My mother was found wandering a hospital parking lot, looking for her car. She had a urinary tract infection nobody had caught. Urgent care missed it, the hospital missed it, and she went home with severe delirium and a new diagnosis of mild cognitive impairment. It took two weeks, and a personal connection to two clinicians willing to look again, before anybody landed on the right answer.
Nobody in that story was careless. What it took was somebody willing to step back and ask whether the first answer was the right one, and that person does not exist by default.
This post is about why that job is so often missing. Not carelessness, and not quite cost either, though cost is how it shows up on a page. The job has no output anyone can see, because what it produces is failures that did not happen. On a page comparing line items, a cost with no visible benefit loses.
What that person produces
Nothing you can point to.
If someone had asked the question in week one, my mother would not have wandered a parking lot. There would be no story, no two months in our guest room, no year and a half of finding her way back. There would be a woman who had a UTI and got antibiotics.
That outcome leaves no record anywhere. No chart note, no billing code, no line on anybody's evaluation in March. The product of the job is failures that did not happen, and a failure that did not happen leaves nothing to look at, because the version of events it was prevented from becoming was never observed by anybody.
Work out what the question would have been worth. The infection itself was ordinary and clears quickly when it is caught. What the delay cost was two weeks of missed diagnosis, two months in our guest room, two moves, and a year and a half of a woman finding her way back to herself. All of that is real and none of it appears anywhere as the price of a question nobody was assigned to ask.
Failures are not like that. Failures generate paperwork the way successes do not. Two weeks of missed diagnosis is fully documented.
So the role gets measured entirely on its bad days. It has no good days that leave a trace.
An organization deciding what to cut is comparing line items. This one shows up as a cost with no matching output, and that is not a lapse in anybody's judgment. It is what the ledger actually says.
Where the role does exist, and why that is weaker than I wanted
If that is the mechanism, the job should appear wherever failure is expensive enough to justify a cost with no visible product. I have one case that fits loosely and one that does not fit at all.
Construction fits loosely. The standard form used on most American commercial projects provides that the contractor "shall be solely responsible for, and have control over, construction means, methods, techniques, sequences, and procedures, and for coordinating all portions of the Work under the Contract." Control and responsibility land on the same party in one clause, before anyone knows whether it will matter. But that clause names a party whose main product is a building, which is the most visible output in commercial life, rather than a staffed role whose entire output is invisible.
The one that does not fit sits in Why Nobody Pays for Coordination. Federal regulation requires every Medicare hospice to designate a registered nurse to coordinate the care and keep assessing what the patient and family need. That is this role, staffed, in the slowest and most distributed harm territory there is, which is where I just said it would be missing. It exists because somebody wrote it into a condition of participation. So the honest version is that the role appears where somebody wrote it into a rule, and expensive failure is what usually moves them to write it.
What I cannot show
If a prevented failure leaves no record, the benefit of the role leaves no record either. I have no way to distinguish a job that is undervalued from a job that does little. My mother’s case is one instance, and its counterfactual is mine. I am asserting that a second look in week one would have produced a urinary tract infection and antibiotics. No one can check that.
The two cases move the explanation elsewhere. Construction fits loosely, and hospice contradicts the prediction outright. The honest version is that the role appears where someone wrote it into a rule. Rule-writing does the explanatory work. Roles whose product is failures that did not happen do exist and persist in internal audit, safety, and engineered redundancy. Organizations buy prevention when the failure distribution is legible enough to argue about. The claim I can hold is narrower than the title.
The same ledger, in the discharge room
A prevented mistake leaves no record. A recommendation leaves a record that can be wrong.
In The Weakest Link there is a scene where a hospital case manager lays out four options for a family and stops just short of saying which one she would choose. The family reads her silence as a legal restriction. The rule is narrower than that. Federal regulation bars a hospital from steering a family toward a particular agency and says nothing about what kind of care somebody needs. Other laws bear on who may say what in that room, and silence is not permission, so I am not claiming she was free to answer. What the regulation establishes is that the rule is not what keeps the sentence unsaid.
What keeps it unsaid is that a recommendation can be wrong in a chart somebody pulls a year later, and a neutral list cannot.
Something to check
Think of the person in your organization who is only ever discussed after something goes wrong.
Then work out what their good weeks look like on paper.