1 Introduction

Mortality after an acute myocardial infarction (AMI) has been found to be higher among people with low rather than high socioeconomic status [1–3]. This may be due partly to a higher prevalence of risk factors for cardiovascular disease among patients in lower socioeconomic groups [4–6]. Patients with lower socioeconomic status also appear to receive less secondary preventive treatment. In a recent study of AMI patients in Denmark, we found that those with a lower income were 27% less likely to receive statins and 12% less likely to receive beta blockers after an AMI [7]. Furthermore, in other studies, patients with low socioeconomic status underwent fewer invasive cardiac procedures [8–18]. Only a few of these studies, however, addressed temporal trends in socioeconomic differences in cardiac revascularization [10, 11, 15, 16].

A ‘heart programme’ designed explicitly to improve cardiac care was implemented by the Danish Government in 1993. The programme made intensive services more widely available, increased the financial resources for these procedures and established a heart registry to follow the activity and results of the intensive procedures [19]. Between 1993 and 2003, therefore, the number of invasive cardiac procedures increased from 544 to 2,170 per million total population [20]. In addition, the management of AMI underwent major changes in this period [21]. Before 1999, percutaneous coronary intervention (PCI) and coronary artery bypass grafting (CABG) were performed mainly during the non-acute phase of AMI, after further diagnostic testing, and pharmacological reperfusion was used in the acute phase [22]. During 1999–2002, a large trial of primary PCI (PCI performed during the acute phase of AMI instead of pharmacological reperfusion) was performed in Denmark [23]. By 2002, primary PCI was the first choice of treatment for eligible patients.

Our hypothesis was that this wider access to PCI and CABG procedures would reduce any existing socioeconomic gradient, and that the change in the timing of PCI procedures, from the non-acute to the acute phase, would also reduce the gradient, because non-clinical factors such as socioeconomic status would be less important in the acute setting. Thus, the objective of our study was to examine temporal trends in the association between socioeconomic status and use of invasive revascularization procedures in both the acute and the non-acute phase of AMI in Denmark during the period 1996–2004 when use of these procedures increased.

2 Material and methods

2.1 Setting

Denmark (total population, 5.3 million) has a tax-financed universal health care system, with free access for all citizens to hospitals and essential operations, including PCI and CABG procedures [24]. The privately financed hospital sector is small, e.g. approximately 1% of hospital beds were private in 2001 [25]. Invasive cardiac care is organized such that 5 of the 73 hospitals in Denmark perform PCI and CABG procedures.

2.2 Study population

Every citizen in Denmark is given a unique, permanent civil registration number, allowing linkage of individuals among registers. We obtained information from the Danish National Patient Registry, which contains administrative data on every hospitalization in Denmark since 1978 [26], and identified all patients aged 30–74 years who had been hospitalized with AMI as the primary diagnosis [International Classification of Diseases (ICD) version 10, codes I21–I22] between 1 January 1996 and 30 June 2004 (n = 54,184). In order to limit our observation to first AMIs, we excluded any patient who had been hospitalized for this condition in the 17 years before the index admission (24%), with ICD-8 code 410 for the period 1978–1994. The accuracy of AMI coding in the National Patient Registry has been reported to be good (positive predictive value, 93%) [27]. We also excluded any patient who died within the first day of admission, in order to avoid the bias that patients would not be eligible for revascularization due to early death after arrival at the emergency room (3%). Furthermore, patients were categorized by type of admitting hospital, by classifying the 73 hospitals providing acute care for AMI in Denmark into three types: (1) local community hospitals, which are small hospitals with departments of surgery and medicine (n = 35); (2) main regional hospitals, which are larger hospitals with a range of medical specialties, including cardiology (n = 33) and (3) cardiac care centres, which are university-affiliated hospitals performing PCI and CABG procedures (n = 5). The patients’ vital status (alive or date of death) was obtained from the Civil Registration System.

2.3 Income and education

From the statistics on tax-related income provided by Statistics Denmark, we obtained information on the gross income of each patient and cohabiting partner, comprising all income subject to taxation (wages and salaries, all types of benefits and pensions). Income was corrected for inflation since 1990 on the basis of components from Statistics Denmark’s price index. To account for yearly variations in income, we calculated the average income of the patient and cohabiting partner for the 5 years before admission. For patients with a cohabiting partner, income was combined and divided by 2. Because of the high average age of the patient population, we did not adjust for the number of children in the household. Patients were divided into tertiles of increasing income, a method used in previous studies [28].

Information on the highest completed educational level was retrieved from the Integrated Student Registry of Statistics Denmark, which assembles individual-based information on educational level for all residents of Denmark from the administrative registries of educational institutions. Patients were categorized into three groups according to length of education: >12 years (short-, medium- and long-term higher education), 10–12 years (vocational education and upper secondary school) and <10 years (primary and lower secondary school). We excluded patients for whom information on income or education was missing (3%).

