Background Acute myocardial infarction (AMI; ICD9-CM 410*) is a leading cause of morbidity and mortality all over the world, and its community surveillance is essential to monitor variation in the occurrence of the disease. Between the late 1990s and the early 2000s more sensitive and specific biomarkers of myocardial necrosis (ie, troponins) were introduced and new diagnostic criteria, emphasising the role of biomarkers, have been developed for clinical and epidemiological purposes.
Methods Tosc-AMI is a population-based registry based on the record linkage between hospital and mortality databases; it provides trends of coronary events in Tuscany, Italy. Two random samples of patients admitted to hospital in 2003 were validated according to the American Heart Association (AHA; 2003) and the Multinational MONItoring of trends and determinants in CArdiovascular disease (MONICA) (1983) criteria. Sample 1 (380 cases) was represented by patients admitted to hospital for AMI and sample 2 (380 cases) for other coronary diagnosis.
Results Tosc-AMI attack rates increased from the period 1997 to 2005 (men: +17%; women: +30%) and then they decreased in the following 2 years (men: −8%; women: −13%). The rise of AMI hospital admissions was due to cases with ICD9-CM code 410.7 (largely representing non-ST elevation MI). According to the AHA criteria, 94.6% events of sample 1 and 29.8% events of sample 2 fulfilled the most extensive criteria for definite, probable or possible AMI. As expected, the more updated AHA definition identified as definite AMI an additional 33.3% when compared to the MONICA criteria (86.0% vs 52.7%).
Conclusions The study suggests an influence of the new diagnostic criteria on the rising AMI trend observed in the early 2000s, an increase of less severe cases and a decreasing trend of forms with a more extended myocardial damage.
- diagnostic criteria
- attack rate
- heart disease
Statistics from Altmetric.com
Acute myocardial infarction (AMI) is a leading cause of morbidity and mortality all over the world, and community surveillance is essential to monitor the burden of disease, the effects of primary and secondary prevention and of changes in treatment. Surveillance data allow us to describe differences in incidence within and between countries and to study time trends; thus, enabling health professionals and policy makers to make better decisions on planning and evaluation of prevention programmes, healthcare delivery, resource allocation and research.1
Since the middle of the 1980s, the WHO MONICA project 2 introduced standardised diagnostic criteria to classify AMI events (based on symptoms, ECG changes and rise of biochemical markers) and coronary deaths, which have been widely applied in epidemiological studies. In the late 1990s, advances in diagnostic technology brought new, more sensitive and specific biomarkers (ie, cardiac troponins) and more precise imaging techniques. Therefore, diagnostic criteria of AMI have been reconsidered for clinical3–5 and epidemiological purposes.6 The new AMI definitions emphasise the key role of necrosis biomarkers in the setting of ischaemic symptoms or ECG findings, and adds imaging evidence of myocardial damage to the diagnostic criteria. Troponins higher sensitivity allows for the detection of very small myocardial necrosis, which would have gone undetected by previous markers (ie, creatine kinase and its MB fraction). The redefinition of diagnostic criteria had a huge influence on the frequency of AMI diagnosis and obscured the incidence trend of the disease.7 8
Time trend of coronary events should be provided by community registries, such as in the MONICA project, based on routine validation of suspected cases.2 The MONICA project, although very informative, was quite complex and too expensive to be maintained over time. Surveillance systems based on current hospital and mortality databases have been developed in Italy and in other European countries and represent a reliable alternative.1
As discharge and mortality diagnoses are often not validated on a routine basis, validation studies are necessary to check the diagnostic quality, in particular when the adoption of new diagnostic techniques, such as troponin, and new diagnostic criteria may cause major changes in hospital discharge data.1
In Tuscany, Italy, a population-based registry (the Tosc-AMI Registry) based on the record linkage between hospital database and the Regional Death Registry has been active since 1997.9 The present study has two purposes. First, to validate hospital discharge diagnoses for ischaemic heart disease according to the American Heart Association (AHA; 2003)6 and the MONICA (1985) criteria.2 Second, to evaluate the effect of diagnosis validation on AMI hospitalised events identified by the Tosc-AMI Registry and to present and discuss recent AMI attack rates trends.
