Death certification and coding for ischaemic heart disease in Tasmania P. T. Sexton NH&MRC Public Health Training Fellow, Department of Medicine, University of Tasmania, Hobart, Tas.
K. Jamrozik Senior Lecturer in Public Health Unit in Clinical Epidemiology, University of Western Australia, Perth, WA.
J. M. Walsh Research Nurse, Department of Medicine, University of Tasmania, Hobart, Tas
Abstract: Official records show that the rates of mortality from ischaemic heart disease (IHD) in Tasmania have been the highest of all the Australian states for most of the past decade. This study assesses the accuracy of the official Tasmanian mortality data for IHD in 1987 and 1988 for males aged 25 to 74 years using routinely available clinical and pathological data supplemented by information from the attending doctor. Our findings show that a death officially coded to ICD 9 rubrics 410-414 (IHD) in Tasmania has 94% sensitivity and a positive predictive value of 90% for fatal definite acute myocardial infarction or possible coronary death as defined by the WHO. Comparison of our results with those of two earlier studies undertaken in Australian mainland centres indicates that differences in the official statistics for coronary mortality between Tasmania and the mainland states reflect true differences in the risk of coronary death. While the results from three Australian studies suggest that the routine system of death certification is reasonably accurate, careful monitoring of death certification and coding practices need to be undertaken regularly in all states of Australia if secular changes in regional patterns of coronary mortality are to be regarded as credible. (Aust NZ J Med 1992; 22: 114-118.) Key words: Death certificates, ischaemic heart disease, cause of death coding.
INTRODUCTION Mortality from ischaemic heart disease (IHD) has declined at a steady rate in Australia' over the past 20 years. However, the decline has not occurred at equal rates among all Australian states.' In Tasmania, the rate of decline was less over the past 20 years than in mainland states and consequently, for most of the past decade, mortality from IHD in Tasmania has been the highest observed among the six Australian states.' While differences in occupation, social class and risk factors for coronary mortality have been demonstrated between states,'.2 other factors such as differences in recording of mortality data on death certificates and subsequent assignment of codes for the underlying
cause of death from the International Classification of Diseases - (eighth revision until 1978 then ninth revision [ICD 91 from 1979 onwards) may contribute to the observed interstate variation in mortality rates. Earlier studies undertaken in mainland Australian states have found that mortality data for IHD are reasonably accurate and reliable3,4and that any variation due to either death certification or coding is unlikely to be ~ignificant.~ This study was undertaken to assess the validity of official mortality data for IHD in Tasmania and to contrast these findings with those of earlier studies undertaken in Australian mainland states.
Reprinr requests to: Dr P. T. Sexton, Department of Medicine, University of Tasmania, Clinical School, 43 Collins Street, Hobart, Tasmania 7000, Australia.
114
Aust NZ J Med 1992; 22
SEXTON ET AL.
METHODS Death Records Death certificates completed by attending doctors or coroners in Tasmania are sent to the Registrar of Births, Deaths and Marriages. Copies are then sent to the Australian Bureau of Statistics (ABS) where a single underlying cause of death (UCD) is derived from the multiple causes of death recorded on the death certificate, in accordance with the World Health Organization (WHO) manual of the International Classification of Diseases, Injuries and Causes of Death. As part of a study supported by the National Heart Foundation and the National Health and Medical Research Council on sudden cardiac death in Tasmanian males, copies of all the death certificates completed by doctors in the state of Tasmania during 1987 and 1988 were provided by the Registrar of Births, Deaths and Marriages. Death certificates of males aged between 25 and 74 years where any CARDIAC cause of death (equivalent to ICD 9 codes: 401-405,410-414,426-429,440) was recorded on the death certificate were regarded as ‘potential IHD’ deaths and were identified for further study. Data derived directly from the death certificate included the death certificate number, the place of death (in or out of hospital), the name of the attending doctor, and conditions which may have contributed to the death. Clinical and Pathological Data Supplementary information on all ‘potential IHD’ deaths was obtained from necropsy reports, ambulance and hospital records and a questionnaire which was sent to the attending doctor. Necropsy reports were obtained from hospital records and from the State Crown Law Department. In coronial cases, the name of the attending doctor was provided by the Crown Law Department. Hospital records were sought when the place of death was given as a hospital and when the deceased was not a regular patient of the attending doctor. Data derived from hospital records included any history of serious illnesses including IHD. Additional information about previous history of IHD and other serious illness was obtained by questionnaire from the attending doctor if the death certificate was signed by a non-hospital based medical practitioner and when hospital records were inadequate. Validation Criteria All ‘potential IHD’ deaths were reviewed in the light of the available supplementary diagnostic data and classified according to the criteria adopted by WHO for the MONICA project. Thus, following the method of Dobson et ~ l . a, death ~ in Tasmania was accepted as ‘true I H D if there was no other likely cause of death and the case met either of the following criteria: IHD IN TASMANIA
‘1. Definite acute myocardial infarction. Electrocardiogram showed unequivocal serial changes; or equivocal electrocardiogram results with elevated enzymes; or typical history of pain with elevated enzymes; or, for fatal cases, positive post-mortem evidence of fresh myocardial infarction or recent coronary occlusion. 2. Possible acute myocardial infarction. Remaining surviving cases with typical pain whose electrocardiogram and enzyme results did not place them in the previous category but in whom there was no good evidence for another diagnosis; or fatal cases with a history of pain or evidence of chronic ischaemic heart disease.’ Either well established coronary symptoms in life or the presence at necropsy of subtotal coronary stenosis or healed myocardial infarction was accepted as sufficient evidence to make a diagnosis of chronic ischaemic heart disease.
