Death and the devil in the detail

Do death certificates really capture all the information we need to accurately inform public-health policies?

---

As lists go, they don’t come any more extensive. But then it has taken the best part of 125 years to accumulate. What started with 161 entries in the 1890s has expanded to upwards of 14,000 today.

The list in question is the International Statistical Classification of Diseases and Related Health Problems, more commonly referred to as the ICD-10.

Most GPs will be familiar with the content.

In it, every conceivable disease or condition known to man is listed, detailed and individually coded in what has become, since its inception in 1893, a mammoth medical directory overseen since 1948 by the World Health Organisation.

It is an illuminating illustration of just how much we have learned about the human body, and what can go wrong with it.

The ICD-10 – the figure relates to its number of revisions, with an ICD-11 update due this year – is, among other things, the set of medical classifications used to code and standardise cause of death around the world.

It is also the compendium mined by the Australian Bureau of Statistics (ABS) in its annual examination of what kills us here, in Australia. Through sophisticated technology, the ABS assigns an ICD code to the reasons behind every death and uses it to analyse our major causes of mortality.

There are, approximately, 50,000 “coding rules”, all designed to unravel information and identify underlying causes of death. What emerges from the data shapes public health planning and policy.

Yet despite the head-spinning mass of information in the ICD, the exhaustive coding combinations and the resulting forensic analysis of our deadliest conditions, diseases and illnesses, the “system” inherently relies on something far more rudimentary; the manual completion of that most final of forms – the death certificate.

This is where the collection of death data starts. Without it there is no system at all, no way of recording who died, let alone who died from what.

As drilled home in numerous manuals which offer instructions on how to properly complete the certificate, it is only through comprehensive and accurate information on cause of death that we can obtain reliable and worthwhile data. And that responsibility falls squarely on the medical practitioner writing the certificate.

But this thirst for knowledge, this quest to delay the inevitable for as long as possible by deciphering what is killing us, and taking steps to eradicate it, is far from flawless.

Last year, the University of Melbourne published a booklet in collaboration with Bloomberg Philanthropies and the Data for Health Initiative titled Improving Cause of Death Information.

In it, the authors suggest inaccurate or incomplete information in too many certificates present problems for health bodies.

“The quality of these statistics depends on the accuracy with which individual doctors fill out death certificates,” the authors state. “Unfortunately, the accuracy is poor in many countries. This reduces the quality of national and international mortality statistics and limits their value for health planning and policy.”

The oft-cited assessment in Australia is that roughly 30% of death certificates contain some form of inaccuracy. In the UK, an audit published in the Journal of Clinical Pathology several years ago found as many as 50% contained errors or sub-standard prognosis.

An Information Paper periodically drawn up by the ABS also spells out the importance of cause of death data in shaping public health policy, and emphasises the decisive role played by medical practitioners.

Statistical data obtained from certificates, it warns, “is only as accurate and complete as the information on the actual certificate”.

Unfortunately, the detail required for certain groups of diseases and conditions is “often lacking”, it adds.

All of which leads us to a fundamental and troubling question. If detail on these final forms is absent, if the cause of death is not clearly articulated – or is simply incorrect – how can we trust the annual ABS statistics which have such a role in underpinning public health policy?

Critically, is our number one killer, heart disease, really that? Or are we missing the real root cause of many deaths because of the way death certificates are written?

Mortality statistics are, in the main, based on the underlying cause of death which is determined by the sequence of events which led to a person dying. In Part 1 of the certificate’s cause of death section, doctors are told to first enter the “disease or condition relating directly to death”, and not only – these words are spelt out in block capitals on the certificate – the mode of death.

Heart failure and asthenia alone do not cut it, not does frailty or respiratory failure.

This immediate, or direct cause of death, should appear on the first line, 1a. Three further lines are designed for the antecedent cause of the condition above, so 1b should be the cause of 1a and so on.

The final line –  1d if there are four events in the sequence leading to death – should be reserved for the underlying cause or, as the ASB states “the disease or condition which initiated the train of morbid events leading directly to death”.

