Over the past decades impressive progress has been achieved in the study of work stress-related health outcomes (Cooper and Payne, 1991; French et al., 1982; House, 1981; Johnsson and Johansson, 1991; Kahn, 1981; Karasek and Theorell, 1990; La Ferla and Levi, 1993; Sauter et al., 1989). This progress concerns both the development of theoretical con- cepts and methods to adequately define and assess stressful experience at work and the empirical dem- onstration of the validity of these concepts. In mod- ern working life, more and more occupations are characterized by a high prevalence of a variety of psychomental and socio-emotional stressors, either at the expense of, or in combination with, physical and chemical hazards. To compare the level of psy- chomental and socio-emotional stressors among a broad range of occupations and to identify critical thresholds of stressful experience at work analytical concepts are needed which overcome the obvious limitations of descriptive approaches (e.g. in terms of job titles).
*Author for correspondence.
In recent past, two such concepts were more widely discussed and tested than others, the per- son environment fit model (French et al., 1982), and the demand-cont ro l model (Karasek, 1979; Karasek and Theorell, 1990). The person-environ- ment fit model puts its main emphasis on the stress- fulness of experienced incongruence between a person's abilities and the demands of his or her job, and the incongruence between a person's goals or aspirations and the supplies offered by the work en- vironment. Whereas this model takes into account the interaction between work environment and per- son characteristics, the demand-cont ro l model explicitely restricts itself to the assessment of job task characteristics. More specifically, jobs defined by high demands in combination with low control (in terms of decision latitude and skill discretion) are considered stressful as they induce autonomic arousal instead of enabling experiences of learning, stimulation and satisfaction.
Both models were tested in a variety of studies, using different indicators of health outcomes, but obviously the validity of the demand-cont ro l model so far seems more convincing, especially so if two types of outcome criteria are considered: (1) stress-
1111
1112 Richard Peter and Johannes Siegrist
related physical disorders such as cardiovascular risk factors and cardiovascular (esp. coronary artery) disease (Karasek and Theorell, 1990; Schnall et al., 1994; Theorell and Karasek, 1996), and (2) sickness absence behavior (Kristensen, 1991; North et al. , 1993). Despite its contribution to the statisti- cal explanation of work stress-related health out- comes the demand-control model leaves researchers with some important unresolved questions, es- pecially those concerning the meaning of "control", the role of personal coping with work demands, and the problem of sensitivity in predicting health outcomes. This latter problem is of special import- ance as it raises the more fundamental question of pathogenic processes which are implicated in the theoretical model (for discussion see Kasl, 1996; Siegrist, 1996a).
In this paper, we discuss and, to some extent, challenge the notion of a general explanatory func- tion of work stress models which disregard the issue of specificity in predicting health outcomes. We challenge this notion both on theoretical and empirical grounds. In stress-theoretical terms, it is difficult to understand how exposure to the same stressful condition ("strain job" as defined by high demand and low control) may be linked to such different phenomena as suffering from pathophysio- logic consequences of sustained autonomic arousal (e.g. "neurogenic" hypertension or coronary artery disease) or as behavioral decision-making (the de- cision to stay away from work: "sickness absence"). Clearly, part of sickness absence may be directly due to the long-term health costs of autonomic arousal, but the proportion of variance in sickness absence explained by these disorders is probably low. Alternatively, one may argue that exposure to a "strain job" results in an increase of non-specific "'susceptibility" or "vulnerability" and that the specific health consequences may depend on individ- ual risk factors and predispositions. However, in this case, the predictive power of the model con- cerning specific health outcomes is limited, and ad- ditional explanatory constructs need to be introduced. Interestingly enough, the potential offered for differential prediction of health out- comes by the demand-control model has not been explored and tested so far. Although the model clearly distinguishes between four psychological states related to "high strain job", "low strain job", "passive job" and "active job" the different types of stressful states were not related to different health outcomes. Rather, the authors maintain that "'for the passive job we hypothesize only an average level of psychological strain and illness risk (as in the case of active work) . . . " (Karasek and Theorell, 1990, p. 38).
