91av

A 49-year-old man is admitted to hospital in Japan with chest pains and a partially paralysed arm. Doctors diagnose a simultaneous heart attack and stroke and the patient seems to respond well to treatment. The next day, however, he has a cardiac arrest, and later dies. The autopsy reveals that all along he’d had an aortic dissection, a tear in the lining of the major artery from the heart.

In the US, a previously healthy and active 79-year-old man is found confused and incapacitated. He is diagnosed with pneumonia and dehydration, and after treatment seems to be recovering well. After three days he starts breathing rapidly and his condition declines. Six days after admission he dies. The autopsy reveals rampant TB.

A 37-year-old woman who is six months pregnant is admitted to hospital in Italy with severe abdominal pain. The pain is attributed to kidney stones and after treatment the woman goes home. One week later, she vomits and loses consciousness, and despite doctors’ best efforts, she and her baby die. The autopsy reveals massive internal bleeding caused by a rare blood disorder.

It is no secret that doctors occasionally kill their patients instead of curing them, whether by failing to wash their hands or prescribing the wrong drug. In many countries, serious efforts are now being made to reduce medical errors. The focus, though, is almost entirely on avoiding mistakes in treatment, rather than in the original diagnosis.

But as the cases above illustrate, major mistakes in diagnosis do happen, and they are surprisingly common. The causes range from medicine’s inherent limitations, through flaws in hospital systems, right down to individual doctors seemingly forgetting what they learned in medical school. It is estimated that as many as 1 in 20 patients who die in hospital, do so because their illness was misdiagnosed.

Shockingly, our best way of uncovering diagnostic errors – the autopsy – is in steep decline. If no one suspects a wrong diagnosis, the evidence will be buried or cremated with the body, and nobody will be any the wiser, so there is nothing to stop the same mistakes being made over and over again. “Diagnostic errors do not receive the attention they deserve,” says Kaveh Shojania of the University of Ottawa in Canada, who studies medical errors. “It is a big part of the problem.”

The value of autopsies was gradually established during the 18th and 19th centuries. Today they remain the gold standard as a way for doctors to identify and learn from their mistakes. It is much easier to find out for sure what was wrong with someone after their death, when pathologists can cut open the body, examine any part in detail and take samples for testing (see “Anatomy of an autopsy”).

Suspicious circumstances

Some autopsies have to be done for legal reasons. These forensic or coroners’ autopsies are often required after violent, accidental or suspicious deaths, or where the cause is unclear. In many countries autopsies must also be carried out on patients who die during surgery or within 24 hours of admission to hospital. Sometimes, however, doctors just want to know more about why someone died; for these hospital autopsies doctors usually need permission from the next of kin.

It was in 1912 that a Harvard University doctor called Richard Cabot did one of the first large studies comparing hospital autopsy results with the initial diagnosis. After looking at 3000 cases, he concluded – to the stunned disbelief of his colleagues – that nearly half the diagnoses had been wrong.

At first glance, today’s performance seems little better. A recent review of all the studies like Cabot’s done since the 1960s concluded that the certified causes of death were wrong in at least a third of cases. Not all the errors would have affected survival (though they still matter, as health policies are often based on death certificate statistics), but some would. At least 10 per cent of autopsies show patients might have lived had their diagnosis been right.

You do need to treat these figures with caution. It is impossible to work out the true misdiagnosis rate for all patients, not least because autopsies are obviously not done on people who survive. Plus the rate of mistakes may appear artificially high if doctors are now requesting a hospital autopsy only if they suspect something went wrong.

To find out the true misdiagnosis rate, Shojania analysed 53 studies published over the past four decades, involving more than 13,000 autopsies in North America, Europe and Australia. Crucially, he took into account the falling autopsy rate and the possibility that autopsies were more likely if misdiagnosis was suspected.

In a paper published in 2003, his team concluded that the accuracy of diagnoses has been improving steadily, with the rate of major discrepancies affecting survival falling by a third each decade (Journal of the American Medical Association, vol 289, p 2849). Even so, the rate remains shockingly high: at least 4 per cent of all US patients who die in hospital might have survived had their diagnosis been right. The figure is higher in other countries. “It’s a big deal,” says Shojania.

So why are mistakes still so common? Even in today’s era of high-tech medicine, some errors are inevitable. Doctors have limited knowledge, limited tools and limited time to make a diagnosis. Even well-studied diseases can produce strange symptoms unlike those in the textbooks, and patients can have several diseases at once. For example, the Japanese man with aortic dissection had an extremely rare collection of symptoms for such a case. It is sometimes impossible to work out what is wrong with a patient while they are alive, and not always possible when they are dead. “It’s a miracle how often doctors get it right,” says Mark Graber of the Veterans Affairs Medical Center in Northport, New York.

Lost X-rays

The crucial question, then, is not how many deaths are due to a major misdiagnosis. It is, how many can be avoided? “Half are preventable and half are not,” Shojania suggests. He was the only doctor 91av asked who was prepared to make an estimate. The few studies to have investigated the causes of misdiagnosis suggest even more than half are preventable.

