Article Text

Download PDFPDF

Changes of occupational class differences in physical functioning: a panel study among employees (2000–2007)
  1. Olli Pietiläinen,
  2. Mikko Laaksonen,
  3. Janne Pitkäniemi,
  4. Ossi Rahkonen,
  5. Eero Lahelma
  1. Hjelt Institute, Department of Public Health, University of Helsinki, Finland
  1. Correspondence to Olli Pietiläinen, Hjelt Institute, Department of Public Health, PO Box 41, FIN-00014 University of Helsinki, Finland; olli.k.pietilainen{at}helsinki.fi

Abstract

Background Panel studies on changes of occupational class differences in health have given varying results. The aim of this study was to examine changes of occupational class differences in physical functioning and the factors that explain these changes.

Methods A cohort of middle-aged employees of the City of Helsinki was followed up for an average of 6 years in two surveys from 2000–2002 and 2007. Hierarchical linear random effects models were fitted to analyse the changes of occupational class differences in SF-36 physical functioning, as well as the contribution of physical and psychosocial working conditions, material conditions, health behaviours and employment status to these changes.

Results Lower occupational classes had worse physical functioning at baseline: among women, the SF-36 scores ranged from 50.5 in the highest class to 47.1 in the lowest one, and among men from 52.2 to 48.9, with higher scores indicating better health. Occupational class differences widened during the follow-up due to stronger decline of physical functioning in the lower occupational classes than in the higher occupational classes. The largest difference in the decline of functioning between classes was 1.2 scores among women and 1.5 scores among men. Among women the widening of the class differences could be explained partly by health behaviours and employment status and among men by material conditions.

Conclusion Occupational class differences in physical functioning widened due to a faster decline of physical functioning in the lower occupational classes. Health behaviours, employment status and material conditions explained the widening class differences in physical functioning.

  • Health inequity
  • social class
  • longitudinal studies
  • employees
  • physical functioning
  • inequalities
  • physical function
  • social class

Statistics from Altmetric.com

Introduction

Socioeconomic differences in health have been well-documented, but few studies have examined how these differences change when the same individuals are followed up. The few existing panel studies on the changes of socioeconomic differences across various domains of health have given inconclusive results. Most panel studies focussing on employed populations have indicated that the decline of health and functioning tend to be more pronounced in the lower socioeconomic positions resulting in widening socioeconomic differences.1–3 However, other panel studies have found no occupational class differences in change of health.4 5

While most studies suggest that socioeconomic health differences tend to widen by age among employees, two competing hypotheses regarding the changes in late middle-age can be distinguished. The hypothesis of cumulative disadvantage predicts that the socioeconomic differences in health widen over the life-course due to a differential accumulation of health damaging exposures across socioeconomic positions. This hypothesis has received support from several panel studies examining differences by occupational class, education or income.6–9 In contrast, the hypothesis of age-as-leveller predicts that socioeconomic differences in health tend to widen until late middle-age but then begin to narrow. This hypothesis has also received support from several studies.10–14 The narrowing has been proposed to be explained by reduced work-related exposures,12 selective mortality or attrition,9–11 and by the fact that even those in better socioeconomic positions will eventually face the inevitable biological effects of ageing.11 15

The reasons for the changes in socioeconomic differences in health are not well-known. In cross-sectional studies, socioeconomic differences in health and functioning have been partly explained by differences in physical working conditions,16 17 material factors, such as income or housing conditions,18 19 and health behaviours.20 21 Psychosocial working conditions have explained the differences in some studies22 but not in others.23 24 Similar factors may also contribute to the changes of socioeconomic differences in health. In the few panel studies explaining the changes of socioeconomic differences in health, working conditions and health behaviours have partly explained the changes1 25 26 but there is heterogeneity in the results.

The aim of this study was to examine changes of occupational class differences in health-related physical functioning in a cohort of middle-aged employees of the City of Helsinki followed on the average 6 years from 2000–2002 to 2007. Furthermore, we examined whether the changes of the differences can be explained by physical and psychosocial working conditions, material conditions, health behaviours and transition to retirement.

Methods

The data are based on two mail surveys among the Helsinki Health Study cohort, a cohort of employees of the City of Helsinki aged 40–60 years at baseline, conducted in 2000–2002 (n=8960, response rate 67%). The follow-up survey was conducted in 2007 among the respondents to the baseline survey (n=7332; response rate 83%). The analyses were conducted on those who participated in both surveys, separately for women and men.

