Mortality inequalities by occupational class among men in Japan, South Korea and eight European countries: a national register-based study, 1990–2015

Background We compared mortality inequalities by occupational class in Japan and South Korea with those in European countries, in order to determine whether patterns are similar. Methods National register-based data from Japan, South Korea and eight European countries (Finland, Denmark, England/Wales, France, Switzerland, Italy (Turin), Estonia, Lithuania) covering the period between 1990 and 2015 were collected and harmonised. We calculated age-standardised all-cause and cause-specific mortality among men aged 35–64 by occupational class and measured the magnitude of inequality with rate differences, rate ratios and the average inter-group difference. Results Clear gradients in mortality were found in all European countries throughout the study period: manual workers had 1.6–2.5 times higher mortality than upper non-manual workers. However, in the most recent time-period, upper non-manual workers had higher mortality than manual workers in Japan and South Korea. This pattern emerged as a result of a rise in mortality among the upper non-manual group in Japan during the late 1990s, and in South Korea during the late 2000s, due to rising mortality from cancer and external causes (including suicide), in addition to strong mortality declines among lower non-manual and manual workers. Conclusion Patterns of mortality by occupational class are remarkably different between European countries and Japan and South Korea. The recently observed patterns in the latter two countries may be related to a larger impact on the higher occupational classes of the economic crisis of the late 1990s and the late 2000s, respectively, and show that a high socioeconomic position does not guarantee better health.


-3. International Statistical Classification of Diseases (ICD) codes for the causespecific death
Cause of death ICD-10 European countries and South Korea Japan

Appendix 2 -Definition of occupational class
We categorized occupational class into five categories: upper non-manual workers (e.g. professionals, managers), lower non-manual workers (e.g. clerical, service, sales workers), manual workers (e.g. craft and related trades workers, semi-skilled and unskilled manual workers), farmers and self-employed. These occupational classes were defined following the Erikson-Goldthorpe-Portocarero (EGP) scheme which was specifically developed for international comparisons. In some countries (Finland, France Switzerland, and Italy(Turin)), data were pre-harmonized by applying the EGP scheme to national data on mortality by occupation before supplying the harmonized data to the coordinating center. For other countries (Denmark, England, Estonia, Lithuania, Japan, and South Korea), detailed data on mortality by occupation that were not yet classified according to the EGP scheme were sent to the coordinating center and post-harmonized there. Occupations in European countries were originally classified according to the International Standard Classification of Occupations (ISCO), whereas occupations in Japan and South Korea were originally classified according to the Japanese Standard Occupational Classification (JSOC) and the Korean Standard Classification of Occupations (KSCO), respectively.
Appendix Table 2-1 shows the definitions of the occupational classes used in our analysis according to ISCO, JSOC and KSCO. Appendix Table 2-2 provides the distribution of occupational classes by level of education in each of the participating countries. Appendix Table 2-3 shows percentages of self-employed by occupational class from the population census in Japan and South Korea. Because self-employed was lacking in Japan, and South Korea (and Estonia), we conducted sensitivity analysis using an indirect estimation of mortality among the self-employed in Japan and South Korea (using Appendix Table 2 -3). The sensitivity analysis is presented in Appendix 5. We need to interpret with caution, especially comparison between European countries and Japan, South Korea because Japanese Standard Occupational Classification (JSOC: https://unstats.un.org/unsd/cr/ctryreg/ctrydetail.asp?id=1155) and Korean Standard Classification of Occupations (KSCO: https://unstats.un.org/unsd/cr/ctryreg/ctrydetail.asp?id=1160) do not directly correspondent to International Standard Classification of Occupations (ISCO: http://www.ilo.org/public/english/bureau/stat/isco/).    1990-1999 2000-2009 2010-2014 Educational level Low  Self-employed ------

