She is not herself

Martha Mills was thirteen when she died of sepsis in 2021, after a pancreatic injury from falling off her bike. Her family raised concerns about her condition repeatedly and they were not acted on. A coroner concluded she would probably have survived had she been moved to intensive care earlier. Her mother campaigned, and the result is a change to how English hospitals are required to work.

The thing worth extracting from that, as an operator rather than as a reader, is not the tragedy. It is the mechanism, because the mechanism is general and it is not about anybody being careless.

A hospital watches for deterioration with an early warning score. Observations are taken, points are assigned to each one for being outside a normal range, the points are summed, and a threshold triggers escalation. It is a good instrument. It is also, by construction, a comparison against a population. It asks whether this respiratory rate is abnormal for a human being.

The family is running a completely different instrument. They are not comparing the patient to a population. They are comparing this person to this person, against a baseline of years, and they are reading things no monitor records: the phrasing, the eyes, the way they responded to a joke, whether they are lying the way they usually lie. When they say she is not herself, they are reporting a deviation from an individual baseline, which is precisely the thing a population instrument cannot see.

Two instruments, two different detection profiles. They are not rivals and the family one is not softer or less scientific. It catches the patient whose numbers are still inside the range and who has nonetheless changed, and the published data now says how often that happens.

Before the policy, the sampling rates, because they decide everything else.

A nurse on a ward is not watching one patient. Divide a realistic share of the shift spent at a bedside by a realistic number of patients and the attention available to any individual is a few minutes an hour, distributed in fragments, mostly while doing something else. That is not a criticism of nursing. It is a ratio, and it is the ratio the establishment was funded for.

The relative in the chair is watching one patient, continuously, for as long as they are allowed to stay. The difference is not marginal. It is more than an order of magnitude, and it arrives with a decade of baseline attached at no cost.

Which reframes the whole subject. The question is not whether to be nice to families. It is whether a hospital intends to use the highest frequency monitoring in the building, or to leave it uninstrumented in a corridor outside visiting hours.

Three components, rolled out across English NHS sites from 2024. Patients are asked at least daily how they are and whether they feel better or worse, and the answer is recorded. Any member of staff can request a review from a different team at any time if they are worried. And that same escalation route is always available to patients, families and carers.

The standard objection is immediate and it sounds reasonable: open an escalation route to frightened relatives and you will drown in calls about nothing, exhaust the team that answers them, and end up less safe rather than more.

The published data does not support it. Across nearly two years, a meaningful minority of calls concerned acute deterioration, and of those, six in ten led to a change in treatment and more than one in eight to a transfer into high dependency or intensive care. Those are not anxious relatives. That is a diagnostic yield most investigations in a hospital would be pleased with.

And then the figure that settles the argument made two chapters ago. More than four in five of those deterioration calls would not have triggered an alert from the early warning score on its own. The family instrument is not duplicating the clinical one. It is finding a different population, and it is finding them before the score does.

It is worth being precise about what this does and does not prove. It shows the calls are clinically substantial and largely not redundant. It does not on its own prove lives saved, and the honest position is that the mechanism is well justified by yield rather than by a mortality figure it is too early to have.

Stop treating this as a healthcare policy for a moment and it becomes something this practice has seen in factories for decades. Anybody on the line can stop the line. The cord is available to the least senior person present, the response is immediate and comes from somebody with authority, and pulling it is not a mark against you.

That matters because the failure modes of an andon cord are thoroughly documented, and they are all cultural rather than technical. A cord nobody pulls is a decoration. A cord that is answered slowly is worse than no cord, because it has spent the credibility as well. A cord that gets you a sigh is pulled once. And an organisation that congratulates itself on how rarely the cord is pulled has, without ever deciding to, built a system that punishes pulling it.

Four conditions. Read the last column of each, because three of the four fail invisibly, and the one that fails visibly is the response time, which is why it is the one worth publishing.

Here is the part that is genuinely difficult, and it is the reason this belongs in a study about the relationship between a family and a provider rather than in a patient safety appendix.

An escalation route that goes round the bedside team exists because the bedside team might not listen. That is its purpose and it cannot be dressed up. So every call is, structurally, a statement that the people at the bed were not sufficient, delivered to their colleagues, about them, while they are standing there.

Follow that through. A team that feels undermined becomes defensive. A defensive team listens slightly less well to the next family, partly out of self protection and partly because every concern now arrives with an implied threat attached. Listening slightly less well increases the need for the cord. The cord gets pulled more. The team feels more undermined. It is a reinforcing loop and it runs in the wrong direction, and none of the people in it are behaving badly at any point.

There is one way out and it is entirely about framing, which makes it sound weak and is not. The call has to be a normal event rather than an exceptional one. If a ward gets several a month, each one is a review. If a ward gets one a year, that one is an accusation, and everybody on the ward will experience it as such no matter what the policy document says. Volume is what makes it safe, which is the exact opposite of the instinct every manager will have when they see the number.

Which produces the single most important rule on this page, and it is a rule about measurement rather than about care. Never appraise a ward on how rarely the cord is pulled. The moment call volume appears as something to be minimised, the wards will minimise it, nobody will announce that they are doing so, and the dashboards will show an improvement.

