Division of Occupational Health, Department of Public Health Sciences, Karolinska Institute, SE-171 76 Stockholm, Sweden
Received 14 July 2000; in revised form ; accepted 5 December 2000
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Several workplace-related explanations, such as placement in stressful or non-rewarding jobs, participation in job-based drinking networks, and absence of work-based social support have been offered for why some occupations are more associated with alcoholism than others (Martin et al., 1996). We have previously found low work control to be related to alcoholism among young men (Hemmingsson and Lundberg, 1998
). Low work control (or decision latitude) was, according to Karasek and Theorell (1990), defined as a combination of skill discretion, meaning the degree of variation concerning job tasks, and decision authority, the opportunity for making autonomous decisions at work.
The individual risk of developing alcoholism depends on both individual predisposition (biologically, genetically or socially determined) and environmental factors (Schuckit, 1987; Skog, 1991
; Edwards et al., 1994
). Accordingly predisposed persons may be at elevated risk of developing alcohol abuse if exposed to unfavourable work environments, such as work characterized by low work control. The relationship may be synergistic, i.e. some persons develop misuse only under the influence of both factors acting together.
We employed data from Swedish men, born during the period 19491951, gathered at time of conscription to compulsory military service at age 1820 years. Information on individual risk factors for alcoholism from late adolescence were available. Information on specific psychosocial work environment risk factors was based on information on occupation in 1975. Our hypothesis was that those risk factors, separated in time, would interact to increase the risk of alcoholism.
![]() |
SUBJECTS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Risk factors established in late adolescence
At conscription, all men were asked to complete two questionnaires. The first concerned social background, behaviour and adjustment, psychological factors, and health. The second dealt specifically with substance use, e.g. alcohol consumption and tobacco smoking. All the conscripts were seen by a psychologist for a structured interview, assessment of intellectual capacity and emotional control, and ratings on a few other predetermined scales.
The variable heavy use of alcohol', as measured at the examination at conscription, was previously shown to be a strong risk factor of alcoholism diagnoses in this study group (Hemmingsson et al., 1998). Heavy use of alcohol is a composite variable including at least one of the following indicators of problem drinking: consumption of at least 250 g of 100% ethanol/week, to have taken an eye-opener during hangover, to have been apprehended for drunkenness, or to have often been drunk. Each factor included in the composite variable has previously been shown to be related to an alcoholism diagnosis in this study group (Hemmingsson et al., 1997
; Upmark et al., 1997
).
Five other variables were chosen for inclusion in the analyses on the basis that they were previously known predictors of alcoholism diagnoses (see Appendix 1). We have shown that the predictors included are more common among those with unfavourable psychosocial work characteristics (Hemmingsson and Lundberg, 1998). The variables are used in a multivariate model to control for confounding effects.
Occupation and social class
Information on occupation for each conscript was obtained by record linkage with the National Population and Housing Census of 1975, for which there was a response rate of 99%.
Occupations were coded according to the Nordic modification of the three-digit International Standard Classification of Occupations. A classification into two social classes, non-manual employees and manual workers (i.e. white-collar and blue-collar occupations) was also made on the basis of information on reported occupation in 1975.
Job characteristics
Low work control (or decision latitude) was defined according to Karasek and Theorell (1990) as a combination of skill discretion (meaning variation concerning job tasks) and decision authority (referring to the scope at work for making autonomous decisions).
A job exposure matrix (JEM) was used to rate all occupations on a 10-digit scale for work control and other work-related factors (Johnson et al., 1990; Karasek and Theorell, 1990
). The JEM was developed for application on population-based data. The source of information was the annual Swedish Surveys of Living Conditions 197583 (ULF) (Statistiska centralbyrån, 1985
), which included questions on work control (12 questions); 242 occupations, including 42 001 individuals, were attributed values for work control.
Based on the value attributed to each occupation, ranked from low to high, all subjects were grouped as being below or above the median value for work control. In the analyses of the relationship between heavy use of alcohol, work control and alcoholism diagnoses, subjects with values of work control under the median (exposed) were compared with subjects with values above the median (reference category). Common low control jobs (jobs within the lowest quartile) included toolmakers, welders, woodworking machine operators, and truck drivers. Common high control jobs (jobs within the highest quartile) included engineers, bank employees, buyers, and officers in the armed forces.
