The magnitude of socioeconomic inequalities in AMI incidence varies between countries and periods, and is related to socio-demographic factors such as age and gender [2,11,16]. In general, age and gender are the most important factors that affect health. Some studies have suggested that living in a deprived neighbourhood may affect coronary health of women to a greater extent than men [6,11-13]. Furthermore, a decrease in the socioeconomic gradient with age has been described [6,11]. However, most of the studies on the association between SES and AMI were too small or limited in age range to allow for explorations of the varying magnitude of socioeconomic inequalities across a wide range of age- and gender-groups [7,8,10,12].

SES was determined at the neighbourhood-level by mean income of the neighbourhood of residence using the regional income register (RIO). [18] The RIO is a national study that contains information on income levels of neighbourhoods based on the tax information of about one third of Dutch inhabitants. The Netherlands (N=15.6 million in 1997) is subdivided into 11,412 neighbourhoods with a mean population of 1,364 inhabitants (range: 1 – 32,786). Mean equivalent household income - a measure of disposable income in proportion to the household composition - was used as SES indicator in our study. Less than 0.1% of Dutch inhabitants had missing information on neighbourhood SES. The study population was divided into socioeconomic quintiles according to the ranking of neighbourhoods in the RIO of 1997 weighted by the number of inhabitants per neighbourhood (Q1 is the most affluent quintile and Q5 the most deprived). In this manner each quintile contained about 20% of the person-years at risk from the Dutch population.

The number of person-years at risk from the Dutch population was calculated by age, gender and socioeconomic quintile from 1997 to 2007 and used as denominator for the incidence rate calculations. The numerators of AMI incidence cases and denominators of person-years at risk from the uniquely identifiable part of the total Dutch population were used to study socioeconomic differences in AMI incidence in relative terms. Thus, for incidence rates non-unique persons were excluded from both AMI incident cases and person-years at risk. A pilot study has suggested that non-uniqueness on the linkage variables relates to large cities, foreign origin and age, however differences are small for the determinants large cities and foreign origin [19]. To examine the impact of socioeconomic inequalities on differences in the absolute number of AMI events in the Netherlands we have inflated the number of AMI events to correct for the exclusion of non-unique persons. Adjustment for the 16% of Dutch inhabitants who were not included in our study enabled the presentation of a nationally representative number of AMI events that was attributable to socioeconomic disadvantage. We used age-gender specific inflation factors, which were calculated on the population registry of the mid-year of the study period 2002.

Our results on incidence rates and the socioeconomic gradient are consistent with other studies of AMI incidence based on neighbourhood-level socioeconomic status [4-6,11,22]. Socioeconomic relative risks were modest, as expected when considering the Netherlands as a relatively small and homogeneous country. The relation between age, gender and the effect of neighbourhood SES on AMI incidence in the Netherlands corresponded to that found in Scottish, Swedish, French and Italian studies [6,11-13], with a steeper socioeconomic gradient in women compared to men, and a decrease in the socioeconomic gradient with increasing age. We can think of several explanations for this age pattern in relative socioeconomic inequalities. Firstly, premature CHD disproportionately affects the most deprived groups. Simultaneously with the increase in the number of AMI events with increasing age, at middle age in men and at early old age in women, the socioeconomic relative risk of AMI started to decrease. High socioeconomic relative risks could be related and limited to premature AMI events. Secondly, socioeconomic inequalities in cardiovascular risk factors are observed to be larger among younger than among older persons, especially regarding smoking [23]. Thirdly, a healthy survivor effect may partly explain the observed decrease in the socioeconomic gradient with age. Selective mortality could narrow socioeconomic inequalities with age since disadvantaged people die younger leaving relatively robust survivors [24,25]. Fourthly, in very old age a substantial part of the population is institutionalized, for whom the neighbourhood of residence might not accurately represent SES.

The apparent “contradiction” between relative and absolute perspectives on socioeconomic inequalities in AMI incidence can be explained by considering the factors that determine the absolute number of excess events attributable to socioeconomic inequalities. Although the socioeconomic gradient in AMI incidence is larger in women and at younger ages, the socioeconomic effect is diluted by the increasing absolute incidence rates with increasing age and male gender. The age-gender structure of the population is the third contributing factor to the absolute number of excess events. We have provided age-gender pyramids to place our findings in the perspective of the demography of the Dutch population. The combination of the three factors resulted in the largest absolute number of excess AMI events attributable to socioeconomic inequalities being found in middle-aged men and middle-aged and elderly women.

