Article Text

Socioeconomic differentials in cause-specific mortality among 1.4 million South Korean public servants and their dependents
1. K Jung-Choi1,
2. Y H Khang2,
3. H J Cho3
1. 1Department of Preventive Medicine, Ewha Womans University School of Medicine, Seoul, Korea
2. 2Department of Preventive Medicine, University of Ulsan College of Medicine, Seoul, Korea
3. 3Department of Family Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea
1. Correspondence to Professor Young-Ho Khang, Department of Preventive Medicine, University of Ulsan College of Medicine, 388-1 Pungnap-2dong Songpa-gu, Seoul, 138-736 Korea; youngk{at}amc.seoul.kr

## Abstract

Background A growing number of investigations have explored the contribution of cause of death to socioeconomic inequalities in mortality in Europe and North America, but few such studies have been performed on Asian populations.

Objectives To analyse the socioeconomic inequality of cause-specific death rates from both an absolute and relative perspective, and to evaluate the contribution of cause of death to total mortality inequality in South Korea.

Methods Data were obtained from public servant health insurance beneficiary records. 1 403 297 subjects aged 35–64 years were followed for 9 years. Health insurance premium levels were used as a socioeconomic position indicator. The outcome variables were all-cause, 11 broad causes and 41 specific causes of death. Mortality differentials were examined using cause-specific age-adjusted mortality, relative indices of inequality, and slope inequality indices.

Results Graded inverse associations between income and mortality were found for most, but not all, specific causes of death. The major contributors to income differentials in total mortality in men were liver disease (15.4%), stroke (12.8%), land transport accidents (10.0%), lung cancer (7.1%) and liver cancer (7.0%). In women, stroke (30.7%), diabetes (9.1%), land transport accidents (6.6%), liver cancer (6.0%) and liver disease (5.1%) were important.

Conclusions The contribution of the cause of death to socioeconomic inequality in mortality in South Korea differed from Western countries. To develop a policy to reduce the magnitude of socioeconomic inequality, an understanding of the major causes of death that contribute to mortality inequality is required.

• Mortality
• socioeconomic inequality
• socioeconomic factors
• contribution
• public health epidemiology
• public health policy
• social inequalities
• socioeconomic

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

## Introduction

A growing number of investigations in Europe and North America have explored the contribution of cause of death to socioeconomic inequalities in mortality. Many previous studies considered that coronary heart disease was a major cause of death and explained the greatest proportion of absolute socioeconomic mortality differentials.1–4 However, the magnitude of the contribution of a specific cause may vary between countries with different cause-of-death structures. For example, in many Asian countries, absolute mortality from ischaemic heart disease (IHD) is much lower than in Western countries.5 Several studies on Asian populations have shown no clear socioeconomic inequalities in IHD mortality, in contrast with northern European countries, the UK and USA, especially among older people.6–10 Only a few studies have investigated the contribution of cause of death to socioeconomic inequalities in mortality among Asian populations.11 Moreover, many studies have used relatively broad classifications of cause of death, which limits the utility of the information to research and policy efforts.

The aims of this study were to analyse the socioeconomic inequalities in cause-specific death rates from both an absolute and a relative perspective, and to evaluate the contribution of cause of death to total mortality inequalities in South Korea (hereafter called ‘Korea’).

## Materials and methods

### Study subjects

This study was approved by the Institutional Review Board of the Asan Medical Center, Seoul, Korea. Data were obtained from public servant health insurance beneficiary records, gathered at the end of 1994 by the National Health Insurance Corporation, Korea. These data covered all Korean government employees, teachers and their dependents, as of 1994. From a total of 1 403 688 subjects aged 35–64 years, the present study included 1 403 297 civil servants and their dependents aged 35–64 years (692 471 men and 710 826 women), excluding missing and inaccurate data (391 subjects, 0.03%). The duration of mortality follow-up was 9 years from January 1995 through December 2003. Unique 13-digit personal identification numbers were linked to information on cause-specific mortality obtained from the Korean National Statistical Office (NSO). By law, all deaths of Koreans must be reported to the NSO within a month of their occurrence. Death certification by a physician is crucial to improve the accuracy of reporting of cause of death.12 13 The proportion of deaths certified by physicians was 81.1% in this study. Deaths occurring during 12 386 011 person-years of follow-up numbered 63 903 in total (43 186 men and 20 717 women).