The same argument applies inside the establishment, and more sharply. A junior nurse escalating past a registrar, or a registrar past a consultant, is doing the same thing across a steeper gradient and with a career attached. Martha's Rule includes staff for exactly this reason, and the staff side is harder than the family side rather than easier.

Any mechanism that requires somebody to initiate selects for the people willing and able to initiate. That is not a flaw in this particular design, it is a property of the category, and it is worth stating plainly because the consequence runs opposite to the intention.

Calling for a review over the heads of the team treating your mother requires you to believe you are entitled to, to have the words, to be confident that it will not rebound on her care, and to be willing to be thought difficult by people you need. Confidence, language, education and status all bear on every one of those. The families who possess them are already the families whose concerns get taken seriously first time, and the families who do not are the ones whose concerns were discounted in the first place.

So an escalation route on its own widens the gap it was built to close. The confident get a second instrument. The rest get a poster.

Which is why the first component of Martha’s Rule matters more than it looks, and why it is the one most likely to be quietly dropped as an administrative burden. Asking every patient and family, every day, as a matter of routine, how they are and whether they are better or worse removes the need for courage entirely. Nobody has to decide to escalate. Somebody comes and asks. It is the only part of this design that is equity preserving by construction, and it is the part that does not appear in any dramatic account of the policy.

Escalation is one thread of a much larger relationship, and it is worth setting the whole of it out, because a hospital that gets the others wrong will not get this one right.

Seven roles. Read the middle field of each, which is what the hospital actually provides for that role. The pattern is stark, and the seventh is the one to finish on: the family is thoroughly instrumented, with a named department, a published procedure and a legal framework, at precisely the point where it is too late to help the patient.

Everything upstream of that, where they could help, is provided for with a chair and goodwill.

All of the above is policy, and policy is the cheap tier. This chapter is the expensive one, because the family’s usefulness as an instrument turns out to be largely a property of the floor plan, and the floor plan closes at the second gate.

Start with the obvious. If there is nowhere for a relative to sit for fourteen hours, they will not sit for fourteen hours. They will go home, and the highest frequency monitoring in the building goes with them. A chair that can actually be slept in, next to every bed, is not hospitality. Measured against what it detects, it is the cheapest piece of monitoring equipment a hospital can install. And it is not free, because it is bay width, and bay width is floor area, and floor area is fixed at massing.

Then the rest of it, all of which is space. A family room on every ward, because the conversation about the scan result happens somewhere, and if no room exists it happens in a corridor, which is a decision the building has made on the clinician’s behalf. Overnight accommodation for a parent, or a paediatric ward loses its best observer every night at the hour when deterioration is least well covered. Somewhere to wait near the emergency department and near intensive care that is not a corridor with chairs against a wall.

Then the parts nobody writes down. Wayfinding, because a family who cannot find the ward is not at the bedside. Parking and the journey, which is the catchment constant arriving in a different form. Mobile coverage and network inside a deep plan concrete building, because an escalation route that is a telephone number is worthless in a dead spot. And a power socket at the bedside for a relative’s phone, which costs almost nothing, is omitted routinely, and is the difference between a family who can stay and be reachable and one who leaves to find a charger.

Every item in those two paragraphs is floor area, services or position. Every one of them is cheap on a drawing and unavailable afterwards at any sensible price. Which means the decision about whether this hospital will use its families as a safety instrument is not taken by a clinical director in year six. It was taken at massing, by somebody who was probably thinking about bed numbers, and it is now a constant.

One honest qualification before the conclusion. Martha’s Rule is an English NHS policy. A private hospital in Lagos is under no obligation to operate it, and no regulator there will ask.

Three reasons to adopt it anyway, and the third is the commercial one.

The group’s existing hospital is in London and is inside the policy, so the operating standard exists already and the cost of writing it twice is not the cost of writing it once. A standard that only applies in the jurisdiction that compels it is not an operating standard, it is compliance, and this group’s product is supposed to be performance that travels between sites.

The failure mode the rule addresses is worse, not better, in the target market. Fewer staff per patient, families already acting as unpaid logistics and therefore already at the bedside, and a steeper hierarchy gradient between junior staff and consultants than exists in the hospital the policy was written for.

And the third. The family testimony on this site lists six things that would keep a Nigerian family from flying abroad, and two of them are answered directly by this: tell me what happens when it goes wrong, and tell me who to complain to. A published escalation route, available to families, with a stated response time and the call volume reported annually, is a verifiable commitment of exactly the kind that page says is missing from the entire market. It is not a compliance cost imported from London. It is a trust instrument, and trust is the binding constraint on the whole demand case.

Four, and the first is the one that will actually happen if nobody is watching for it.

The last one deserves its own sentence because it is the one this study is least comfortable with. In a self pay market with litigation attached, a careful record of a family raising a concern that was not immediately acted upon is exactly the document a claim is built from, and every clinician writing it knows that. The pressure to record less, and less specifically, is real, it is never stated out loud, and no policy removes it. The only counter is an organisation that visibly treats a well documented escalation as evidence that its system worked, and that has to come from the top, repeatedly, for years, before anybody believes it.