Dear Editor

In response to our study,[1] Kivimäki et al suggested that reported sickness absence frequencies were underestimates of the total sickness absence burden in European Union (EU) member countries.[2] This concern about the veracity of these estimates led Kivimäki et al to caution policy makers to not use this data to inform policy. While we agree that more research is needed to establish potential biases associated with different approaches to ascertain accurate asking sickness absence data, we consider the European Survey on Working Conditions (ESWC) to be useful to inform the cross-national policy debate. Country-specific studies contribute knowledge to the evidence base but cross-national studies such as ours help to provide a stronger basis upon which to make cross-national inferences. Furthermore, cross-national studies become more relevant as data accumulate and the data collection quality improves. We hope that Kivimäki and colleagues are not suggesting the ESWC be discontinued.

We consider the studies by Kivimäki et al to be some of the most relevant epidemiological studies of sickness absence predictors.[3][4][5] Although informative, these studies raise several issues in the context of cross-national comparisons. First, epidemiological cohorts in Finland and the United Kingdom represent very homogenous and specific working populations (ie municipal employees, hospital workers and civil servants) with unknown generalizability to the national representative surveys studied in our paper or the ones referenced by Kivimäki et al.[6][7] Second, a fundamental advantage of national workforce surveys is the ability to capture all workers, whereas registries may lead to an under representation of marginal work groups typically not included in national registries. Indeed, Kivimäki et al are not arguing that the Finnish and British cohorts are representative of the countries’ workforces. Even so, labor market inequalities may cause temporary and less protected workers to be underrepresented in the type of well-designed cohort studies they have referenced.[8] Temporary and less protected workers are important in the EU economy and lack of knowledge about their labor market experiences as related to sickness absence could lead to their further marginalization in the policy debate. Third, Kivimäki et al criticised the data collection method employed in the ESWC. We are not aware of any cross-national study comparing the reliability, validity and performance of different sickness absence data collection methods. Concerns have been raised about who is placed on a sickness absence registry. Registered data are very conditioned by the country’s social security system criteria for sickness absence, which complicates between-countries comparisons.[9] Therefore, whether registries are the gold standard in sickness absence studies remains a point of debate yet to be closed.

In addition, Kivimäki et al compared our results to two survey based studies from Finland and Britain, but differences in sample selection and questionnaire design between these studies may limit comparisons. Our study included people aged 15 years and older who had any paid job during the reference week, or who had a job but were temporarily absent. The recall period for sickness absence was 12 months. The Finnish study was based on employees aged 25 to 64 using a 6 month recall period for sickness absence.[10] The British survey investigated the psychiatric morbidity prevalence among the British adult population. This study sampled workers aged 16-64 years and excluded workers with a psychosis diagnosis. Workers who were currently working or had been working in the last year were asked to report absence days due to their health or feelings.[11] For these reasons, caution is needed if a direct comparison between these three studies is intended.

Finally, we agree with Kivimäki et al that potential bias in the ESWC could be present (see page 868-9 in our article). However, we would argue that the best sources of data to inform policy are derived from systematic efforts to collect sickness absence data in a clear and consistent fashion from a representative sample of the labor force within each country. We consider the evidence presented by Kivimäki et al to support our argument of the difficulty in establishing between-country comparisons due to the fragmented and insufficient sickness absence data available at the European Union level. We consider our results useful. Although the results are preliminary and may be subjected to scientific scrutiny, the comparative findings may provoke researchers to develop standards for sickness absence studies to facilitate between-country comparisons. In addition, we hope the observed differences will promote further investigation into root causes of between-country differences, especially between northern and southern EU members, as well as within-country gender differences. We certainly welcome cross-national collaborative efforts among the EU sickness absence researches to address all these issues.

References

1. Gimeno D, Benavides FG, Benach J, Amick BC III. Distribution of sickness absence in the European Union countries. Occup Environ Med 2004;61:867-9.

2. Kivimäki M, Vahtera J, Head J, Ferrie JE. Are sickness absence frequencies in the study of EU countries underestimates? [electronic response to Gimeno D et al. Distribution of sickness absnce in the European Union countries] occenvmed.com 2004 URL direct link to eLetter.

3. Vahtera J, Pentti J, Kivimäki M. Sickness absence as a predictor of mortality among male and female employees. J Epidemiol Community Health 2004;58:321-6.

4. Kivimäki M, Virtanen M, Vartia M, Elovainio M, Vahtera J, Keltikangas-Järvinen L. Workplace bullying and the risk of cardiovascular disease and depression. Occup Environ Med 2003;60:779-83.