### Socioeconomic position (SEP) indicators and outcome variables

Health insurance premium levels (53 grades) were used as SEP indicators. Premium levels, which are proportional to the monthly salary of public servants, have been used previously as a proxy measure of income.8 9 14 15 Gender-specific income quartiles were created from 53 premium levels in each 5-year age stratum. For example, in 1994, the monthly salary for the lowest and highest quartiles of 45–49-year-old men was <745 000 Won (US$643.9) and >1 265 000 Won (US$1093.3), respectively. The SEP of a public servant was assigned to his/her dependents. The outcome variables for this study were all-cause and cause-specific mortality. Causes of death were encoded using the International Classification of Disease, 10th Revision (ICD-10). A total of 11 broad and 41 specific (37 for men and 40 for women) causes of death were considered.

We also analysed the magnitude of socioeconomic inequalities for alcohol-related diseases and smoking-related conditions—both of which are amenable to medical intervention. Alcohol-related diseases included conditions associated with alcohol misuse.16 Alcohol-related illness included mental and behavioural disorders caused by alcohol use, alcohol polyneuropathy, degeneration of the nervous system due to alcohol, alcoholic cardiomyopathy, alcoholic gastritis, alcoholic liver disease, alcohol-induced chronic pancreatitis, accidental poisoning by alcohol, intentional self-poisoning by alcohol, and an excessive blood alcohol concentration. A list of smoking-related diseases was constructed by selecting conditions associated with a high RR of mortality (>6.5) in current male smokers, based on the Smoking-Attributable Mortality, Morbidity, and Economic Costs (SAMMEC).17 18 Smoking-related causes of death were chronic obstructive pulmonary disease and cancer of the buccal cavity, pharynx, oesophagus, larynx, trachea, bronchus and lung. Lists of conditions amenable to medical intervention were derived from review papers.3 19–21 Causes of death amenable to medical intervention included tuberculosis and other infectious and parasitic diseases, cervical cancer, breast cancer, Hodgkin's disease, leukaemia, hypertension, stroke, pneumonia or influenza, appendicitis, hernia, peptic ulcer, cholelithiasis and cholecystitis, and complications of childbirth.

### Analyses

Cause-specific age-adjusted mortality by gender was calculated using a direct standardisation method. The Korean census population of the year 2005 by gender was used as a standard. CIs of mortality were estimated assuming a Poisson distribution of deaths. In this study, relative indices of inequalities (RII) and slope indices of inequality (SII) were used as summary measures for health inequalities. Cause-specific RIIs were estimated using Cox's regression, and SIIs were calculated by linear regression analysis. A relative income quartile indicator (a value between 0 and 1), which was assigned by computing the relative position in the cumulative population distribution of the central subject in the income hierarchy, was used to estimate RIIs and SIIs. For example, if the lowest income quartile reflected the bottom 10% of the population, the group was assigned a relative rank of 0.05 (0+0.5×0.10). Next, the relative rank figure was used as an independent variable in Cox's regression (to obtain an RII value) and linear regression (for SII) analysis. Thus, the RII and SII values theoretically reflect the relative and absolute risk difference, respectively, of dying at the lowest end of the income hierarchy compared with the mortality risk at the very top of the income level. RII has been used previously as a measure of relative health inequality.3 7 Previous studies on health inequality have used SII, based on linear regression analysis.2 22 With the use of SII, the contribution of a specific cause of death to absolute mortality inequality was calculated from ((SII of a specific cause of mortality)/(SII of all-cause mortality)×100).

## Results

Table 1 presents the number of subjects and deaths, person-years and distribution by age group and income quartile. Public servants made up 46.0% of the total and their dependents (mostly consisting of spouse or parents) made up the rest. Public servants were more likely than their dependents to be men and were younger than their dependents.

