In this study, we estimate the number of deaths, proportions of deaths, and years of life lost attributable to educational inequality using data from Spain. During the study period, there were approximately 420,000 average annual deaths in Spain, of which more than 80,000 were estimated to be attributable to educational inequality (AF = 19,1%). We find important differences between education groups, sexes, age groups, and causes of death. In absolute terms, older groups contributed most to the mortality attributable to educational inequality, while in relative terms, younger age groups had a higher share of deaths attributable to inequality (e.g., 50% in males aged 35–39). This inequality is explained by specific causes of death, led by circulatory diseases in females and by neoplasms and circulatory diseases in males.
Strengths and limitations
Using recent data for the period 2016-21, containing more than 2.5 million deaths, we give details on mortality attributable to educational inequality for broad and granular sets of causes. However, our study relies on an assumption that deserves attention. In our counterfactual, we assume that the whole population would die at the same rate as the highly educated, as has been done similarly in previous studies [14,15,16,17]. This is somewhat unrealistic in the short term, but it is possible in the medium term. Such a massive change does not imply population shifts between nominal education groups but rather the elimination of circumstances leading to mortality differences across educational groups. In other words, educational attainment does not necessarily directly kill (or protect) per se, but it is a proxy for or determinant of other variables that are linked to mortality risks (e.g., income, occupation, health behaviours, racism, and myriad others). Our findings are sensitive to the education definitions used: The number and distribution of population and death counts across education groups have a strong impact on the potential number of deaths attributable to inequality. For instance, if we could distinguish more socioeconomic groups, the number of estimated deaths attributable to inequality would tend to increase, and vice versa. Finally, we did not analyse trends over our study period, and we suggest these should be evaluated in further studies.
Comparison and explanation of results
Our results showing that 19% of mortality is attributable to educational inequality are similar to those obtained for the populations of South Korea and the USA [14, 16], but the differences in study settings and educational groupings are a barrier to accurate comparison of estimates. Our estimates of the attributable fractions for ages 35–74 were 19% and 31% for females and males, respectively, and appear to be lower than those derived using area-level socioeconomic indicators in the UK in 2013-18, which found that this explained 33% and 37% [17], respectively, and compared with the Japanese study for ages 30–79, which found that 11–16% of mortality was attributable to educational inequality [15]. Direct comparison of these figures may be misleading due to differences in the choice of socioeconomic variables and groupings, age ranges, group-specific mortality conditions, and population age structures. Yet, our results suggest that in Spain, a country with relatively low (educational) inequalities in mortality [2, 7], the magnitude of these inequalities and the potential for improving mortality at the population level mortality are substantial.
To give a better sense of the magnitudes we report, we translate these results into other intuitive metrics. Estimated mortality due to educational inequality would represent, for instance, the third cause of death for both females and males, with comparable results to those from COVID-19 causes in 2020, which accounted for 16.4% of deaths according to official statistics [26]. Counterfactual remaining life expectancy at age 35 (that of the high education group) would be 52.9 and 48.6 years for females and males, respectively, 1.6 and 5.5 years higher than that observed in 2019 (see Table S4). These life expectancy gaps are equivalent to 10 and 30 years of recent life expectancy progress before the pandemic (2009–2019, and 1989–2019)for females and males, respectively. Additionally, our estimates are higher than those from the COVID-19 pandemic in 2020 [27], mainly due to the younger age profiles for deaths due to education inequality, particularly for males.
Additionally, our results can be put into context by comparing them with estimates of attributable fractions for major risk factors, for instance smoking. According to recent estimates, the AFs of smoking-related mortality were 4.3% and 14.1% in females and males, respectively [28]. These figures contrast with the higher figures (19% for both sexes) we estimate for education, suggesting that reducing educational inequalities in mortality would potentially have a greater beneficial effect on mortality than eliminating smoking-related mortality. Nevertheless, we should acknowledge that the AF difference between smoking and education owes to the relative sizes between age-specific population groups (current smokers vs. low educated) rather than to the relative mortality penalty, which is higher for smoking. Again, although the underlying assumptions used to obtain these estimates are strong and both factors tend to be correlated, these figures help contextualize the large impact of educational inequalities on mortality in Spain.
