Jump to comment:

• Re: Mortality inequalities by occupational class among men in Japan, South Korea and eight European countries: a national register-based study, 1990–2015
  Hirokazu Tanaka, Jung Su Lee, Yasuki Kobayashi and Johan P. Mackenbach
  Published on: 23 December 2019
• The surprising result of manual workers in Korea enjoying lower mortality than non-manual workers is likely due to numerator-denominator bias
  Young-Ho Khang
  Published on: 19 July 2019

• Published on: 23 December 2019

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

  • Hirokazu Tanaka, JSPS Overseas Research Fellow Department of Public Health, Erasmus University Medical Center
  • Other Contributors:
    • Jung Su Lee, Project Associate Professor
    • Yasuki Kobayashi, Professor
    • Johan P. Mackenbach, Professor

  As Prof. Young-Ho Khang points out, numerator-denominator bias may affect the estimation of mortality for the Korean and Japanese populations, because we used a cross-sectional unlinked design.[1] We mentioned the possibility of this bias in our paper, citing a study from Lithuania, which suggests that the mortality of persons with high socioeconomic status may be underestimated as a result of this bias.[2] However, based on a national validation study Prof. Khang suggests that the direction of this bias may work the other way around in the Korean population.[3] Furthermore, because – according to his information – the registration of occupation has changed in South Korea, Prof. Khang also claims that the deterioration of the mortality rates among upper non-manual workers observed in our paper is likely to be an artefact.
  While we agree with Prof. Khang that the direction of the numerator-denominator bias may be different in South Korea as compared to Lithuania, we do not agree that the ‘reverse’ manual/non-manual mortality rate ratio that we found in South Korea can be explained by this bias, or that the unfavourable mortality trends among upper non-manual workers that we observed in South Korea can be explained by a change in registering occupation. Our findings prior to 2005 are similar to those of a longitudinal study that followed participants between 1995 and 2008 and reported low mortality among male managers and professional workers in South Korea.[4] Our stu...

  Show More

  As Prof. Young-Ho Khang points out, numerator-denominator bias may affect the estimation of mortality for the Korean and Japanese populations, because we used a cross-sectional unlinked design.[1] We mentioned the possibility of this bias in our paper, citing a study from Lithuania, which suggests that the mortality of persons with high socioeconomic status may be underestimated as a result of this bias.[2] However, based on a national validation study Prof. Khang suggests that the direction of this bias may work the other way around in the Korean population.[3] Furthermore, because – according to his information – the registration of occupation has changed in South Korea, Prof. Khang also claims that the deterioration of the mortality rates among upper non-manual workers observed in our paper is likely to be an artefact.
  While we agree with Prof. Khang that the direction of the numerator-denominator bias may be different in South Korea as compared to Lithuania, we do not agree that the ‘reverse’ manual/non-manual mortality rate ratio that we found in South Korea can be explained by this bias, or that the unfavourable mortality trends among upper non-manual workers that we observed in South Korea can be explained by a change in registering occupation. Our findings prior to 2005 are similar to those of a longitudinal study that followed participants between 1995 and 2008 and reported low mortality among male managers and professional workers in South Korea.[4] Our study, however, also finds that after 2010 mortality among upper non-manual workers rapidly increased, causing a reversal of the upper non-manual/manual rate ratio. It is unlikely that this is due to a change in the recording of occupation. According to our information, there was a slight change in the method of recording occupation on Korean death certificates during the 1990s, but there has been no further changes since the year 2000.[5-6] Therefore, we believe that the discrepancy between some Korean studies and our results mentioned by Prof. Khang is due to a difference in study period.
  Prof. Khang’s letter implicitly also suggests the possibility that numerator-denominator bias explains our findings on Japan. In contrast to South Korea, there are no longitudinal studies in Japan in which occupational class at baseline can be related to mortality during follow-up. We therefore can only speculate about the magnitude and direction of this bias. However, the method of recording occupation on death certificates or the census has not changed during the study period in Japan, and the deterioration of the mortality rate among upper non-manual workers is therefore unlikely to be an artefact.
  Reference
  1. Tanaka H, Nusselder WJ, Bopp M, et al. Mortality inequalities by occupational class among men in Japan, South Korea and eight European countries: a national register-based study, 1990–2015. J Epidemiol Community Health 2019;73:750-758. doi:10.1136/jech-2018-211715
  2. Shkolnikov VM, Jasilionis D, Andreev EM, et al. Linked versus unlinked estimates of mortality and length of life by education and marital status: evidence from the first record linkage study in Lithuania. Soc Sci Med 2007;64:1392–406. doi:10.1016/j.socscimed.2006.11.014
  3. Kim HR, Khang YH. [Reliability of education and occupational class: a comparison of health survey and death certificate data]. J Prev Med Public Health. 2005;38:443-448. (in Korean)
  4. Lee H-E, Kim H-R, Chung YK, et al. Mortality rates by occupation in Korea: a nationwide, 13-year follow-up study. Occup Environ Med 2016;73:329–35. doi:10.1136/oemed-2015-103192
  5. Vital Statistics Division, Social Statistics Bureau, Statistics Korea (KOSTAT) [Internet]. Available: http://kostat.go.kr/portal/eng/aboutUs/3/2/9/2/index.static (Accessed 23 Sep 2019)
  6. Supreme Court of Korea. Regulation on Document Form of Family Related Registration (in Korean) [Internet]. Available: https://glaw.scourt.go.kr/wsjo/gchick/sjo330.do?contId=2258861&q=%EA%B0%... (Accessed 24 Sep 2019)