Table 1

Total numbers, person-years, number of deaths and distribution by age group and income quartile of study subjects (public servants and their dependents)

Cause-specific mortality differentials by income quartile in men aged 35–64 years are presented in table 2. Analysis of the results showed a graded inverse association between income and mortality in all broad groups of cause of death and for most specific conditions. However, for colon cancer, prostate cancer and kidney cancer, mortality in men with the highest incomes was higher than in those with the lowest incomes.

Table 2

Number of deaths (N=43 186), age-adjusted mortality by income quartile, relative indices of inequalities (RII values), and slope indices of inequalities (SII values), among South Korean men aged 35–64 years (N=692 471)

Among broad groups of cause of death, RII values were >5 for diseases of the digestive system, infections and parasitic diseases, and all external causes. RII scores for accidental poisoning (20.42), alcoholic liver disease (20.12), alcoholism (17.00), tuberculosis (9.54), suicide (7.04) and liver cirrhosis (6.91) were all relatively high. Those for death from kidney cancer and airplane accidents were 0.62 and 0.04 (data not shown here), respectively; these values are statistically significant.

The broad groups of cause of death that were the main contributors to total mortality inequality in men were cancers, external causes, cardiovascular diseases and digestive conditions (25.7%, 20.8%, 18.0% and 16.7%, respectively). Among specific causes, the major contributions to inequality in total mortality were from liver disease (15.4%), stroke (12.8%), land transport accidents (10.0%), lung cancer (7.1%), liver cancer (7.0%) and stomach cancer (5.6%). IHD accounted for 0.8% of absolute inequality in mortality in men. Alcohol-related causes of death showed the greatest relative inequality; RII was 19.3. However, causes amenable to medical intervention were more important from an absolute viewpoint and accounted for 18.6% of absolute mortality inequality. The contribution of smoking-related conditions to total mortality inequality was 11.8%.

In women, graded inverse patterns between income and mortality were less clear than in men when either broad groups of cause of death or specific causes were examined (table 3). Causes of death such as tuberculosis, stomach cancer, liver cancer, diabetes mellitus, hypertensive disease, stroke, liver disease, land transport accidents and suicide showed clear graded inverse relationships between income and mortality.

Table 3

Number of deaths (N=20 717), age-adjusted mortality by income quartile, relative indices of inequalities (RII values), and slope indices of inequalities (SII) among South Korean women aged 35–64 years (N=710 826)

Generally, RII and SII values were smaller in women than in men. In women, endocrine, nutritional and metabolic disease (due to diabetes), and diseases of the circulatory system showed higher RII values than did diseases of the digestive system (arising from liver disease), which, in men, showed the highest RII. Among specific causes, the RII values for tuberculosis (2.96) and accidental poisoning (2.79) were relatively high in women—similar findings to those in men. Accidental drowning (4.25) and unspecified uterine cancer (2.53) also showed significant and relatively high RII values. Significant RII values favourable to poorer women were noted for non-Hodgkin's lymphoma and pneumonia. The RII values for breast cancer and ovarian cancer were 0.82 and 0.66, respectively.

Cardiovascular disease accounted for 40.4% of the absolute inequality in mortality of women, whereas cancers and external causes explained 18.5% and 13.9%, respectively, of absolute inequality. The contribution from diseases of the digestive system was relatively small, whereas endocrine and metabolic disease yielded a relatively large contribution. In women, specific causes contributing greatly to absolute socioeconomic inequality in mortality were stroke (30.7%), diabetes mellitus (9.1%), land transport accidents (6.6%), liver cancer (6.0%), liver disease (5.1%) and stomach cancer (4.9%). The contribution of liver disease, which explained 15.4% of total inequality in men, was only 5.1% in women. IHD only explained 1.7% of absolute mortality inequality. In women, causes amenable to medical intervention contributed 35.5% of total mortality inequality.