Data from the Australian Bureau of Statistics The Australian Bureau of Statistics (ABS) provided a computer record of death certificate numbers of all Tasmanian males aged between 25 and 74 years who died during 1987 and 1988 and whose cause of death was coded to ICD 9 rubrics 410-414 (ischaemic heart disease). Deaths coded by ABS to ICD 9 rubrics 410-414 were then compared with those identified in this study as ‘true I H D deaths. ‘Potential IHD deaths’ for which the UCD was not coded to ICD 9 rubrics 410-414 by the ABS were identified by death certificate number and the ICD 9 coding was provided by the ABS. RESULTS Death Records In the official records compiled by the ABS, there were 342 deaths from IHD recorded for males aged 25-74 years in Tasmania in 1987 and 304 in 1988. Of the 646 certificates coded to ICD 9 rubrics 410-414, 47 were excluded from further study because they related to deaths of Tasmanian residents that occurred interstate, deaths of overseas visitors or Islanders, or deaths which occurred in 1986 but were registered and coded in 1987. Supplementary diagnostic data were not available for a hrther 55 otherwise eligible cases. Thus 91% (5441599) of eligible deaths in 1987 and 1988 of males aged between 25 and 74 years which were coded by the ABS to ICD 9 rubrics 410-414 were included for validation.
‘Potential IHD Deaths’ All 3,876 death certificates issued for Tasmanian males who died in 1987 and 1988 were reviewed. Of the fatalities occurring in Tasmania among men aged 25 to 74
Aust NZ J Med 1992; 22
115
TABLE 1 ‘Potential IHD’ Deaths (Combined Data for Years of Death 1987 and 1988) with Breakdown of Supplementary Diagnostic Data* ‘Potential IHD’ deaths Total number of necropsy reports available Total number of hospital records available and reviewed Total number of questionnaires received from attending doctors Supplementary data not available
729 206 (28%) 313 (43%) 314 (4300) 80 (11O/o)
*Percentages add up to more than 100% because supplementary data were obtained from more than one source for some deaths.
years, 376 in 1987 and 353 in 1988 were classified as ‘potential IHD’ deaths. Table 1 shows data were available from necropsy for 28% of the 729 ‘potential IHD deaths’. Additional information was obtained from hospital records for 43% of cases and from attending doctors for 43% of cases, but no further information could be obtained for 80 (11%) of the ‘potential IHD’ deaths. Further analysis of ‘potential IHD’ deaths was therefore limited to 649 cases for which at least some supplementary information was available. Table 2 presents a 2 x 2 comparison of ‘potential IHD’ deaths with deaths classified to ICD 9 rubrics 410-414. The sensitivity of an official underlying cause of death coded to any of ICD 9 rubrics 410-414 for ‘true I H D defined by WHO criteria was 9470, with a corresponding specificity of 59% and positive predictive value of 90%. The positive predictive value of code 410 considered alone was 90% as compared with 89% for code 414 considered alone. Only one fatality was coded to rubric 41 1, ‘sub-acute ischaemic heart disease’. The official underlying causes of death for the 30 ‘potential IHD’ deaths identified by review of death certificates and that satisfied the criteria for ‘true IHD’ but were not coded to IHD by the ABS are shown in Table 3. In half of these cases, the official underlying cause of death was coded to a condition known to be implicated in the aetiology of atherosclerosis, for example, diabetes mellitus (11 cases), raised blood pressure or blood lipids, or ‘atherosclerosis’ itself. The coding rules incorporated in the ICD are open to an interpretation that would allow identification of such TABLE 2 Comparison of 729 ‘Potential IHD’ Deaths with Underlying Cause of Death Given by ABS Final diagnosis ‘True Not Subtotal NO additional Total IHD’ IHD data Coded by ABS
116
410-4 492
52
544
55
599
Other 30 Total 522
75 127
105 649
25 80
130 729
TABLE 3 ‘True IHD’ Deaths NOT Officiallv Coded to ICD 9 Rubrics 410-414 ICD code
Disease
250 272 402 425 427 429 429 440 496 73 1 799 Total
Diabetes mellitus Hypercholesterolaernia Hypertensive heart disease Cardiomyopathy Cardiac arrest Cardiovasc disease (unspec) Myocarditis Atherosclerosis COAD Paget’s disease Not known
11 2 1 2 8 1
1 1 1 1 1 30
abnormalities as the primary reason for the death. Of the remaining cases in Table 3, more than half were ascribed to ‘cardiac arrest’, a diagnostic label bearing no aetiological information. Table 4 sets out the alternative diagnoses, based on our review of the supplementary information, for 52 cases ascribed to IHD by ABS but not fulfilling the WHO criteria for ‘true IHD’. In four of the cases, evidence for the alternative diagnosis was available from necropsy but the death certificate had not been amended in the light of this information.
