Underlying cause of death (UCoD) and injury prevention

Most countries (eg, Australia,1 USA2) have used UCoD to identify cases of injury death. Volume 1 of the WHO's International Classification of Diseases, 10th Revision (ICD10) manual3 describes the UCoD as:

‘(a) the disease or injury which initiated the train of events leading directly to death, or (b) the circumstances of the accident or violence which produced the fatal injury’.

As injury prevention researchers and practitioners, we are interested in preventing death by preventing injury. This leads to an interest in all cases where injury lies on a causal pathway that leads to death, even if the external cause that resulted in the injury is not the UCoD, according to WHO coding rules.3 Intervening anywhere on the causal pathway can prevent the outcome, injury death. The challenge, therefore, is to identify and agree on a theoretical definition of injury death, and an operational definition of injury death that is consistent with it, since the use of UCoD alone for identifying injury deaths results in potentially lost opportunities for prevention.

Concerns regarding UCoD accuracy

Concerns had been expressed about the difficulties of accurate classification of UCoD for older people,4–8 particularly when they had fallen and died. Our concerns were subsequently heightened by the results produced as part of our recent project aimed at investigating enhancements to the ICD-based Injury Severity Score.9

This ICD-based Injury Severity Score project involved linking the New Zealand Ministry of Health's National Minimum Data Set of hospital discharge data to the Ministry of Health's Mortality Collection. The latter collection is based on the Medical Certificates of Cause of Death, coroners' reports, etc. This work revealed apparent contradictions between hospital discharge and Mortality Collection data. For 52% of the individuals who died in hospital, there was a gross mismatch (ie, at the ICD10 chapter level) between hospital principal diagnosis (PDx) and UCoD recorded on the Mortality Collection. That is, many who had been recorded by the hospital with an injury principal diagnosis (PDx) were recorded on the Medical Certificates of Cause of Death to a medical UCoD. Only 48% of in-hospital injury deaths had an UCoD coded to an external cause of injury. This is consistent with a number of studies including those from Australia and from Sweden.1 7 This problem is largest for older people.10 11

Reasons for mismatches between hospital diagnosis and UCoD

For these mismatches, and for a given death that was preceded by an injury event, the appropriate choice for UCoD was either:

1. an external cause of injury (ie, UCoD was incorrectly coded) or

2. a medical UCoD (ie, the UCoD was classified and coded correctly, but there was a concomitant injury event).

We consider these in turn.

1. In regard to 1, for example, Johansson and Westerling investigated differences between the UCoD and the main condition from the hospital discharge record for people who had died within 1 year of discharge from hospital. They found that the UCoD and the main condition at discharge differed, at the ICD chapter level, for 54% of deaths. They also found that adding the hospital discharge data to the cause of death classification process resulted in a 58% increase in the number of falls classified as the UCoD. The authors indicated that these inaccuracies in the classification of UCoD were often due to omissions from the death certificate.7

2. If the appropriate choice of UCoD is a medical cause, it is also conceivable that the injury and the external cause before it either:

   • A. were not part of the same causal pathway as the UCoD or

   • B. were intermediate steps in the causal pathway, without which death would not have occurred when it did.

In respect of 2A, for example, a woman fell and was admitted to a hospital with a traumatic subdural haemorrhage. This was confirmed on a computed tomographic scan. She died 2 months later; she was discharged 2 days before death. The UCoD was coded to myocardial infarction. Her injury was listed in Part II of the death certificate (ie, it was judged to be a contributory cause of death). She was readmitted the same day as the death; only cerebrovascular disease (CVD) was listed on the hospital discharge record. Although not part of the same causal pathway, the parallel influences of the head injury and the underlying heart disease were both judged in the death certification process to have contributed to this person's death.

In respect of 2B, for example, a man had a motor vehicle traffic crash (MVTC), sustained a fractured sternum and was admitted to hospital. He was discharged dead 9 weeks after his MVTC with the following diagnoses recorded on the electronic record: fractured sternum (PDx), coma, CVD, pneumonia. The UCoD was classified to CVD or stroke. If the stroke led to the MVTC and coma, which in turn led to death, it is a ‘necessary cause’. That is, without the MVTC, the death would not have occurred when it did. So, irrespective of the UCoD, this is a case of interest to injury prevention professionals.

Even though others have identified this problem before us, none have proposed new theoretical and operational definitions of injury death. This paper addresses the first of these—the theoretical definition of injury death.