5. Kivimäki M, Head J, Ferrie JE, Shipley MJ, Vahtera J, Marmot MG. Sickness absence as a global measure of health: evidence from mortality in the Whitehall II prospective cohort study. BMJ 2003;327:364-8.

6. Kauppinen T, Hanhela R, Heikkilä P et al. Work and Health in Finland 2003. Finnish Institute of Occupational Health, Helsinki, 2004.

7. Stansfeld SA, Head J, Rasul F, Singleton N, Lee A. Occupation and mental health: Secondary analyses of the ONS Psychiatric Morbidity Survey of Great Britain. Research Report 168, Health and Safety Executive Books, 2003.

8. Moss NE. [Review of the book Why Are Some People Healthy and Others Not? The Determinants of Health of Populations, by Robert G. Evans, Morris L. Barer, and Theodore R. Marmot] Health Affairs 1995;(14) 2:318- 21.

9. Gründemann RWM, van Vuuren CV. Preventing absenteeism at the workplace. Dublin: European Foundation for the Improvement of Living and Working Conditions, 1997

10. Kauppinen T, Hanhela R, Heikkilä P et al. Work and Health in Finland 2003. Finnish Institute of Occupational Health, Helsinki, 2004.

11. Stansfeld SA, Head J, Rasul F, Singleton N, Lee A. Occupation and mental health: Secondary analyses of the ONS Psychiatric Morbidity Survey of Great Britain. Research Report 168, Health and Safety Executive Books, 2003.

Dear Editor

Pattani and colleagues present some interesting and useful findings in an area of great importance to the delivery of health care within the United Kingdom.[1] They note that “doctors were nearly four times more likely to return to work as health care assistants and support staff” and that “this may reflect more flexible working opportunities for highly skilled staff”. I would suggest that it might also reflect the nature of the reason for retirement and how the workplace can be adjusted to deal with this. Having been involved in the assessment of NHS ill-health pension applications during the study period my impression (with observer bias) was that the major reason for doctors applying was either stress related or where a cardiovascular problem was the primary problem the rationale was often that it was exacerbated by workplace stress. Finding work where the stressors are diminished by say reduced working hours or workloads for doctors would seem to be easier than finding work with reduced muscoloskeletal demands for those jobs which essentially involve significant amounts of manual handling (eg health care assistants) who have musculoskeletal disease.

They also note “there are no formal arrangements to redeploy staff whose ill health prevents them from continuing in such posts”. It would be useful if the NHS Pensions Scheme, and indeed any occupational health pension scheme, considered whether provisions of the Disability Discrimination Act in terms of reasonable adjustments had been appropriately applied before concluding that the individual was permanently incapable of doing their job. A Treasury report on ill health retirement in the public sector, published in 2000, stated, “the evidence suggests that redeployment and rehabilitation are not always investigated actively” and that “if there is no requirement to consider alternative duties, redeployment may not be actively considered. The employee’s skills and experience can thus be unnecessarily lost to the service.”[2]

Pattani’s paper gives us a useful insight into the potential of those who have retired from the NHS due to ill health. The next step is to go back to the workplace and examine the factors that lead to an application for ill-health retirement.

References

1. Pattani S, Constantinovici N, and Williams S. Predictors of re-employment and quality of life in NHS staff one year after early retirement because of ill health; a national prospective study. Occup Environ Med 2004; 61: 572-576.

2. HM Treasury. Review of ill health retirement in the public sector. London: HM Treasury, July 2000

Dear Editor

Gimeno, Amick, Benavides and Benach [1] raise a number of issues with a paper recently published in Occupational and Environmental Medicine.[2] It is important that researchers cross-examine others’ findings and conclusions, as well as explain and defend their own findings and conclusions, so that debate can proceed that is both informed and useful for policy and practice.

In the original paper, I present findings that suggest the perceived risk from occupational stress varied across countries in the EU. I interpreted the findings as suggesting an element of social construction is present in such perceived risk, and that this element of social construction could have implications for policy, in respect of risk communication, for example.

There are two important things to note about these conclusions.

First, I analysed perceived risk from occupational stress, and not objective risk. These are recognised as two separate entities.[3] Whilst they are arguably intertwined closely in occupational stress and emotional health,[4] these data do not speak to this issue, and I drew no conclusions concerning objective risk. However, perceived risk is an important issue that can have practical implications, as noted.[2]

Second, it is uncontroversial in many areas concerned with risks to health (e.g. nuclear power, food) that risk is partially socially constructed, at this social construction has important practical implications.[5] There is no reason I know of why there should be controversy concerning the practical implications of a socially constructed element to risk from occupational stress.