One such study, carried out by Graber and published last year, analysed 100 cases where diagnostic mistakes had injured a patient or led to their death (Archives of Internal Medicine, vol 165, p 1493). Its findings may not be widely representative, as it identified cases through voluntary reports and other methods as well as autopsies, but it does at least start to give us some idea of why errors are made. What Graber found is that it is typically not just one thing that goes wrong, but five or six.

A recurring theme was system failures at hospitals, such as X-rays getting lost, or a lack of qualified staff around on a holiday evening. An even more important cause of error was mistakes by individual doctors. These ranged from lack of medical knowledge to using flawed reasoning to reach their diagnosis. The commonest error of this sort is “premature closure”: a doctor arrives at a diagnosis that seems to fit the facts, then stops considering other possibilities. “When you come up with an answer, you are happy,” Graber says. “You stop thinking about the problem.”

Some doctors gave every sign of sheer incompetence, such as failing to pass on test results or even skipping parts of a physical examination. One failed to notice that a patient’s toes were gangrenous.

Just 7 out of the 100 misdiagnoses were identified as “no-fault” errors that staff had no part in. Some of these cases involved a disease presenting in an unusual way. Some patients missed their hospital appointments or told lies to their doctors. A case of AIDS went undiagnosed because the patient did not tell his doctors he had engaged in high-risk sex. More often, however, patients are just not very good at telling doctors what they need to know to make an accurate diagnosis.

Whatever the cause of misdiagnoses, nothing can be done about them if they are never discovered. And the only sure way to detect more diagnostic errors is to do more autopsies. Of course, they cannot help the patients in question, but they can help correct whatever it was that led to the error, be it bad organisation, flawed reasoning or faulty equipment. “No lesson is as powerful as seeing your own mistakes,” says Graber.

But systems for alerting doctors to their errors tend to be patchy and unreliable. Who wants to tell a colleague that they got things horribly wrong?

And the number of hospital autopsies continues to fall in most countries despite repeated calls to reverse the trend. “There are constant pleas, but it’s not happening,” says Graber. “It’s a losing battle.” In the 1960s, an autopsy was done on around 60 per cent of patients who died in hospitals in Europe and the US. Today the rate of hospital autopsies is thought to be less than 10 per cent in Europe, and less than 5 per cent in the US.

Why? “Clinicians don’t think it’s necessary any more,” says Shojania. “It’s no longer part of training.” Cash-strapped public healthcare systems often decide the money is better spent elsewhere, and private hospitals cannot charge relatives for autopsies so they have little incentive either.

“One doctor failed to notice that a patient’s toes were gangrenous”

Another possible cause is increasing fear of litigation. Some argue that such worries are groundless. A recent study of US appeals court records showed that the crucial factor in law is not whether an autopsy reveals a discrepancy, but whether the misdiagnosis was due to negligence. “It is not necessary to be right,” says lead author Kevin Bove of Children’s Hospital Medical Center in Cincinnati. “You just have to do the right thing.”

But others argue that doctors may not request autopsies in cases where they suspect they could be held liable for negligence. “If you say ‘don’t worry, you will never get sued’, that’s just not realistic,” says Lee Goldman, now at Columbia University in New York City. “You have so much selection over autopsies that of course no one gets sued.”

Then there is the issue of getting consent from relatives. In the UK, there was public revulsion in 1999 on the discovery that a pathologist at Alder Hey Children’s Hospital in Liverpool had stored thousands of organs from children’s autopsies without their parents’ knowledge. There have been similar public outcries about stored organs in Australia and Ireland. “Some doctors are now frightened to ask for consent,” says Emyr Benbow, a pathologist at the University of Manchester in the UK.

An audit at University Hospitals of Leicester before and after the Alder Hey scandal revealed the hospital autopsy rate had dropped from 10 per cent to less than 1 per cent. The main cause was not that relatives were refusing consent; it was that doctors were less likely to ask for it.

So what can be done to change matters? “There should be a minimum autopsy rate, a requirement for feedback and doctors should not be subject to malpractice [lawsuits] if they do an autopsy,” says Goldman.

Such protection from lawsuits would be unlikely to go down well with an increasingly litigious public. But if doctors keep quiet about misdiagnoses, as may happen now, there is no chance of improving matters. “When you are not happy with what you are getting from people who want to do their best, the system is all messed up,” Goldman says.

Ideally, autopsies would be carried out on a random sample of people who die. In the US, hospitals once had to have a minimum autopsy rate of 20 per cent, but this was abandoned in 1970. In the UK, the Royal College of Pathologists once considered trying to push for a minimum 10 per cent random autopsy rate, but the Alder Hey scandal kicked the idea into touch. “There would be a substantial outcry,” says Benbow.

Nearly a century after Cabot’s 1912 autopsy study, it seems we have forgotten his most valuable lesson. But you can do something about this. If one of your family dies and a doctor suggests an autopsy, give permission. Or consider requesting one yourself.

It will not help your relative, but it might help save someone else’s life. And since there’s a fair chance of you succumbing to the same illness as your father, mother, or siblings, that someone might even be you.