Mortality rates from coronary heart disease (CHD), and from acute myocardial infarction (AMI), in particular, have been declining steadily since the 1970s.1 2 Data from the MONICA study suggest that two-thirds of this decline can be attributed to changes in the incidence of first CHD events.1 Since the end of the MONICA study, CHD incidence has been monitored through national registries based on continuous updating of routinely collected data from hospital records.3 These data, however, are not without limitations. In this issue of Heart, Chan et al4 show that data from total hospital admissions can be misleading (see page 1589). Previous reports based on the same data showed a marked increase in hospital admissions for AMI, signalling a new epidemic of CHD.5 In a thorough reanalysis, Chan et al demonstrate that this is solely due to an increase in AMI readmissions and changes in diagnostic practices, rather than an increase in the incidence of first AMI.4 In fact, as in several other Western populations,1 2 the incidence of first AMI has been declining steadily in New Zealand since 1993.

Figure 1 shows data from the European World Health Organization (WHO) regional office on total CHD hospital admissions in selected European countries, based on national hospital registries.6 These data do not routinely distinguish first from recurrent admissions. The figure would appear to point to relatively small changes in CHD admissions during the past 5 years in the Netherlands, Spain and Italy, an increase in Denmark and a decline in Sweden. To what extent do these trends reflect changes in first CHD events, readmissions or diagnostic practices? To understand the relevance of this question, it is essential to reflect on the rationale for CHD surveillance.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Unadjusted total (first and recurrent) hospital discharges in selected European countries in the period 1989–2006 based on data from the European health for all database (HFA-DB).6

RATIONALE FOR MONITORING TRENDS IN HOSPITAL ADMISSIONS

Trends in CHD admissions are usually used as a proxy for trends in first incidence. They reflect the impact of primary prevention, and contribute to decisions on the development and maintenance of public health interventions.3 7 On the other hand, changes in recurrent CHD events are likely to reflect changes in the prevalence of CHD survivors, which point towards the impact of secondary prevention and treatment.3 The distinction between first and recurrent events is critical because information conveyed by them leads to fundamentally different policies either on primary prevention, or on secondary prevention and treatment. Needless to say, it is also essential to differentiate real trends in CHD incidence from artefactual trends exclusively attributable to changes in diagnostic practices, hospital transfers or population ageing.

HOSPITAL ADMISSIONS AND INCIDENCE OF FIRST CHD

Findings from Chan et al4 are puzzling. Studies in other populations indicate that total admissions can accurately reflect true trends in first incidence. Although information from routine hospital data (which includes both recurrent and first events) overestimates the actual incidence of first AMI, relative trends in incidence are not substantially changed by excluding recurrent events and double counts in the Netherlands8 and Denmark.3 Results from these European studies contrast with the findings by Chan et al4 of a dramatic increase of 106% in total AMI hospitalisations from 1993 to 2005, but a decline of about 15% in the age-adjusted incidence of first AMI in the same period.

Why do trends in total AMI admissions reflect trends in the incidence of first AMI in some European populations, but not in New Zealand? To a large extent, this is owing to the fact that about half of all admissions in New Zealand were readmissions, and this proportion increased from 52.6% in 1993 to 60.2% in 2005. In contrast, in several other Western populations only about 15% of all admissions are readmissions.8^–10 This proportion does not appear to have changed substantially,8^–10 and readmission rates have remained relatively stable over time in these populations3 8. Thus, the growing number of readmissions makes total hospitalisations a poor proxy for AMI incidence trends.3 8

Several factors may account for the large proportion of readmissions and double counts out of all admissions in New Zealand.4 First, the adoption of treatment technologies may have occurred more rapidly in some hospitals, leading to more hospital transfers and rising double admission counts. This phenomenon may be less marked in the Netherlands or Denmark, where most hospitals may have benefited equally from improvements in treatment technology. Second, survival after AMI may have changed faster in New Zealand than in other countries, leading to a larger increase in the pool of survivors and readmissions. Alternatively, hospital registration methods and readmission thresholds may differ between New Zealand and other countries, and may have changed over time in a different way in New Zealand than in other populations.

