Occupational variation in healthy worker effects: self-reported - - PDF document

Conference paper International Population Conference 2017 1

Occupational variation in healthy worker effects: self-reported health in Belgium

By Laura Van den Borre & Patrick Deboosere

Affiliation: Interface Demography, Sociology Department, Vrije Universiteit Brussel, Belgium Introduction Occupational health research is complicated by the Healthy Worker Effect (HWE), referring to the continuous selection process of healthy individuals in the workforce. As a result, the comparison of workers’ health with the general population is biased.[1] Due to different hiring policies and specific job requirements, the HWE is likely to be different across industries. Physically demanding occupations should exhibit the HWE more clearly than non-manual labour jobs.[2] If so, will this persist when making the transition to the inactive population? So far occupational differences within the HWE have scarcely been

• investigated. To our knowledge, this is the first study to examine specific occupational groups. This study

follows the total Belgian work force of 1991 and investigates variations in self-reported health for specific

• ccupational groups 10 years later.

Methodology Data were derived from an anonymous record linkage between the Belgian censuses of 1991 and 2001. Linkage at the individual level is possible because each citizen has a unique identification number. An additional linkage with the population register was performed to account for migrations or deaths between the census dates. The total Belgian working population aged 25 to 55 years was selected from the 1991 census and followed up until the 2001 census. A total of 1,773,345 men and 1,176,913 women were employed on 1 March 1991. In the period between the two censuses, 3.2% of male workers and 1.5% of female workers died. Loss to follow-up due to emigration was 2.1% and 1.5% in the male and female working population, respectively. As shown in table 1, the study population consists of approximately 1.6 million men and 1.1 million women who were at work on 1 March 1991 and resided in Belgium on 1 October 2001. Health information was derived from the 2001 census using the question ‘How is your health in general?’ Self-reported health was dichotomized into good (very good/good coded 0) and poor (fair/bad/ very bad coded 1) health. Health questions were not included in the 1991 census. Occupational groups were composed using the 2-digit codes from the International Standard Classification

• f Occupations (ISCO-88) as recorded in the 1991 census. Although detailed occupational information is

not available for 2001, the dataset does include information on the activity status in 2001. This allows us to determine who is still active, unemployed, (pre)retired or inactive due to personal, health or familial reasons.

Conference paper International Population Conference 2017 2 In order to determine which occupational groups experienced stronger HWEs, the health situation in 2001 was compared with three possible scenarios for the 1991 working population. Healthy worker effects entail a comparison problem between the active and total population. As a result, we considered whether members of the 1991 workforce were still active 10 years later. In addition to the activity status in 2001, we also took mortality between the two census dates into account assuming deaths to be the result of poor health. Analyses were performed for men only, as inactivity among women in prime working ages may be motived by caregiving and familial reasons rather than their own health situation.[3]

Table 1 Number and percentage of study subjects with the mean age at the time of the 2001 census and the percentage of subjects experiencing fair or (very) poor health in 2001