Material and methods
The Tosc-AMI Registry
The Tosc-AMI Registry, on the basis of record-linkage between hospital discharge diagnoses and mortality databases, identifies hospitalised AMI events (International Classification of Diseases (ICD)-9-CM code 410 as primary discharge diagnosis) and out-of-hospital coronary deaths (ICD9 codes 410–414 as underlying cause of death) as first and recurrent events. A recurrent event is defined as a new AMI hospital admission or a coronary death after 28 days of a previous admission for MI.1
Events are identified by the personal identification number, based on the fiscal code, which allows us to perform anonymously the electronic record linkage among multiple information sources. All patients resident in Tuscany (about 3 500 000 inhabitants) are included in the Tosc-AMI Registry.
AMI attack rates trend
Age-standardised AMI attack rates (standard: Tuscany population in 2001) in the 1997–2007 period, derived from the Tosc-AMI Registry, are presented by gender for total events, hospitalised AMI and out-of-hospital coronary deaths. AMI hospitalised events were then subdivided according to the ICD9-CM fourth-digit subcategories into cases discharged with the 410.7 code and with other 410 codes (410.1–410.6 or 410.8–410.9). The two fractions of cases were presented for the 2001–2007 period. The first group largely identified AMI without ST-segment elevation (NSTEMI) in ECG and the second group largely identified cases with ST-segment elevation (STEMI).3 4
Validation of hospital discharge diagnosis for AMI
A random sample of 380 patients, all ages and both genders, was selected from the 7910 AMI hospitalised cases (all with a ICD9-CM code 410 as primary discharge diagnosis in the hospital database) recorded in the Tosc-AMI Registry during 2003 (sample 1). A second random sample of 100 cases was extracted from the 13 986 hospital admissions during 2003 with a principal discharge diagnosis of other forms of ischaemic heart disease—acute, subacute or chronic (ICD9-CM codes 411, 413 and 414)—and without any other discharge diagnosis of codes 410–414 in the previous 28 days (sample 2).
Trained nurses, under the supervision of a physician, collected the clinical information from hospital documents using a standardised data collection form. Events were classified on the basis of symptoms, ECG and biomarker findings in accordance with the 2003 AHA definitions 6 and the WHO-MONICA project 2 definitions (both considered as ‘gold standards’). Biomarkers were classified as positive on the basis of the limits established by the laboratory of the hospital to which patients had been admitted. Three ECGs per episode (when available) were collected and classified according to the Minnesota code by a trained investigator (FD). Some details of 2003 AHA and of MONICA criteria are shown in table 1. The main characteristic of the AHA definition is the emphasis on the role of biomarkers in a setting of signs, symptoms and ECG findings suggestive of acute ischaemia.
The proportions of patients discharged with code 410 (sample 1) and codes 411, 413 and 414 (sample 2) who had a confirmed diagnosis of AMI according to the 2003 AHA definitions and to the previous WHO-MONICA criteria were assessed (AMI validated diagnosis; AVD). The proportions of AVD are presented with 95% CIs.
Effects of hospital discharge diagnosis validation on 2003 AMI events estimates
The proportions of AVD were applied to current data to assess the number of estimated AMI events in Tuscany according to the MONICA and the 2003 AHA criteria. The proportions of AVD derived from sample 1 were applied to AMI hospitalised events recorded by the Tosc-AMI Registry to estimate the number of patients discharged with 410 codes and truly affected by AMI. The proportions of AVD derived from sample 2 were applied to the hospital admissions with discharge diagnosis of ischaemic heart disease (ICD9 codes 411, 413, 414) to estimate the number of AMI patients discharged with a diagnosis other than 410 code.