2.4 Comorbidity

Primary and secondary diagnoses, both at the index admission and at admissions up to 1 year before the index admission, were used to define morbidity [29]. Diagnoses of congestive heart failure (ICD-10; I50), cardiogenic shock (ICD-10; R57, A41), arrhythmia (ICD-10; I46–I49) and pulmonary edema (ICD10; J81, J18.2) gave indications of the severity of heart disease, while diagnoses of malignancy (ICD-10; C00–C97), diabetes with complications (ICD-10; E10–E14 except E10.9, E11.9, E12.9, E13.9 and E14.9), cerebrovascular disease (ICD-10; I60–I69), acute renal failure (ICD-10; N17, N19, R34) and chronic renal failure (ICD-10; T85.8–9, Z99.2) indicated comorbidity. This method is an extension of the Ontario AMI mortality prediction rule. [30] Thus, gender, age, period and each risk factor in the Ontario AMI mortality prediction rule were retained in the model regardless of their significance level. To further determine comorbidity at the time of admission for AMI, we recorded use 1 year before admission of selected drug classes: antihypertensive drugs (including beta blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors and angiotensin 2 receptor blockers (anatomical therapeutic classification (ATC) C02, C07–9), heart therapy drugs (including digitalis and antiarrhythmic drugs classes I and III, ATC C01), lipid-lowering drugs (ATC C10), diuretics (ATC C03), drugs for chronic obstructive pulmonary disease (ATC R03) and anti-diabetic drugs (ATC A10). Each drug class was also kept in the model regardless of significance level.

2.5 Outcomes

First, we estimated the effect of socioeconomic status on total revascularization (time to either PCI or CABG, whichever came first, within 6 months of admission; patients were censored if they died). Secondly, we stratified the composite outcome into time to CABG within 6 months of admission, time to acute PCI (performed on day 1 or day 2 after admission) and non-acute PCI (performed between day 3 and day 180 after admission) (Fig. 1). We did not stratify CABG into acute and non-acute because only a small number received acute CABG.

Fig. 1
figure 1

Chart illustrating the sub-division of end-points

Full size image

2.6 Statistical analyses

To analyse differences in baseline characteristics across income and education levels, we tested for linear trends in logistic regression models (categorical variables) or linear regression models (continuous variables), where appropriate. In a previous study of the same population we found a socioeconomic gradient in mortality after AMI [1]. Consequently, in this study we used Cox proportional hazards models, with time to revascularization as the outcome variable and censoring for death, to account for socioeconomic differences in observation time, a method used in similar studies [31]. Besides income and education, adjustment was made for age, gender, year of admission, cohabitation status, prior PCI or CABG (within 5 years of the index AMI), comorbidity, residence at county level and type of admitting hospital. Robust estimation was used, on the assumption that observations within individual hospitals were correlated. As reperfusion with thrombolytic medication is not recorded in the National Patient Registry, we were not able to adjust for this procedure. We used the ‘forced entry method,’ in which all the variables are entered into the regression model simultaneously. Model assumptions—linearity of continuous variables, the proportional hazard assumption and lack of interactions—were tested and found valid unless otherwise indicated. For each outcome, we first present the results from a regression with no stratification on time period; then, because of interaction between calendar year and education, we present the results in which the time period is stratified into 1996–1998, 1999–2001 and 2002–2004.

As a sensitivity analysis of whether censoring for death introduced ‘informative censoring’, we reanalyzed the data, performing competing risk analyses with death or revascularization within 6 months of admission as the outcome [32]. Further, we performed a sensitivity analysis in which we allowed for economies of scale for couples, by dividing their combined income by 1.8 instead of 2.0. Neither of these analyses altered the results. All analyses were performed with SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA).

2.7 Ethics

The Danish Data Protection Agency approved the design of the study, and data were made available to us with anonymized but unique personal identification numbers, such that individuals could not be identified. According to Danish law, retrospective register studies do not require approval by the regional committee on scientific ethics.

3 Results

3.1 Baseline data

The population in the present study consisted of 38,803 patients. Table 1 illustrates the univariate relation between baseline patient characteristics and income or educational level. Patients with an income in the lowest third or with <10 years of education were older, had more comorbid conditions and used more drugs than patients with income in the highest third or with >12 years of education.

