A speech-language pathologist finishes a bedside swallow assessment and writes a clear recommendation: this patient takes thickened liquids only, and nothing by mouth without supervision. The assessment was expert, the recommendation is correct, and it is about to fail, not because anyone disagrees with it, but because of how many people need it and how few will receive it.
Trace who actually depends on that recommendation. The nurse who sets up the meal. The aide who refills the water jug. The dietary office that builds the tray. The tray-line cook who plates it. The family member who arrives at visiting hours with a milkshake. The float staff who cover the weekend and have never met this patient. Six consumers of one piece of information, and if you map how the recommendation travels to each of them, you will usually find it reliably reaching two. The rest depend on a chain of manual re-entries and verbal relays, and every link in that chain is a place for it to quietly drop.
This is a distinct kind of failure, and naming it correctly is half the cure. The recommendation does not fail because it was wrong or because anyone ignored it. It fails by non-arrival. It is an information product, needed in many places at once, produced once and then expected to distribute itself through a hospital that has no reliable mechanism for fan-out. Call it what it is, a distribution problem, and the pointless fixes fall away immediately. Re-educating the SLP does nothing; the recommendation was fine. Reminding staff to check for swallow precautions does a little, and leans on the same overloaded vigilance that fails at the water jug at two in the morning.
The fixes that hold are single-source fixes. One authoritative field that every downstream system reads, so the tray card, the bedside sign, and the electronic record all draw from the same place rather than from six separate transcriptions. Bedside signage generated from that field, not handwritten beside it by whoever passed last. The aim is to make the recommendation arrive everywhere it is needed without depending on a person carrying it, because a person carrying it is exactly the link that breaks.
The wider point applies far beyond dysphagia. A great deal of clinical work produces information that many hands need and one hand writes. Allergy status, isolation precautions, code status, weight-bearing limits, all of them fan out from a single source to a crowd of consumers, and all of them fail the same way when the distribution is left to relay and memory. Whenever you find a recommendation that is correct and still not being followed, count the consumers before you blame the followers. The number that actually received it is usually the whole story.
This article accompanies the book Before Harm Happens.