Separate analyses were conducted for blue-collar workers alone. In those analyses, group limits, concerning level of exposure for work control, for the entire material were employed for the purpose of categorization.
Men who received an alcoholism diagnosis during the period 19761983
We utilized the unique civic registration numbers assigned to every Swedish resident at birth to perform record linkages with the Register of Diagnoses at Discharge from In-patient Psychiatric Care, to determine whether or not men in the cohort had received an alcoholism diagnosis [alcohol psychosis (ICD-8 291), alcoholism (ICD-8 303), and alcohol intoxication (ICD-8 980)] during the period 19761983. Primary and secondary diagnoses at any discharge were included. In 1984, the register of psychiatric diagnoses was discontinued, and such data were no longer available for research. Subjects who received an alcoholism diagnosis at conscription or in the Register of Diagnoses at Discharge from In-patient Psychiatric Care 19731975 were excluded from the study population.
Calculation of interaction effects
In the analyses, relative risks (RRs) for combinations of risk factors, i.e. heavy use of alcohol and an unfavourable work environment, were calculated with those jointly unexposed as reference category. The hypothesis is that the RR estimate among those exposed for both risk factors will be at least additively increased. The interaction effect was calculated according to Rothman (1986) and the synergy index (SI) is reported. Synergistic interaction refers to a situation where two exposures are component causes in the same sufficient cause. If some persons develop alcohol abuse only under the influence from both factors acting together (i.e. heavy use of alcohol and low work control) synergistic interaction occurs. An interaction effect is proven by departure from additivity of absolute effects, i.e. the relative excess risk among those with combined exposure should exceed the sum of the relative excess risks for each of the component causes where those unexposed to both causes are used as reference category (Rothman, 1986; Hallquist et al., 1996). Synergy index is defined as:
![]() |
where denotes unexposed and RR(A B) the RR among those exposed to both factors where RR(
) is used as reference category (RR = 1.0). SI >1.0 indicates more than additive interaction, i.e. synergy. SI = 2.0 denotes a RR among those exposed to both factors twice as high as would be expected from additivity and implies that 50% of the increased disease occurrence among those jointly exposed is attributable to the interaction (Rothman, 1986
).
A program developed in the SAS computer package for the calculations of SI with confidence intervals was used (Lundberg et al., 1996). Confidence intervals were calculated as proposed by Hosmer and Lemeshow (1992).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Combined effects of heavy use of alcohol in late adolescence and later exposure to low control at work in relation to alcoholism diagnosis
Persons who in late adolescence reported heavy use of alcohol more often received an alcoholism diagnosis if they were later exposed to a work environment characterized by low control when compared to those jointly unexposed (Table 1a). When, for each risk factor, the other risk indicators were taken into consideration, the risk estimates in each cell decreased markedly, indicating a confounding effect (Table 1b
). However, the estimate of the synergy index changed only marginally and remained significantly increased while its confidence intervals became wider.
|
|
Of the individuals who were included in the analyses in this paper, i.e. those who held an occupation according to the 1975 census, only 64% could be attributed a value for level of work control from the 1970 census. At this point in time, some subjects were still studying, whereas others were doing their military service. In the remaining group, 75% of those who were classified as having low work control in 1975 were also classified as having low control in 1970. Those who changed from low to high work control, or from high to low work control, did not show an increased RR, compared with those with long-term high work control (reference category). But we found a significantly increased RR among persons in jobs with low work control in both censuses when other risk indicators were taken into account (Table 3). The same pattern was seen when only manual workers were investigated. Accordingly, it seems as if long-term exposure to low work control was related to high level of alcohol consumption.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The study has several advantages. It has a longitudinal design; data on individual characteristics (predisposing factors) were collected when the subjects were 1820 years of age, work characteristics when the subjects were 2426 years of age, and alcoholism diagnoses when the subjects were 2534 years of age. Information was available on several known risk indicators for alcoholism; outcome data were diagnoses from in-patient care; the study population was likely to be highly representative of not only the young male working population in Sweden in one birth cohort but of quite a few adjacent birth cohorts as well, and a large number of occupations were also represented. Weaknesses of the study are that only men were considered, and that the follow-up period was relatively short (8 years).
