Could time trends in myocardial infarction incidence be due to diagnostic inconsistency? A study of the validity of hospital discharge diagnoses.

Abstract

Dear Sir, since the beginning of the 1980s there has been a steady decline in coronary heart disease mortality in Sweden. Less is known about time trends in acute myocardial infarction (AMI) incidence. In Sweden, studies from Stockholm and Gothenburg indicate declining trends regarding AMI incidence as well as AMI mortality [12]. Methods for using routinely collected hospital discharge data and mortality data in combination to identify incident cases of AMI in a defined population have been developed previously [3]. These methods may be used to shed further light on time trends in AMI incidence in Sweden. A national AMI register based on these methods has recently been set up at the Swedish National Board of Health and Welfare [4]. In an epidemiological investigation of AMI in relation to drinking water hardness, the AMI incidence was followed between 1981 and 1990 in Umeå, a university city in northern Sweden [5]. Incident cases of AMI were identified using the methods based on routinely collected information about hospital discharges and deaths. During the study period a decline in AMI incidence for men of about 30%, and for women of about 15%, was observed. In order to evaluate whether this observed decline could be explained by changes in the quality of the diagnostic information, medical records and autopsy protocols were examined using defined criteria for fulfilment of the diagnosis of AMI. This examination was carried out for all males aged 30–64 years and a random sample of those aged 65–89 years with a hospital discharge diagnosis of AMI from the University Hospital in Umeå during 1981–82 and 1984–85. These years were chosen for the examination because the change in water hardness occurred in 1982–83. Altogether medical records for 205 subjects were examined by one investigator (GB). Our diagnostic criteria for definite or possible AMI were a slight modification of the principles employed in the MIAMI (metoprolol in acute myocardial infarction) trial [6]. These criteria were based on information regarding typical symptoms and enzyme and ECG changes in reasonable time for infarct definition. For a diagnosis of definite AMI, typical conditions regarding at least two of these indicators were required or conclusive findings at autopsy. Patients who did not fulfil the criteria for definite AMI were classified in the categories of possible AMI, previous AMI and others. The age-adjusted relative risk amongst men aged 30–89 years for AMI incidence in Umeå during 1984–85 compared with 1981–82 was 0.85 (95% confidence interval 0.73, 0.98). There was essentially no difference in the percentage of patients fulfilling the diagnostic criteria of definite AMI between the two time periods (86%, 1981–82; 87%, 1984–85) (Table 1). The proportion of patients classified as possible AMI was also similar during the two time periods (12%, 1981–82; 11%, 1984–85). Almost all patients (98%) were thus classified as definite or possible AMI. There was no substantial difference in autopsy rate between the two time periods. Section was performed in 17 of 27 (63%) fatal cases during 1981–82 and nine of 17 (53%) during 1984–85. In the present study, the proportion of patients not fulfilling the diagnostic criteria for definite AMI was similar to corresponding proportions previously observed in the county of Stockholm and in Central Sweden [789]. The results of this study seem to confirm that there is a reasonable diagnostic quality of the AMI diagnosis in Sweden. In addition, the results suggest that the observed decline in AMI incidence in Umeå was not due to diagnostic inconsistency. Sincere thanks are given to Anders Wahlin, Head of Department of Medicine, University Hospital of Northern Sweden, Umeå, who kindly gave us permission to examine medical records. The study was supported by The National Environment Protection Board.