AMI attack rates trend based on current data
In both genders, age-standardised attack rates of AMI hospitalised cases, based on Tosc-AMI registry, were substantially stable until 2000, then they significantly rose until 2005 and, thereafter, significantly decreased (figure 1). In particular, between 1997 and 2005, rates showed a 17.0% increase in men and a 30.7% in women (from 272.3/100 000 (95% CI 264.3 to 281.2) to 318.6/100 000 (95% CI 309.4 to 327.7) and from 108.2/100 000 (95% CI 103.4 to 112.9) to 141.4/100 000 (95% CI 136.4 to 146.4), respectively). In contrast, between 2005 and 2007, rates decreased by 7.7% in males and 13.0% in women (2007 rates: 294.1/100 000 (95% CI 285.6 to 302.5) and 123.0/100 000 (95% CI 118.4 to 127.5), respectively).
Between 2001 and 2007, the number of AMI hospitalised cases, based on the Tosc-AMI registry, with the code 410.7 (largely representing NSTEMI) reported on discharge diagnosis (figure 2) showed a marked increase (+135.2%); about four-fifths of this growth occurred between 2002 and 2003. Over the same period, the number of hospitalised AMI discharged with the other 410 codes (largely representing STEMI) progressively diminished by 13.9%.
Age-standardised rates of coronary deaths outside hospital decreased progressively between 1997 and 2007 by 4.8% in men and by 6.8% in woman (from 130.5/100 000 (95% CI 123.8 to 137.1) to 124.3/100 000 (95% CI 118.3 to 130.3) and from 70.5/100 000 (95% CI 66.7 to 74.2) to 65.7/100 000 (95% CI 62.5 to 69.0), respectively).
Quality of AMI hospital discharge diagnosis
A total of 372 (97.9%) hospital clinical records have been reviewed from sample 1 and 97 (97.0%) from sample 2.
Table 2 shows the results of hospital records validation according to the 2003 AHA definition and to the MONICA criteria. In sample 1, 86.0% of events were classified as definite AMI according to the 2003 AHA definition. Altogether, 94.6% fulfilled the criteria for definite, probable or possible AMI, and 5.4% were classified as not AMI. According to the MONICA criteria, 52.7% of events were classified as definite AMI, and 65.3% fulfilled the criteria for definite or possible AMI, whereas the diagnosis of AMI was not confirmed for 34.7%.
In sample 2, 29.9% of suspected events met the criteria for definite, probable or possible AMI according to the 2003 AHA definition. According to the MONICA criteria, a considerably lower proportion of events (23.7%) met the criteria for definite or possible AMI.
Effects of AHA and MONICA hospital discharge diagnosis validations on AMI events estimates
According to the 2003 AHA definition (figure 3A), on the basis of the proportions of AVD obtained from the validation samples, out of the 7910 AMI hospitalised events in Tosc-AMI, 6803 were classifiable as definite, 103 as probable and 577 as possible AMI, whereas 427 were not AMI. Considering the 13 986 patients discharged with a ICD9-CM code 411, 413 or 414 as primary diagnosis and without another discharge for 410–414 in the previous 28 days (not included in Tosc-AMI registry), a further 2308 cases were classifiable as definite, 727 as probable and 1147 as possible AMI. Altogether, the number of AMI hospitalised events increases by 15.2% (to 9111 cases; 95% CI 7777 to 10 444) for definite, by 25.5% (to 9926 cases; 95% CI 8805 to 11 378) for definite or probable and by 47.5% (to 11 665 cases; 95% CI 10 182 to 12 266) for definite, probable or possible AMI.
Considering the proportions of AVD obtained according to the MONICA criteria (figure 3B), the number of AMI hospitalised events decrease by 43.6% (to 4462 cases; 95% CI: 3767 to 5397) for definite or moderately increase by 7.4% (to 8494 cases; 95% CI: 6897 to 10 070) for definite or possible AMI.
Our results provide several insights from the health system perspective into the population trends of AMI events.