Gimeno et al. raise four concerns.

1. The first issue is related to sample selection. The data were drawn from the Third European Survey on Working Conditions (ESWC), and then a more homogenous sample selected of people working in the same country as their nationality. Albeit imperfect (as noted by Gimeno et al), this procedure excludes workers who are more likely to have recent extensive experience of different working cultures. As noted in the paper, then, ‘variation within individuals due to multiple influences of work cultures was minimized’. Note I use the term minimized, rather than excluded. Notwithstanding, this procedure provides a more accurate means of assessing socio-cultural influences on perceived risk of occupational stress than was possible just by comparing responses across country of data collection or by comparing people of different nationalities.

Gimeno et al. explain that there is an influx of foreign workers into the EU, and that these workers may be exposed to more hazardous working conditions. Gimeno et al. make the argument that ‘removing workers who may be experiencing the greatest stress and hazardous working conditions from the sample may represent a severe selection bias’. I would agree with Gimeno et al.had the purpose of the paper been to determine objective risk in all areas of the labour market in the countries surveyed. Further, I agree that such workers are an important section of EU labour markets deserving of research in their own right with a view to informed intervention to enhance their working conditions – and such research should include assessment of perceived and objective risk in due course.

However, the purpose of the paper was to determine whether there is a socially constructed element in the perceptions of stress that can be attributed to socio-cultural variables at the national level. Including such workers may well have skewed the data by introducing a level of bias related to the nationality of immigrant workers in a country.

Gimeno note ambiguity with the statement "individuals who were working in a different country as their nationality". As well as re- reading the paper, I have conducted an electronic search of the paper, and cannot find this quote. Therefore I am unsure of the context of the quote, and so cannot be sure of the source of the perceived ambiguity. However, Gimeno et al. do raise issues not covered in the ESWC data concerning assimilation of social constructs and life-course data, all of which are relevant to propagation of cultural values at an individual level [6] and possibly also risk perception. However, by including only those people who list their nationality as being the same as the country they work in, it is a reasonable assumption that they will be familiar with the dominant cultural values of that country (even if they do not subscribe to those values themselves). Other data would be desirable of course, but unfortunately, they are not available.

2. The second concern raised by Gimeno et al concerns the use index of perceived risk of occupational stress. They raise an issue concerned with principal components analysis (PCA) conducted on the whole sample, rather than the sub-set used to compare the EU countries. It seems that the implication of this argument is that a different cluster of symptoms would be uncovered by PCA conducted on the sub-sample. I re-analysed the data, using only those people that worked in the same country as their nationality. PCA indicated the same pattern of relationships between perceived risk of symptoms.

What does this analysis mean though? It seems that a cluster of symptoms – which may be labelled stress – are commonly perceived as co- occurring. This might be because when such symptoms occur, they commonly occur together. This could reflect a physiological process, or heightened cognitive processing of physical sensations during emotional episodes.[7] However, as other analyses in [2] show, the extent to which the risk of experiencing these symptoms is attributed to work has a socially constructed component.

Gimeno et al. do raise the point that important differences might exist between individual health-related outcomes, and that these health- related outcomes might vary between factors such as type of employment, gender etc. First, in [2], I did control for industry, demographic factors, job dissatisfaction and perceived working conditions. In this way, factors that might represent cross-national variations in working practices (such as working hours) and individual level factors (such as job dissatisfaction) could not be said to account for the pattern of findings. Second, the ESWC contains many individual level questions assessed by single item indicators. By looking at health outcomes assessed by single indicators, there is danger that any findings reflect the error of measurement embedded in single item indicators of indeterminate reliability - rather than the true variable of interest. This threat is magnified in analyses involving large samples, where even small differences due to error can be labelled statistically reliable. By producing a scale – with an acceptable reliability coefficient (a=0.73), this possibility is lessened.[2]

Gimeno et al. also bring attention to the nature of the questions in the ESWC that I used to examine cross-national differences. It is important to note that the question I used asked participants to attribute various symptoms to work, not whether those symptoms were currently being experienced. Therefore, the question reflects a perception of the impact of work on symptoms – and not symptoms themselves or any underlying pathology. Therefore, any conclusions drawn from the data cannot reflect anything other than factors influencing the perceived influence of work on various symptoms.