  Conflict of Interest
  None declared

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 19 July 2019

  The surprising result of manual workers in Korea enjoying lower mortality than non-manual workers is likely due to numerator-denominator bias

  • Young-Ho Khang, Professor and Chair Department of Health Policy and Management, Seoul National University College of Medicine

  I read with great interest the article by Tanaka and colleagues [1], which examined occupational inequalities in mortality in Korea and reported the surprising result that manual workers in Korea enjoyed lower mortality than non-manual workers. The authors employed unlinked data from Japan and Korea, with population denominators from census data and mortality numerators from death certificates. This type of unlinked data is prone to numerator-denominator bias. A prior Korean study examined the reliability of occupational class between survey and death certificate data using individually linked data from the Korea National Health and Nutrition Examination Survey (KNHANES), clearly showing this possibility [2]. Among 104 deaths of KNHANES participants aged 30-64, the number of deaths among non-manual workers increased from 8 in the survey data to 12 in the death certificate data, while the number of deaths among manual workers decreased from 59 in the survey data to 41 in the death certificate data [2]. The number of deaths in other groups (corresponding to ‘inactive or class unknown’) increased from 37 to 51. Therefore, using unlinked data may result in increased mortality estimates among non-manual workers and other groups and reduced mortality estimates among manual workers [2]. It should be noted that, in Appendix Table 1-2 of the article by Tanaka and colleagues [1], the ‘inactive or class unknown’ group accounted for 44%-51% of total deaths in the most recent 10 years...

  Show More

  I read with great interest the article by Tanaka and colleagues [1], which examined occupational inequalities in mortality in Korea and reported the surprising result that manual workers in Korea enjoyed lower mortality than non-manual workers. The authors employed unlinked data from Japan and Korea, with population denominators from census data and mortality numerators from death certificates. This type of unlinked data is prone to numerator-denominator bias. A prior Korean study examined the reliability of occupational class between survey and death certificate data using individually linked data from the Korea National Health and Nutrition Examination Survey (KNHANES), clearly showing this possibility [2]. Among 104 deaths of KNHANES participants aged 30-64, the number of deaths among non-manual workers increased from 8 in the survey data to 12 in the death certificate data, while the number of deaths among manual workers decreased from 59 in the survey data to 41 in the death certificate data [2]. The number of deaths in other groups (corresponding to ‘inactive or class unknown’) increased from 37 to 51. Therefore, using unlinked data may result in increased mortality estimates among non-manual workers and other groups and reduced mortality estimates among manual workers [2]. It should be noted that, in Appendix Table 1-2 of the article by Tanaka and colleagues [1], the ‘inactive or class unknown’ group accounted for 44%-51% of total deaths in the most recent 10 years in Korea. The percentage in Japan was even greater. Moreover, the methods of recording occupation in Korean death certificates have changed during recent decades. In the 1990s, ‘occupation before death’ was recorded, which was later changed to ‘occupation at the time of occurrence of disease or accident’. This change might have influenced the trends reported by Tanaka and colleagues [1] in mortality according to occupational class between 1990 and 2015. In fact, several national studies in Korea employing individual mortality follow-up have shown clear mortality inequalities unfavorable to manual workers compared with non-manual workers [3, 4, 5]. A recent KNHANES long-term mortality follow-up study revealed that male manual workers aged 30-64 had 3.85 times higher (95% confidence interval: 2.25–6.60) mortality risks than their non-manual counterparts [3]. In conclusion, the surprising reverse pattern in the relationship between occupational class and mortality in Korea reported by Tanaka and colleagues [1] is likely due to numerator-denominator bias, and therefore may not reflect the true situation in Korea.

  1. Tanaka H, Nusselder WJ, Bopp M, Brønnum-Hansen H, Kalediene R, Lee JS, Leinsalu M, Martikainen P, Menvielle G, Kobayashi Y, Mackenbach JP. Mortality inequalities by occupational class among men in Japan, South Korea and eight European countries: a national register-based study, 1990–2015. J Epidemiol Community Health doi: 10.1136/jech-2018-211715
  2. Kim HR, Khang YH. [Reliability of education and occupational class: a comparison of health survey and death certificate data]. J Prev Med Public Health. 2005;38:443-8. (in Korean)
  3. Khang YH, Kim HR. Socioeconomic Inequality in mortality using 12-year follow-up data from nationally representative surveys in South Korea. Int J Equity Health 2016;15:51.
  4. Khang YH, Lee SI, Lee MS, Jo MW. Socioeconomic mortality inequalities in Korea Labor & Income Panel Study. Korean J Health Policy Admin 2004;14:1-20. (in Korean)
  5. Lee HE, Kim HR, Chung YK, Kang SK, Kim EA. Mortality rates by occupation in Korea: a nationwide, 13-year follow-up study. Occup Environ Med 2016;73:329-35.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top