Men Women Occupation (Census 1991) N (% of total) Mean age (years) % of N in poor health N (% of total) Mean age (years) % of N in poor health Legislators and senior officials 4427 (0.28%) 53.18 16.38% 1269 (0.12%) 50.97 16.57% Corporate managers 134047 (8.37%) 50.94 15.87% 41217 (3.86%) 49.06 17.57% Managers of small enterprises 62531 (3.91%) 50.37 24.44% 39141 (3.67%) 50.16 26.28% Physical, math. and engineering science professionals 44553 (2.78%) 46.63 10.91% 6876 (0.64%) 42.34 8.70% Life science and health professionals 29762 (1.86%) 48.20 11.34% 73403 (6.87%) 45.67 13.78% Teaching professionals 66821 (4.17%) 51.45 19.07% 113020 (10.58%) 49.02 17.34% Other professionals 56647 (3.54%) 49.18 16.05% 41848 (3.92%) 46.66 15.24% Physical and engineering science assoc. professionals 118132 (7.38%) 49.64 20.69% 19564 (1.83%) 46.97 17.11% Life science and health assoc. professionals 10549 (0.66%) 47.80 13.26% 21365 (2%) 46.18 13.66% Teaching associate professionals 8161 (0.51%) 47.90 18.41% 16065 (1.5%) 46.86 18.73% Other associate professionals 60868 (3.8%) 49.17 17.75% 38950 (3.65%) 47.55 15.90% Office clerks 188232 (11.76%) 49.10 20.64% 240961 (22.57%) 46.88 17.86% Customer services clerks 6566 (0.41%) 48.11 19.52% 24995 (2.34%) 46.70 21.62% Personal and protective services workers 60747 (3.79%) 48.17 23.53% 86914 (8.14%) 47.24 26.03% Salespersons and demonstrators 23881 (1.49%) 47.31 20.47% 68484 (6.41%) 46.88 21.68% Skilled agricultural and related workers 39462 (2.46%) 49.97 25.19% 13726 (1.29%) 51.57 25.70% Extraction and building trades workers 141225 (8.82%) 48.40 30.65% 1700 (0.16%) 47.85 28.98% Metal, machinery and related trades workers 136747 (8.54%) 48.10 25.97% 11611 (1.09%) 47.44 26.45% Precision, (handi-)craft and related trades workers 20746 (1.3%) 49.19 25.27% 4665 (0.44%) 46.38 22.47% Other craft and related trades workers 45314 (2.83%) 48.32 25.74% 29739 (2.79%) 47.27 24.85% Stationary plant and related operators 23831 (1.49%) 48.41 26.13% 2086 (0.2%) 47.55 28.51% Machine operators and assemblers 41942 (2.62%) 47.58 26.01% 29982 (2.81%) 46.33 25.76% Drivers and mobile plant operators 103341 (6.45%) 48.75 28.36% 2695 (0.25%) 47.36 27.30% Services elementary occupations 63630 (3.97%) 48.74 31.04% 119135 (11.16%) 48.89 32.41% Agricultural and related labourers 444 (0.03%) 43.31 22.27% 2242 (0.21%) 52.90 29.82% Labourers in mining, constr., manuf. and transport 83206 (5.2%) 47.97 28.37% 14189 (1.33%) 47.08 28.50% Armed forces 25228 (1.58%) 46.35 17.89% 1950 (0.18%) 43.68 18.20% Total 1601040 (100%) 48.95 19.84% 1067792 (100%) 47.53 17.99%

Conference paper International Population Conference 2017 3 Standardized Morbidity Ratios (SMR) were calculated for (A) the total 1991 workforce, entailing active and non-active members in 2001 as well as workers who died between 1 March 1991 and 1 October 2001; (B) active and non-active members of the 1991 working population; (C) active workers in 1991 and 2001. We used the health distribution of the 2001 male population aged 35 to 65 years as reference for each scenario. The share of men reporting poor health in 2001 were obtained by 5-year age group. These proportions were applied to the 1991 male workforce by 5-year age group and occupational group for scenario A, B and C. Using the occupational information from the 1991 census, the observed number of workers in poor health were divided by the expected number based on the health distribution of the total male population in 2001. The result was multiplied by 100. An SMR of 100 indicates that the observed distribution of poor health is similar to the distribution in the reference category. An SMR (below) above 100 indicates that the study population has (lower) higher proportions of poor health than can be expected based on the 2001

• distribution. Large differences between SMRs for the three scenarios indicate strong HWEs.

Figure 1 shows the percentage of men experiencing poor health in 2001 according to their age at the time

• f the 2001 census. Lines represent the total male population in 2001 and the 1991 male workforce in

three scenarios. The health situation in the 1991 workforce (A) and among active and non-active members in 2001 (B) are masked when comparing the self-reported health among the active population (C) with the health situation in the 2001 population. Scenario C represents the ‘ideal’ working population, assuming no-

• ne had left employment or no-one had died between the two census. Scenario A simulates a situation

where the unhealthiest workers were still at work. The comparison of scenario A and scenario C thus represents the impact of poor health on the formation of the active population in 2001 net of ageing. As can be seen in figure 1, the difference between the share of men in scenario C in poor health is still considerably lower than the share of men in the 2001 total population. This can be explained by health selection processes prior to the 1991 census, which have not been taken into account in this study design. Figure 1 Percentage of men experiencing poor health in 2001 total population and 1991 workforce by age in 2001.