The detailed cause-of-death information we use in this study allows us to discuss potential health determinants. Cardiovascular causes accounted for an important number of deaths due to inequality for both males and females and were led by ischaemic heart disease and stroke. Although both males and females have relatively similar estimates of mortality attributable to educational inequality, important differences in age patterns and causes of death exist. For females, deaths attributable to inequality occurred at older ages and were led by cardiovascular diseases both in terms of deaths and years of life lost, while male deaths tended to occur at relatively younger ages, and there was more variability in the leading causes (neoplasms and circulatory causes accounted for 56% of all deaths attributable to inequality). Beyond cardiovascular causes, deaths attributable to inequality among females were dominated by Alzheimer’s disease (1,800 annual deaths) and other dementias (3,350 annual deaths), genitourinary diseases (over 3,000 annual deaths), and due to diabetes, were dominant (see Table S3). A glance at the cause-of-death contributions to the observed AF reveals these to be largely considered amenable to healthcare or avoidable through primary intervention (e.g. lung cancer, diabetes, hypertensive heart disease, and ischemic heart disease), highlighting the significance of socioeconomic inequality for reducing mortality.
For males, the figures for Alzheimer’s disease, other dementias, and genitourinary diseases were less than half the estimates compared to those for females, which is consistent with evidence of sex differences in the prevalence of dementia [29]. On the other hand, male educational gradient is important for causes that are related to smoking behavior: lung cancer or chronic lung diseases, as well as liver diseases and external causes of death (Table S3). Beyond smoking, the magnitude of the described socioeconomic inequalities in mortality suggests that lower socioeconomic classes are exposed to additional risks. Well-established cardiovascular risk factors, such as the prevalence of obesity, hypertension, diabetes, or high blood cholesterol in blood, are among the mechanisms through which socioeconomic gradients exist [30]. In both absolute terms and in attributable fraction terms, hypertension and diabetes have a greater impact on females compared to males. Other mechanisms that may discriminate more against lower socioeconomic groups include food choices, housing quality or health care [3], although their quantification is beyond the scope of this study. Overall, the differential impact of well-established risk factors between males and females seems to be explained by the different composition of mortality, as the differences between males and females in these main factors disappear when focusing on premature mortality (see Table S3).
The AF age patterns, higher at young ages, discernable in Fig. 2 merit some speculation. First, the decreasing pattern over age is driven entirely by the underlying pattern of relative risk (see Figure S1). Second, education outcomes partly depend on health [31], which has the effect of increasing relative risk, and by extension AF, in younger ages. This pattern is seen more broadly over a range of acquired health conditions or social statuses, where interactions between socioeconomic status and health are postulated as key drivers [32, 33]. Third, we also know that health selection in the low-education group is increasing over time [34]), such that the age pattern we see could be partly driven by cohort patterns. The health conditions that might co-determine education outcomes and later contribute to elevated younger-age AFs are themselves, to a large degree, socially determined and therefore are indirectly accounted for within our counterfactual.
The rapid educational expansion of the Spanish society in recent decades suggests that, in absolute numbers (i.e. deaths attributable to inequality), the inequalities described in this paper are unlikely to be increasing, and that absolute sex differences seem to be decreasing. That is, the observed and expected changes in the educational composition of the Spanish population [35] are favourable for further reducing absolute educational inequalities and sex differences. However, unhealthy lifestyles among young generations remain a major concern for current and future health dynamics. For instance, several studies have demonstrated the social importance and its impact on the adoption of lifestyles by adolescents, such as binge drinking [36], or the increase in the prevalence of obesity [37, 38], which is associated with parental education and social deprivation [39, 40]. Life course approaches accounting for understanding the role of (cumulative) risk factors and lifestyle exposures and interactions on health outcomes have great potential to influence mid- and long-term population health and mortality outcomes [41]. Further studies should monitor unhealthy lifestyles in younger generations and the impact they may have on current and future all-cause mortality and mortality inequalities.
Education is one of many social determinants of health, and we have shown its power to be substantial in examining socioeconomic inequalities in Spain, in line with findings from previous studies focusing on other countries [14,15,16]. Our results imply that interventions that shift the education distribution upward represent an indirect form of prevention for various causes of death. Yet, we acknowledge that the potential health effects of further educational shifts may be more important in the mid- and long-term. Similar mechanisms should be expected for any socially equitable intervention, including those that do not contemplate health as an outcome, as has been the case for contemporary educational expansion. Increases in social equality, a goal that is valuable in itself, also act to improve population health and contribute to overall mortality reduction. In the short term, policy interventions improving access and quality of public health care or strengthening the social security system may contribute to narrowing socioeconomic health gaps.