Appendix 3 -A procedure for adjusting estimates of occupational mortality differences for the exclusion of economically inactive men
The last occupation is unknown of economically inactive men in most datasets. When these men are excluded from the calculation of mortality by occupational class, this may lead to bias because economically inactive men tend to have higher mortality than economically active men, and because men in lower occupational classes have a greater likelihood of being economically inactive than men in higher occupational classes. Therefore, we applied the procedure originally developed by Kunst et.al 1,2 to account for the exclusion of inactive men when computing occupational class differences in mortality. This algorithm was also applied by Toch-Marquardt et.al 3 on data that are also partly included in our analysis.
This algorithm aims to compute a "correction factor" that represents the extent to which the mortality rate among active men in each occupational class would have to be raised in order to correct for the exclusion of economically inactive men from that class. For occupational class z, the formula for the correction factor is: with # $ %&'()%*+ being the proportion of inactive men by occupational class, and --$ %&'()%*+/'()%*+ being the mortality rate ratio of inactive compared to active men by occupational class. These proportions and relative risks by occupational class are not available in most mortality dataset, but can be estimated using information from other datasets and some simplifying assumptions. 1. The proportion of inactive men by occupational class was estimated as follows: This formula assumes that the proportion of inactive men in occupational class z is equal to some multiplication factor M times the national average proportion of inactive men. The latter is known in all mortality datasets, whereas the multiplication factor for each occupational class was computed from EuroStat, 4 the Japanese Population Census, 5 and the South Korean social survey. 6 Appendix Table 3-1 shows the multiplication factor (M z ) by occupational class in all countries and time-periods. Here, the multiplication factor (M z ) means that the proportion of inactive men in occupational class z is equal to M z times the national average proportion of inactive men. Appendix Table 3-2 shows percentages of occupational class unknown (# '33 %&'()%*+ ).
2. Under the simplifying assumption that there is no interaction between occupational class and activity status, the mortality rate ratio of inactive compared to active men for each occupational class can be estimated as follows: The (age-adjusted) mortality rate ration for inactive as compared to active men in the whole population could be computed in each mortality dataset. Substituting (A-3.2) and (A-3.3) into formula (A-3.1) then gives: The observed mortality rates in each occupational class were multiplied by the correction factor as estimated in formula (A-3.4) to derive mortality rates adjusted for the exclusion of economically inactive men. Formula (A-3.4) can be extended to a formula for the comparison of relative inequality (class y to class z), as follows: Formula (A-3.5) was used as a correction factor for mortality rate ratios.  1990-1994 1995-1999 2000-2004 2005-2009

Appendix 5 -Sensitivity analysis using an indirect estimation of mortality among the selfemployed in Japan and South Korea
Employment status (employed versus self-employed) is not registered on the death certificate in Japan and South Korea, and it was therefore impossible to calculate mortality rates for the self-employed in these two countries, in contrast to most European countries where the self-employed could be distinguished as a separate occupational class. Because this could have biased our results, we conducted a sensitivity analysis, in which we indirectly estimated the mortality rates for the self-employed in Japan and South Korea, and removed the estimated numbers of deaths and person-years of the self-employed in Japan and South Korea from the other occupational classes. We also re-estimated the average inter-group difference (AID absolute and relative version) for five occupational class (upper non-manual, lower non-manual, manual, farmer, and selfemployed) using the new estimates for Japan and South Korea. In order to estimate mortality among the selfemployed in Japan and South Korea, we proceeded as follows; 1. Percentages of self-employed by occupational class in Japan and South Korea were obtained from the census (in contrast to the death certificate, the census did register employment status; shown in Appendix Table 2-3). These percentages were used to estimate population numbers and person-years for the self-employed in Japan and South Korea, and these were then subtracted from the person-years for the other occupational classes. 2. We estimated the rate ratio of mortality among the self-employed as compared to the whole population in European countries from the available datasets (Appendix Table 5 -1). This showed that this rate ratio has an average value of around 0.80. We assumed that the same rate ratio applies in Japan and South Korea. 3. We re-estimated mortality by occupational class using the results of steps 1 and 2 as follows; where proportion z,self-employed is proportion of self-employed among occupational class z (data from Appendix Table 2-3). Here, occupational class z refers to upper non-manual, lower non-manual, or manual workers (according to the EGP scheme, the occupational class of farmers should include both employed and selfemployed farmers).
Appendix Table 5-1. Rate ratios of mortality among self-employed (with mortality in whole population used as a reference) 1990-1994 1995-1999 2000-2004 2005-2009  Appendix Figure 5-1 and 5-2 show that levels and trends of mortality by occupational class in Japan and South Korea, re-estimated using the indirectly estimated mortality among the self-employed, were not essentially different as compared to those presented in the main text of this paper. We also calculated the mortality rates for the self-employed in Japan and South Korea using different rate ratios for their mortality as