We shall discuss below the limitations of the study and the generalizability of the results.
In survey investigations, considerable under-reporting of alcohol consumption is likely, and under-reporting is assumed to occur in proportion to actual consumption level (Kühlhorn, 1998). Although the other risk indicators used in this study, to control for confounding, are subject to misclassification, the factors employed are strongly related to an alcoholism diagnosis and explain a large proportion of the social class differences in alcoholism in this material (Hemmingsson et al., 1998
).
A differential misclassification of outcome could occur if persons from lower social strata, where low job control is more prevalent, more easily receive an alcoholism diagnosis and are given comparable symptomatology, than persons from higher social strata. It has been suggested that such misclassification does occur, e.g. due to the attitudes of physicians (Wolf et al., 1965). If such differential misclassification of outcome occurs, it is more likely that it occurs in in-patient care, where the patient is present, and less likely for diagnoses at death. However, socio-economic differences in alcohol-related mortality in Sweden show the same pattern as for alcoholism diagnoses from in-patient care.
Johnson and Stewart (1993) have found that JEM scores accounted for 63.5% of individual variance in work control. These authors concluded that job control is a work characteristic. Given the strength of the relationship between social class and control, it seems unlikely that job changes would have a major impact on degree of control over shorter time periods. There may be some misclassification of individuals in our study with regard to work control. Such misclassification is independent of both background factors and outcomes, and would thus result in a bias towards unity in the RRs associated with work control.
Heavy use of alcohol was a composite variable based on four indicators of problem drinking (see Appendix 1). Those who were classified as exposed to at least one of those indicators were classified as heavy users of alcohol. There were no differences in average exposure level for the separate indicators among those classified as exposed to heavy use of alcohol between low and high control workers in 1975. The proportion exposed to more than one indicator of heavy use of alcohol was only marginally higher among low control workers (19%) compared with high control workers (15%).
A possible criticism of our findings is that they could reflect a selection of heavy drinkers to low control jobs, rather than a combined effect of early established heavy drinking and the low control in low control jobs. For several reasons, we argue that the interaction effect cannot to any major extent be explained by selection. First, in the analyses we controlled for several other previously known risk factors for later alcoholism that were already established at late adolescence. Those risk factors were important to decrease the RR in each group, but there were still significantly increased RRs in almost all groups, as were the synergy indices in most groups. Second, as shown in Table 3, the RR of alcoholism was not increased among those who changed from high to low control jobs between 1970 and 1975. This indicates that persons with heavy alcohol consumption at conscription were not selected from high to low control jobs during this period. Third, we did not find a tendency for persons who developed problem drinking between the conscription survey, where data on heavy alcohol consumption were collected, and the census in 1975, where data on work control were collected, to move to low control jobs. Persons with no heavy alcohol consumption at time of conscription, but who later increased their alcohol consumption, and with later low work control would be classified into the less than heavy drinking category, with low job control in our analyses (see Table 2
) and there contribute to an increased RR. Those categories showed only very small increased RRs of an alcoholism diagnosis. Our interpretation is that the increased RR of an alcoholism diagnosis in low control jobs could be explained only to a very limited extent by a selection effect when other risk factors at conscription are controlled for.
It has been suggested that the individual risk of developing alcoholism depends on both individual prediposition and environmental factors (Schuckit, 1987; Skog, 1991
). The results of this study are in accordance with such suggestions: heavy use of alcohol in late adolescence interacted synergistically with low control at work to increase the risk of alcoholism. The consequence is that some cases of alcoholism diagnosis occurred only because of these factors acting together. From a health policy perspective, knowledge of processes of interaction is important. If risk factor A interacts with risk factor B, intervention directed at risk factor A, besides its effect on A, also contributes to lowering the risk from risk factor B.