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 Total male population 2001 A Male working population 1991 - Active & non-active in 2001 plus deceased in 1991-2001 B Male working population 1991 - Active & non-active in 2001 C Male working population 1991 - Active in 2001

Conference paper International Population Conference 2017 4

Figure 2 SMRs and 95% confidence intervals for poor self-reported health among the 1991 male workforce in three scenarios. Reference population: 2001 male population. Results sorted on largest to smallest difference between scenario A and C 25 50 100 200 Extraction and building trades Agricultural and related labourers Drivers and mobile plant operators Services elementary occupations Mining, constr., manufact. and transport Other craft workers Machine operators and assemblers Managers of small enterprises Precision, (handi)craft and related workers Salespersons and demonstrators Metal, machinery and related trades workers Personal and protective services workers Stationary plant and related operators Skilled agricultural and related workers Customer services clerks Other assoc. Office clerks Physical and engin. science assoc. Corporate managers Life science and health assoc. Other professionals Teaching assoc. Armed forces Teachers Legislators and senior officials Life science and health Physical, math. and engin. science

A Male working population 1991 - Active & non-active in 2001 plus deceased in 1991-2001 B Male working population 1991 - Active & non-active in 2001 C Male working population 1991 - Active in 2001

Conference paper International Population Conference 2017 5 Results Figure 2 shows SMRs and 95% confidence intervals for poor self-reported health in the 1991 male workforce by occupational group in 1991. Poor health for three scenarios was compared to the general health situation in 2001. An overview of the results is presented in supplementary table A1. In order to determine which occupational groups experienced the strongest HWEs, we compared SMRs for the population with the healthiest individuals (C active population in 1991 and 2001) with the total 1991 male workforce (A active and non-active workers in 2001 plus persons deceased between the census dates). SMRs for scenarios C and A differed most among extraction and building trades workers; unskilled agricultural and related labourers; and drivers and mobile plant operators. The smallest HWEs were found among scientific professionals; health professionals; and legislators and senior officials. Results for scenario A show active workers in 1991 and 2001 had a significantly lower likelihood to report poor health for all

• ccupational groups compared to the 2001 health distribution.

The example of extraction and building trade workers shows important differences in interpretation. SMRs for this group of workers who were active in 2001 were significantly lower than expected based on the 2001 health distribution (SMR 92 CI 91-94). When taking also non-active and deceased building trade workers into account, SMRs were significantly higher compared to the reference population (SMR 126 CI 125-127). Conclusion This study demonstrates clear occupational differences in the strength of potential HWEs among the 1991 Belgian workforce with a time lag of 10 years. Large differences in HWEs were identified in the comparison between SMRs for the healthiest population (i.e. the active population in 2001) and the original study population (i.e. active and non-active workers including deceased persons in 1991-2001). The strongest HWEs were found in occupations with a relatively high degree of manual labour, such as construction work and elementary occupations. Previous studies have documented similar findings.[4–6] However, our results show a more nuanced pattern with considerable differences in HWEs between manual jobs. For example, extraction and building trade workers experiences larger HWEs than skilled agricultural workers with the difference between scenario A and C being 33.5 and 23.9, respectively. The work force consists of the healthiest individuals, but large variation in perceived health exists across the occupational spectrum. These differences seem to persists even when making the transfer to the inactive population. Funding Laura Van den Borre is a PhD fellow at the Research Foundation-Flanders (FWO). Acknowledgements We would like to thank Statistics Belgium for performing the linkage procedure. Competing interests The authors declare that they have no conflict of interest.

Conference paper International Population Conference 2017 6 References

[1] Steenland K, Stayner L. The importance of employment status in occupational cohort mortality studies. Epidemiology. 1991;2:418–423. [2] Li CY, Sung FC. A review of the healthy worker effect in occupational epidemiology. Occup. Med. 1999;49:225–229. [3] Carmichael F, Ercolani MG. Unpaid caregiving and paid work over life-courses: Different pathways, diverging outcomes.

• Soc. Sci. Med. 1982. 2016;156:1–11.

[4] Ihlebaek C. Occupational and social variation in subjective health complaints. Occup. Med. 2003;53:270–278. [5] Gueorguieva R, Sindelar JL, Falba TA, et al. The Impact of Occupation on Self-Rated Health: Cross-Sectional and Longitudinal Evidence from the Health and Retirement Survey. J. Gerontol. B. Psychol. Sci. Soc. Sci. 2009;64B:118–124. [6] Morefield B, Ribar DC, Ruhm CJ. Occupational Status and Health Transitions. BE J. Econ. Anal. Policy [Internet]. 2012 [cited 2017 Jan 27];11. Available from: https://www.degruyter.com/view/j/bejeap.2011.11.issue-3/1935-1682.2881/1935- 1682.2881.xml?format=INT.

