How Sick She Had to Be

Savita Halappanavar came into University Hospital Galway on a Sunday morning in October 2012 with back pain, was told to take paracetamol and see a physiotherapist, and went home. She came back that afternoon saying she had felt something coming down. She was seventeen weeks pregnant and losing it. The notes called the miscarriage inevitable and the plan, in the record's own phrase, was to await events.

Her membranes ruptured after midnight. She and her husband asked for a termination; the consultant's account to the investigation team was that she told them Irish law did not allow it while there was still a fetal heartbeat. Her husband's evidence was that they asked three times over two days. On the Wednesday afternoon she miscarried on her own, by which point the team had documented infection of the membranes and possible sepsis. She died at nine minutes past one on Sunday morning, a week after she first walked in, of septic shock.

I went into this expecting to write that a rule with no exceptions had killed her. That is not what happened, and what did happen is worse.

Ireland had an exception. It had had one since 1992, when the Supreme Court held that a termination was lawful where there was a real and substantial risk to the life, as distinct from the health, of the mother. The doctors' own professional code repeated it, and went further: it said that in rare complications where the baby has little or no hope of surviving, it may be necessary to intervene to protect the life of the mother. That is a description of her situation, written down in advance, by her own regulator.

What no document anywhere said was how sick she had to be first.

I read rules for a living, and I have rarely seen a gap that precise. Her regulator had thought about her case in advance. It described the shape of it, wrote down that intervention might be necessary, and then left out the only number anyone standing at that bedside would need.

What the reports actually found

I want to be careful here, because the version of this story everyone knows is wrong in a way that lets people throw the whole thing out. The easy version is one I am glad to be rid of.

Galway is a public hospital. The coroner said in terms that Irish public hospitals follow no religious tenets or dogma. The famous line about Ireland being "a Catholic country" was not said by the consultant, who denied on oath saying anything about religion; it was said by a midwife manager, Ann Maria Burke, who confirmed it to the coroner's face, explained she had been trying to answer a patient who had raised her own Hindu faith, and said it came out the wrong way and she was sorry.

Nor did any official body find that the law killed her. The health service's investigation listed three key causal factors and the first one is not the law: it is inadequate assessment and monitoring that would have let the team spot that she was deteriorating. The law appears second, and hedged — the failure to offer her all the options appears to have been either due to the way the law was interpreted or due to a lack of appreciation of the rising risk. The word Catholic does not appear in that report at all. The health regulator's separate 250-page investigation never uses the words termination, abortion or Catholic even once. Neither report says the death was preventable. When the inquest jury returned medical misadventure, the coroner went out of his way to warn that this did not mean the failures in her care had caused her death.

I labour this because the cheap version of the story is also the disposable one. Tell people a priest killed her and everyone who likes priests stops reading, and they are right to, because it is not true. The honest summary is not that a religious rule killed a woman. It is that a rule with a door in it killed a woman, because nobody would say where the door was. And the person who put that best was the coroner, whose first recommendation was that the Medical Council spell out exactly when a doctor may intervene, in order to remove doubt and fear from the doctor.

Doubt and fear. Not doctrine.

The traditions built the door themselves

This is the part that changed my mind about what the essay was.

Three weeks after she died, the Irish Catholic bishops issued a statement with her name in it, saying that the Church has never taught that the life of a child in the womb should be preferred to that of a mother, and that where a seriously ill pregnant woman needs treatment which may put her baby's life at risk, such treatment is ethically permissible. They noted, not without a certain defensiveness, that they had already published this three weeks before she was admitted.

The American bishops' rulebook for Catholic hospitals is more specific still. Directive 47 permits treatments aimed at curing a proportionately serious pathological condition of a pregnant woman even if they will result in the death of the unborn child, and that sentence has not changed a word in twenty-five years. Two years before Galway, the Catholic bioethics centre in the United States had already applied that reasoning to exactly her clinical picture: when the membranes rupture early and the placenta becomes infected, emptying the uterus is permitted.

Now hold that against the tradition that thought hardest about this problem and reached the opposite conclusion about who should decide. I did not expect to find the best answer in a body of religious law, and I did not expect it to be this blunt.

Jewish law treats saving a life as overriding almost the whole Torah, and having established that, it spends its energy on a different worry: that people will hesitate. The Talmud says you act, and you need no permission from a court. If a door has shut on a child, you break the door. Maimonides says it is forbidden to delay. The codes are blunt about the pious man who wants a ruling first: the one who hurries is praised, and the one who asks is a spiller of blood. A later commentary calls a patient's refusal on religious grounds foolish piety, and says he is to be overruled.