James Eynstone-Hinkins, director of health and vital statistics at the ABS, told The Medical Republic the manner of data collection, and the information supplied on death certificates, worked well in “lots of cases”.

But he acknowledged there were pitfalls.

“Certification is really important, and we need to get better at it. We are never going to argue with that,” he said.

“Sometimes we see doctors put something that can’t have caused some of the conditions in front of it.

“But in those cases, even if the sequence that led to death isn’t quite right, we have 50,000 coding rules in the decisions tables that help us deal with that and enable us to come up with the appropriate underlying cause of death.

“Can we rely on the stats? Hand on heart, I say we can. We have one of the best cause-of-death data sets in the world. But just because we can rely on them doesn’t mean it can’t be better.

“I want to see as much as possible written on death certificates. A doctor’s best opinion is still our best way of getting information about deaths. The more detail they give us, the more we know and the more can be done with it.”

Yet death, as any GP will tell you, is often riddled with complexity, rendering accurate completion of the death certificate often tricky, sometimes all-but impossible.

Another guide, the pragmatically titled How to Complete a Death Certificate, was penned for GPs by Sara Bird, manager of Medico-Legal and Advisory Services at medical defence organisation MDA National.

Bird, a GP by training, said medical practitioners can be confronted with unenviable circumstances where they must make professional judgments with very limited knowledge of a patient or their medical background.

She cited a recent example where a GP sought advice from MDA.

“It concerned a very elderly patient, close to 100, who died in a nursing home. The GP rang us and said ‘I have no idea what to write on the death certificate’,” Bird told The Medical Republic. “What do you put? It was one of those cases where you want to write ‘old age’ but you can’t put that on a death certificate.

“I’m sympathetic to doctors in those situations. The GP is always going to be challenged by ill-defined deaths, when you are trying to come up with a disease or condition that directly led to death.”

Bird continued: “It’s all very well to say they need to be more accurate, but in some situations it must be tempting to write something like heart failure. You can’t send in a blank death certificate.

“You just have to write what is the probable cause of death. You’re not purporting to say with 100% certainty, but you have to be comfortably satisfied.”

This example, and there will be several like it, is one that raises the specific question of heart-death data and whether it could be over-represented.

In cases where multiple, complex conditions are present, where there is very limited or zero medical information, or simply where the medical practitioner does not, for whatever reason, give the form as much attention as the ABS would like, could there be a tendency to conclude the heart was the catalyst for death?

Of further consideration is that responsibility for many death certificates invariably falls to medical interns or junior doctors in palliative care wards. Aside from their lack of experience, it would not be unfair to suggest their priority may not lie in exploring the medical history of a newly-deceased patient in order to meet the requirements of the ABS.

That same argument, in fact, could be applied to any medical practitioner charged with completing the certificate.

“Playing devil’s advocate, if you weigh up your day in that setting, do you want to spend more of your time looking after patients and saving lives, or ensuring the ASB data is correct?” Bird said.

“I would imagine that goes through the minds of junior doctors in training or interns when they are writing a death certificate.

“It’s also difficult for doctors in hospital because, unlike GPs, they don’t have that longitudinal history [of a patient] and certainly won’t necessarily know all the significant conditions that may or may not have been listed on the admission notes.”

In a palliative-care setting that became even more pertinent, she said.

“Unless the doctor writing the certificate takes the trouble of contacting the patient’s GP to get their input, there will be limited medical history,” Bird added.

“In an ideal world they probably should, but I can’t imagine that happens very frequently.”

Eynstone-Hinkins acknowledged the potential issue with junior doctors in settings where in-depth knowledge of the patient could be absent. But he put his faith in “any doctor” to come up with a probable cause, without resorting to generalisations.

“We know that in some cases the junior doctors are left to fill out death certificates, the degree to which something might get skewed by that … I would hope that any doctor can pick up the clipboard at the end of the bed and see what the person has been admitted for and still give us a reasonable estimation of the cause, even if we don’t get quite as much of the medical history.”