DISCUSSION During most of the past two decades official rates of mortality from IHD have declined more slowly in Tasmania than in the mainland Australian states.’ For TABLE 4 ‘Potential IHD’ Deaths Officially Coded to ICD 9 Rubrics 410-414 Classified as NOT ‘True IHD’ Deaths Disease Amiodarone pneumonitis Asbestosis Blood loss and hypotension Bronchiectasis Cardiac valve disease Chronic renal failure Chronic respiratory disease Ernbohc cerebral infarct Haemorrhagic bronchopneumonia infarction of small bowel Infective cardiac tamponade Malignancy Panhypopituitarism Pneumococcal septicaemia Polycythaemia rubra Vera Post aortic aneurysm repair Post op bowel obstruction Viral pneumonitis Embolic cerebral infarct* Cerebral infarct” Pulmonary embolus’ Total
1 1 1 1 7 1 8 2 1 1 1 15 1 1 1 1 1 3 1 1 2 52
‘Death certificate not amended after necropsy
Aust NZ J Med 1992; 22
SEXTON ET AL.
TABLE 5 Results from Three Australian Studies when Officially Coded Underlying Cause of Death (UCD) of IHD was Validated against ‘True IHD’ Perth (3)
Sensitivity Positive predictive value
Newcastle (4) Tasmania
1971
1978
1979
1987-88
93%
89%
91 o/o
94%
88%
89%
89%
90%
such findings to be of value to the process of planning and implementation of future health services in Tasmania, routinely collected mortality data must be shown to be valid and reliable. Like those of previous Australian ~ t u d i e s , ~our ,~ findings suggest that the official mortality statistics for IHD are reasonably accurate. The two previous reports, from Perth and Newcastle, come from centres in which there has been an active programme of epidemiological research into IHD for at least ten years, the existence of which might have altered local practice in the area of death certification. Nevertheless, as may be seen from Table 5, our results from Tasmania are consistent with those of the mainland. Taken together, these three studies show that the routine system of death certification and coding performs well. Furthermore, they suggest that the higher rate of mortality from IHD observed in the official statistics from Tasmania is unlikely to be an artefact. Over the two years of our study, 1987 and 1988, information obtained from clinical and necropsy records supported the official Tasmanian data for deaths from IHD in approximately 90% of cases. As with other studies, both in A u ~ t r a l i aand ~ . ~overseas5 the number of false positive cases (those where UCD was identified as IHD but additional evidence refuted this diagnosis) was largely compensated for by the number of false negative cases. Like Martin et uZ.,~ we found that diabetes mellitus contributed significantly to this imperfect sensitivity of an official UCD of IHD. On the other hand, while ‘cardiac arrest’ (code 4275) was also an important source of false negative cases, this term appears unsatisfactory as a final pathological diagnosis. In the absence of any further information, it would seem more ‘honest’ to code such cases to ‘sudden unexplained death’ (code 798). While we have confidence in those diagnoses which were supported by evidence from necropsy, we acknowledge that earlier studies have shown the accuracy of clinical data such as hospital records to be ~ariable.~,’ It has been argued that the only acceptable test of the validity of the ‘underlying cause of death’ is a necropsy.B However, it is neither practical nor appropriate for a necropsy to be performed for all IHD IN TASMANIA
deaths. Moreover, as Martin has pointed out,’ it is never likely to be feasible to separate ‘definite’ from ‘possible’ coronary deaths, despite some of the latter being diagnosed on the basis of coronary symptoms in life and the absence clinically of a competing cause of death. In any event both the American College of Cardiology and American Heart Association agree that routine non-invasive diagnostic procedures provide sufficient data to confirm or exclude AM1 or angina.’O.“ Given that supplementary clinical data can be used to determine the validity of official mortality statistics, it is essential that such information is both available and of high quality. In Tasmania, no additional information could be obtained for 11% of the fatalities that we assessed. Low necropsy rates (28%) may reflect the priorities of the Tasmanian coronial system, which requires no medical review of death certificates. The decision to release a body for burial is made on the absence of evidence of ’unnatural causes’ with no overriding need to establish an underlying cause.12The lack of a state-wide computerised hospital morbidity data system, equivalent to those used in the compilation of population-based registers of IHD in Perth and Newcastle, limits access to diagnostic data held in hospitals to which a Tasmanian patient had been admitted in the past, before his final illness. Under the extreme assumption that confirmatory clinical evidence might have been obtained from this source for all 55 patients with an UCD of I H D for whom we could not obtain any supplementary data, the sensitivity of this UCD for ‘true IHD’ would have been 95’3’0, with a positive predictive value of 91%. Alternatively, if a clinical diagnosis of IHD could have been positively excluded using information from hospitals, the specificity of this UCD would have been 71%. Our study has provided cross-sectional data to suggest that routine certification and coding of deaths in Tasmania is sufficiently accurate at present for us to be confident that mortality from IHD truly is higher in Tasmania than elsewhere in the country. Whether this is due to differences in the incidence of major coronary events or in their outcome cannot be discerned from the data that we have collected. Nor are we in a position to comment whether the present separation between Tasmanian and mainland rates for coronary mortality has arisen through progressive changes in either or both of the incidence and case fatality of acute myocardial infarction, or through longitudinal changes in death certification and coding. Given that these practices are reasonably comparable now, progressive emergence of an apparent excess in coronary mortality in Tasmania could only have resulted from artefact if there had been selective elimi-