In accordance with recent physiological (Henry and Stephens, 1977; Mason, 1968; Weiner, 1992) and psychological stress theories (Lazarus, 1991) we maintain that adverse health effects produced by
the exposure to chronically stressful conditions lar- gely depend on the results of individual coping pro- cesses, and especially on the degree of control obtained over threatening stimuli. Coping efforts are exerted at different levels including cognitions, emotions, autonomic nervous system activation and overt behavior. In stress-theoretical terms the dis- tinction between problem-focused (active) and emotion-focused (passive) coping was shown to be particularly important (Carver et al. , 1989; Lazarus and Folkman, 1984). Evidence indicates that these two patterns of coping clearly differ at the levels of overt behavior and of physiologic response (Henry and Stephens, 1977; Mason, 1968). While active coping with challenge under conditions of limited control is characterized by excessive striving and struggling and by a simultaneous activation of at least two stress axes within the organism, the sym- patho-adrenomedullary axis and the pituitary-adre- nocortical axis (Henry and Stephens, 1977; Manuck et al., 1995), passive coping involves avoidance or withdrawal behavior and an excessive activation of the pituitary-adrenocortical stress axis (Axelrod and Reisine, 1984; Levine, 1987; Mason, 1968; Seligman, 1975). Elsewhere it has been argued that active coping under conditions of low control ("sus- tained active distress") in the long run adversely affects the cardiovascular system whereas passive coping under limited control is associated with de- pressive mood, depressed immunocompetence, and "'giving up" behavior (Ballieux et al. , 1984; Beamish et al., 1985; Kiecolt-Glaser and Glaser, 1995; Lazarus, 1989: Manuck et al., 1995). This argument suggests that chronically stressful experience at work followed by active coping is more likely to be associated with pathophysiological conditions of active distress (such as cardiovascular risk and dis- ease), whereas stressful job experience followed by passive coping is more likely to be associated with depressive mood and withdrawal behavior (such as sickness absence behavior).
To test this hypothesis a theoretical model of work stress is needed by which differential predic- tion of health outcomes according to the criterion of active versus passive coping can be made. We propose that the model of effort reward imbalance at work allows for such differential prediction (Siegrist, 1996a,b). In this model, the work role in adult life is considered a basic tool to link import- ant emotional and motivational self-regulatory needs of a person, such as self-esteem and self-effi- cacy, with the social opportunity structure. The availability of an occupational status is associated with recurrent options of contributing and perform- ing (job tasks), of being rewarded and esteemed (salary, income), and of belonging to some signifi- cant group (e.g. work collegues). Yet, these poten- tially beneficial effects of the work role on self- regulation are contingent on a basic prerequisite of exchange in social life, that is, reciprocity. Effort at
Chronic work stress and sickness absence 1113
work is spent as part of a socially organized exchange process to which society at large contrib- utes in terms of rewards. These rewards are distrib- uted by three transmitter systems to the working population: money, esteem, and status control. The model of effort-reward imbalance claims that lack of reciprocity between costs and gains (i.e. high cost/low gain conditions), define a state of emotion- al distress with special propensity to autonomic arousal and associated strain reactions. This holds especially true if poor reward is experienced in terms of low job stability, forced occupational change, downward mobility or lack of promotion prospects (low occupational status control). Therefore, having a demanding, but unstable job, achieving at high level without being offered any promotion prospects, are examples of a particularly stressful working context.
In summary, the model of effort reward imbal- ance defines a state of sustained active coping under conditions of limited control. But why should people engage in such unfavorable tradeoffs in their working life? Why should they decide to pursue their active coping with work demands instead of withdrawing from these constraints? According to our theory, two factors operate against reduced engagement. Firstly, the possible costs produced by disengagement (e.g. the risk of being laid off or of facing downward mobility) may outweigh the costs of accepting inadequate benefits. Secondly, in our model we define two sources of high effort at work, an extrinsic source, the demands on the job, and an intrinsic source, the motivations of the individual worker in a demanding situation. In this latter regard, we introduced the concept of "need for con- trol" as a personal pattern of coping with the demands at work (see Methods section). It is likely that persons with high need for control spend high costs in terms of energy mobilization and job invol- vement even under conditions of relatively low gain. This may be explained partly by the character- istics of their perceptual and attributable style, partly by the self-gratifying experience of "being in control" of a challenging situation. Yet, under cer- tain conditions (e.g. lack of excessive "need for con- trol", fair job stability) workers may react to the experience of effort-reward imbalance by passive rather than active coping. Reduced expectancy, dis- tancing and calculated disengagement is likely to occur, and sickness absence behavior becomes part of these strategies of calculated disengagement (see also Kristensen, 1991; Lazarus, 1989; Malmgren and Andersson. 1984: Melamed et al., 1992; North et aL, 1993).