HOW SHOULD WE MONITOR TRENDS IN CHD INCIDENCE?

Regardless of the factors that explain the peculiar trends in hospital readmissions in New Zealand, the main lesson to be learnt from Chan et al4 is that monitoring of CHD incidence cannot rely exclusively on crude numbers of total hospital admissions. This is particularly important in the context of increases in survival after CHD,1 2 11 which will inevitably lead to a growing number of survivors at risk of being readmitted.3 As a consequence, total admissions will include a growing proportion of readmissions, making them inadequate to examine trends in first incidence and assess the impact of primary prevention.

To the extent possible, surveillance of CHD admissions should meet a set of basic criteria. The first admissions should be distinguished from recurrent admissions. This can be achieved by linkage of hospital admission records with population registries, or by using unique individual identifiers across hospitals.8 9 Second, owing to population ageing, analyses of trends need to use age-adjusted estimates. In Chan et al,4 the yearly average decline in the age-adjusted rate of first CHD admissions between 2000 and 2005 (5.0%) was twice as big as the decline in unadjusted rates of first CHD admissions (2.5%).4 This suggests that ignoring demographic changes can result in an underestimation of the decline in CHD incidence rates.

Third, trends in the incidence of AMI need to be interpreted in the context of changes in other coronary conditions, particularly angina. In Britain, for example, a substantial decline in the rate of major coronary events over the past two decades has been offset by an increase in the incidence of diagnosed angina.12 As a consequence, the incidence of first diagnosed CHD has declined little over time. Although at first this would appear to suggest little impact of primary prevention, the reciprocal trends for angina and major coronary events raise the possibility that the pattern of CHD is shifting from major events to less severe manifestations of disease.12 Therefore, surveillance should monitor admissions separately for different conditions across the entire spectrum of acute coronary syndrome.13 This poses challenges for the development of strategies to account for changes in the definition of AMI and other forms of CHD.13

Finally, Chan et al4 examined trends in non-fatal first CHD events that resulted in hospitalisation, but did not include first CHD events that led to out-of-hospital death. It is estimated that of all first AMI events, about 40% are fatal but do not lead to hospitalisation, and are therefore not included in hospital registries.7 Estimates of AMI incidence based solely on hospital admissions are thus insufficient to examine trends in both fatal and non-fatal admissions.4 A promising approach to this problem is to link hospital admission records with death registries using unique personal identifiers, or combinations of demographic characteristics that can uniquely identify patients.7 9 Such linkages offer a more comprehensive picture of trends in the incidence of first CHD.7 As a minimum, changes in CHD incidence should be interpreted in the broader context of changes in CHD mortality rates.

Do we have an alternative for monitoring trends in CHD that overcomes the limitations of routine hospital admission registries? Clinical databases such as the national audit of myocardial infarction project (MINAP) database in the United Kingdom are a hopeful alternative.14 These databases contain more accurate diagnostic and clinical information than routine hospital registries. Although their availability is currently limited, existing clinical databases could be used to develop adjustment factors to correct CHD incidence trend estimates obtained from hospital registry data. Nevertheless, clinical databases are expensive to compile and maintain; they are only available for recent years; and they have only been used to examine the quality of hospital care.14 Thus, although clinical databases are well suited for monitoring performance,14 their feasibility for monitoring international trends in the incidence of CHD is yet to be assessed.

In conclusion, hospital admission registries remain the main source for monitoring trends in CHD. They do not always accurately reflect changes in incidence of first CHD, but they can deliver accurate estimates of CHD trends if they meet a set of basic criteria. Trends in CHD hospital admissions vary widely across countries (fig 1). Determining to what extent they reflect true differences in CHD incidence trends remains the crucial challenge in monitoring CHD world wide in the coming years.