## Discussion

In this study, we calculated mortality differentials by income for 11 broad and 41 specific causes of death, and further analysed the contribution of each cause of death to total inequality using 9-year mortality follow-up data from 1.4 million Korean adults. Major contributors to income differentials in total mortality in men were liver disease, stroke, land transport accidents, lung cancer, liver cancer and stomach cancer, whereas, in women, stroke, diabetes mellitus, land transport accidents, liver cancer, liver disease and stomach cancer were important. These results are consistent with previous studies in which mortality inequality by cause of death was examined in the general Korean population.23

Although direct comparison with other studies is difficult because of the different study subjects (general population versus occupational groups), age groups, study designs and study periods, the patterns in cause-specific contributions in Korea seem to differ from those of the West. In northern European countries and the UK, the contribution of IHD was significant,3 4 24 whereas it was negligible (0.8% in men and 1.7% in women) in this study. The contribution of stroke, as well as IHD, was low in southern European countries, whereas stroke was the leading cause of both absolute mortality and absolute mortality inequalities in Korean men and women. Moreover, in contrast with Western countries, liver disease and liver cancer contributed 22.4% and 11.1% to mortality inequality in Korean men and women, respectively. Different cause-of-death structures may account for these variations in cause-specific contributions. As seen in tables 1 and 2 (based on number of deaths), the leading causes of mortality in Korea were stroke, liver cancer, lung cancer, stomach cancer, liver disease and land transport accidents in men, and stroke, stomach cancer, diabetes, liver cancer, lung cancer and land transport accidents in women. The contributions of these leading causes of death to total mortality inequalities were significant.

Stroke, liver disease, liver cancer and stomach cancer can be related to adverse childhood conditions.25–28 Adult height, a proxy measure of childhood socioeconomic conditions, has been found to be inversely associated with stroke mortality in both Korean men and women.29 30 Early infections with hepatitis B virus and Helicobacter pylori are important in the development of liver cancer and cirrhosis and stomach cancer, respectively.28 31 As deaths from these diseases are decreasing with improvements in living conditions,32 the contributions of these diseases to mortality inequalities should decrease in the future. The risks of lung cancer and diabetes are related to health behaviours such as smoking, diet and obesity, which have changed substantially in Korea with rapid industrialisation and Westernisation.33 34 However, considering that the female smoking rate is still very low in Korea, unlike in most Western countries, passive smoking at home and in other environments may contribute to mortality and mortality inequalities from lung cancer in Korean women.35

Liver disease was found to be the leading cause significantly contributing to socioeconomic mortality inequality in men. In addition to hepatitis B viral infection, gender-specific health behaviours such as alcohol consumption and misuse need to be considered to explain the liver disease inequalities. From previous Korean studies,36–39 the relationship between SEP and drinking habits varies with the study subjects and definitions of drinking habits and SEP indicators. From the finding that alcohol-related death has a high RII, alcohol misuse may have contributed to absolute mortality inequalities for liver disease.

The magnitude of relative and absolute inequalities was greater in men than in women. The contribution of cardiovascular disease (40.4%) was much higher than that of cancer (18.5%) in women, whereas cancer (25.7%) played a greater role than did cardiovascular disease (18.0%) in men. A previous study showed that, in women, the contribution of cardiovascular disease was greater than that of cancer in all European countries.3 However, in some regions (France and certain southern European countries), the contribution from cancer exceeded the influence of cardiovascular diseases in men.3

On the basis of a previous investigation,3 we grouped causes of death into alcohol-related illness, smoking-related causes, and conditions amenable to medical intervention. The results of our analysis reflect the importance of causes of death related to amenability to medical intervention in explaining total mortality inequality, especially in women. This is primarily because stroke has been classified as a cause of death related to medical intervention. We found smaller contributions from smoking-related diseases. However, considering the increasing mortality from smoking-associated causes,40 and rising socioeconomic inequalities in smoking behaviour,33 41 the contribution of causes of death related to smoking is expected to increase.