Within the constraints of a small-scale cross-sec- tional study we test the hypothesis that employees who exhibit active coping behavior in terms of high effort at work in combination with low reward are more likely to suffer from cardiovascular risk (as evidenced by the prevalence of manifest hyperten-
sion), whereas employees who exhibit passive cop- ing are more likely to experience episodes of withdrawal and "giving up" behavior in terms of sickness absence. Passive coping is characterized by stressful experience of low occupational rewards in the absence of high extrinsic or intrinsic effort at work. These associations are assumed to exist after statistically adjusting for relevant confounders.
DATA AND M E T H O D S
Study sample
A cross-sectional study was conducted in a large car-producing enterprise as part of a cardiovascular screening offered to the total group of middle man- agers by the company's Occupational Health Department. Based on epidemiologic information on prevalence of cardiovascular risk factors we restricted our investigation to men aged 40-55 years. During the study period a total of 189 middle managers in this age category were screened. Roughly 95% (n = 179) of them agreed to partici- pate in our investigation. Of the remaining 10 men, six refused to participate and four suffered from long-term sickness absence.
In this study middle managers have been choosen for two reasons. Firstly, we decided to study a population as homogeneous as possible in terms of education, training and occupational status to rule out reported confounding effects of these variables (Kristensen, 1991; Marmot et al., 1995; North et al., 1993, 1996). Secondly, due to exposure to specific job characteristics, we expected a rather high proportion of middle managers to exhibit high chronic work stress in terms of high effort and/or low reward. The present analysis is restricted to middle managers with complete data. This leaves 170 subjects for analysis with regard to hyperten- sion and 146 subjects for analysis with regard to sickness absence (90 and 82%, respectively, of the study sample). When considering possible sample bias within the reduced study group we did not find systematic differences on core variables between the full and the reduced sample with the exception of status inconsistency: a significantly higher pro- portion of middle managers suffering from status inconsistency was found in the full as compared to the reduced sample (sickness absence data). However, as status inconsistency is one of the indi- cators of low occupational rewards, the reported findings (see Results section) may underestimate rather than overestimate the observed effect.
Psyehosoeial and behavioral measures
Data on sickness absence covering the year in which we conducted this study were obtained from the company's office. These records included first and last date of all sickness absence episodes. In ad- dition, the officially registered reason for sickness
SSM 45 7 F
1114 Richard Peter and
absence, illness or accident, was available. However, due to reasons of personal data protection, we had no access to the specific medical diagnosis. To inter- pret the distinction made between short-term and long-term sickness absence it should be noted that in Germany, for sickness absence up to three days (short-term) there is no need for a medical certifi- cate. Rather, staying away from work can be justi- fied by the employees themselves. Medical certificates are required for long-term sickness absence (four and more days only). Information on important confounding factors concerning associ- ations with sickness absence rates such as preva- lence of a diagnosed chronic disease, subjective health, mood, health detrimental behaviors (smok- ing, alcohol consumption, lack of physical exercise, overweight) and age (Kristensen, 1991; Malmgren and Andersson, 1984; Marmot et al., 1993, 1995) was recorded from personal interview (see below). Information on prevalence of chronic disease (not directly related to sickness absence data) was obtained from the company's Occupational Health Department. The records included information on most prevalent chronic diseases such as cardiovas- cular disease, gastrointestinal disease, musculoscele- tal disorders, etc.
The variables of the model of effort-reward imbalance at work were measured by a set of Likert-scaled items which were previously tested and which were assessed by means of a structured interview and a self-administered questionnaire (Peter, 1991; Siegrist, 1996b). The experience of extrinsic effort (demands) was measured by asses- sing the intensity of being distressed by time press- ure, frequent interruptions, inconsistency of demands, and severity of work problems (six items). Measures of occupational rewards covered the three dimensions of income, esteem, and occupational status control (11 items). In this latter regard the stressfulness of job instability, low promotion pro- spects, forced job change and status inconsistency was explored. Status inconsistency was conceptual- ized as status discrepancy measuring the inconsis- tency between educational level and job position, and status incongruence measuring the mismatch between efforts and aspirations on the one hand and career achievements on the other hand. Again, the intensity was assessed by Likert-scaled items.
Intrinsic effort was assessed by a self-adminis- tered questionnaire termed "need for control" which defines a critical personal style of coping with work demands. The questionnaire contains 29 dichotomous items describing excessive job involve- ment, positive and negative feelings and attitudes related to work commitment as measured by four unidimensional scales: (1) need for approval; (2) competitiveness and latent hostility; (3) impatience and disproportionate irritability; and (4) inability to withdraw from work obligations. These four sub- scales were repeatedly found to load on one latent
Johannes Siegrist
factor termed "immersion" (Siegrist, 1996b). In a previous study we found that persons scoring in the upper tertile of immersion were at excess risk of suffering from coronary heart disease (Siegrist et al., 1990).
