The population attributable fraction of low education for mortality in South Korea with improvement in educational attainment and no improvement in mortality inequalities

In 2010, the PAFs of low education for total mortality were 43.7% and 38.3% in men and women aged 30–44 years, respectively, compared with 43.5% and 32.0% in men and women aged 45–59 years, respectively. This means that if all adults aged 30 years old and over enjoyed the advantages of those with a college education or higher, approximately 32-44% of all deaths could be prevented.

In the United States [25], 11.5% of total mortality was attributed to a low education level in 2000; this figure is 20-30% points lower than the comparable figure in our study. Much of this difference can be ascribed to differences in the definition of low education level. The United States study divided education level into two groups (< high school, ≥ high school diploma or equivalent), and the group with at least a high school education was used as the theoretical minimum risk group. In contrast, we divided education level into three groups (middle school graduate or less, high school graduate, college graduate or higher), the latter of which was used as the counterfactual group. If we had used the same definition of low education level as that used in the U.S. study (< high school, ≥ high school diploma or equivalent), one-fifth of our study subjects would have fallen into the less-educated group, which may have reduced the PAF.

The PAF of low education for total mortality decreased in our study, even though the RRs for total mortality increased over the same period. The reduction in the PAF of low education was mainly due to the improvement in educational attainment in both genders and age groups. Educational opportunities in Korea have steadily increased across levels of study, including elementary, secondary, and tertiary education [10]. In 1945, when Korea was liberated from Japanese colonial rule, only 65% of primary school-aged children were enrolled in school, and the enrollment rate for secondary schools was less than 20%. Thanks to the strong government commitment to the “six years of compulsory primary plan,” the enrollment rate for elementary school increased to 96.4% in 1959, and the goal of universal primary education was fulfilled around the early 1960s [26]. In the 1970s, the implementation of the middle school screening system without examination and the introduction of a high school standardization system provided an opportunity for the spread of secondary education. By 1970, 66% of primary school graduates entered middle school, and 70% of middle school students continued on to high school. In the 1980s and 1990s, college enrollment increased substantially as a result of changes in government policies related to the relaxation of entrance quotas. By 1999, almost all students graduating from elementary and middle schools continued on to the next level of schooling, with 85% of graduates from academic high schools continuing on to higher education. This substantial improvement in educational attainment in Korea can be attributed to state-led educational reforms and a parental emphasis on their children’s education derived from the importance placed by Confucianism on scholarship and education [26-29].

Individuals born between 1967 and 1981 have received the greatest benefits as a result of changes in Korea’s educational policies. Further reflecting changes in social attitudes regarding gender equality, the increase in the proportion of college graduates was higher among women than men aged 30–44 years from 1995 to 2010. However, the most prominent decrease in the PAFs of low education for total mortality was observed among men, not women, during this period. The relatively low reduction in the PAF of low education among women despite greater improvements of educational attainment relative to their male counterparts was due to the ascending trend in RRs among those with low education. Total mortality inequality among Korean women between the ages of 25 and 44 also increased, a difference primarily attributable to changes in suicide rates within this group [18]. The suicide rate for women aged 30–44 years showed an increase in both the RR and PAF for low education groups, consistent with that seen in previous studies. During the financial crisis, the RR for suicide among those with the lowest degree of educational attainment increased in women aged 30–44 years [30], with the strongest effects on mortality seen in those aged 34–44 years [15]. Taken together, these data show that as educational attainment improved, women with the lowest degree of education attainment were at the greatest disadvantage in many respects. Social policies aimed at further helping these disadvantaged individuals are therefore urgently needed.

Note that the suicide rate among individuals in the lowest educational group increased about tenfold between 1995 and 2010 in women aged 30–44 years. Korea’s suicide rate was the highest among OECD countries for 10 consecutive years beginning in 2002 [31], characterized by a relatively low gender ratio (men vs. women), especially among younger adults [32]. Suicide rates among women aged 25–44 years are also the highest among five Asian countries—Korea, Hong Kong, Japan, Singapore, and Taiwan—and have remained that way since 1997 [33]. Factors contributing to these relatively high suicide rates in women include the feminization of poverty and labor market marginalization, especially in women with a low education level [34].

Among older age groups, the PAF of low education for cerebrovascular and heart diseases in men, and for heart disease in women, increased by 3-9% points between 1995 and 2010. This increase was due, in part, to the widening of inequalities in terms of disease-specific mortality. Many studies have reported inequalities in mortality related to cardiovascular disease (CVD), while many others have sought to identify the root causes underlying these inequalities [35-37]. The magnitude and direction of educational inequalities in CVD mortality varied among the United States and 11 European countries, as each country was characterized by a different distribution of cardiovascular risk factors such as cigarette smoking, alcohol consumption, obesity, and lack of fresh vegetables [35]. Numerous studies have reported socioeconomic inequalities in health behaviors related to CVD in Korea [38-42]. In particular, cigarette smoking was the leading factor driving inequality in CVD mortality among men aged 30–64 years [43]. The smoking rate of Korean men aged >15 years of age was the second highest among OECD countries at 48.4% in 2010 [44]. These rates are even more pronounced when stratified based on education, with 64.7% of men aged 25–64 years having a middle school education or less describing themselves as smokers compared to 49.1% of individuals with a college degree or higher. Moreover, the RIIs of current cigarette smoking increased from 1995 to 2006 among both men and women [45], which may have contributed to the increase in the RR of low education for mortality related to CVD. One of the most important factors in the treatment of CVD is time from symptom onset to treatment, and a study conducted in South Korea found inequality in the access to emergency medical services according to education level [46], which may also contribute to the widening of inequalities in terms of CVD mortality.

The PAF is composed of three elements: the RRs for each cause of death as a function of exposure level, the current levels of exposure, and the counterfactual distribution of exposure. According to the equation for the PAF, the accuracy and relevance of the estimated PAF depend on these three elements. Thus, both the limitations and strengths of our study can be discussed in this context.

First, because PAF research usually estimates RRs based on case–control or prospective cohort studies, information about the RRs for every risk factor-disease pair is difficult to obtain. Consequently, RRs are assumed to be universal, and the RR from one population has been applied to many other populations, albeit cautiously [43]. However, the RR may differ across populations and time periods, which can be characterized by varying exposure levels. In this study, we directly calculated the RRs of disease-specific mortality rates by age, gender, and time period based on nationally representative data rather than by applying the RRs of other populations. Furthermore, we considered differences according to the level of exposure. We divided the lower education level into two groups, middle school graduates or less and high school graduates, because those with a middle school education or less are expected to have a higher risk for mortality than high school graduates. These methods allowed us to estimate the PAF of low education more precisely. However, a numerator-denominator bias may affect education-specific mortality rates based on unlinked data for the numerator and the denominator, possibly yielding biased educational mortality differentials. Second, in terms of current levels of exposure, we chose periods of analysis based on census years, and the distribution of the population in terms of educational level was estimated directly for a given census year. Since a census is an official survey of the population of a country, it is suitable for estimating the current levels of educational attainment of a population. This approach constitutes a strength of this study. Third, with regard to the counterfactual distribution of exposure, education, which was used as the indicator of SEP, is an ordinal variable. Thus, the theoretically minimum risk distribution would be found among those who had achieved the highest level of education, a doctorate. However, we chose to use the group with at least a college education as reflective of the theoretically minimum risk distribution because of limitations in the available data for levels higher than college graduation. As a result, we might have underestimated the PAF of low education.