Read that next to a Galway coroner asking, in 2013, that doubt and fear be removed from the doctor. Fifteen centuries apart, the same engineering problem, and the rabbis got there first. What I take from them is not piety. It is that they saw the failure coming and designed against it: if you write an exception and leave the person who needs it exposed to judgement afterwards, you have not written an exception. You have written a trap.

What happens to people who use the door

Ask the nun.

In November 2009 a 27-year-old mother of four arrived at St Joseph's in Phoenix, eleven weeks pregnant, with pulmonary hypertension. Her chart, as quoted in the moral analysis her hospital's parent later commissioned, put her risk of death at close to a hundred per cent if the pregnancy continued, and said surgery was absolutely contraindicated. The ethics committee approved ending it. Sister Margaret McBride, the hospital administrator on that committee, signed off.

Bishop Thomas Olmsted's public position was that an unborn child is not a disease. He declared that McBride had incurred automatic excommunication, and when the hospital would not back down he stripped it of its Catholic status. What I find most revealing is a line from his ultimatum letter, because it says out loud what the fight was actually about. It was not medicine and it was not doctrine. It was that the hospital had not acknowledged my authority to settle this question but had only provided opinions of ethicists who agreed with them and disagreed with him.

Olmsted never denied that Directive 47's exception exists. His finding was that this case did not meet it. And the tradition split down the middle on that: the Catholic Health Association backed the hospital within a day, the moral theologian it hired concluded the Directives had been followed correctly, and the bioethics centre that condemned the decision admitted it had never been given the clinical facts.

McBride's excommunication was lifted about a year later. The hospital announced it in an email and said nothing else, and no explanation has ever been published. So the record now shows a committee that used a written exception, a punishment for using it, and a quiet reversal that nobody will account for. If you are the next administrator on the next committee, what have you learned? I know what I would have learned, and it has nothing to do with what the rulebook permits.

The exception written into the statute

Texas is where you can watch this happen with the law in front of you rather than reconstructed from an inquest.

The Texas ban carries an exception in the statute itself. When doctors and patients went to court to ask what it covered, the state Supreme Court told them, unanimously, that a physician who believes the law will not let him perform a life-saving abortion is simply wrong in that legal assessment, and that the law does not require the woman's death to be imminent. Then it declined to say which conditions qualify, on the ground that courts do not issue injunctions based on lists of hypothetical future possibilities.

Six months earlier the same court had shown what that costs. Kate Cox was about twenty weeks pregnant with a fetus diagnosed with trisomy 18 and had had two prior caesareans. She won an order permitting a termination; the court stayed it the next day and then vacated it, and the reasoning turned on wording. Her doctor had stated a good-faith belief, and the statute asks for a reasonable medical judgment. The court did not hold that Cox failed to qualify. It held that her doctor had not said it in the statute's words. By then she had left the state.

And the doctors were not confused about the law, which is the detail that settles it for me. In a national survey of obstetricians, sixty-eight per cent of those working under bans said they understood very well when abortion was legal where they practised. Sixty-one per cent were worried about their own legal risk anyway. They knew what the rule said. They were frightened of being judged wrong afterwards, which is a different thing entirely, and no amount of clarifying the text touches it.

My own reading is that this is the most damning fact in the essay, because it cannot be blamed on anybody's theology or anybody's malice. The rule was clear. The people were competent. They still would not move.

The legislature eventually noticed. In 2025 the same body that wrote the ban passed a law explaining its own exception, spelling out that the risk need not be imminent and that a condition need not be actively injuring the patient, and ordering the medical board and the state bar to run free courses teaching doctors and lawyers how to read it. Then it added a section insisting all of this was merely consistent with what the courts had already said. Both things are true at once, and that is the whole problem: the exception was always there, and it took three years, several deaths and a training programme to make it reachable.

The same machine, without any religion in it

If you think this is about faith, look at sentencing, where nobody claims divine authority for anything.

On two afternoons in November 1995 Leandro Andrade walked out of Kmart with videotapes stuffed down his trousers, $84.70 the first time and $68.84 the second. He told the probation officer he was selling them for heroin and had been an addict since 1977. Under California's three-strikes law he got two consecutive terms of twenty-five years to life. He was thirty-seven, so his first parole date was at eighty-seven. I have read that sentence a dozen times now and it still will not go in.

That statute had two escape valves written into it. A judge could downgrade the offence, and a prosecutor could move to strike a prior conviction in the furtherance of justice. Justice Souter, dissenting, pointed straight at the second one and said that in Andrade's case the safeguard had simply failed. Congress did the same thing federally in 1994: it wrote an exception into its own mandatory minimums and everyone calls it the safety valve. By 2010 nearly half of eligible defendants were getting out from under the minimum through it. Black defendants reached it least often of all — eleven per cent, against twenty-seven for white and forty-three for Hispanic defendants.

