Mistakes don’t happen for a reason, they’re usually the product of a string of them. The incorrect prescription is caught by the dispensing pharmacist or administering nurse; the wrongly prepared intravenous medication is countersigned before being given. These are part of the checks and balances modern medicine has developed to prevent errors, but despite this errors are thought to contribute to over 1700 avoidable deaths each year. When things do go wrong, they often make the news, and the resulting investigations often reveal a series of flaws that build into the critical event. The Swiss Cheese Model is commonly trotted out to explain how systems can breakdown, leading to a mistake.

Swiss Cheese Model of mistakes. Credit:[Ben Aveling - Own work, CC BY-SA 4.0](https://commons.wikimedia.org/w/index.php?curid=91881875)
Swiss Cheese Model of mistakes. Credit:[Ben Aveling - Own work, CC BY-SA 4.0](https://commons.wikimedia.org/w/index.php?curid=91881875)

The general gist is that a single error can be caught or mollified, but mistakes occur as a results of several coincident errors. Ensuring absolute unfailing perfection is unfortunately unrealistic, but we can try to effect systemic changes to make mistakes less likely and their consequences less grave. However, to do this, we need to know what actually happened, and regrettably medicine has long had a culture of denial and shame (see this talk by Brian Goldman). So consider what follows to be an honest example of how the system can break down despite best intentions, although some details are altered/obscured in the name of confidentiality.


To set the scene it’s a busy night in our local DGH (District General Hospital); there are several ambulances queueing up outside A&E, and our PACS system is down. PACS systems are the means we look at the various X-Rays, CTs, and MRIs we request for our patients. Smaller hospitals, such as ours, use them to send images to larger institutions that have 24 hour cover for specialist services like Cardiology, Neurosurgery, and Cardiothoracics. They’re also used to send images to outside radiologists to report if hospitals are not busy enough to warrant an on call radiologist on site overnight. So without our PACS system we’re diverting trauma and stroke patients to other hospitals because we wouldn’t be able to seek specialist input for them, and one of our radiologists has been in the building over 18 hours and is set up with a camp bed in their office to report urgent overnight scans on the machine that is physically plugged into the scanner.

We see patients in order of need, this is the source of perpetual ire in A&E, especially when the person who has been waiting half as long as you gets called through first. So when I see a gentleman labelled as “Acute Right Subdural” pop up on the medical list, I pick him up next. This is referring to a Subdural Haematoma, a kind of bleed between the brain and some of its coverings, that if it progresses untreated, can lead to permanent brain damage and potentially death. Suffice it to say, this man is in need of urgent care.

Example of a Subdural Haematoma on CT. Credit:[James Heilman, MD - Own work, CC BY-SA 3.0](https://commons.wikimedia.org/w/index.php?curid=19364350)
Example of a Subdural Haematoma on CT. Credit:[James Heilman, MD - Own work, CC BY-SA 3.0](https://commons.wikimedia.org/w/index.php?curid=19364350)

One of the luxuries of being part of the medical team, as opposed to A&E, is that someone has usually seen our patients first, whether that be A&E or a GP. On reviewing the existing notes I learn that this nonagenarian gentleman had a sudden onset of left sided weakness and slurred speech earlier this afternoon. The paramedics had phoned through to department that they were coming, so his head had been imaged early and reported as an acute right sided subdural haematoma.

As an aside, the nerves of sensation and muscle control cross the body in the spinal cord and base of the brain. Therefore damage to one side of the cerebrum (the bit that looks like a walnut) typically causes symptoms on the other side of the body. So broadly speaking, this man’s left sided weakness fits with his right sided subdural.

The team in A&E have already started the referral to the local neurosurgical services, from this referral I learn that our patient has had a previous unprovoked subdural 4 months ago, which was conservatively managed (read no surgery required), and a traumatic subdural (read bleed after banging his head) about 18 months ago. He had been on blood thinning medications for his atrial fibrillation (an abnormal heart rhythm that predisposes to clots and stroke), but these had been stopped after the first event because bleeding on the brain and medications that increase bleeding don’t mix well. In summary, this is an unfortunate chap, with an extensive neurological history.

The referrals system works a little like a forum thread, the specialist posts replies or asks further questions, and we can respond. At this point, we’re still unable to see images on any machine not physically plugged into the scanner, and we can’t transfer images to the specialist. As it is their decision whether or not to operate on this man’s head, the neurosurgeon wants to see these images, but in the meantime asks for some information regarding his functional state and clarification about his history. I think it’s partly often misunderstood by patients and family members, but if a doctor, surgeon, or anaesthetist starts asking you how far you can walk, they’re not trying to guilt you into doing more exercise. This information is used to estimate your physiological reserve, or how likely your body is to cope with acute illness, surgery, or escalation to intensive care.

So in my first mistake of the evening, instead of starting from the beginning, I ask targeted questions, trying to elicit answers to the surgeons questions. I learn that he lives with his equally aged wife, that he has no carers, and a handful of comorbidities that are well managed. He performs well in our usual tests for confusion, he knows the current Prime Minister, the dates of World War II, and his details. He got the name of the hospital wrong, instead picking the next town over, but give him the benefit of the doubt as probably no one told him which one the paramedics took him to.