Table 1 Baseline characteristics of patients with a first acute myocardial infarction by income and education
Full size table

3.2 Revascularization (composite of PCI and CABG)

A higher proportion of patients with high income or education received revascularization during the period (Fig. 2). Table 2 shows the associations of income and education with revascularization procedures, adjusted for all baseline characteristics. For income, there was no statistical interaction with calendar year (p = 0.26), and patients with a high income (HR, 1.12; 95% CI, 1.07–1.17) or a medium income (HR, 1.07; 95% CI, 1.02–1.11) were more likely to receive revascularization than patients with a low income. For education, there was a significant interaction with calendar year (p < 0.001), and in the period-stratified regression the educational gradient reversed during the period (Table 3). When data for women and men were analyzed separately, statistical significance was not achieved for women, possibly owing to the sample size, but the analyses yielded similar trends to those found for men, and there was no interaction between gender and income or education in any of the strata.

Fig. 2
figure 2

Unadjusted relation between income or education level and revascularization within 6 months (composite of CABG and PCI) after first AMI (proportions determined by unadjusted Kaplan–Meier method with censoring for death). *P value for log-rank test <0.001

Full size image
Table 2 Associations of income and education with revascularization (composite of CABG and PCI), CABG, non-acute PCI and acute PCI among patients with a first AMI in the period 1996–2004 (adjusted for all baseline characteristics)
Full size table
Table 3 Association of income and education with revascularization (composite of CABG and PCI) after first AMI stratified into time periods (adjusted for all baseline characteristics)
Full size table

3.3 CABG

CABG was performed within 6 months of hospitalization in 6.8% of patients in 1996–1998, 12.9% in 1999–2001 and 11.8% in 2002–2004. The median time between admission and the procedure was 89 days in 1996–1998, 41 days in 1999–2001 and 21 days in 2002–2004. There was no interaction between income and calendar year, and patients with a high income were more likely to receive this procedure throughout the period (Table 2). No interaction with calendar year was seen with education either, and there was no association with receiving CABG (Table 4). There was a significant interaction between age and education, such that the effect of education was larger for patients aged 65–74 years than patients aged 30–64 years, although the gradient was in the same direction for the two age groups.

Table 4 Association of income and education with CABG after first AMI, stratified into time periods (adjusted for all baseline characteristics)
Full size table

3.4 Acute and non-acute PCI

Acute PCI was performed in 2.4% of patients in 1996–1998, 8.1% in 1999–2001 and 29.1% in 2002–2004; and non-acute PCI was performed in 9.3, 26.6 and 36.1%, respectively. The median time between admission and non-acute PCI was 51 days in 1996–1998, 19 days in 1999–2001 and 12 days in 2002–2004.

For acute PCI, there was no interaction between income and calendar year, and no income gradient was observed throughout the period (Table 2). An interaction was seen between education and period, and an inverse gradient was seen in 2002–2004 (Table 5). For non-acute PCI, patients with a high income were more likely to receive this procedure throughout the period (Table 2). An interaction was seen between education and period, such that the gradient declined during the period (Table 5).

Table 5 Association of income and education with acute and non-acute PCI after first AMI stratified into time-periods (adjusted for all baseline characteristics)
Full size table

4 Discussion

In this population-based cohort study of socioeconomic differences in invasive cardiac revascularization procedures after AMI during 1996–2004, patients with an income in the highest tertile were more likely to receive revascularization (composite of PCI and CABG) than patients with incomes in the lowest tertile throughout the period. The gradient for education reversed during the period: in 1996 patients with >12 years’ education were more likely to receive revascularization than patients with <10 years of education, while the opposite was true at the end of the period. When revascularization was stratified into CABG, non-acute and acute PCI, the income gradient was seen for CABG and non-acute PCI but not acute PCI. The decrease in the gradient for education was present for all end-points, although more pronounced for acute PCI.

4.1 Comparison with other studies

Socioeconomic differences in invasive cardiac revascularization have been found in other countries with universal health insurance, such as Australia [8], Canada [9, 10], Finland [11, 12], Italy [13, 14], Sweden [15], and the United Kingdom [16, 17], but also in the USA [18]. For example, of patients admitted for AMI between 1994 and 1996 in Ontario, Canada, 16% in the lowest income quintile and 20% in the highest received revascularization within 1 year [9]. Of patients admitted for a first acute coronary event in Rome, Italy, between 1998 and 2000, 12.9% of those with ≥13 years of education and 6.7% of those with <8 years of education received PCI within 28 days of hospitalization (odds ratio, 1.27; 95% CI, 1.10–1.42) [14]. Only a few studies have specifically addressed temporal trends in socioeconomic differences in revascularization [10, 11, 15, 16]. A study in Canada showed that access to cardiac catheterization after AMI varied according to socioeconomic status in both 1989 and 1994 [10]. Researchers in England found that additional resources for tertiary cardiology during the 1990s might have reduced socioeconomic inequities in angiography, even though the resources were not specifically targeted at needier, more deprived groups [16]. A Finnish study on the effect on socioeconomic inequity of the 2.5-fold increase in coronary revascularization between 1988 and 1996 showed that, although disparities in use of the operation diminished, some socioeconomic inequities persisted [11].