Studies in the US have certainly given credence to the argument that death certificate information can be unreliable and skewed towards common causes of death.

An excellent analysis of death certificates in the New Yorker magazine (which drew the conclusion that the question of what kills us is incredibly complex) pointed to The Framington Heart Study in Massachusetts which found death certificates over-stated coronary heart disease by 24%.

According to pathologist David Ranson, deputy director and head of forensic services at the Victorian Institute of Forensic Medicine, it’s a hypothesis not without merit. “It’s an interesting question and there is some evidence to say that occurs and people default to heart disease because it’s the easiest option,” he said.

“That said, many people who die in later life do have some form of heart disease. And you have to ask yourself, how much of a skewing or error in those statistics would have a significant impact on healthcare planning?

“Statisticians who have done work on this believe that a significant number of death certificates could be wrong, but not lead to a significant skewing of healthcare data planning.”

While heart failure in isolation is a frustratingly vague and unsatisfactory term as far as the ABS is concerned – it is mode of death rather than a condition – there are certain, albeit limited circumstances, where it could be a credible and reasonable conclusion to draw.

In addition, where such sparse information is supplied, no assumption is made by the ABS that the heart failure was associated with heart disease.

In 2016 for example, (the most recent statistics available) ischemic heart disease was the leading cause of mortality, accounting for 19,077 deaths, 12% of the 158,504 total.

Yet heart failure and complications and “ill-defined heart disease” sat ninth, accounting for just under 3400 deaths.

“A lot of people would say heart failure is a poorly described cause of death and we are not going to disagree with that,” Eynstone-Hinkins said.

“That said, if someone dies aged 98 and the doctor writes heart failure on the death certificate, they are probably referring to the fact the patient’s heart had been failing for several years and had eventually given up. That’s not always the case, but you need to take into account the age structure around death where people who are very elderly are dying often of multiple chronic conditions.

“The concept of an underlying cause is perhaps a lot easier to apply when someone dies prematurely from something.

“It’s harder when the person has died from a combination of vascular degeneration, organ failure and all sorts of things that are happening concurrently.”

As with most things surrounding the gathering of mortality data, nothing is as straightforward as it seems.

While detecting the underlying cause is the focus, WHO’s ICD coding rules are also designed to capture the most “pertinent condition” on a death certificate, even if, somewhat confusingly, that overlooks the very underlying cause it is generally seeking to identify.

That, Eynstone-Hinkins explained, was to ensure data had public-health benefits.

One example the statistician cited was selecting self-harm or suicide over mental health, despite the latter’s possible cause of the former.

Another concerned matters of the heart.

“A person can suffer chronic heart disease and eventually die from a heart attack. In this case the ICD rules select the acute event – the heart attack – as the underlying cause despite the fact the chronic heart disease may have contributed to the heart attack.

“The logic is that the acute event may be preventable, and without it the person may have lived for many years despite the chronic heart condition.”

For those responsible for completing death certificates, the issues are varied. Not least is the very design of the certificate itself, which some feel is too prescriptive by allowing for only one sequence leading to death. Such a rigid format almost backs doctors into a corner and compels them to identify one single pathway.

Professor Mark Nelson, professorial research fellow at the Menzies Institute for Medical Research at the University of Tasmania, told The Medical Republic there was often ambiguity in death.

He revealed a recent major study he helped conduct, unrelated to cause of death data, involved the scrutiny of death certificates and associated clinical notes. Causes of death were “frequently” overturned by two adjudicators, he said, with even causes of death attributed following an autopsy sometimes reversed.

“The gold standard for cause of death is the autopsy, but even then we found statements saying someone died of coronary artery disease even though there was no evidence of thrombosis, so nothing to suggest they had an acute event,” he said. “It made us think that even pathologists are making probabilistic diagnosis.

“I don’t think it’s doctors being overwhelmed or too lazy or whatever else. What we are seeing is related to complexity.

“Because of survival rates you get competing causes of diseases. I work as a GP and most people over 75 have often more than one chronic disease, and with competing causes attributing a single cause of death does become more problematic.