This is the first Dutch study to estimate the proportion of AMI incidence attributable to socioeconomic inequalities by using population attributable risk methods [22,26,27]. Hallqvist et al. [27] compared relative and absolute differences in AMI risk according to socioeconomic status in Swedish men and women, based on individual-level SES derived from self-reported occupation. In their study, which concerned a comparison between manual workers and low-level employees with high- and middle-level employees in the age range of 45–64years, they found a population attributable risk proportion of 17% in men and 30% in women over the years 1992–94. The present study found similar results with proportions of 17% in 45–64year old men and 32% in 45–64year old women attributable to socioeconomic inequalities between 1997 and 2007 in the Netherlands. A study of Ramsay et al. [22] estimated population attributable risks in a population of British men aged 60–79years old between 1998 and 2000. The population attributable risk for AMI incidence of manual versus non-manual social classes was estimated at 12%. Men aged 65–74 in the present study showed a population attributable risk of 14% based on neighbourhood socioeconomic status. However, population attributable risks are difficult to compare across studies. The population attributable risk depends on both the socioeconomic relative risk and on the prevalence of the exposure, in this case the distribution over socioeconomic category. In addition, studies vary widely in their definition of SES. Both individual-level measures (e.g. income, education and occupation) and neighbourhood-level aggregated data or deprivation indices are frequently used to study socioeconomic health differences.

Mean equivalent household income at the neighbourhood-level was used as SES indicator in our study. Income levels have shown to be a good indicator and determinant of SES [28], even in more egalitarian countries [29]. It has been claimed that neighbourhood income has an impact above and beyond the effect that personal income itself exerts on individual health [8,30,31]. The neighbourhood socioeconomic context is thought to contribute to the disadvantage of individuals through material, psychological, physical and social mechanisms [4,32,33]. In addition, the effect of neighbourhood SES can be in part either due to or mediated through conventional risk factors [34]. For example, prevalence of smoking, obesity and physical inactivity were found to be higher among more deprived populations in Sweden, independent of individual-level socioeconomic status [3]. Approximately 50% of the relative and absolute socioeconomic difference in CHD risk can probably be explained by the four behavioural and biological risk factors - hypertension, smoking, high cholesterol and diabetes [26,35,36].

This nationwide study has several strengths but also some limitations. Strengths are its large size, population-based nature and the wide range of age- and gender-groups studied. A limitation is that the Dutch hospital discharge register was digitally available for record linkage only from registration year 1995 onwards. Most recurrent events occur within one year after the first AMI events [37], although some AMI events, particularly in the beginning of the study period could have been misclassified as being incident events. Because the socioeconomic gradient in AMI risk has been reported to be of similar magnitude in recurrent and incident AMI events [38], we did not consider this limitation a problem. A second limitation to address is that neighbourhood SES was assessed only at a single point in time based on the first place of residence in the study period. Neighbourhood of residence may have changed during follow-up. Moving out of areas might have diluted the effect of neighbourhood to a small extent, although most residential mobility in the Netherlands is to neighbourhoods with comparable neighbourhood socioeconomic status [39]. Besides the effect neighbourhood itself exerts on health, neighbourhood SES might also serve as proxy for individual-level SES. Unfortunately, our nationwide study did not allow us to disentangle the neighbourhood, individual and behavioural and biological effects captured by neighbourhood-level SES.

Population attributable risks can inform policy makers in planning public health interventions [40]. Nonetheless, some caution should be taken in the interpretation of population attributable risks associated with socioeconomic inequalities. The starting point of population attributable risk calculations is the assumption that there is a causal relationship between exposure and disease. Since conventional risk factors may mediate rather than confound part of the effect of neighbourhood SES on CHD [10,41], we considered the method appropriate to estimate the number of potentially preventable AMI events attributable to socioeconomic inequalities. With our estimates of the relative risks, population attributable risks and absolute numbers of excess events due to socioeconomic inequalities we have provided information which can be used in prevention at the individual level, and ultimately, to improve population health. It is not realistic to expect that the total estimated population attributable risk proportion, that was attributable in the past, could be avoided entirely in the future. This would essentially mean to eliminate all inequality. Therefore we have also estimated the potential impact of a population shift in the risk for AMI associated with socioeconomic inequalities, adopting a population approach. Public health policies aimed at reducing socioeconomic inequalities in AMI incidence should take note of the considerable benefit of shifting the population distribution, even in seemingly egalitarian countries.