The Tosc-AMI registry data, based on hospitalisation discharge diagnosis, show that AMI attack rates increased between 1997 and 2005 by 17% in men and 31% in women (in particular between 2000 and 2003). Thereafter, in the two subsequent years, they decreased by 8% in men and 13% in women. The rise of hospitalised AMI was due to the increase of cases discharged with ICD9-CM code 410.7 (largely represented by NSTEMI). In contrast, AMI cases coded with other ICD9-CM codes (largely represented by STEMI), as well as out of hospital coronary deaths, lowered progressively.
From a clinical point of view, the STEMI/NSTEMI classification, based on initial EGCs findings, is very relevant for treatment and prognosis, and different guidelines have been developed for the management of STEMI 10 and of NSTEMI.11
The majority of patients with STEMI develop a full thickness necrosis of the affected myocardial wall segment (transmural infarction), whereas the majority of patients with NSTEMI develop a more limited necrosis (non-transmural or sub-endocardial infarction).3 4 In particular, STEMI is a condition for which early reperfusion with fibrinolysis or percutaneous coronary intervention should be performed unless clear contraindications are present.10 Furthermore, after correcting for age and comorbid conditions, STEMI shows a higher mortality.12
Our results agree with previous studies performed in the USA showing a reduction of STEMI incidence13 and an increasing proportion of AMI cases diagnosed as NSTEMI.14 In particular, a recent analysis of the Nationwide Inpatient Sample, a very large database of hospital admissions, showed a steady linear reduction of incidence of STEMI since 1996.13 The data from the National Registry of Myocardial Infarction showed that the proportions of STEMI and of NSTEMI were equal in the year 2000 and, afterwards, NSTEMI rose to 59% in 2006.14
Our validation study helps explain time trend in AMI incidence observed in Tuscany on the basis of the Tosc-AMI registry data. Hospital discharge diagnoses for ischaemic heart diseases have been compared to the 2003 AHA6 and to the 1985 MONICA2 diagnostic criteria and, as expected, different definitions produce different AMI rates. First of all, results indicate that AMI discharge diagnoses (ICD9-CM code 410) largely correspond to definitions based on the more up-to-date AHA clinical criteria6 showing the reliability of current data to monitor the epidemiology of disease. In fact, 86% of AMI cases, identified on the basis of hospital current data, were classifiable as definite AMIs according to the 2003 AHA classification.6 Altogether, about 95% of cases fulfilled the most extensive criteria for definite, probable or possible AMI. Second, in our series, the 2003 AHA definition, which emphasises the role of necrosis biomarkers, identified more definite cases than WHO-MONICA criteria (+33%) and lowered the number of cases in which the diagnosis of AMI was excluded (about 5% vs 35%). This result is consistent with many studies evaluating the impact of new diagnostic tests and of new case definition criteria on the diagnosis of AMI.15–20 Besides, as observed in other studies,7 8 results of discharge diagnosis validation indirectly support the hypothesis that the increasing trend of hospitalised AMIs is largely due to the widespread adoption in clinical practice of new diagnostic guidelines and new diagnostic markers. A community study conducted in Minnesota, USA, confirmed that the application of new definitions resulted in a large increase in the number of AMIs.17 In a study performed in Ontario, Canada, the 2003 AHA definition results in a 62–84% increase in the frequency of AMI compared to the WHO-MONICA criteria.18 In the UK, among patients admitted to hospital with a suspect cardiac pain, the 2000 European Society of Cardiology/American College of Cardiology criteria identified an additional 26% of patients as having AMI compared to the WHO criteria.19 Data from the FINAMI study, a population-based AMI registry operating in Finland, showed that the 2003 AHA criteria identified 83% more definite cases than the MONICA definition and detected a sizeable number of definite AMI otherwise classified as not AMI according to the MONICA.16 Data from the FINAMI study also showed that the real declining trend in MI incidence was partly hidden by the effect of the troponin test especially among women and older patients.7 In Perth, Australia, a rapid rise in the use of troponin has been observed in the late 1990s causing an apparent increase in MI hospital admission rates of 42% in men and 21% in women compared to rates expected on the basis of the previous declining trend.8
Finally, our validation analysis suggests that, in 2003, a not negligible proportion of patients discharged with other diagnoses of cardiac ischaemia (ICD-9-CM codes 411–414) met the diagnostic criteria for AMI according to the 2003 AHA rules. Our results agree with previous studies showing an incomplete acceptance of new diagnostic criteria in current clinical practice.8 21 A progressively widespread adoption of the new diagnostic definition in the current clinical practice could help to explain the AMI increasing trend in subsequent years and the opposite pattern for 410 fourth-digit subcategories.