Conference paper International Population Conference 2017 7

Table A1 Standardized morbidity ratios and 95% confidence intervals for poor self-reported health among the 1991 male workforce * A Total & deceased† B Total population‡ C Active population§ Occupational group SMR LL UL SMR LL UL SMR LL UL A-C**

Extraction and building trades workers 125.7 124.6 126.9 116.9 115.8 118.1 92.2 91.0 93.5 33.5 Agricultural and related labourers 113.0 91.8 137.6 106.1 85.4 130.3 80.0 60.2 104.1 33.0 Drivers and mobile plant operators 116.5 115.2 117.8 106.7 105.4 108.0 84.2 82.8 85.6 32.3 Services elementary occupations 126.2 124.4 127.9 116.1 114.4 117.8 93.9 92.0 95.8 32.3 Labourers in mining, construction, manufacturing and transport 120.7 119.2 122.2 110.9 109.4 112.4 88.5 86.8 90.2 32.2 Other craft and related trades workers 107.0 105.1 108.9 98.4 96.6 100.3 75.9 73.9 78.0 31.1 Machine operators and assemblers 111.7 109.7 113.8 103.4 101.4 105.4 83.4 81.1 85.7 28.4 Managers of small enterprises 94.2 92.7 95.7 84.5 83.1 85.9 66.9 65.5 68.4 27.3 Precision, handicraft, craft printing and related trades workers 102.1 99.4 104.9 93.8 91.2 96.4 75.1 72.2 78.1 27.0 Salespersons and demonstrators 91.0 88.5 93.5 82.0 79.7 84.4 64.5 62.0 67.0 26.5 Metal, machinery and related trades workers 109.4 108.3 110.5 101.2 100.2 102.3 83.3 82.0 84.5 26.1 Personal and protective services workers 101.1 99.5 102.7 90.8 89.3 92.4 75.1 73.5 76.8 26.0 Stationary plant and related operators 108.9 106.2 111.5 100.3 97.7 102.9 83.6 80.5 86.6 25.3 Skilled agricultural and related workers 98.0 96.1 99.9 89.3 87.5 91.2 74.1 72.2 76.1 23.9 Customer services clerks 85.6 81.2 90.3 77.1 72.9 81.6 63.1 58.5 68.0 22.5 Other associate professionals 74.7 73.3 76.1 65.7 64.4 67.0 53.5 52.2 54.9 21.2 Office clerks 86.6 85.8 87.4 77.5 76.7 78.3 65.9 65.0 66.8 20.7 Physical and engineering science associate professionals 83.3 82.3 84.3 75.7 74.7 76.6 63.4 62.3 64.5 19.9 Corporate managers 62.6 61.8 63.4 54.5 53.7 55.2 45.4 44.6 46.3 17.2 Life science and health associate professionals 60.3 57.3 63.4 52.7 49.9 55.6 43.9 41.2 46.7 16.4 Other professionals 67.4 66.1 68.8 59.3 58.0 60.5 52.1 50.8 53.4 15.4 Teaching associate professionals 80.9 77.0 84.9 73.3 69.5 77.2 65.7 61.7 69.9 15.2 Armed forces 84.4 82.0 86.8 75.5 73.2 77.8 70.6 67.8 73.5 13.8 Teaching professionals 72.0 70.8 73.2 65.0 63.8 66.2 59.4 58.0 60.8 12.6 Legislators and senior officials 58.4 54.3 62.7 51.7 47.9 55.7 46.3 42.0 50.8 12.1 Life science and health professionals 51.6 49.9 53.2 44.2 42.7 45.8 39.5 38.1 41.1 12.0 Physical, mathematical and engineering science professionals 52.1 50.7 53.5 45.7 44.4 47.1 41.1 39.7 42.4 11.0 * Reference population: total male population of 2001 census † Active and non-active members of 1991 male workforce in 2001+ members deceased between censuses. ‡ Active and non-active members of 1991 male workforce in 2001. § Active members of 1991 male workforce in 2001. ** SMR(Active and non-active members of 1991 male workforce in 2001+ deaths between censuses) – SMR(Active members of 1991 male workforce in 2001)