Aust NZ J Med 1992; 22
117
nation of false positive cases from the official statistics for other states or a selective increase in such cases in the figures for Tasmania. Neither of these seems likely since, if the sensitivity and specificity of the process of compiling statistics for coronary deaths remains constant, the proportion of false positive cases should rise as coronary mortality falls, and should rise fastest in the communities that are experiencing the most rapid fall in true coronary mortality. Martin et a1.3 were unable to find any evidence for such an increase in Perth, at least. In conclusion, we have shown that an official UCD of IHD for Tasmanian men aged 25-74 years has 94% sensitivity, 59% specificity and a positive predictive value of 90% for fatal definite acute myocardial infarction or possible coronary death as defined by the WHO. Comparison of our results with those of similar validation studies undertaken previously in Perth and Newcastle indicates that differences in official statistics for coronary mortality between Tasmania and the mainland states reflect true differences in the risk of coronary death. We cannot say whether this is due to variation between the states in the risk or outcome of acute myocardial infarction. However, our method of screening all death certificates to identify a smaller number of ‘potential IHD’ deaths for detailed review potentially does provide an efficient way of monitoring death certification and coding practices. Such monitoring is necessary if we are to detect longitudinal drift in diagnostic and nosological performance that could lead to artefactual trends in the official figures for coronary mortality. Where detailed review of all death certificates is not practicable, such as in the more populous states, death certificates could be sampled at regular intervals for validation of the official ICD coding. Coded causes of deaths other than 410-414 most likely to conceal true IHD deaths could also be identified e.g. nonspecific cardiac codes (426-429), sudden death (798) and diabetes mellitus (250). Changes in the proportion of deaths falling into these groups could be monitored over time to detect shifts in diagnostic allocation. Reviews of this kind need to be undertaken regularly in all states of Australia if secular changes in
118
regional patterns of coronary mortality are to be given any credence. Acknowledgements We would like to acknowledge the assistance of Mrs T-L Sexton in computing and proof reading; Mr Ray Smithurst, Australian Bureau of Statistics, Hobart, in data collection; the Medical Records departments, Royal Hobart Hospital, Launceston General Hospital, North-West General Hospital, Mersey General Hospital; and the Registrar, Births, Deaths and Marriages, Hobart. In addition we wish to thank the referees for their valuable suggestions. Date of submission: 15 March 1991
References 1. Sexton PT, Woodward DR, Gilbert N, Jamrozik K. Interstate differences in trends in coronary mortality and risk factors in Australia. Aust NZ J Med 1990; 152: 531-4. 2. Gibberd RW, Dobson AJ, Florey C du Ve, Leeder SR. Differences and comparative declines in ischaemic heart disease mortality among subpopulations in Australia 1969-1978. Int J Epidemiol 1984; 13: 25-31. 3. Martin CA, Hobbs MST, Armstrong BK. Estimation of myocardial infarction mortality from routinely collected data in Western Australia. J Chron Dis 1987; 40: 661-9. 4. Dobson AJ, Gibberd RW, Leeder SR. Death certification and coding for ischaemic heart disease in Australia. Am J Epidemiol 1983; 117: 397-405. 5. Kircher T, McIlwaine WJ, Donnelly MDI er al. Certification of death from ischaemic heart disease in Belfast. Int J Epidemiol 1985; 14: 560-5. 6. Kircher T, Nelson J, Burdo H. The autopsy as a measure accuracy of the death certificate. N Engl J Med 1985; 313: 1263-91. 7. Cameron HM, McGoogan E. A prospective study of 1152 hospital autopsies I. Inaccuracies in death certification. J Pathol 1981; 133: 273-83. 8. Stehbens WE. An appraisal of the epidemic rise of coronary heart disease and its decline. Lancet 1987; ii: 606-1 1. 9. Martin CA, Hobbs MST, Armstrong BK. Measuring the incidence of acute myocardial infarction: the problem of possible acute myocardial infarction. Acta Med Scand 1988; Suppl 728: 40-7. 10. Ross J, Brandenburg RO, Dinsmere RE er al. Guidelines for coronary angiography: a report of the American College of Cardiology/American Heart Association Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures (Subcornmi& on Coronary Angiography). Circulation 1987; 76: 963A-977A. 11. Chaitman BR, Bourassa MG, Davis K er al. Angiographic prevalence of high risk coronary artery disease in patient subsets (CASS). Circulation 1981; 64: 360-7. 12. Coroners Act 1957, Tasmania; Australian Government Printing Service.
