Error & Discrepancy in Radiology
- An error is a miss or a wrong call; a discrepancy is just two readers disagreeing — and not every discrepancy is an error.
- Most misses are perceptual (you never saw it) rather than cognitive (you saw it but reasoned wrong). Both happen to everyone.
- Error in radiology is common, baked into the job, and largely systemic — blaming one tired human rarely fixes it.
- The grown-up response to error isn't shame; it's a blameless system that catches misses and learns from them.
Here's an uncomfortable truth they don't print on the recruitment posters: radiologists miss things. Not occasionally, not only the lazy ones — all of us, routinely, as a measurable feature of the work. The eye is a brilliant, deeply unreliable instrument, and we point it at thousands of images a day looking for a lump that might be the size of a grape seed. So this page isn't about whether errors happen. They do. It's about naming them honestly and building a system that catches them anyway.
Error vs. discrepancy — not the same word
These two get used interchangeably, and they shouldn't be.
A discrepancy is simply a disagreement: two radiologists look at the same study and report it differently. That's it. Sometimes one of them is wrong (that's an error), but sometimes the finding is genuinely ambiguous and reasonable people differ. Imaging is full of gray, literally and figuratively.
An error is a failure that, in hindsight, was a wrong answer — a missed cancer, a misclassified nodule, a fracture called normal. The cleanest way I've heard it: a discrepancy is two opinions; an error is the gap between an opinion and the truth.
Why the distinction matters: if you treat every disagreement as an error, peer review becomes a blame engine and everybody stops being honest. The goal is to surface the real misses, not to keep score on judgment calls.
Where misses come from: your eyes vs. your brain
Most diagnostic errors fall into two big buckets, and they fail in different ways.
Perceptual errors — you simply never registered the finding. The pixels were there; your visual system skated right past them. This is the majority of misses, and it's humbling, because it isn't about knowledge. It's like reading a sentence and not noticing the the doubled word. (Go back. It's there.)
Cognitive errors — you saw it but reasoned your way to the wrong conclusion. You spotted the nodule and called it benign. You found the fracture and attributed the pain to something else. This is where bias lives: anchoring on the first thing you see, satisfaction of search (finding one thing and stopping), and getting steered by an incomplete clinical history.
| Type | What failed | Classic flavor |
|---|---|---|
| Perceptual | You never saw it | Missed lung nodule hiding behind a rib |
| Cognitive | You saw it, misjudged it | "Satisfaction of search" — found the obvious thing, stopped looking |
| Communication | Right call, message lost | Critical finding reported but never reaches the team |
Satisfaction of search is the sneaky one. You spot the dramatic finding — the big mass — feel that little hit of got it, and your search quietly switches off before you've checked the rest. The second fracture, the missed second nodule: that's where it bites. A disciplined search pattern is the antidote — you finish the checklist even after you've found something.
A miss is not always negligence
This is the part that takes new readers a while to sit with. A finding can be retrospectively visible — painfully obvious once someone circles it — and still have been a reasonable miss given how it looked the first time, with no clinical history, buried in a stack of normal studies at 2 a.m. Hindsight is a liar; it makes everything look bigger and brighter than it was.
That's why a single miss, on its own, usually tells you very little about a radiologist. Patterns tell you something. Individual cases mostly tell you that humans were involved.
When you find someone else's miss, the professional move isn't a victory lap — it's a quiet, factual notification through the proper channel, with the patient's care as the only agenda. You will be on the other side of this exact moment, and sooner than you think.
Fixing systems, not flogging people
The mature approach to error borrows from aviation: assume competent people will still make mistakes, and design the system so those mistakes get caught before they reach a patient. This is the blameless mindset, and it's not softness — it's the only way to get people to report errors instead of hiding them.
Practically, that means a few things working together:
- Double reading and over-reading — a second set of eyes on high-stakes studies catches perceptual misses.
- Structured peer review — sampling cases to find and learn from discrepancies, with the focus on patterns and education rather than punishment.
- Closing the communication loop — making sure a correct finding actually lands. The best diagnosis on earth is worthless if it never reaches the team, which is why communicating critical results is its own discipline.
- Honest follow-through when a real error reaches a patient — that crosses into error disclosure, a deliberate, patient-centered conversation.
The single most useful reframe: error isn't a character flaw to be eliminated by trying harder. It's a baseline property of human perception under load. You don't beat it with willpower — you beat it with systems, second looks, and a culture safe enough that people actually report it.
If you remember one thing, make it this: the question is never whether radiologists err — they do, all of them, every week. The question is whether the system around them is built to catch the miss, learn from it, and protect the next patient. That system, not flawless vision, is what good radiology actually runs on.