3. The third concern raised by Gimeno et al regards the empirical status of my conclusion that workers in lower ranked countries might be less likely to attribute minor or non-specific psychosomatic symptoms to work when consulting with health practitioners. Whilst Gimeno et al are aware of no data that support such a conclusion, this conclusion – however conditional the language within which it is phrased - cannot be dismissed without further work. There is evidence that indicates emotional displays and discussion of life stress during consultation might cause misdiagnosis.[8] This potential for misdiagnosis arguably warrants further investigation.

4. The fourth concern raised by Gimeno et al. concerns the discussion section, especially in relation to conclusions with respect to the UK. I conclude that the media and socio-cultural factors – specifically cultural attitudes – might influence perceptions of risk from occupational stress. There is an extensive literature relating to these issues in relation to other risks – and some of the more prominent examples were referenced in [2]. There is also discussion elsewhere of socio-cultural factors and the media in influencing perceptions of risk from occupational stress [9] and in politicising stress as a social issue that subsequently influences reporting of stress.[10]

Gimeno et al. suggest that classifying countries according to common characteristics could provide an a priori basis for explaining the pattern of findings. The cultural approach to risk perception [11] provides one such approach. In this framework, countries with an individualist orientation (Ireland and Great Britain in this sample) would be expected to be ranked lower than other countries. This was indeed the case in [2]. However, countries with a more egalitarian orientation, such as the Nordic countries and Germanic countries, would be predicted to be ranked near the top, and countries with more hierarchical work cultures – such as France – would be expected to be ranked lower than the Nordic and Germanic countries. The results clearly do not bear such predictions out (see 12 for a cultural typology of European countries).

Therefore, the ranking of countries may reflect more complex processes than simple categorisation of countries on cultural or other factors. This conclusion is supported by some of the literature cited in [2] that has been used to discuss the British case [9,13-15]. In this literature, a number of suggestions have been made, in which, factors such as the media, socio-cultural factors and cognitive processes in the perception of risk have been advocated as explanations or partial explanations for perceived risk from occupational stress. I do not know the extent to which the scientific and policy literature in other countries covers such issues, but the fact there is open debate in the scientific community concerning such issues in Britain makes Britain a useful focus for analysis. Moreover, the British case with respect to monitoring stress and intervention is now more interesting, given the introduction of Stress Management Standards by the UK Health and Safety Executive. In due course, the British case will provide a useful focus for analysis of how advisory rather than coercive processes influence reporting and perception of occupational stress. It was for these reasons that the discussion section had some focus on the British case.

So rather than further cross-national comparisons on perceptions of stress – now we know they exist but not why - researchers might find it useful to examine individual countries on a case-by-case basis, so as to build rich pictures of the factors that influence perceived risk from stress in any one societal context and explain in more detail the pattern of findings in [2]. Once such exploratory work has been completed, research can move to hypothesis testing.

Finally, Gimeno et al. note that different rankings are apparent if different dependent variables are used. This should not be surprising – I can think of no reason why sickness absence has to necessarily mirror findings with attributions of stress-related symptoms to work – although there might be several reasons why they could be loosely coupled. Stress is not the only reason for absence from work.

Whether the ranking in [2] is more or less accurate and the reasons for the rankings, only further research can tell. However, the socially constructed elements of stress have real policy implications. They are worthy of further debate and research. To ignore them or deny them ultimately will be to the detriment of policy, intervention and occupational health.

References

1. Gimeno, D, Amick III BC, Benavides FG, Benach J. Bias and caution. Occup Envirn Med On-line 2004, http://oem.bmjjournals.com/cgi/eletters/61/5/467

2. Daniels, K. Perceived risk from occupational stress: a survey of 15 European countries. Occup Environ Med, 2004; 61: 467-70.

3. Royal Society. Risk: Analysis, Perception and Management. London: Royal Society, 1992.

4. Daniels K, Harris C, Briner R. Understanding the Risks of Stress: A Cognitive Approach. Sudbury: HSE Books, 2002.

5. Kasperson RE, Kasperson JX, Renn O. The social amplification of risk: progress in developing an integrative framework. In S Krimsky, D Golding (eds) Social theories of risk. London: Praeger, 1992.

6. Berry JW. Immigration, acculturation, and adaptation. Applied Psychology: An International Review, 1997; 46: 5-34.

7. Kirmayer LJ, Robbins JM, Paris J. Somatoform disorders: personality and the social matrix of somatic distress. Journal of Abnormal Psychology, 1994; 103: 125-136.