Self reported quality of health was assessed by a single five-point Likert-scaled item. Emotional mood was assessed by two Likert-scaled items exploring the intensity of feelings of sustained anger and of sustained helplessness and hopelessness during the past 12 months. A further part of the structured interview which preceeded the medical examination concerned health-related behaviors. In a number of well tested questions the frequency and duration of cigarette smoking, alcohol consumption and physical exercise were assessed.
Biomedical measures
Independent blood pressure readings were taken by occupational physicians as part of the cardiovas- cular screening program using sphygmomanometry according to WHO criteria. In addition, body weight and height were assessed. Manifest hyperten- sion was defined according to WHO criteria (_>160mmHg systolic and/or >_95 mmHg diastolic blood pressure). Effects of antihypertensive medi- cation and of well established biobehavioral deter- minants of high blood pressure were controlled for in statistical analysis.
Blood samples were obtained under standardized conditions, i.e. after 30 minutes of rest in supine position at fixed diurnal time. A fasting time of 12 hours was not required for the current measurement approach (Cremer et al., 1991). Total serum choles- terol and triglycerides were determined enzymati- cally. Lipoproteins were measured according to an analytical schedule consisting of precipation tech- niques, quantitative lipoprotein electrophoresis and ultracentrifugation procedures as described else- where in more detail (Wieland and Seidel, 1983).
Statistical analyses
Univariate and bivariate analysis was performed to characterize middle managers in terms of sick- ness absence, biobehavioral and psychosocial vari- ables. Three types of absence-related information are used in the statistical analyses: (1) long spells, (2) short spells, and (3) number of absence episodes. Overlapping between these three sickness absence measures is possible. All analyses were performed comparing middle managers with short spells, long spells and at least two absence episodes to the remaining group without any absence days. With regard to blood pressure middle managers with manifest hypertension were compared to the remaining group. The chi-squared test was used to define the significance of the associations explored in bivariate analysis. In a second step, variables which discriminated between the respective sub- groups of middle managers entered multivariate
Chronic work stress and sickness absence 1115
logistic regression analysis in order to estimate the independent effect of each variable. The statistical significance of model parameters was tested by two- sided t-tests including main effects as well as two- way interaction terms. The model fit of the most parsimonious model was tested by the likelihood ratio difference test (Efron, 1975). To minimize the risk of erraneously accepting the zero hypothesis a level of significance <0.15 was defined (Andress, 1986). In all remaining statistical tests a level of sig- nificance of P_< 0.10 was accepted. This was done in view of the small sample size.
Our decision to dichotomize all Likert-scaled or continuous variables was based on results from a non-linear principal component analysis indicating statistical justification of a median split strategy (Girl, 1981). Moreover, the dichotomization of the Likert-scaled psychosocial variables measuring effort reward imbalance was shown to be sucessful in explaining cardiovascular risk and disease (Peter, 1991; Siegrist, 1996b; Siegrist and Peter, 1996). Finally, by this procedure we avoided zero cells in statistical analyses.
To test the model of effort reward imbalance at work we recommend two strategies of analysis, a "weaker" one and a "stronger" one. In the first strategy effects of single components on the criteria under study are explored. According to this strat- egy, the hypothesis is accepted if at least one indi- cator of high effort and at least one indicator of low reward is significantly associated with the out- come criterion of active coping, i.e. hypertension, after adjusting for confounding effects. In turn, at least one indicator of low reward, but no indicator of high effort, either intrinsic or extrinsic, is expected to be significantly associated with the pos- tulated outcome criterion of passive coping, i.e. sickness absence behavior, after adjusting for con- founding. The second strategy includes the compu- tation of a ratio of a sum score of items measuring extrinsic efforts (nominator) and of a sum score of items measuring rewards (denominator) as derived from factor analysis. Subsequently, single and com- bined effects of the ratio between extrinsic effort and reward and of the sum score of immersion (intrinsic effort) are analyzed in appropriate models of multivariate statistical analysis. While the appli- cation of the latter strategy is strongly rec- ommended in large-scale studies, small-scale data sets as the one reported here may be too vulnerable for this second strategy. Thus, the present data analysis is confined to the first approach mentioned.