He tells me he can walk about 10m before his legs start feeling a bit weak, but that he used to be better before his stroke a few months ago. I think to myself “Subdural vs Stroke”, it’s a complex area and he’s probably just a bit muddled, the A&E consultant and his discharge summary from a few months ago say it was a subdural. Mistake number two.

A discharge summary gets sent out to patients and their GP after each admission. It serves as a record of what happened during that visit, and as the primary means of communication to the GP if there’s anything they need to do as a result. His from several months ago is terse, and one doesn’t exist from his admission before that. To paraphrase, it says that he presented with left sided weakness, that his head CT showed a left sided subdural which was treated conservatively, that he made a good recovery, and that was about it.

Now back to our patient. On examination the left side of his face was drooping, he was having some difficulty getting his words out (known as dysarthria), and could barely move his left arm, although he could move his leg more than they’d given him credit for. In short, all his symptoms point to an issue with the right side of his brain.

I faithfully relay these findings back to the neurosurgeon, but doing my best to sell him as a potential surgical candidate if needed. However, we still can’t transfer images, though radiology have assured us they’re trying to fix the issue. As drilling into someone’s skull is a major undertaking, it helps to be prepared, so while we’re waiting for the images he wants to know more about the prior decisions not to operate when he’s had the previous subdurals. But as I’ve noted, his previous discharge summary was not helpful so I needed to look a layer deeper, and this is where everything began to unravel.

Most NHS hospitals, particularly smaller ones, exist in a limbo between paper and digital record keeping. Ours is no different. Some notes are scanned, but on the whole, it’s difficult to find notes from previous admissions quickly. What we do have are the imaging reports from previous admissions, and the record of his previous neurosurgical referrals.

From these I gather that yes, this man did have a traumatic left sided subdural haematoma about 18 months ago after falling from a ladder. He was fortunate enough to have no major symptoms, and so after discussion with the neurosurgeons they treated him conservatively. His blood thinning medication was stopped to prevent it getting worse, his head was rescanned a month later which showed a stable appearance, and that was that.

Fast forward about 12 months and he is admitted with an acute unprovoked left sided facial weakness. His CT head shows a left sided subdural haematoma, which was dutifully discussed with the neurosurgeons. They appear to mistake the previous episode as though it was only a few weeks ago, and suggest continuing to treat him conservatively according to that plan. A couple of days later and after a bit of back and forth it’s suggested that his left sided symptoms don’t fit with a left sided subdural. Sure enough a repeat scan shows evidence of a right sided stroke that’s barely visible on the original CT, and that the subdural is likely chronic from the previous year. His discharge summary from that admission says none of this.

Now back to the present. We have our patient presenting again with left sided symptoms, he’s had a stroke relatively recently, and is no longer on medication to reduce the risk of stroke. His head CT is reported as a right sided subdural, but we know that he’s had left sided subdural that’s been visualised on two occasions a year apart, but is not mentioned on his scan tonight.

The PACS system is still down, so the only way to see today’s images is directly on the machine controlling the CT, and not even the radiologist has access to images not taken in the last few days. I went to the radiographers, and begged them to show me the images. I am not a radiologist, my ability to interpret brain imaging is pretty much at the extent of ‘is there a big badness’ and what side is it on. The quick witted are probably ahead of me, and sure enough the subdural is visible on the left side of this man’s brain. We tell the radiologist, and sure enough the report is corrected, but that leaves us with an incongruent left sided stable pathology, with new left sided symptoms.

It appears we’ve fallen into the exact same trap as several months ago; that this man has likely had another stroke, and that we originally blamed his stable left sided subdural. After filming myself scrolling through the images and emailing it to the neurosurgeon because the image transfer still doesn’t work, our patient understandably has no imminent surgical indication.

Onto the treatment of stroke. There are two broad categories, ones that occur due to bleeding, and ones that occur due to blockage, known as ischaemic stroke. The treatment for ischaemic stroke boils down to thinning the blood, and potentially giving medications to dissolve clots if recognised within a certain time. However this man’s history of prior bleeds is a contraindication to most forms of this therapy. So after this long protracted rigmarole of misdiagnosis, none of it changes immediate management.

Unfortunately I do not know the end to this story, but we can say that he’s been the victim of a string of complications over a course of months.

  1. His initial diagnosis 4 months ago was wrong
  2. The neurosurgeons at the time mistook the chronology of the events
  3. His discharge summary at the time was wrong
  4. This lead to his referral to neurosurgery this night containing errors
  5. I began with a targeted history, instead of starting with a clean review of events
  6. I trusted a discharge summary over the patient on what had happened 4 months ago
  7. The radiologist couldn’t review prior images as the PACS was down
  8. The subdural was reported to be on the wrong side of his head
  9. We weren’t able to send images to the specialists in a timely manner

Quite frankly he is both exceedingly unlucky, and perversely lucky that he hasn’t suffered as a result.