The algorithm currently in use may underestimate the incidence of acute coronary events in Australia, according to this validation study.
Monitoring the incidence of acute coronary events is critical to assess the health and economic burden of coronary heart disease (CHD) on the Australian population and for health service planning. There are currently no reliable national or jurisdictional data on the number of acute coronary events. Hence, the AIHW has developed a proxy measure that combines unlinked hospitalisation data from the AIHW National Hospital Morbidity Database and deaths data from the AIHW National Mortality Database. An algorithm (a problem-solving method) is required to avoid double counting. It makes a number of assumptions about how acute coronary events are recorded in the 2 data sets. For instance, it assumes that acute coronary events ending in death in hospital will have an acute CHD underlying cause of death, and that hospitalisations for acute coronary events ending in transfer have a subsequent episode with an acute myocardial infarction (AMI) or unstable angina (UA) principal diagnosis.
This working paper presents the results of a validation study using linked hospitalisation and deaths data from WA and NSW. Analysis of these data sets indicates that the assumptions underlying the method are largely valid and it provides a reasonable measure of the incidence of acute coronary events. Nonetheless, there are some limitations:
- Around 36% and 34% of people in WA and NSW, respectively, who died in hospital with a principal diagnosis of AMI or UA recorded for the hospitalisation did not have a cause of death of acute CHD in the deaths data.
- Around 5% of hospitalisations in WA and 11% of hospitalisations in NSW coded with a separation mode of 'transferred to another acute hospital' did not have a subsequent episode identified in the linked data.
- Following a transfer to another hospital, not all identified subsequent episodes are coded with a principal diagnosis of AMI or UA (4% and 7% of all AMI/UA hospitalisations in WA and NSW, respectively).
- Clinical practices and coding are changing over time. The proportion of deaths in hospital reported as acute CHD in the deaths data is declining and hospital transfer rates are increasing.
A comparison of age-standardised event rates calculated using the linked and unlinked data, suggests that the unlinked data appears to underestimate the rate of acute coronary events (by 6% in WA and 11% in NSW in 2007). The results also suggest that trends over time are comparable within jurisdictions, but, given the difference in the level of underestimation between WA and NSW, comparisons between jurisdictions should not be made.
In the absence of an acute CHD event register or national linked hospitalisation and deaths data, the current methodology for estimating the incidence of acute coronary events is at present the best approach, despite the limitations outlined. This linked data study has broadened our understanding and awareness of key issues influencing the estimation of rates of acute coronary events using unlinked administrative data.