Heavy use of alcohol in late adolescence could for some persons be a link in a chain of negative circumstances during early life (Glendinning et al., 1995). Such circumstances could include poor school achievement, which restricts the opportunities to avoid unfavourable work environments in adult life (Koivusilta et al., 1998
). It has been suggested that working conditions may influence the total life situation. Men in jobs with low control appear to apply the lessons of their job to their behaviour off the job (Kohn, 1976
; Parker and Farmer, 1990
). Long-term exposure to poor working conditions combined with low economic reward may lead to feelings of hopelessness and depression as well as poorer behavioural risk factor profiles (Lynch et al., 1997
). The results in this study suggest that young men with an early established heavy alcohol consumption more often respond to such work environments by increasing their alcohol consumption compared with young men without an early established heavy alcohol consumption.
Although risk factors probably can accumulate over the entire life course, it has been suggested that the transition from adolescence into adulthood and work life is a critical period for the establishment of adult drinking behaviours (Roman and Blum, 1992). How this age-related window exerts influence on future alcohol habits probably depends on background factors, such as individual predisposition.
In the first step of the analyses, we investigated if the risk factors, measured at conscription, and the later work conditions acted independently of each other, or synergistically, to increase the risk for alcoholism. In the second step of the analyses, we also controlled for other predictors of alcoholism established in late adolescence. Since these predictors decreased the RR differences among the groups with different combinations of risk factors/working environments, the findings indicate a differential accumulation of risk factors for alcoholism in the exposure groups.
In conclusion, this longitudinal study indicates that heavy use of alcohol in late adolescence interacts with later working conditions characterized by low control to increase the risk of a later alcoholism diagnosis. Also other risk factors for alcoholism seem to accumulate among those with low work control.
![]() |
APPENDIX: VARIABLES INCLUDED IN THE DATA ANALYSES* |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Smoking cigarettes/day: (1) >20; (2) 1120; (3) 610; (4) 15; (5) not smoking
Emotional control: This variable was assessed by a psychologist at conscription as a summary assessment of mental stability, emotional maturity, and tolerance for stress and frustration
Psychiatric diagnosis at conscription: (1) Yes; (2) No
Contact with police or childcare: Have you had any contact with police or child-care authorities? (1) >1 time; (2) 1 time; (3) 0 times
*Underlining denotes exposure.
Ranking on psychometric test: The psychometric test performed included tests on general intellectual ability, verbal ability, visuospatial ability, and formal reasoning. Ratings were on the scale 19. We used three categories where 1, 2 and 3 formed the lowest 20%, and 7, 8 and 9 formed the highest 20%.
![]() |
ACKNOWLEDGEMENTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
FOOTNOTES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Andréasson, S., Allebeck, P. and Brandt, L. (1993) Predictors of alcoholism in young Swedish men. American Journal of Public Health 83, 845850.[Abstract]
Clapper, R. L., Buka, S. L., Goldfield, E. C., Lipsitt, L. P. and Tsuang, M. T. (1995) Adolescent problem behaviors as predictors of adult alcohol diagnoses. International Journal of Addiction 30, 507523.[ISI][Medline]
Donovan, J. E. and Jessor, R. (1978) Adolescent problem drinking. Psychosocial correlates in a national sample. Journal of Studies on Alcohol 39, 15061524.[ISI][Medline]
Edwards, G., Anderson, P. and Babor, F. (eds) (1994) Alcohol Policy and the Public Good. Oxford University Press.
Glendinning, A., Hendry, L. and Shucksmith, J. (1995) Lifestyle, health and social class in adolescence. Social Science and Medicine 2, 235248.
Hallqvist, J., Ahlbom, A., Diderichsen, F. and Reuterwall, C. (1996) How to evaluate interaction between causes: a review of practices in cardiovascular epidemiology. Journal of Internal Medicine 239, 377382.[ISI][Medline]
Hawkins, J. D., Catalano, R. F. and Miller, J. Y. (1992) Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin 112, 64105.[ISI][Medline]
Hemmingsson, T. and Lundberg, I. (1998) Work control, work demands, and work social support in relation to alcoholism among young men. Alcoholism: Clinical and Experimental Research 4, 921927.