Aust NZ J Med 1992; 22
SEXTON ET AL.
K. Jamrozik Senior Lecturer in Public Health Unit in Clinical Epidemiology, University of Western Australia, Perth, WA.
J. M. Walsh Research Nurse, Department of Medicine, University of Tasmania, Hobart, Tas
Abstract: Official records show that the rates of mortality from ischaemic heart disease (IHD) in Tasmania have been the highest of all the Australian states for most of the past decade. This study assesses the accuracy of the official Tasmanian mortality data for IHD in 1987 and 1988 for males aged 25 to 74 years using routinely available clinical and pathological data supplemented by information from the attending doctor. Our findings show that a death officially coded to ICD 9 rubrics 410-414 (IHD) in Tasmania has 94% sensitivity and a positive predictive value of 90% for fatal definite acute myocardial infarction or possible coronary death as defined by the WHO. Comparison of our results with those of two earlier studies undertaken in Australian mainland centres indicates that differences in the official statistics for coronary mortality between Tasmania and the mainland states reflect true differences in the risk of coronary death. While the results from three Australian studies suggest that the routine system of death certification is reasonably accurate, careful monitoring of death certification and coding practices need to be undertaken regularly in all states of Australia if secular changes in regional patterns of coronary mortality are to be regarded as credible. (Aust NZ J Med 1992; 22: 114-118.) Key words: Death certificates, ischaemic heart disease, cause of death coding.
INTRODUCTION Mortality from ischaemic heart disease (IHD) has declined at a steady rate in Australia' over the past 20 years. However, the decline has not occurred at equal rates among all Australian states.' In Tasmania, the rate of decline was less over the past 20 years than in mainland states and consequently, for most of the past decade, mortality from IHD in Tasmania has been the highest observed among the six Australian states.' While differences in occupation, social class and risk factors for coronary mortality have been demonstrated between states,'.2 other factors such as differences in recording of mortality data on death certificates and subsequent assignment of codes for the underlying
cause of death from the International Classification of Diseases - (eighth revision until 1978 then ninth revision [ICD 91 from 1979 onwards) may contribute to the observed interstate variation in mortality rates. Earlier studies undertaken in mainland Australian states have found that mortality data for IHD are reasonably accurate and reliable3,4and that any variation due to either death certification or coding is unlikely to be ~ignificant.~ This study was undertaken to assess the validity of official mortality data for IHD in Tasmania and to contrast these findings with those of earlier studies undertaken in Australian mainland states.
Reprinr requests to: Dr P. T. Sexton, Department of Medicine, University of Tasmania, Clinical School, 43 Collins Street, Hobart, Tasmania 7000, Australia.
114
Aust NZ J Med 1992; 22
SEXTON ET AL.
METHODS Death Records Death certificates completed by attending doctors or coroners in Tasmania are sent to the Registrar of Births, Deaths and Marriages. Copies are then sent to the Australian Bureau of Statistics (ABS) where a single underlying cause of death (UCD) is derived from the multiple causes of death recorded on the death certificate, in accordance with the World Health Organization (WHO) manual of the International Classification of Diseases, Injuries and Causes of Death. As part of a study supported by the National Heart Foundation and the National Health and Medical Research Council on sudden cardiac death in Tasmanian males, copies of all the death certificates completed by doctors in the state of Tasmania during 1987 and 1988 were provided by the Registrar of Births, Deaths and Marriages. Death certificates of males aged between 25 and 74 years where any CARDIAC cause of death (equivalent to ICD 9 codes: 401-405,410-414,426-429,440) was recorded on the death certificate were regarded as ‘potential IHD’ deaths and were identified for further study. Data derived directly from the death certificate included the death certificate number, the place of death (in or out of hospital), the name of the attending doctor, and conditions which may have contributed to the death. Clinical and Pathological Data Supplementary information on all ‘potential IHD’ deaths was obtained from necropsy reports, ambulance and hospital records and a questionnaire which was sent to the attending doctor. Necropsy reports were obtained from hospital records and from the State Crown Law Department. In coronial cases, the name of the attending doctor was provided by the Crown Law Department. Hospital records were sought when the place of death was given as a hospital and when the deceased was not a regular patient of the attending doctor. Data derived from hospital records included any history of serious illnesses including IHD. Additional information about previous history of IHD and other serious illness was obtained by questionnaire from the attending doctor if the death certificate was signed by a non-hospital based medical practitioner and when hospital records were inadequate. Validation Criteria All ‘potential IHD’ deaths were reviewed in the light of the available supplementary diagnostic data and classified according to the criteria adopted by WHO for the MONICA project. Thus, following the method of Dobson et ~ l . a, death ~ in Tasmania was accepted as ‘true I H D if there was no other likely cause of death and the case met either of the following criteria: IHD IN TASMANIA
‘1. Definite acute myocardial infarction. Electrocardiogram showed unequivocal serial changes; or equivocal electrocardiogram results with elevated enzymes; or typical history of pain with elevated enzymes; or, for fatal cases, positive post-mortem evidence of fresh myocardial infarction or recent coronary occlusion. 2. Possible acute myocardial infarction. Remaining surviving cases with typical pain whose electrocardiogram and enzyme results did not place them in the previous category but in whom there was no good evidence for another diagnosis; or fatal cases with a history of pain or evidence of chronic ischaemic heart disease.’ Either well established coronary symptoms in life or the presence at necropsy of subtotal coronary stenosis or healed myocardial infarction was accepted as sufficient evidence to make a diagnosis of chronic ischaemic heart disease.