“Doctors are doing the best they can with this complexity, and the fact is they are forced to provide one sequence which may not give the whole picture.

“Something may be a contributing factor but it’s not the only one when you have heart failure in the setting of renal failure in the setting of cachexia.”

There is, however, a Part II in the cause of death section of the death certificate which allows for the inclusion of other significant conditions that “contributed to death, but not related to the disease or condition causing it”.

This does not sit well with pathologist David Ranson who suggested there was an “illogicality” to the structure. “If it is contributing and stopping them recovering or overcoming the disease, then surely it is part of the cause?” he said. “If you presented this to a logician who knew about the medical issues they would probably say it doesn’t make sense.”

Ranson painted the scenario of a patient who fractures their neck of femur, is bedridden for six to eight weeks, was a smoker for 40 years, contracts pneumonia and dies.

“You could say the underlying process of the fractured neck of femur led to them becoming immobile which led to hypostatic pneumonia and to death,” he explained. “However, the smoking and lung disease contributed to the patient developing pneumonia, so you can’t put that in part II because it’s part of the cause of death.”

Echoing the words of Professor Nelson, Ranson added: “The problem is, the death certificate only lets you have one pathway to death. Clinical diagnostics are so much better these days, so we have a complex and rich knowledge of a person when they come to the end of their life. We know so much about their disease processes that the streamlined approach in the way the death certificate is laid out doesn’t meet the realities of everyday practice.”

Dr Tony Carbone, vice president of the Australian Medical Association (AMA), said the system needed to be streamlined, with doctors provided with electronic access to patients’ medical history.

Until that happens, “it is always going to be problematic”.

“We are snowed under with an enormous amount of requirements and red tape and other non-income producing activities. And we have patients to see,” he said.

“We have to find a solution that assists doctors fill their forms in a diligently and timely manner.”

Too often, GPs were operating “in a vacuum”, he added, which made it clinically impractical to offer anything more than a high-level assumption about the cause of death.

Citing an example similar to that of Sara Bird, Dr Carbone said: “If you have been called to a routine death in a nursing home how can you determine the exact cause of death? Are you going to order an autopsy? No you’re not. There will most likely be a number of conditions and it could have been any one of them.

“Sometimes it’s just not practical to give an exact pathological premise. It’s not an exact science.”

Despite the difficulties – be they pressure of time, lack of information or multiple possible causes of death – Carbone rejected any notion GPs might default to descriptions such as heart failure as an easy “get out”.

“I don’t believe so,” he said when asked about that prospect.

“I believe all my colleagues are exceptionally diligent. It will be based on their own observations, the observations of others or through an examination of the individual’s clinical history.”

Nevertheless, he suggested it was “quite possible” heart issues were over-represented.

“But how do you get around that? How do you become more precise?” Carbone asked, arguing the benefits of overhauling a system which may deliver only incremental benefits would not justify the “enormous amounts of work for already over-stretched GPs”.

The multiple-cause argument is not lost on the ABS.

Describing the issue as one of his “passions”, Eynstone-Hinkins acknowledged the difficulties faced by medical practitioners and, while the ASB does have multiple-cause data, he suggested the time was approaching when the process might need to evolve.

“The fact is that, as people accumulate multiple conditions, the underlying-cause construct can become more and more complicated and eventually it will become less relevant to deaths which are from multiple morbid conditions in very old people,” he said.

“We have to keep thinking, what are the public-health benefits that we are trying to highlight?

“We are working with WHO and international colleagues to ask, what do we need to change to ensure we are capturing the right construct of information that is going to get the best public-health benefits?”

Yet when all is said and done, however the data is assimilated, coded and dissected, whether doctors gain access to electronic medical histories or not, the inescapable fact is that the professional judgment of a GP, pathologist or junior doctor will continue to underpin the system.

It will remain up to them to determine, to the best of their ability, the reasons behind that most profound of questions; why did we die?

“Every man’s life ends the same way,” observed Ernest Hemmingway. “It is only the details of how he lived and how he died that distinguish one man from another.”