Physical functioning

Physical functioning was measured by the Short Form 36 (SF-36) inventory, which consists of 36 questions concerning physical and mental functioning.27 The measure is summarised into eight subscales depicting different domains of health, and these can be further summarised to physical and mental component summaries. This study examines the physical component summary, which is a measure of general health-related physical functioning. The component summary score ranges from 0–100 with higher values indicating better functioning. The component summaries are scaled to have a mean of 50 and SD of 10 in the general US population. In our data, the mean score at baseline was 48.7 among women and 50.3 among men and the scores decreased somewhat during the follow-up. The SF-36 reflects many clinical conditions and it can be considered a reliable and valid measure of general physical functioning.28 SF-36 is well-suited for analysing changes of functioning, because as a continuous measure it captures small changes in functioning better than commonly used dichotomous health measures.

Occupational class

Occupational class was measured at baseline and divided into four hierarchical classes: professionals, semi-professionals, routine non-manual employees and manual workers. Manual workers and non-manual employees were distinguished using the socioeconomic classification of Statistics Finland,29 and the occupational classification of the City of Helsinki was used to further divide non-manual employees into three groups on the basis of proficiency requirements and supervisory status.30 The professionals include employees with a university degree as well as managers who have subordinates and who do managerial or administrative work. Semi-professionals include intermediate level white-collar employees. Routine non-manual employees include non-professional clerical employees and non-professional employees within social and health care. Manual employees do mainly physical work.

Working conditions

Physical working conditions were measured by 18 questions on work and working conditions, which were summarised to three measures on the basis of factor analysis. The three factors that emerged were environmental exposures in work, physical job strenuousness and computer display/terminal-aided work. Psychosocial working conditions were measured by the Framingham31 version of Karasek's job-demand-control inventory.32

Material conditions

To yield a measure of household-income adjusted for consumption units, self-reported household income after removing taxes and adding any welfare benefits was divided by household size using weights of the Organisation for Economic Cooperation and Development (OECD) equivalence scale. Housing tenure was divided to owner occupiers, renters from the municipality of Helsinki, renters from the open market and others. Financial difficulties were measured by the questions: ‘How often do you have enough money to buy the food and clothes needed by you or your family?’ and ‘How much difficulty do you have in paying your bills?’ Financial satisfaction was measured by a question on how satisfied the respondent is with his/her standard of living.

Health behaviours

Volume of alcohol use was measured by adding the reported weekly consumed amounts of beer, wine and spirits. Frequency of binge drinking was measured by asking how often the respondent drinks more than six alcoholic drinks. Drinking problems were assessed with the CAGE questionnaire.33 Tobacco smoking was divided to those having never smoked, ex-smokers, current moderate smokers and current heavy smokers. Physical activity was measured by asking how many hours a week the respondent engages in physical activity corresponding to walking, vigorous walking, jogging and running, and the answers were converted to metabolic equivalent tasks (METs).34 Healthy food habits were measured by following the recommendations in the Finnish national dietary guidelines.35 Relative weight was measured by body mass index (BMI).

Employment status

Employment status at the follow-up was divided to those who remained in employment, those who retired due to disability, those who retired due to age and others, including unemployed, housewives and those not employed for other reasons. While all respondents were employed by the City of Helsinki at baseline, by the follow-up 21% had retired and 3% had left employment for other reasons.

Statistical analyses

First we assessed whether there were cross-sectional differences between occupational classes in physical functioning at baseline and follow-up by calculating age-adjusted means and 95% Bayesian credible intervals of SF-36 physical component summary scores for each occupational class.

Next we applied Bayesian linear hierarchical random effects modelling36 to assess the effect of occupational class on individual-specific differences in SF-36 scores at baseline and on changes of the SF-36 scores between baseline and follow-up measurements. Random effect modelling was used because it enables the use of all information relevant to the problem in a statistically coherent framework, including individual missing values, varying length of follow-up period and the effects of the predictors separately for baseline functioning and change of functioning.37 Bayesian inference was preferred because it incorporates parameter uncertainty with sampling uncertainty 38 and the resulting measures of statistical significance have a natural interpretation of the probability of the tested hypothesis being true.

The hierarchical model at level 1, describing variation within individuals, can be written as follows:

yitN(αi+βiti,σ2)

where yit is the SF-36 score of individual i at time t, αi is the constant indicating individual-specific baseline SF-36 score and βi stands for individual-specific slope indicating linear change from baseline in the SF-36-score.

The level 2 equations for individual-specific intercepts and slopes can be written as follows:

αiN(β1αsi+β2αz1iα+,σα2)βiN(β1βsi+β2βz1iβ+,σβ2)

where s stands for occupational class, covariates z1α,z2α stand for assumed predictors of the differences—namely age, physical and psychosocial working conditions, material conditions and health behaviours asked at baseline—and z1β,z2β stand for the same covariates as well as employment status asked at follow-up. First a model where only age is adjusted for was fitted and in the further models each of the predictors was added in turn to this base model. If there is a significant difference in change of physical functioning between occupational classes between the measurements, this means that the differences measured at baseline have changed during the follow-up period. To facilitate judging how much the explanatory factors explain the differences, we calculated expected SDs of the group-wise means of change in physical functioning for each model.