Data from the Australian Bureau of Statistics The Australian Bureau of Statistics (ABS) provided a computer record of death certificate numbers of all Tasmanian males aged between 25 and 74 years who died during 1987 and 1988 and whose cause of death was coded to ICD 9 rubrics 410-414 (ischaemic heart disease). Deaths coded by ABS to ICD 9 rubrics 410-414 were then compared with those identified in this study as ‘true I H D deaths. ‘Potential IHD deaths’ for which the UCD was not coded to ICD 9 rubrics 410-414 by the ABS were identified by death certificate number and the ICD 9 coding was provided by the ABS. RESULTS Death Records In the official records compiled by the ABS, there were 342 deaths from IHD recorded for males aged 25-74 years in Tasmania in 1987 and 304 in 1988. Of the 646 certificates coded to ICD 9 rubrics 410-414, 47 were excluded from further study because they related to deaths of Tasmanian residents that occurred interstate, deaths of overseas visitors or Islanders, or deaths which occurred in 1986 but were registered and coded in 1987. Supplementary diagnostic data were not available for a hrther 55 otherwise eligible cases. Thus 91% (5441599) of eligible deaths in 1987 and 1988 of males aged between 25 and 74 years which were coded by the ABS to ICD 9 rubrics 410-414 were included for validation.
‘Potential IHD Deaths’ All 3,876 death certificates issued for Tasmanian males who died in 1987 and 1988 were reviewed. Of the fatalities occurring in Tasmania among men aged 25 to 74
Aust NZ J Med 1992; 22
115
TABLE 1 ‘Potential IHD’ Deaths (Combined Data for Years of Death 1987 and 1988) with Breakdown of Supplementary Diagnostic Data* ‘Potential IHD’ deaths Total number of necropsy reports available Total number of hospital records available and reviewed Total number of questionnaires received from attending doctors Supplementary data not available
729 206 (28%) 313 (43%) 314 (4300) 80 (11O/o)
*Percentages add up to more than 100% because supplementary data were obtained from more than one source for some deaths.
years, 376 in 1987 and 353 in 1988 were classified as ‘potential IHD’ deaths. Table 1 shows data were available from necropsy for 28% of the 729 ‘potential IHD deaths’. Additional information was obtained from hospital records for 43% of cases and from attending doctors for 43% of cases, but no further information could be obtained for 80 (11%) of the ‘potential IHD’ deaths. Further analysis of ‘potential IHD’ deaths was therefore limited to 649 cases for which at least some supplementary information was available. Table 2 presents a 2 x 2 comparison of ‘potential IHD’ deaths with deaths classified to ICD 9 rubrics 410-414. The sensitivity of an official underlying cause of death coded to any of ICD 9 rubrics 410-414 for ‘true I H D defined by WHO criteria was 9470, with a corresponding specificity of 59% and positive predictive value of 90%. The positive predictive value of code 410 considered alone was 90% as compared with 89% for code 414 considered alone. Only one fatality was coded to rubric 41 1, ‘sub-acute ischaemic heart disease’. The official underlying causes of death for the 30 ‘potential IHD’ deaths identified by review of death certificates and that satisfied the criteria for ‘true IHD’ but were not coded to IHD by the ABS are shown in Table 3. In half of these cases, the official underlying cause of death was coded to a condition known to be implicated in the aetiology of atherosclerosis, for example, diabetes mellitus (11 cases), raised blood pressure or blood lipids, or ‘atherosclerosis’ itself. The coding rules incorporated in the ICD are open to an interpretation that would allow identification of such TABLE 2 Comparison of 729 ‘Potential IHD’ Deaths with Underlying Cause of Death Given by ABS Final diagnosis ‘True Not Subtotal NO additional Total IHD’ IHD data Coded by ABS
116
410-4 492
52
544
55
599
Other 30 Total 522
75 127
105 649
25 80
130 729
TABLE 3 ‘True IHD’ Deaths NOT Officiallv Coded to ICD 9 Rubrics 410-414 ICD code
Disease
250 272 402 425 427 429 429 440 496 73 1 799 Total
Diabetes mellitus Hypercholesterolaernia Hypertensive heart disease Cardiomyopathy Cardiac arrest Cardiovasc disease (unspec) Myocarditis Atherosclerosis COAD Paget’s disease Not known
11 2 1 2 8 1
1 1 1 1 1 30
abnormalities as the primary reason for the death. Of the remaining cases in Table 3, more than half were ascribed to ‘cardiac arrest’, a diagnostic label bearing no aetiological information. Table 4 sets out the alternative diagnoses, based on our review of the supplementary information, for 52 cases ascribed to IHD by ABS but not fulfilling the WHO criteria for ‘true IHD’. In four of the cases, evidence for the alternative diagnosis was available from necropsy but the death certificate had not been amended in the light of this information.