4.2 Socioeconomic differences in revascularization

The Danish Multicentre Randomized Study on Fibrinolytic Therapy versus Acute Coronary Angioplasty in Acute Myocardial Infarction (DANAMI-2) found a beneficial effect of primary PCI among patients with AMI with elevated ST, and this finding profoundly affected revascularization practice in Denmark [23]. Furthermore, several randomized trials have shown a beneficial effect of early invasive procedures in patients with unstable coronary artery disease or AMI with no ST elevation [33–35]. The increase in use of PCI in the acute phase of AMI has somewhat changed who decides that the procedure is to be performed, from non-specialists or primary-care physicians to sub-specialists. This might have affected the socioeconomic gradient, as Barnhart et al. found that physicians’ perceptions of non-clinical factors, such as low socioeconomic status, unhealthy lifestyle, financial barriers and lack of social support, are taken into account in decisions about coronary revascularization and could preclude its use [36]. They found that family physicians were more likely than sub-specialists to perceive that such factors played a role in decision-making for their patients. Once patients with coronary disease reached cardiothoracic specialists, non-clinical factors were less likely to influence decisions. Furthermore, strict national criteria for performing primary or sub-acute PCI have been implemented in Denmark since 2001, in accordance with the evidence obtained in the randomized clinical trials, promoting a more uniform clinical approach to PCI. We observed a socioeconomic gradient for non-acute revascularizations (CABG and non-acute PCI), the proportion of patients receiving CABG increasing slightly during the period, whereas the proportion receiving non-acute PCI increased considerably. The increase in non-acute PCI was not associated with a decrease in both of the two socioeconomic determinants; the income gradient persisted through out the period, whereas the educational gradient decreased. As revascularization has the greatest benefit for high-risk patients [37], patients with a low socioeconomic status should theoretically receive more revascularizations, because of their higher comorbidity, than those with a high status. A study in the United Kingdom showed considerable underuse of coronary revascularization procedures in patients considered appropriate candidates for revascularization, with those not receiving revascularization having more comorbid conditions [38]. Further, they found that underuse was associated with adverse clinical outcomes, emphasizing the importance of offering revascularization to all appropriate candidates.

As the universal health care system in Denmark provides free access to revascularization for all patients found to be candidates for the procedure, the socioeconomic differences observed are not likely to stem from financial barriers, and factors such as patient and physician preferences and knowledge might play a larger role.

4.3 Limitations

Our study has several limitations. We did not have information on key clinical variables, such as type of infarction (with or without ST elevation), symptoms, smoking habits, hypertension, left ventricular ejection fraction, renal function, number of diseased coronary vessels and left main coronary stenosis. Nonetheless, we adjusted for several important factors, including comorbidity and pre-existing use of a number of medications. Even though we did not have comprehensive clinical information on the extent of coronary artery disease, nothing indicates that patients with a low income have less severe coronary artery disease than those with a high income. Furthermore, information on whether coronary angiography had been performed (both in patients who did and did not subsequently undergo revascularization) was available only for the period 2002–2004. Analysis of these data revealed an inverse gradient between coronary angiography and socioeconomic status similar to that found for revascularization, whereas among persons who underwent coronary angiography there was no relation between socioeconomic status and the likelihood of receiving revascularization. This finding is in accordance with that of a Canadian study on the effect of socioeconomic status on access to invasive cardiac procedures, which showed that access to coronary angiography was the rate-limiting step in access to revascularization [9].

We also had no information on whether patients received pharmacological instead of invasive reperfusion, as revascularization might have been rejected because of admission to a facility with no capacity for invasive reperfusion. A Canadian study found, however, that socioeconomic status was as important a predictor of angiography use in hospitals with ready access to cardiac catheterization facilities as it was in those without [39]. Furthermore, as we adjusted for clustering at the hospital level, hospital type and patients’ residence at county level, we believe that these factors had little effect on the socioeconomic gradients observed. As only a small percentage of the patients who underwent revascularization procedures were treated at private hospitals (3.3%; 493/15,007), it is unlikely that patients with a high socioeconomic status had better access to health-care facilities, as is the case in countries with other health-care systems [40]. While the extent to which our findings are generalizable to other jurisdictions and disease settings is unknown, our study is comprehensive, consisting of all AMI patients aged 30–74 years in Denmark over a period of more than 8 years.

5 Conclusion

We found a persistent income gradient for CABG and non-acute PCI procedures performed after AMI in the period 1996–2004, even though the number of non-acute PCIs increased considerably. For acute PCI, we found no socioeconomic differences in use, except for an inverse relation with education at the end of the period. More research is needed on the reasons why socioeconomic differences in cardiac revascularization exist in a universal health care system such as that in Denmark.