8. Ellington L, Wiebe D. Neuroticism, symptom presentation, and medical decision making. Health Psychology, 1999; 18: 634-643.

9. Daniels, K. Why aren’t managers concerned about occupational stress? Work & Stress, 1996; 10: 352-366.

10. Barley SR, Knight DB. Toward a cultural theory of stress complaints. In B.M.Staw & L.L. Cummings (Eds.), Research in organizational behavior, 1992; 14: 1-48.

11. Thompson M, Ellis R, Wildavsky A. Cultural Theory. Boulder: Westview, 1990.

12. Ronen S, Shenkar, D. Clustering countries on attitudinal dimensions: a review and synthesis. Academy of Management Review, 1985; 10: 435-454.

13. Wesseley S, Hotopf M. Are some public health issues better neglected? Lancet 2001; 357: 976-7.

14. Newton T. ‘Managing stress’ emotion and power at work. London: Sage, 1995.

15. Rick J, Hillage J, Honey S, Perryman S. Stress: big issue, but what are the problems? Brighton: The Institute for Employment Studies, 1997.

16. Gimeno D, Benavides FG, Benach J, Amick III BC. Distribution of sickness absence in the European Union Countries. Occup Environ Med (in press).

Dear Editor

A recent article by Daniels [1] in this journal presented occupational stress data from the 15 European Union (EU) countries.

Cross- national comparisons contribute to our scientific understanding of how and why health-related indicators (i.e. stress) are unequally distributed across countries and provide clues and guidelines for researchers, policy makers and trade unions at the EU level. However, we are concerned the data’s strengths and limitations are not explained appropriately to inform the policy debate. Specifically we have four concerns.

Our first concern is related to sample selection. The data were drawn from the Third European Survey on Working Conditions (ESWC) conducted in 2000.[2] The sampling strategy was designed to obtain nationally representative samples of the total active population, including non- Europeans if they could be interviewed in the national language of the country where they worked. Daniels asserted occupational stress is socially constructed and thus used nationality to obtain a more homogeneous sample. However, removing workers who may be experiencing the greatest stress and hazardous working conditions from the sample may represent a severe selection bias.[3]

The inflow of foreign workers (especially the so called low-skilled immigrants) to the EU advanced industrial countries is increasing.[4] Immigrant workers are becoming a significant part of the active EU population and are likely to rise sharply with the addition of the ten new EU countries.[5,6] Compared to native workers, non-EU nationals have greater problems obtaining jobs.[7] They may obtain employment in contractual jobs where exposure to work-related stressors is high.[8] It is plausible foreign workers are more frequently employed in ‘bad’ jobs exposed to more hazardous work environments.[9]

Furthermore, because of the ambiguity of the statement "individuals who were working in a different country as their nationality" (i.e. immigrants) – it is unclear who is being excluded and whether the issue of the social construction of stress can be redressed through the exclusion. The ESWC does not identify whether a person was born in the country of present nationality, lived in the same country since birth with or without the nationality of that country, or lived, or was born, elsewhere before coming to the country of present nationality. Neither does the ESWC provide data on when, in a life course perspective, the arrival of a person to a country occurred which may be relevant for assimilating social constructs by an individual. All these factors can influence health variability and without such detailed data deductions based only on a person’s nationality seems uncertain.

Second, information bias can affect the results.[3] By means of principal component analyses, Daniels grouped a number of items (stress, anxiety, irritability, sleeping problems, stomach ache, headache, or overall fatigue) to create an "index of perceived risk from occupational stress". However, two methodological issues have to be considered. First, if, as Daniels argues, stress is socially constructed and thereby biased by nationality, how can be the occupational stress index be created using the total sample instead of only those workers who were working in the same country as their nationality. We would agree this was the appropriate approach if Daniels would not have argued for a social construction of stress. Second, in the ESWC questionnaire, a filter question existed before the selection of the symptoms that the author did not mention. Only after replying that their work affected their health, which some 60% of the workers did, could participants endorse any of 22 symptoms.[2] In analyses using the same data set,[8,10,11] we found differences in various health-related outcomes (stress, overall fatigue, job dissatisfaction and sickness absence) by country, type of employment, and gender.[12] So, grouping items without exploring individual symptom variation might hide important differences.

