Exercise · about twenty minutes
Reading a System-Drafted Note - Paediatrics II
A twenty-minute exercise for clinicians
The fourth in the series, and the second from paediatrics.
This note was assembled by a system from two sources: the audio of the consultation, and the triage entry typed by the nurse on arrival. That is how many of these tools now work. They do not only listen. They read what is already in the chart and fold it in.
The clinical decision in this case was sound. Keep that in mind as you read. The exercise is not about whether the right thing was done. It is about whether the record still shows why.
Part 1
The note
Paediatric emergency department. Assessment.
Infant aged 3 months, brought by mother. Three days of coryzal symptoms and cough, worse over the past 24 hours. Feeding reduced to around half of usual volumes.
No fever or cyanotic episodes.
Triage: heart rate 158, respiratory rate 52, SpO2 96%, temperature 37.4°C. Blood glucose within normal limits.
Examination: mild subcostal recession, no nasal flaring. Widespread fine crackles on auscultation, no wheeze. Well hydrated. Alert and responsive.
Impression: Bronchiolitis, mild.
Plan: Admit for observation. Continuous saturation monitoring. Nasogastric feeds if intake falls further. Parents informed and in agreement.
Signed: ____________________
Read it once at the pace you would read it on a busy shift.
Then read it again with the questions below.
Part 2
The questions
-
Is anything negated?
What is said to be absent, and is the negation doing only what the clinician meant it to?
-
Is anything hedged?
Where is uncertainty, and where has it gone?
-
Whose is it?
Does every finding belong to this patient?
-
When was it?
Now, or at some point before?
-
Which side?
Where it matters, is it there?
-
What does this number mean at this age and this weight?
And under what conditions was it taken?
And one more, for any note that draws on something typed in shorthand: what did the abbreviation mean to the person who typed it?
Write down what you find before going further.
Part 3
The key
Do not read this until you have written your answers.
A negation that reached too far
No fever or cyanotic episodes.
What the mother said was this: no fever. The cough has been there three days. And two days ago, during a feed, the baby went blue for a few seconds.
The clinician heard it, and it was the reason she admitted. The system heard a negative near the start of the sentence and applied it to everything that followed. The one positive finding in the history, and the most important, is now recorded as its opposite.
What follows from it.
In a young infant with bronchiolitis, a reported episode of cyanosis or apnoea is a red flag. It changes the level of care. Anyone reading this note tomorrow, at handover or on review, will see a baby with no such history. If she goes blue again on the ward, the record will say it is the first time.
This is the failure the series keeps returning to, in a new shape: not a word that was dropped, but a word that was stretched.
A number without its conditions
SpO2 96%.
In air, or on oxygen?
It was in air, and in air 96% is reassuring. On oxygen it would mean something entirely different. The note does not say, and a saturation without its conditions cannot be interpreted by anyone who was not there.
What follows from it.
This is the paediatric version of a problem that runs through every measurement in the record. A number is only as meaningful as the circumstances attached to it: what the child was breathing, whether they were feeding or asleep, how long the probe was on. The system carried the figure and dropped the part that made it readable.
A test that was never done
Blood glucose within normal limits.
No glucose was measured.
The triage nurse typed BS equal bilat - breath sounds, equal on both sides. The system expanded BS using the meaning that appears most often across all the text it was built on, which is blood sugar. Then it rendered that as a normal result.
What follows from it.
The record now contains a laboratory result that does not exist. It is plausible, it is normal, and it sits among other normal findings, so nothing about it invites a second look. If this infant later becomes unwell and someone asks whether glucose was checked, the record will say yes.
The obvious version of this failure is easy to spot: an abbreviation like MS expanded to multiple sclerosis in a three-month-old's chart would be caught at once. The quiet version is the one that survives - an expansion that is wrong for this patient and entirely normal for the page.
The age that does not tell the whole story
Infant aged 3 months.
She was born at 34 weeks. Her age from birth is three months; her age corrected for prematurity is closer to six weeks.
The clinician discussed it with the mother, and it was the second reason for admission. Prematurity and a young corrected age both raise the risk of apnoea and of a more severe course in bronchiolitis. The note records neither.
What follows from it.
An age with no gestation attached is a number without its denominator, in a specialty where the denominator is often the point. Every calculation that depends on this child's age - risk, dosing, what counts as a normal value - now starts from the wrong place.
Part 4
What to take away
Put the four together and look at what the note now describes.
A three-month-old with mild bronchiolitis. Normal saturations. No fever, no cyanosis, no apnoea. A normal glucose. No risk factors. And an admission.
The decision was right. The record no longer shows why.
The two findings that justified admission - the cyanotic episode and the prematurity - are both gone, one reversed and one omitted. What remains is a set of reassuring findings, one of which was never measured, followed by a decision nothing in the note supports.
Now think about who reads it next. A reviewer auditing admissions for mild bronchiolitis. A payer deciding whether the stay was necessary. A quality programme counting avoidable admissions. A model trained on this institution's records, learning that infants with this presentation are admitted for no documented reason. Every one of them reads the clinician's sound judgement as an error - and none of them can see that the error is in the note, not in the care.
Notice which of the four you found first. For many readers it is the glucose, because it is the one that is simply wrong. The negation and the prematurity are harder, because nothing on the page is incorrect. They are missing, and a missing thing looks exactly like nothing.
One question for your institution
If a payer questioned this admission, could your institution show what the clinician actually heard, and establish that the system - not the clinician - removed it?
The answer is diagnostic of a great deal else.
This exercise draws on Chapters 11 and 12 of Clinical Meaning in the Age of Health Data & AI, both published in full on this site. The clinical note is composite and illustrative, and is not drawn from any real patient record. It is an exercise in reading records, not a guide to clinical management.
Clinical meaning is not generated by AI. It is captured, governed, and made computable.