METHODS

At selected intervals since 1973, the NHAR has collected extensive information on clinical care and outcome on all patients (irrespective of age) referred acutely to Nottingham’s two district general hospitals and presenting with symptoms thought by the admitting medical team to be suggestive of acute myocardial infarction. Details of its operation have been reported.⁸ Patients in all acute settings are included in the NHAR.

Subjects and setting

For this exercise, a cohort of patients who were admitted in 1992 was identified from the register. We have previously reported on postdischarge management and on the survival of those with confirmed⁹ and unconfirmed⁷ myocardial infarction.

Case definitions used in the NHAR

• Definite myocardial infarction—Symptoms compatible with acute myocardial infarction accompanied by both definite ECG changes of myocardial infarction and rises in both of two cardiac enzymes (lactate dehydrogenase and creatine phosphokinase) to above twice the upper limit of normal.

• Probable myocardial infarction—Symptoms compatible with acute myocardial infarction and either definite ECG changes of myocardial infarction or rises in cardiac enzymes to above twice the upper limit of normal.

• Possible myocardial infarction—Symptoms compatible with acute myocardial infarction and either an abnormal ECG (but not characteristic of myocardial infarction) or abnormal cardiac enzyme rises to less than twice the upper limit of normal.

• Ischaemic heart disease—Symptoms compatible with acute myocardial infarction and only old ECG changes (old Q waves on ECG) with no new sustained ECG or cardiac enzyme changes.

In this study, definite myocardial infarction and probable myocardial infarction are jointly analysed as confirmed myocardial infarction.

Survival

The methods used to identify and subsequently trace the original cohort have been published.^7,9 We report the complete survival history for a minimum of four years following discharge from the index event for patients in all the above categories.

Initial working diagnosis

The NHAR database records both the final case definition above, and also the initial working diagnosis of the admitting medical team.¹⁰ Although it is a primary requirement of registration that all patients are investigated for a suspected myocardial infarct, the initial working diagnosis reflects the clinical certainty of myocardial infarction based on the history and presenting ECG. The initial working diagnosis is based on the admitting doctor’s differential diagnosis list:

• chest pain of uncertain cause

• definite myocardial infarction: the diagnosis is not in doubt

• chest pain—rule out myocardial infarction; differential diagnosis: cardiovascular system only

• other symptoms: tests requested, predominant symptom not chest pain

• heart failure: main symptom is breathlessness or, clinically, pulmonary oedema, and an underlying cardiac cause is sought.

Statistical methods

Kaplan–Meier and Cox proportional hazards analyses were used to explore survival. Alternative case definitions used in published series of acute cardiovascular events were applied to the NHAR clinical data. Reclassified cases were then analysed in a similar manner. The primary outcome was survival status at 30 days from admission and four year survival following discharge alive.

Final classifications were based on cardiac enzyme changes and ECG changes. Data were also available on the initial clinical working diagnosis at admission and other clinical variables. These were examined in a multivariate model that took account of age, sex, a past history of myocardial infarction, the presence of abnormal ECG or cardiac enzyme results, Killip score on admission, whether receiving a diuretic on discharge, and the classification into confirmed myocardial infarction or not. These factors were chosen on the basis of the previous demonstration of their prognostic value.¹¹ Agreement between the different classification systems was calculated using the κ statistic. Diabetes mellitus comorbidity was not accurately recorded in this dataset in 1992.

Reclassification of NHAR cases

Modification of the NHAR classification

Two approaches were used:

1. The case definition of “confirmed myocardial infarction” was expanded to include patients with any of the following: a rise of a single enzyme to more than twice the upper limit of normal; sequential ST and T wave changes in the absence of Q wave development; new T wave inversion.

2. To address more recent changes in clinical practice, a diagnosis of “acute coronary syndrome” was assigned to all confirmed cases of acute myocardial infarction, and those with an initial working diagnosis of “definite myocardial infarction” or “chest pain—rule out myocardial infarction” where either the ECG or cardiac enzymes were found to be abnormal.

Use of alternative classification systems

MONICA classification

The NHAR dataset was recoded to fulfil as nearly as possible the MONICA criteria.² Patients over 64 years of age were excluded in MONICA. The hospital admitted cases included in the NHAR were equivalent to the Definite non-fatal myocardial infarction cases identified in MONICA.