DISCUSSION During most of the past two decades official rates of mortality from IHD have declined more slowly in Tasmania than in the mainland Australian states.’ For TABLE 4 ‘Potential IHD’ Deaths Officially Coded to ICD 9 Rubrics 410-414 Classified as NOT ‘True IHD’ Deaths Disease Amiodarone pneumonitis Asbestosis Blood loss and hypotension Bronchiectasis Cardiac valve disease Chronic renal failure Chronic respiratory disease Ernbohc cerebral infarct Haemorrhagic bronchopneumonia infarction of small bowel Infective cardiac tamponade Malignancy Panhypopituitarism Pneumococcal septicaemia Polycythaemia rubra Vera Post aortic aneurysm repair Post op bowel obstruction Viral pneumonitis Embolic cerebral infarct* Cerebral infarct” Pulmonary embolus’ Total
1 1 1 1 7 1 8 2 1 1 1 15 1 1 1 1 1 3 1 1 2 52
‘Death certificate not amended after necropsy
Aust NZ J Med 1992; 22
SEXTON ET AL.
TABLE 5 Results from Three Australian Studies when Officially Coded Underlying Cause of Death (UCD) of IHD was Validated against ‘True IHD’ Perth (3)
Sensitivity Positive predictive value
Newcastle (4) Tasmania
1971
1978
1979
1987-88
93%
89%
91 o/o
94%
88%
89%
89%
90%
such findings to be of value to the process of planning and implementation of future health services in Tasmania, routinely collected mortality data must be shown to be valid and reliable. Like those of previous Australian ~ t u d i e s , ~our ,~ findings suggest that the official mortality statistics for IHD are reasonably accurate. The two previous reports, from Perth and Newcastle, come from centres in which there has been an active programme of epidemiological research into IHD for at least ten years, the existence of which might have altered local practice in the area of death certification. Nevertheless, as may be seen from Table 5, our results from Tasmania are consistent with those of the mainland. Taken together, these three studies show that the routine system of death certification and coding performs well. Furthermore, they suggest that the higher rate of mortality from IHD observed in the official statistics from Tasmania is unlikely to be an artefact. Over the two years of our study, 1987 and 1988, information obtained from clinical and necropsy records supported the official Tasmanian data for deaths from IHD in approximately 90% of cases. As with other studies, both in A u ~ t r a l i aand ~ . ~overseas5 the number of false positive cases (those where UCD was identified as IHD but additional evidence refuted this diagnosis) was largely compensated for by the number of false negative cases. Like Martin et uZ.,~ we found that diabetes mellitus contributed significantly to this imperfect sensitivity of an official UCD of IHD. On the other hand, while ‘cardiac arrest’ (code 4275) was also an important source of false negative cases, this term appears unsatisfactory as a final pathological diagnosis. In the absence of any further information, it would seem more ‘honest’ to code such cases to ‘sudden unexplained death’ (code 798). While we have confidence in those diagnoses which were supported by evidence from necropsy, we acknowledge that earlier studies have shown the accuracy of clinical data such as hospital records to be ~ariable.~,’ It has been argued that the only acceptable test of the validity of the ‘underlying cause of death’ is a necropsy.B However, it is neither practical nor appropriate for a necropsy to be performed for all IHD IN TASMANIA
deaths. Moreover, as Martin has pointed out,’ it is never likely to be feasible to separate ‘definite’ from ‘possible’ coronary deaths, despite some of the latter being diagnosed on the basis of coronary symptoms in life and the absence clinically of a competing cause of death. In any event both the American College of Cardiology and American Heart Association agree that routine non-invasive diagnostic procedures provide sufficient data to confirm or exclude AM1 or angina.’O.“ Given that supplementary clinical data can be used to determine the validity of official mortality statistics, it is essential that such information is both available and of high quality. In Tasmania, no additional information could be obtained for 11% of the fatalities that we assessed. Low necropsy rates (28%) may reflect the priorities of the Tasmanian coronial system, which requires no medical review of death certificates. The decision to release a body for burial is made on the absence of evidence of ’unnatural causes’ with no overriding need to establish an underlying cause.12The lack of a state-wide computerised hospital morbidity data system, equivalent to those used in the compilation of population-based registers of IHD in Perth and Newcastle, limits access to diagnostic data held in hospitals to which a Tasmanian patient had been admitted in the past, before his final illness. Under the extreme assumption that confirmatory clinical evidence might have been obtained from this source for all 55 patients with an UCD of I H D for whom we could not obtain any supplementary data, the sensitivity of this UCD for ‘true IHD’ would have been 95’3’0, with a positive predictive value of 91%. Alternatively, if a clinical diagnosis of IHD could have been positively excluded using information from hospitals, the specificity of this UCD would have been 71%. Our study has provided cross-sectional data to suggest that routine certification and coding of deaths in Tasmania is sufficiently accurate at present for us to be confident that mortality from IHD truly is higher in Tasmania than elsewhere in the country. Whether this is due to differences in the incidence of major coronary events or in their outcome cannot be discerned from the data that we have collected. Nor are we in a position to comment whether the present separation between Tasmanian and mainland rates for coronary mortality has arisen through progressive changes in either or both of the incidence and case fatality of acute myocardial infarction, or through longitudinal changes in death certification and coding. Given that these practices are reasonably comparable now, progressive emergence of an apparent excess in coronary mortality in Tasmania could only have resulted from artefact if there had been selective elimi-
