The supplies arrived on a Friday night. My father-in-law was coming home to die, and his infusion care was going to be my mother-in-law's job. The hospital had trained her before discharge, in the compressed hour or two while everything else is also happening, and by the time she stood in her own kitchen with the boxes she had a fraction of it. Nobody from the supply company could come out until Monday.
I drove over and set it up and went through the training again, step by step, until she had it.
Everyone in that story did their job. The training happened, the supplies arrived, somebody came out, and nothing failed.
Here is the claim, and then what each part of this post is doing.
Information dies at a handoff because the receiving form decides what survives. Not because somebody was careless. If the form has no box for a fact, the fact does not travel, even when it was said out loud and heard.
That story is where I came to the question. It is not evidence for the claim, and I will say why before the end. The evidence is one study, and the illustration in the middle is invented.
The record was accurate
Somewhere in a discharge summary there is a line saying caregiver education was provided, and that line is true.
There is no line for whether she would still have it at eight o'clock on a Friday night in her own kitchen, alone, with a man dying in the next room.
A form is a list of the things the next person is expected to act on. By listing them it also announces, without saying so, which facts are not going to travel.
I wrote a book about how information gets lost between people and framed it with the children's game of telephone, where the sentence degrades and nobody in the line did anything wrong. That framing is right about the fact and wrong about the cause. It suggests the loss is noise, some fraction shaved off at random each pass. Losses at real handoffs are not random.
The thing you can say and still lose
The cleanest illustration I have is invented, and I would rather say so up front than let it pass as reporting.
Picture a plating shop. An operator finishing second shift tells the man coming on that number three tank has been running two degrees warm all night. Not faulted. Never tripped anything, held steady there, he had been watching it.
The form had a box for the temperature. The temperature was in spec, so he wrote the temperature. The incoming operator worked his shift off the form, because that is what the form is for, and when the finish came back wrong on Thursday nobody could reconstruct where it started.
Nothing failed in the conversation. The outgoing operator said the tank was running warm and the incoming operator heard him say it. The form had no field for a temperature that was in spec and drifting, so by Thursday the warning existed nowhere but in one man's memory of a two-minute exchange.
Why the family story is not the evidence
These are two different failures and I should not blur them.
My mother-in-law's problem was retention under pressure. She was told the thing and could not hold it, which is a real problem and a common one and not the problem this post is about.
The plating shop is the other kind, where the information survives the telling and dies for want of a field.
Only the plating shop demonstrates the claim, and the plating shop is a thing I made up in order to demonstrate it. That is the weakest joint in the argument.
What was actually measured
A program called I-PASS put a standardized handoff structure into nine pediatric residency programs in the United States and Canada, and measured before and after across 10,740 admissions, 5,516 of them before the change and 5,224 after.
Medical errors went from 24.5 per hundred admissions to 18.8, a drop of 23 percent. Preventable harm, meaning the subset of those errors that injured somebody, went from 4.7 per hundred to 3.3, a drop of 30 percent.
The obvious objection is that people who know a safety program has started go looking harder, and find more of what they are looking for. The study has an answer to that, and it is the best thing in the paper.
Adverse events that were not preventable did not change. They ran at 3.0 per hundred admissions before and 2.8 after. If the drop were an artifact of who was looking and how hard, that number should have moved too. It did not.
Two other things push the same direction. Every suspected incident was classified by two physicians who did not know which period it came from. And the case-finding included voluntary reports from residents and nurses, which a culture campaign would tend to increase, biasing the count toward finding more errors afterward rather than fewer.
What the study cannot establish
The intervention was seven things at once. A mnemonic, a two-hour workshop, an hour of simulation, a computer module, a faculty development program, observation tools for feedback, and a culture-change campaign, with the written handoff built into the medical record at seven of the nine sites.
Nothing in the design separates the structure from the training, and if the gain came from the training then my argument gets no support from this study at all. The authors say it themselves, twice. Bundling "prevents us from determining which elements of the intervention were most essential," and the design "precludes definitively establishing a causal link between implementation of the I-PASS Handoff Bundle and improved patient safety."
There was no control group. The comparison is the same sites before and after.
Six of the nine sites improved significantly. Three did not, and the authors say they do not know why.
The handoffs themselves took no longer, 2.4 minutes per patient against 2.5, and residents spent no less time with patients and families. Getting there was not free, though. The workshop, the simulation, the faculty program and the record build all cost something the study does not price.
So the narrow version is that seven things changed at once in nine teaching hospitals and errors fell by about a quarter, and the non-preventable rate says the fall was probably real.
Something to try
Find a form you hand off on. A shift report, a ticket template, a transfer sheet, a project closeout.
Then ask the last person who filled it out what they needed to say and had nowhere to put.
My mother-in-law got through that weekend because I live twenty minutes away and had set up equipment like it before.