Proposed theoretical definition of injury death

The WHO indicates that the purpose of coding the UCoD, from a public health perspective, is to prevent the precipitating cause from operating (ie, identifying the primary cause of the death in order to prevent/reduce the occurrence in future).3 But why focus solely on precipitating cause? Intervening at any point on any causal pathway can prevent disease and death.

The following is an example of why this is the case. On 13 July 1967, during one of the later stages of the Tour de France cycle race, Tom Simpson, who was cycling up Mount Ventoux in the very hot sun, collapsed and died. Fotheringham12 describes part of the findings from the autopsy report as follows:

‘Death was due to cardiac collapse which may be put down to exhaustion in which unfavourable weather conditions, an excessive workload, and use of medicines of the type discovered on the victim may have played a part. The dose of amphetamines ingested by Simpson could not have led to his death on its own; but on the other hand it could have led him to go beyond the limit of his strength and thus bring on the appearance of certain troubles linked to his exhaustion’. (page 178)

However, this does not tell the whole story. An expanded list of factors that appear to be associated with Tom Simpson's death was as follows:

• amphetamine use

• alcohol use

• unfavourable weather conditions, that is, very hot temperatures

• overexertion

• dehydration associated with:

  • a contemporary belief that to starve oneself of liquid enhances performance during big races

  • the Tour de France rules, in operation at the time, that did not allow riders to take liquid from support cars

  • diarrhoea for 3 days before

• hypoglycaemia associated with inability to eat for 3 days before death.

It is likely that all of these contributed to the death of Tom Simpson. The concept of an UCoD is unhelpful here, since it is likely that to intervene and remove/reduce any one of these contributing factors would have prevented the death. In this, and in many other examples, there is not just one cause of death, but several—many of which are of interest to injury prevention researchers and practitioners.

More useful than the concept of UCoD is the WHO's theoretical definition of ‘cause of death’,3 which is:

‘all those diseases, morbid conditions or injuries which either resulted in or contributed to death and the circumstances of the accident or violence which produced such injuries’.

This can be likened to the definition of ‘necessary cause’. In A Dictionary of Epidemiology, Last has defined a ‘necessary cause’ as follows13:

‘a causal factor whose presence is required for the occurrence of the effect’.

(Some have referred to this as a ‘component cause’.14) Translating this to injury deaths, injury is a necessary cause of death if injury is required for the occurrence of death. We all die sometime; the issue is when we die. So we propose the following theoretical definition of injury death:

An injury death is one in which the injury resulted in premature death.

That is, if the injury had not occurred, the death would not have occurred at that time, or it would have occurred later.

This theoretical definition of injury death is in contrast to the UCoD. The WHO coding rules force the coder to choose one UCoD when there may be several causes each contributing to the death. Typically, UCoD is used by government agencies in producing mortality statistics. Such practice results in an undercount of injury deaths, consistent with the theoretical definition above.1

Summary and way forward

As injury prevention researchers and practitioners, we are interested in preventing serious injury and death by preventing injury. This leads to an interest in all cases where injury lies on a causal pathway to death, even if the external cause that resulted in the injury is not the UCoD, according to WHO conventions. That is, injury prevention researchers and practitioners should be interested in all injuries which are necessary causes of death, that is, that satisfy our proposed theoretical definition of injury death.

If our proposed and expanded theoretical definition of injury death is accepted by the injury prevention community, the challenge would be to identify an operational definition of injury death that goes beyond the UCoD and is consistent with this theoretical definition, since restricting solely to the UCoD results in an undercount and potentially lost opportunity for prevention. This has been recognized by several authors (eg, Kreisfeld and Harrison1) and groups (eg, Centers for Disease Control and Prevention15) who have used multiple cause of death data to operationally define a case of injury death. However, our view is that further work and debate are needed on an appropriate operational definition of injury death. An ideal group to further discussions around the proposed theoretical definition, and the way forward in determining an expanded operational definition of injury death consistent with that theoretical definition, is the International Collaborative Effort on Injury Statistics.16

One final remark is that this commentary is not suggesting that the WHO coding rules be changed. Rather, our proposed change to the theoretical definition of injury death is independent of WHO coding rules. Additionally, we currently envisage that any operational definition of injury death that is agreed upon would work within the current WHO framework for coding cause of death, but would take account of existing common practices.