Since information for occupational class was missing from 125 respondents, we assumed a categorical prior for occupational class, prior probabilities corresponding to the frequencies in the observed data and, therefore, assuming that occupational class is missing at random. Missing values for other predictors were imputed according to the same principles.

We summarise the results by calculating the posterior means and 95% credible intervals of the baseline scores and slopes for each occupational class using the above model. Statistical significance of the differences in change of physical functioning was assessed with posterior probabilities, or ‘Bayesian p values’,36 comparing each possible pair of occupational groups to each other assuming that there was no difference between the groups. The analyses were conducted using the WinBUGS programme.39

Results

Differences in physical functioning between occupational classes were clear at both baseline and follow-up: lower classes had worse physical functioning among both women and men (table 1). Between the surveys, physical functioning declined among women and men in all occupational classes. Among women, the difference in SF-36 scores between the professionals and manual workers was 3.4 at baseline and 4.4 at follow-up. Among men, the corresponding difference was 3.3 at baseline and 3.8 at follow-up.

Table 1

Number of participants and age-adjusted means and 95% Bayesian credible intervals (CIs) of SF-36 scores in baseline and follow-up by occupational class, women and men

Among women, the occupational class differences in physical functioning widened during the follow-up as decline in physical functioning was more pronounced in the two lowest occupational classes with clearly the strongest decline among manual workers (table 2). At maximum, the difference in change on functioning between the groups with the largest and the smallest change was 1.2 points among women and 1.5 points among men. Among women, the change in physical functioning among manual workers differed from that in semi-professionals and professionals and the difference was statistically significant. Among men, physical functioning in the lower two occupational classes declined faster than in the upper classes but the differences were not statistically significant.

Table 2

Age-adjusted mean change in SF-36 scores by occupational class and pair-wise significance of the difference in trajectories between groups, women and men

In order to assess the reasons behind the change of occupational class differences in physical functioning, a series of further models was fitted to examine potential explanatory factors of the differences in change of physical functioning. SF-36 scores by these explanatory factors are presented in online Appendix 1. Among women, physical working conditions, health behaviours and, to a small extent, material conditions and psychosocial working conditions, explained occupational class differences in physical functioning at baseline with physical working conditions having the strongest effect (table 3). The explanatory factors had less effect on occupational class differences in change of physical functioning than on differences in baseline functioning. Differences between occupational classes in change of functioning were clear in the age-adjusted model, and adding baseline physical working conditions to the age-adjusted model had negligible effects on occupational class differences in change of functioning. Adjusting for health behaviours, employment status and, to a small extent, also material conditions and psychosocial working conditions narrowed the differences between occupational classes with employment status having the strongest effect.

Table 3

Baseline mean and mean change in SF-36 scores with 95% Bayesian credible intervals by occupational class after adjusting for predictors, women

Among men, there were clear differences between the occupational classes in functioning at baseline in the age-adjusted model (table 4). Adjusting for baseline physical working conditions, psychosocial working conditions, material conditions or health behaviours in the age-adjusted model narrowed the occupational class differences in functioning at baseline, whereas adjusting for psychosocial working conditions widened them. Among men, the decline in physical functioning was larger in the two lower occupational classes than in the higher ones but the differences were not statistically significant. Adjusting for physical working conditions slightly widened the occupational class differences in change of physical functioning, whereas adjusting for all other explanatory factors slightly narrowed them with material conditions having the strongest effect.

Table 4

Baseline mean and mean change in SF-36 scores with 95% Bayesian credible intervals by occupational class after adjusting for predictors, men

Discussion

This study examined the changes of socioeconomic differences in physical functioning over 6 years by analysing a cohort of employees aged 40–60 years at baseline. In addition to examining the changes of occupational class differences in physical functioning, the possible determinants of the changes were analysed.

The study had two main findings. First, physical functioning declined in all occupational classes over the follow-up, but the decline was larger in the lower occupational classes resulting in widening occupational class differences in physical functioning over the follow-up period. Second, health behaviours, retirement before the follow-up survey and, to a small extent, material conditions and psychosocial working conditions explained a part of the changes of occupational class differences in physical functioning among women, while physical working conditions did not explain the differences. Among men, only material conditions explained the change of occupational class differences.

Our results are in accordance with the cumulative disadvantage hypothesis predicting that socioeconomic differences in health widen over the life-course due to the disparate accumulation of health damaging exposures. While the results suggest widening differences, we cannot say for sure whether this widening results from accumulation of exposures as the explanatory factors explained the changes of occupational class differences in functioning only to a small degree. Unlike our study, some previous studies have observed a narrowing of socioeconomic differences in health towards older ages.10 12–14 It is possible that as we study a cohort of 40–60 year olds as a whole we lose sight of possible age differences within the cohort. It is also possible that our study population is not old enough for the possible decline in the differences to emerge.