Aust NZ J Med 1992; 22
117
nation of false positive cases from the official statistics for other states or a selective increase in such cases in the figures for Tasmania. Neither of these seems likely since, if the sensitivity and specificity of the process of compiling statistics for coronary deaths remains constant, the proportion of false positive cases should rise as coronary mortality falls, and should rise fastest in the communities that are experiencing the most rapid fall in true coronary mortality. Martin et a1.3 were unable to find any evidence for such an increase in Perth, at least. In conclusion, we have shown that an official UCD of IHD for Tasmanian men aged 25-74 years has 94% sensitivity, 59% specificity and a positive predictive value of 90% for fatal definite acute myocardial infarction or possible coronary death as defined by the WHO. Comparison of our results with those of similar validation studies undertaken previously in Perth and Newcastle indicates that differences in official statistics for coronary mortality between Tasmania and the mainland states reflect true differences in the risk of coronary death. We cannot say whether this is due to variation between the states in the risk or outcome of acute myocardial infarction. However, our method of screening all death certificates to identify a smaller number of ‘potential IHD’ deaths for detailed review potentially does provide an efficient way of monitoring death certification and coding practices. Such monitoring is necessary if we are to detect longitudinal drift in diagnostic and nosological performance that could lead to artefactual trends in the official figures for coronary mortality. Where detailed review of all death certificates is not practicable, such as in the more populous states, death certificates could be sampled at regular intervals for validation of the official ICD coding. Coded causes of deaths other than 410-414 most likely to conceal true IHD deaths could also be identified e.g. nonspecific cardiac codes (426-429), sudden death (798) and diabetes mellitus (250). Changes in the proportion of deaths falling into these groups could be monitored over time to detect shifts in diagnostic allocation. Reviews of this kind need to be undertaken regularly in all states of Australia if secular changes in
118
regional patterns of coronary mortality are to be given any credence. Acknowledgements We would like to acknowledge the assistance of Mrs T-L Sexton in computing and proof reading; Mr Ray Smithurst, Australian Bureau of Statistics, Hobart, in data collection; the Medical Records departments, Royal Hobart Hospital, Launceston General Hospital, North-West General Hospital, Mersey General Hospital; and the Registrar, Births, Deaths and Marriages, Hobart. In addition we wish to thank the referees for their valuable suggestions. Date of submission: 15 March 1991
References 1. Sexton PT, Woodward DR, Gilbert N, Jamrozik K. Interstate differences in trends in coronary mortality and risk factors in Australia. Aust NZ J Med 1990; 152: 531-4. 2. Gibberd RW, Dobson AJ, Florey C du Ve, Leeder SR. Differences and comparative declines in ischaemic heart disease mortality among subpopulations in Australia 1969-1978. Int J Epidemiol 1984; 13: 25-31. 3. Martin CA, Hobbs MST, Armstrong BK. Estimation of myocardial infarction mortality from routinely collected data in Western Australia. J Chron Dis 1987; 40: 661-9. 4. Dobson AJ, Gibberd RW, Leeder SR. Death certification and coding for ischaemic heart disease in Australia. Am J Epidemiol 1983; 117: 397-405. 5. Kircher T, McIlwaine WJ, Donnelly MDI er al. Certification of death from ischaemic heart disease in Belfast. Int J Epidemiol 1985; 14: 560-5. 6. Kircher T, Nelson J, Burdo H. The autopsy as a measure accuracy of the death certificate. N Engl J Med 1985; 313: 1263-91. 7. Cameron HM, McGoogan E. A prospective study of 1152 hospital autopsies I. Inaccuracies in death certification. J Pathol 1981; 133: 273-83. 8. Stehbens WE. An appraisal of the epidemic rise of coronary heart disease and its decline. Lancet 1987; ii: 606-1 1. 9. Martin CA, Hobbs MST, Armstrong BK. Measuring the incidence of acute myocardial infarction: the problem of possible acute myocardial infarction. Acta Med Scand 1988; Suppl 728: 40-7. 10. Ross J, Brandenburg RO, Dinsmere RE er al. Guidelines for coronary angiography: a report of the American College of Cardiology/American Heart Association Task Force on Assessment of Diagnostic and Therapeutic Cardiovascular Procedures (Subcornmi& on Coronary Angiography). Circulation 1987; 76: 963A-977A. 11. Chaitman BR, Bourassa MG, Davis K er al. Angiographic prevalence of high risk coronary artery disease in patient subsets (CASS). Circulation 1981; 64: 360-7. 12. Coroners Act 1957, Tasmania; Australian Government Printing Service.
Aust NZ J Med 1992; 22
SEXTON ET AL.