Oxford myocardial infarction study (OXMIS)

This study closely followed the MONICA coding scheme but included patients under 80 years of age.¹²

UK heart attack study (UKHAS)

The classification system used in this study was clinical and based on at least two of the following criteria: a typical chest pain history, a doubling of creatinine kinase, and an abnormal ECG (sequential ECG changes compatible with myocardial infarction—either new Q wave development, or ST and T wave changes, or left bundle branch block (Gaylani E, personal communication). Cases in this study were restricted to patients aged 75 years or less.¹³

Clinical classification

The data were reclassified to reflect more familiar clinical definitions of ST elevation myocardial infarction, non-ST elevation myocardial infarction, unstable angina, classical angina, and “other presentations” (which were generally more atypical, but still conformed to presentation with suspected myocardial infarction and having ECG and/or cardiac enzyme abnormalities). These categories were compared with the cases of confirmed myocardial infarction defined by the epidemiological methods discussed above.

RESULTS

Nottingham patient population

In 1992, 4571 patients were admitted to the two Nottingham hospitals with suspected myocardial infarction. Of these, 81% fell into the four NHAR categories outlined above. A further 850 patients were classified as “chest pain of unknown cause” as ECG and cardiac enzymes were normal. Patients from outside Nottingham were excluded (owing to a lack of follow up data), as were recurrent admissions. The remaining 2739 patients fell into the diagnostic categories outlined above (table 1). Of these, 286 died during admission (in-hospital survival of 90%), leaving 2453 patients for the survival analysis following hospital discharge. Survival of all those admitted was 87% by 30 days and 58% after four years (1130 deaths). Of those discharged alive, 844 (34%) subsequently died during the four year follow up period.

Patients classified as having “confirmed myocardial infarction” were slightly younger and had a substantially lower prevalence of reported previous myocardial infarction than patients classified as not having had myocardial infarction, whichever classification system was used (NHAR, MONICA, or UKHAS).

Survival according to original NHAR diagnosis

The crude (unadjusted) in-hospital mortality of the different diagnostic groups was substantially different (table 1). The 30 day mortality (table 2) was worse in patients of all ages with confirmed myocardial infarction; even after adjustment, the NHAR category was a highly significant predictor of 30 day survival, with confirmed myocardial infarction remaining the group with the worst prognosis.

Survival after discharge from hospital showed a pattern that was distinct from that observed in hospital, patients with confirmed myocardial infarction having a significantly better four year prognosis during this time than patients not classified as having confirmed myocardial infarction, before and after adjustment for available prognostic variables (adjusted hazard ratio range 0.78–0.87).

Reclassification of NHAR cases

The results of the reclassification of NHAR cases are given in table 2.

Modified NHAR classification

Recoding the cases to encompass single cardiac enzyme or ECG changes as defined increased the number of patients classified as having sustained a myocardial infarct by 69% in hospital (from 723 to 1224), and by 89% in those who were discharged alive (from 564 to 1065).

In all, 1661 patients fulfilled the recoding criteria for an “acute coronary syndrome” (an increase of some 1.3-fold compared with the original NHAR confirmed number of myocardial infarcts).

Survival at 30 days was worse in the recoded confirmed myocardial infarction group and correspondingly better in the remainder of the cohort.

MONICA and OXMIS classifications

Recoding of the NHAR cases applying MONICA criteria increased the number of patients classified as having a confirmed myocardial infarct by 69% (from 723 to 1051), and by a similar amount if the analysis was restricted to patients under 65 (as in the original MONICA study) or under 80 (OXMIS).

UKHAS

The number of patients coded as having confirmed myocardial infarction increased by 12% (from 723 to 807). Patients under 76 (the UKHAS study age group) showed the same patterns of 30 day and four year survival as the original NHAR and the MONICA coded groupings (table 2).

Clinical classification

These results are shown in table 3. The numbers of patients classified as ST elevation myocardial infarction (STEMI) or non-ST-elevation myocardial infarction (NSTEMI) were 583 and 344, respectively. This gave a total of 927 cases with a clinically defined myocardial infarct, an increase of 28% on the originally defined NHAR confirmed number. The 30 day survival from admission of patients with STEMI and NSTEMI was 75% and 94%, respectively; the difference between these two groups remained highly significant following adjustment (hazard ratio 3.1, 95% confidence interval 1.9 to 4.9). Patients classified as STEMI were almost all identified by the epidemiological classifications of confirmed myocardial infarction, but there was less agreement for NSTEMI.

Differences in survival of confirmed myocardial infarction cases identified by different coding schemes

The unadjusted 30 day survival in confirmed cases was significantly different between the different classification systems (table 4). This effect was less evident under the age of 65, but this may have reflected smaller numbers and wider confidence intervals. There were no corresponding significant differences in postdischarge four year survival. The κ statistics for comparison between the MONICA and amended NHAR and between the UKHAS and NHAR systems were 0.78 and 0.68, respectively, indicating “substantial” or “very high” levels of agreement in both comparisons.