Our study has several limitations. First, the number of patients included in the validation samples is relatively small, so the CIs of the proportions of AVD are quite wide, in particular in sample 2 (patients discharged ICD-9 codes 411–414). Also, the validation samples included only hospital admissions for cardiac ischaemia; therefore, suspected MIs discharged with other diagnoses are not included in the calculations. Second, the ICD9-CM codes for MI did not completely correspond to the current relevant clinical terminology. In fact, the terms STEMI and NSTEMI were introduced only in the 2005 edition of ICD9-CM and the terms NSTEMI and sub-endocardial or non-transmural infarction coexist in the code 410.7.22 This paradox could at least partly affect the time trends of AMI subcategories. Therefore, it can be difficult to get accurate data on the proportion of STEMI and NSTEMI in the population on the basis of current discharge data. Finally, in Italy, the ICD9-CM classification introducing the 410.7 fourth-digit subcategory was only adopted at the end of 2000 for coding hospital discharge diagnosis. Its gradual introduction in the coding practice during the subsequent years can partly explain the increasing trends of the 410.7 code.
Our study shows that AMI incidence rose at the beginning of the 2000s until 2005 and, thereafter, rapidly decreased. Results of validation of AMI discharge diagnoses indirectly confirms the impact on increasing AMI trend of the new diagnostic criteria introduced in the same years and of the widespread use of the highly sensitive troponine test. In particular, the 2003 AHA criteria,6 which emphasise the role of necrosis biomarkers, identified as definite AMI an additional 33% of cases compared to the WHO-MONICA criteria,2 lowering the number of cases in which the diagnosis of AMI was excluded. Besides, results suggest that this increase is due to less severe forms (ie, NSTEMI). The change in diagnostic criteria implies important consequences for patients and physicians. In fact, many patients that in the past were classified as affected by other forms of cardiac ischaemia are now identified and treated as MI patients.
More generally, our data show the importance, in a health policy perspective, of a regional surveillance system, based on hospital admission and mortality data and validated according to clinical and epidemiological criteria, for the prompt identification of the trend of ischaemic heart disease.
What is already known on this subject?
Acute myocardial infarction (AMI; ICD9-CM 410*) is a leading cause of morbidity and mortality all over the world and community surveillance is essential to monitor the occurrence of disease. In the last decade, more sensitive and specific biomarkers of myocardial necrosis (ie, troponins) and new diagnostic criteria emphasising the role of biomarkers have been introduced. The use of new diagnostic criteria and of new diagnostic markers resulted in a large increase in the number of AMI diagnosis.
What does this study add?
Results confirm the impact of the new diagnostic criteria on the increasing AMI trend observed in the early 2000s and show that this increase is due to less severe forms of disease. A significant reduction in AMI incidence has been observed in the most recent years. The study shows the importance, in a health policy perspective, of surveillance systems, based on current hospital admission and mortality data and validated according to clinical and epidemiological criteria, to identify promptly the trend of ischaemic heart disease.
We thank Dr Diego Vannuzzo and Dr Simona Giampaoli for their help in developing the validation scheme of AMI diagnosis.
Competing interests None declared.
Ethics approval At the time of the patients' enrolment in the validation samples, ethics committee approval and informed consent from patients were not required for observational studies in Italy.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.