5%,” and that number is partly a product of decisions that families like this one are now being asked to make, the framing has done something important—and not quite honest. It has presented a contingent outcome as a fixed fact, and a value-laden choice as a clinical inevitability.

The argument here isn’t that the data is useless, or that everything is relative, or that clinicians should stop offering professional judgment. It’s an argument for something narrower and harder: epistemic humility. The disciplined recognition that what you know is partial, that your confidence should match your evidence, and that early judgments will need to be revised as new information comes in.

What this looks like in practice

At the bedside, epistemic humility translates into a few specific habits:

• Time-limited trials of treatment, with clearly articulated goals, instead of framing early decisions as irrevocable commitments. You can start aggressive care, see how the baby responds over days or weeks, and reassess. The first decision doesn’t have to be the last.
• Honesty about where statistics come from. Not a meta-analysis at the bedside, but a willingness to say: “These numbers reflect what happens at places that treat aggressively. Other places have different numbers, partly because they make different choices.”
• Documentation that records decisions as provisional, subject to reassessment—not as endpoints.
• Recommendations that come with their reasoning attached, so parents can see why a recommendation might change as the clinical picture evolves.

There’s a side benefit here, too. Moral distress among NICU clinicians—the experience of feeling trapped by decisions that no longer fit reality—is often made worse by what might be called premature epistemic closure: treating uncertain prognoses as if they were settled. A practice that legitimizes uncertainty and revision can ease some of that burden. Not all of it. Some moral distress is just what comes with caring for very sick babies, and it shouldn’t be erased. But the distress that comes from being asked to feel certain when you’re not—that part is structural, and it can be changed.

Beyond the NICU

The neonatal case is unusually clean—the uncertainties are vivid, the stakes are immediate, the path dependence is easy to see. But it isn’t actually exceptional. Wherever medicine pushes its limits—neurology after severe brain injury, oncology with experimental therapies, intensive care for the very old—clinicians end up making decisions whose downstream consequences shape the data that future decisions will rely on.

If that’s right, the neonatal paradigm isn’t a special case at all. It’s a preview of how a growing slice of medicine is going to work. Which means the ethical task isn’t to pretend the facts are fixed, or to abandon shared decision making, but to practice it in a way that’s honest about what kind of knowledge we actually have—and what kind of responsibility comes with making choices that help bring the future into being.