Significant and marked inverse gradients were found between income and mortality from tuberculosis and accidental poisoning in both genders and from alcohol-related deaths in men. Tuberculosis is known as a disease of ‘poverty’,42 and socioeconomic inequality among adults with tuberculosis may be the result of exposure to poverty in both childhood (for initial infection) and adulthood (for re-activation of Mycobacterium bacilli). Accidental poisoning is usually attributed to exposure to pesticides (51.9% of deaths), according to detailed analysis of death certificates, and such deaths are likely to be concentrated in rural public servants and their dependents with low income. Alcohol-related deaths have been related to socioeconomic differentials in previous studies.43–45 The very high RII values for alcohol-related deaths in men may in part be related to a difference in drinking habits according to job type (manual public servants versus white-collar public servants) or ‘reverse causation’, in which alcohol-related problems may adversely affect career promotion, thus directly affecting income.

In the present study, we found an inverse relationship between income and mortality for most causes of death, but several positive relationships were also found. Several cancers, including non-Hodgkin's lymphoma, kidney cancer and ovarian cancer, showed higher mortality in those of higher income than in those of lower income. Other studies have also documented increased mortality among higher social classes from diseases of the brain, lymphatic system, urinary system and sex hormones.46 47 In the present study, negative relationships between income and IHD mortality were less clear for both genders. Reports from southern European countries presented positive or no socioeconomic gradients in IHD.48 49 Changes in the relationship of SEP with IHD from positive to negative have been reported in northern European countries and the USA,50–52 although it has also been argued that the positive relationships found in the past may be due to misclassification of cause of death for IHD associated with the prevalent views in the early 20th century that IHD is a disease of the affluent classes.53 Positive relationships between SEP and major IHD risk factors (eg, cholesterol and obesity) may explain these findings.8 In addition, heterogeneities in the pattern of socioeconomic inequalities of various causes suggest that no single factor, such as smoking or general susceptibility,54 can account for the results obtained, and specific mechanisms for particular causes of death need to be considered.

Our study has limitations. Firstly, we used public servant health insurance premium levels as a surrogate for the income of public servants and their dependents. Each premium is proportional to a public servant's monthly salary. However, other sources of income were not considered in determining the income levels of public servants' dependents. Despite this limitation, previous Korean studies used the same insurance information as an SEP indicator and found graded inequalities in mortality and morbidity.14 55 Secondly, our subjects are not representative of the overall Korean population. The magnitudes of relative and absolute inequality may be underestimated, because (1) the poorest people with medical assistance were not included in our sample, (2) the mean premium level (which is proportional to the income level) for public servants is higher than for other employees,56 and (3) workers at high risk because of their job, such as construction workers and farmers, may be under-represented. However, our data represent 9.6% of the total population in the 35–64 age group in 1994. In addition, mortality inequalities found in this study were comparable to those from a previous Korean study in which the whole Korean population was examined.7 23 Thirdly, at the end of the follow-up, age at death included age groups >70 years. Mortality inequality may be underestimated in these older age groups because greater mortality inequalities in premature deaths are likely to be found before the age of 70.57

In conclusion, the magnitude of mortality inequality by income varied with cause of death. The leading causes of death contributing to socioeconomic inequalities in mortality were liver disease in men and stroke in women. The relative contribution of cause of death to socioeconomic mortality inequality in Korea was different from that seen in Western countries. To develop a policy to reduce the magnitude of socioeconomic inequality, an understanding of the major causes of death contributing to inequality is required.

### What is already known on this subject

• A growing number of investigations in Europe and North America have explored the contribution of cause of death to socioeconomic mortality inequality.

• Previous studies have found that coronary heart disease explains the greatest proportion of absolute socioeconomic mortality differentials.

• The contribution of various causes of death to socioeconomic mortality inequality in South Korea was found to be different from that of Western countries; the leading causes of death contributing to socioeconomic inequalities in mortality were liver disease in men and stroke in women. Ischaemic heart disease accounted for <2% of absolute inequality in mortality.

• To develop a policy to reduce the magnitude of socioeconomic inequality, an understanding of the major causes of death contributing to absolute mortality inequality is required.

## Footnotes

• Competing interests None.

• Ethics approval This study was conducted with the approval of the Institutional Review Board of the Asan Medical Center, Seoul, South Korea.

• Provenance and peer review Not commissioned; externally peer reviewed.