Distribution of major prognostic variables between epidemiological classification systems

The distribution of available major prognostic variables in confirmed cases of myocardial infarction between the different classification systems was significantly different (table 5). Although median age was similar between classification systems, there were differences in the frequency of other important prognostic variables.

DISCUSSION

Clinical outcomes of 30 day mortality following myocardial infarction are now being used as measures of hospital performance. The potential for variations in 30 day survival shown here, which are simply due to the way confirmed myocardial infarction cases are defined, reinforces the need for a consistent and robust national classification system before league tables using this outcome should be compared. Other audit activity that relies on establishing a denominator of confirmed cases of myocardial infarction—such as the current Royal College of Physicians myocardial infarction audit work including examination of thrombolysis times, or standards for cardiac rehabilitation after myocardial infarction—may be subject to similar variation unless case definitions are clear and unambiguous. Adjusting for significantly different distributions of prognostic variables appears to be necessary if different classification systems are used. However, this is difficult and reinforces the need for robust and agreed case definitions.

Epidemiological and familiar pragmatic clinical definitions of ST elevation myocardial infarction, non-ST elevation myocardial infarction, and angina may not include or identify the same patients (table 3), particularly for less “typical” presentations. For instance, of the 344 patients defined as having non-ST elevation myocardial infarction in this cohort, many were not identified as confirmed cases of infarction by the rigorously defined original NHAR classification (35%), with greater proportions being identified using MONICA (84%) or UKHAS (89%).

The method used to define myocardial infarction has significant implications for epidemiological monitoring, for those whose work requires a commercial driving licence, and for insurance purposes. Clinically, inappropriate underdiagnosis may deny some patients access to monitoring and interventions of proven benefit. Uncertainty of diagnosis may also be associated with additional psychological morbidity.¹⁴ Numbers of patients defined as confirmed myocardial infarction and survival at 30 days varied considerably, and this study suggests that for these variables rigour in case definition is necessary. Postdischarge four year survival patterns were similar whichever system was used.

The case definition of acute coronary syndrome focused on the identification of patients with “typical” chest pain.⁶ However, NHAR data suggest that significant numbers of suspected cases of myocardial infarction might be excluded because symptoms were “atypical”. Across all classification systems, patients with more atypical presentations not only showed greater variability in the classification accorded to them, but those not classified as “confirmed acute myocardial infarcts” had similar postdischarge mortality to confirmed cases.¹¹

The NHAR eligibility criteria might be considered overinclusive, but if that were the case, the finding of a high postdischarge mortality in unconfirmed cases of myocardial infarction is even more unexpected and a cause for concern. However, all the patients in this study had abnormal ECGs or cardiac enzymes during the index admission, and have been shown to have predominantly cardiovascular causes of death.⁷ It is therefore likely that as a cohort they were generally suffering from coronary heart disease, an observation made in other settings.⁴

The recoding of NHAR data was at best an approximation. The method we adopted to identify patients with acute coronary syndrome was deliberately overinclusive, and troponin measurements were not available to refine the selection. It was unlikely that patients misclassified as “acute coronary syndrome” would have had systematically higher mortality, so the finding that the remaining patients with “suspected myocardial infarction” not in the acute coronary syndrome group had comparable mortality is significant. If the use of troponin measurements in such patients were patchy in clinical practice, risk stratification relying on this measure may miss some with a high subsequent mortality.

The NHAR recoding system identified a group with at least as bad a prognosis when compared with the direct measurement of patients with acute coronary syndromes in other populations, and this appeared to be valid. Postdischarge survival of patients recoded as acute coronary syndrome in this study was 87%, 83%, 78%, and 69% at 6, 12, 24, and 48 months, respectively. In other studies of postdischarge survival in patients with acute coronary syndrome, survival has been reported to be 85% at six months,¹⁵ 89% at 12 months,¹⁶ and 74% at 60 months.¹⁷

Although different diagnostic criteria generated different numbers of infarct events, there was much similarity in survival outcome following hospital discharge over a four year period. The recent Joint European Society of Cardiology redefinition of myocardial infarction (2000), based on the use of troponins as markers of myocardial infarction, may improve both diagnostic precision and risk stratification.¹⁸ Our study suggests that this is very important if short term outcomes that compare performance in different settings are to be credible. It remains to be seen what impact these may have on the numbers of patients diagnosed as having sustained a myocardial infarct, and the characteristics of excluded patients, and some doubts remain.¹⁹ Further work is under way to describe in more detail the patients with non-confirmed myocardial infarction in recent cohorts.