Hemmingsson, T., Lundberg, I., Nilsson, R. and Allebeck, P. (1997) Health-related selection to seafaring occupations and its effect on morbidity and mortality. American Journal of Industrial Medicine 31, 662668.[ISI][Medline]
Hemmingsson, T., Lundberg, I., Diderichsen, F. and Allebeck, P. (1998) Explanations of social class differences in alcoholism among young men. Social Science and Medicine 47, 13991405.[ISI][Medline]
Hosmer, D. W. and Lemeshow, S. (1992) Confidence interval estimation of interaction. Epidemiology 3, 452456.[ISI][Medline]
Johnson, J. and Stewart, W. (1993) Measuring work organization exposure over the life course with a job-exposure matrix. Scandinavian Journal of Work Environment and Health 19, 2128.
Johnson, J., Stewart, W., Fredlund, P., Hall, E. M. and Theorell, T. (1990) Psychosocial job exposure matrix: An occupationally aggregated attribution system for work environment exposure characteristics. Karolinska Institute: Stress Research Reports, No. 221. Stockholm, Sweden.
Kandel, D. B. and Andrews, K. (1987) Process of adolescent socialisation by parents and peers. International Journal of Addiction 22, 319342.[ISI][Medline]
Kandel, D. B. and Yamaguchi, K. (1985) Developmental patterns of the use of legal, illegal, and medically prescribed psychotropic drugs from adolescence to young adulthood. NIDA Research Monograph 56, 193235.[Medline]
Karasek, R. A. and Theorell, T. (1990) Healthy Work. Stress, Productivity, and the Reconstruction of Working Life. Basic Books, New York.
Kilty, K. M. (1990) Drinking styles of adolescents and young adults. Journal of Studies on Alcohol 51, 556564.[ISI][Medline]
Kohn, L. (1976) Occupational structure and alienation. American Journal of Sociology 8, 111130.
Koivusilta, L., Rimpelä, A. and Rimpelä, M. (1998) Health related lifestyle in adolescence predicts adult educational level: a longitudinal study from Finland. Journal of Epidemiology and Community Health 52, 794801.[Abstract]
Kühlhorn, E. (1998) Problem och nya grepp i kartläggningar av alkoholkonsumtion och alkoholvanor. In Svenska alkoholvanor i förändring, Kühlhorn, E. and Björ, J. eds. Sober förlag (in Swedish).
Lundberg, M., Fredlund, P., Hallqvist, J. and Diderichsen, F. (1996) A SAS program calculating three measures of interaction with confidence intervals. Epidemiology 7, 655656.[ISI][Medline]
Lynch, J., Krause, N., Kaplan, G. A., Tuomiletho, J. and Salonon, J. (1997) Workplace conditions, socio-economic status, and the risk of mortality and acute myocardial infarction: the Kuopio Ischemic Heart Disease Risk Factor Study. American Journal of Public Health 4, 617622.
Martin, J. K., Roman, P. M. and Blum, T. C. (1996) Job stress, drinking networks, and social support at work: A comprehensive model of employees' problem drinking behaviors. Sociological Quarterly 37, 579599.[ISI]
Parker, D. A. and Farmer, G. C. (1990) Employed adults at risk for diminished self-control over alcohol use: The alienated, the burned out, and the unchallenged. In Alcohol Problem Intervention in the Work Place, Roman, P. ed. Quorum Press, New York.
Roman, P. M. and Blum, T. C. (1992) Life transitions. Work and alcohol: An overview and preliminary data. Alcoholism: Clinical and Experimental Research 16, 149158.[ISI][Medline]
Rothman, K. J. (1986) Modern Epidemiology. Little and Brown, Boston.
Schuckit, M. A. (1987) Studies of populations at high risk for the future development of alcoholism. Progress in Clinical and Biological Research 241, 8396.[Medline]
Skog, O.-J. (1991) Alkoholisme, arv og miljo. Nordisk Alkoholtidskrift 8, 6372.
Statistiska centralbryån (1985) Ohälsa och sjukvård. Levnadsförhållanden 197583. Rapport 42 (ULF), Stockholm.
Upmark, M., Hemmingsson, T., Romelsjö, A., Lundberg, I. and Allebeck, P. (1997) Predictors of disability pension among young men: The role of alcohol and psychosocial factors. European Journal of Public Health 7, 2028.[Abstract]
Wolf, I., Morris, I. C., Blane, H. T. and Hill, M. J. (1965) Social factors in the diagnosis of alcoholism. II. Attitudes of physicians. Quarterly Journal of Studies on Alcohol 26, 7279.[ISI]