The explanatory factors in our study have not been systematically examined in any previous panel study and, thus, direct comparisons of their effects could not be made. While physical working conditions, material conditions and health behaviours have previously provided substantial explanations for socioeconomic differences in health in cross-sectional studies, in panel studies they have not explained the differences in health change or have explained them only to a lesser degree. Similar health behaviours, as in our study, have explained the socioeconomic differences in health change among US adults to a minor degree,25 which conforms to our results. A study on British civil servants1 found that health behaviours, material problems and job decision latitude explained employment grade differences in change of SF-36 functioning among men but not among women. In another study from the Netherlands,40 behavioural factors did not explain educational or income-based differences in change of physical functioning among 55–70 year olds, although the factors explained the differences cross-sectionally.

While most of the explanatory factors probed in this study did not explain the changes in socioeconomic differences in functioning or explained them only slightly, they explained the differences at baseline. This could be due to several reasons. First, the differences at baseline may result from the total accumulation of health damaging and protecting exposures over the whole life-course, whereas the changes in functioning are confined to the average of 6 years of follow-up, which may be too short a time for the effects to become evident. Second, the differences at baseline may result not only from causation but also from selection, while differences in change of functioning can only result from causation of the predicting factors. Third, the predicting factors were measured at baseline and, thus, may have a stronger effect on functioning at baseline than on functioning at follow-up, because during the follow-up period the predictors may change, attenuating their effects.

Most of those in the oldest age groups of our study underwent a major transition—that is, retirement, during the follow-up. After retirement people are no longer directly exposed to work-related hazards or protected by positive work-related factors and their incomes decline. Our results showed that differences in employment status partly explain the changes in occupational class differences in physical functioning. Whether the socioeconomic health differences change differently among the employed and the retired populations warrants more investigation.

Methodological considerations

Certain aspects of our study give credibility to the results. The data used were relatively large with 7332 participants. The hierarchical model used makes it possible to take into account the individuals who constitute the occupational classes instead of considering the occupational classes only at an aggregate level. The study design, where a cohort is followed over time and the predictors of the changes of the occupational class differences precede the changes under examination, supports the role of the examined factors as having a causal relation to the observed changes.

Our study has also some limitations. The relatively short follow-up period and narrow age-range limit the possibility of interpreting our results in the life-course framework, since part of the burden of different risk factors is already quite fixed in this age group and the effects of exposures on earlier life-course not captured in our data could not be reliably assessed in this study. There may be cohort and period effects behind the change of health over the life course.9 12 As our data included only one follow-up, we were unable to distinguish period or cohort effects. Furthermore, regression to the mean is a potential source of bias but it is unlikely to seriously undermine our results. Regression to the mean may cause only attenuation to the observed differences and, as a result, the true differences might be larger but not smaller than those observed in our study.

Non-response may cause bias to the results. For the baseline survey, a non-response analysis has been made.41 The relation between socioeconomic position and health as measured by register-based sickness absence was not significantly affected by non-response in our data. For the follow-up survey such analysis has not been conducted, but the response rate in the follow-up was relatively high (83%), which reduces the biasing effects of attrition. The baseline health did not differ significantly between those who participated in the follow-up and those who did not. Non-participation was somewhat more common in the lowest occupational class, but as in the whole study population also in this class the baseline health did not significantly differ between the follow-up participants and non-participants. Therefore non-participation is unlikely to cause serious bias to our results.

Conclusions

In conclusion, occupational class differences in physical functioning widened over 6 years of follow-up among employees aged 40–60 years. Our study supports the role of health behaviours, employment status and material conditions as explanatory factors for the widening but more research is warranted to clarify the explanations for the changes of socioeconomic differences in health and functioning.

What is already known on this subject

Higher socioeconomic position is known to be connected to better health. However, the current knowledge on the change of socioeconomic health differences over the life-course is inconclusive and the possible explanations for the changes have not been examined systematically.

What this study adds

Occupational class differences in physical functioning widened due to a faster decline of physical functioning in the lower occupational classes. Health behaviours, employment status and material conditions partly explain the widening occupational class differences in physical functioning.

References

View Abstract

Supplementary materials

  • Web Only Data jech.2010.110270

    Files in this Data Supplement:

  • Web Only Data jech.2010.110270

    Files in this Data Supplement:

Footnotes

  • Funding The Academy of Finland, Vilhonvuorenkatu 6, POB 99, FI-00501 Helsinki. http://www.aka.fi/en-gb/A/.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Helsinki Health Study, which has been approved by ethical committees at the Department of Public Health, University of Helsinki and at the City of Helsinki Health Authorities.

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

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.