An exception that exists and is handed out unevenly is not mercy. It is a lottery with a good name.

The single best document I found in all of this is a sentencing opinion. Weldon Angelos was twenty-four, with no significant adult record, when an informant bought marijuana from him three times at $350 a go. He had a handgun on him, and that made three firearm counts. The prosecution offered fifteen years for a plea and told him it would add counts if he refused. He refused; it added them. Judge Paul Cassell — a Bush appointee, a former Scalia clerk, a former prosecutor, nobody's idea of a soft touch — wrote that the sentence he was required to impose was unjust, cruel, and even irrational. Then he imposed it: fifty-five years and one day.

He also wrote to the jurors afterwards and asked what sentence they thought right. Nine answered. The median was fifteen years. Not one came close to what the statute required of him.

The bit I keep returning to is that Cassell said the extra counts existed only because the defendant would not plead. That is the machine, stated by the man operating it. And it is the same machine here at home: I live in a country that resolves four criminal cases in five by settlement rather than trial. A trial is not forbidden in Estonia. It is an exception you have to be able to afford, which for most people means it is not there. I would not fancy my own chances of using it.

The strongest thing anyone can say against me

Rules like these were not invented by fools, and I would be writing propaganda if I left this out.

In 1974 the federal judges of one American circuit ran an experiment on themselves. Fifty judges sentenced the same paper defendants from the same files. On one case the answers ranged from three years in prison to twenty years plus a $65,000 fine. In sixteen of the twenty cases they could not agree on whether the defendant should go to prison at all. That is what discretion looked like from the inside, and mandatory sentencing was the answer to it.

It gets sharper, and I have not found a way around this one. Upholding a death sentence in the face of statistics showing Georgia defendants were four times likelier to be condemned for killing a white victim, the US Supreme Court wrote that the power to be lenient is the power to discriminate. That sentence is aimed directly at everything I have argued, and I think it lands. Every door in a wall is a door somebody decides who walks through, and the people who get waved through tend to look a certain way and have a certain lawyer. Ireland's real exception before 2018 was a flight to England, and three thousand women a year took it, which is to say the exception was available to anyone who had the fare.

I do not think this refutes the argument, but it does change what the argument can ask for. The answer to a discretionary lottery is not an exceptionless rule, because that is not what anyone actually builds. The answer is an exception that is specified, cheap to reach, and safe to use.

What this looks like in my own trade

I run infrastructure other people's businesses sit on, and my trade has the same structure with the stakes turned down, which is a good place to see it clearly.

The rule is that you do not touch production credentials. There is a documented way to break it, named after the fire alarm: break the glass and you are in, past everything. What makes it work, in my experience, is not the door but the accounting around it. It raises an alarm. It gets reviewed by your peers the following week. You attach a ticket number saying why, and Google's security book, which is where this is written down most plainly, is honest about the failure mode: without cultural reinforcement, audits become rubber stamps and breaking the glass becomes an ordinary Tuesday.

The standards are worse than the practice. NIST wrote the control for this — an audited override of access control, for cases involving a threat to human life — and then put it in none of its three baselines, so nobody has to build it. Sixteen words in the American health privacy rules require an emergency access procedure and say nothing about reviewing it afterwards.

And I am not clean here. Part of my job is enforcing rules I did not write on people who never agreed to them: acceptable use, suspension, a lawful request for data that happens to live on a machine I administer. There is a sentence always available to me, and it is a good one. The terms say so. It is true and it is checkable and it answers whether I was entitled to act, and I have used it to avoid the question of whether I should have. Not once. As a habit, on ordinary days, because it works and because nobody at the far end can tell the difference between a man who has thought about it and a man who has a clause.

So here is what I actually think, and it is narrower than where I started.

Insisting that your rules have no exceptions is not the dangerous part, because you are almost certainly wrong about your own rules. Yours have exceptions. Somebody wrote them down. The dangerous part is the shape of the exception: whether anyone will say where it starts, how much it costs to reach, and what happens to the first person who uses it. Ireland had the door and never gave it a measurement. Phoenix had the door and excommunicated the woman who opened it. California had two doors and let a man go to prison until eighty-seven anyway. My industry wrote the door into a standard and left it out of every baseline.

The rabbis were right and they were right early. If you want people to use the exception, punishing hesitation has to be at least as credible as punishing the act. Otherwise everyone waits, which is a decision too, and the person it costs is never the one who made it.

Savita Halappanavar's doctors were waiting for her to get sick enough. She obliged them, and by the time she qualified it was too late to matter. The expert who reviewed the case for the inquest said the law opened at about half past six on the Wednesday morning, when the signs of severe sepsis were finally there, and that starting even then would not have saved her.

She died qualifying.