Third, in the discussion section, Daniels stated that according to his findings, compared to workers in higher ranked countries, British workers and workers in other lower ranked countries may be less likely to attribute minor or non-specific psychosomatic symptoms to work (or might attribute them erroneously to other life domains) when consulting with health practitioners. We are aware of no data to support such an assertion. Hence, the findings do not support this conclusion, so we suggest prudence in the application of these results to policy and practice decisions.

Fourth, Daniels ranked the EU countries investigated according to the number of stress related symptoms attributed to work. From these differences Daniels’s made inferences about how sociocultural factors and the role of the media can affect health perceptions. Given the concerns expressed above abut selection and information bias we wonder how the results led to these inferences. Furthermore, although Daniels recognised that these processes may not apply to other EU countries, he focused almost the entire discussion section on the UK, building an idiosyncratic and ad hoc explanation for the UK that seems unsupported by the research. A way to facilitate more systematic analyses could be by classifying the countries according to certain common characteristics. In social and policy research, different typologies have been proposed. For instance, Esping-Andersen’s typology has proven to be useful in comparing welfare states and may be helpful in the present study’s context.[13]

Moreover, the Daniels’ ranking is very different, and in some cases completely inverse, than the one obtained by us when analyzing sickness absence data from the same survey.[14] Among other factors that could explain this divergence, it has to be considered that different outcomes could present different between-countries distributions. Compared to Daniels’ occupational stress index, sickness absence is recognised as a public health surveillance measure indicating countries economic performance.[15] Rankings are an important tool scientists can use to communicate with policy makers. The ESWC provides a unique data source for scientists. Differences between our work and Daniels illustrates the need to develop mechanisms to enable researchers using the ESWC to share results prior to publication.

Finally, we do understand that the space available for short papers limits the author’s possibility to explain in detail all relevant issues. However, we think that authors should give the reader as much information as needed to judge the results validity.

References

1.Daniels K. Perceived risk from occupational stress: a survey of 15 European countries. Occup Environ Med 2004; 61: 467-70.

2.Paoli P, Merllié D. Third European Survey on Working Conditions 2000. Dublin: European Foundation for the Improvement of Living and Working Conditions Luxembourg: Office for Official Publication of the European Communities, 2001.

3.Szklo M, Nieto FJ. Understanding lack of validity: Bias. In: Szklo M, Nieto FJ. Epidemiology. Beyond the Basics. Gaithersburg: Aspen Publishers, Inc., 2000; 125-76.

4.Lowell BL, Kemper Y. Transatlantic Roundtable on Low-skilled Migration in the Twenty-first Century: Prospects and Policies. International Migration 2004;42(1):117-40.

5.Hilderink H, van der Gaag N, van Wissen L, Jennissen R, Román A, Salt J, Clarke J, Pinkerton C. Analysis and forecasting of international migration by major groups - Part III. Luxembourg: Office for Official Publication of the European Communities, 2003.

6.Angrist JD, Kugler AD. Protective or counter-productive? Labour market institutions and the effect of immigration on EU natives. Economic Journal 2003; 113(488): F302-F331.

7.Thorogood D, Winqvist K. Women and men migrating to and from the European Union. Statistics in focus 2/2003. European Communities, 2003.

8.Wrench J. Preventing Racism at the Workplace. A summary. European Foundation for the Improvement of Living and Working Conditions. Luxembourg: Office for Official Publications of the European Communities, 1996.

9.Benach J, Gimeno D, Benavides FG. Employment status and health. Dublin: European Foundation for the Improvement of Living and Working Conditions. Luxembourg: Office for Official Publication of the European Communities, 2002.

10.Benach J, Gimeno D, Benavides FG, Martínez JM, Torné MM. Types of employment and health in the European Union: changes from 1995 to 2000. Eur J Public Health (in press).

11.Gimeno D, Benavides FG, Amick III BC, Benach J, Martínez JM. Psychosocial factors and work-related sickness absence among permanent and non-permanent employment. J Epidemiol Community Health (in press).

12.Karjalainen A (Coordinator). Work and health in the EU. A statistical portrait. Data 1994–2002. Luxembourg: Office for Official Publication of the European Communities, 2004.

13.Esping-Andersen G. The three worlds of welfare capitalism. Princeton (NJ): Princeton University Press, 1990.

14.Gimeno D, Benavides FG, Benach J, Amick III BC. Distribution of sickness absence in the European Union countries Occup Environ Med (in press).

15.Kivimaki M, Head J, Ferrie JE, Shipley MJ, Vahtera J, Marmot MG. Sickness absence as a global measure of health: evidence from mortality in the Whitehall II prospective cohort study. BMJ 2003 16;327:364-9.