RESULTS
In Table 1, means and/or frequencies of the three measures of sickness absence, short-term, long-term sickness absence and number of absence episodes are documented. In addition, the prevalence of im- portant biomedical, behavioral and psychosocial
Table 1. Measures of sickness absence and biobehavioral charac- teristics in middle managers (n = 146)
Variable Mean or frequency
Absence days 11.0 +_ 23,1 Long-term absence (mean days _+SD/%) 21.2 + 28.9/39.0% Short-term absence (mean days _+SD/%) 1.6 + 1.0/19.2% No. of absence episodes
l 40.4% 2 13.7% 3 4.8% 4 2.1%
Age (mean yrs _+SD/% > 50 yrs) 48.4 _+ 4.6/39.0% Self-rated health (rather poor, poor) 50.3% Lack of physical exercise (< l/week) 59.6% Recorded medical diagnosis (yes) 57.5% Regular alcohol consumption (yes) 34.9% Regular cigarette smoking (> 10 cig./day) 23.3% Sustained anger (often, very often) 18.3% Sustained hopelessness (often, very often) 5.9%
factors in the study population which may influence sickness absence behavior is shown. Almost 40% of middle managers experienced long-term sickness absence during the year of our screening, and every fifth middle manager had two or more sickness absence episodes. Long-term sickness absence was more Prevalent than short-term sickness absence (39% versus 19%), and this may be explained by a rather high percentage of employees with a diag- nosed chronic disease (57.5%).
As can be seen from Table 2, middle managers who stay away from work are significantly older than the remaining group, they suffer more often from diagnosed chronic disease, and significantly more of them rate their health as being poor. Moreover, they experience sustained anger more often than their healthier colleagues. With regard to the components of the model of effort reward imbalance at work, conditions of low reward only are more prevalent among middle managers who stay away from work, whereas no indicator of extrinsic or intrinsic effort is overrepresented in this group. Status incongruence, i.e. a mismatch between efforts and aspiration on the one hand and career achievements on the other hand, is associated with all three indicators of sickness absence, whereas forced job change is associated with two and status discrepancy (between current job held and edu- cational level) with one sickness absence measure.
In multivariate logistic regression analysis (see Table 5) five variables exert independent effects on short-term sickness absence: self-rated health, sus- tained anger and three indicators of low occu- pational rewards (status incongruence, status discrepancy and forced job change). Similarly, three variables exert independent effects on long-term sickness absence (self-reported health, sustained anger and status incongruence), whereas self- reported health and two indicators of low reward (status incongruence, forced job change) are associ- ated with the number of sickness absence episodes. The effects of age and prevalent chronic disease no
1116 Richard Peter and Johannes Siegrist
Table 2. Biobehavioral and psychosocial factors associated with sickness absence in middle managers
Short-term absence Long-term absence No. of absence episodes ~ 2
(Yes) No Yes No Yes No Variable ( n = 28) (n = 57) ~(2 ( n : 75) (n = 57) X 2 ( n = 30) ( n = 57) X-'
longer remain significant in multivariate analysis. As indicated in the Data and Method section, the most parsimonious model is analysed in a bottom- up procedure where the effect of each variable is adjusted for each other variable in the model.
Finally, Tables 3 and 4 show the respective bivariate and multivariate analyses for the groups
of hypertensive versus normotensive and borderline hypertensive middle managers. Manifest hyperten- sives are significantly older than the remaining group, their body weight is slightly higher (not sig- nificant) and they smoke less cigarettes (which may be due both to older age and diagnosed disease). Two indicators of high extrinsic effort at work
Table 3. Hypertension according to biobehavioral and psychosocial factors in 170 middle managers (frequencies and X-')
Variable Hypertensives Remaining group X 2 (_>160/95 mmHg) ( < 160/95 mmHg)
Time pressure 76.5% 59.7% 4.42* Frequent interruptions 62.7% 46.2% 3.90*
High intrinsic effort Immersion (upper tertile) 33.3% 40.0% 0.56
Low reward Lack of reciprocal support 40.0% 48.3% 0.96 Status incongruence (efforts and
achievements) 67.3% 62.2% 0.40 Status discrepancy (education and training) 53.2% 57.3% 0.23 Forced job change 20.0% 6.7% 6.52**
Effort-reward imbalance index Neither high effort ~ nor low reward b 30.0% 51.3% 8.14" High effort a OR low reward b 58.0% 44.5% High effort ~ AND low reward b 12.0% 4.2%
(time pressure and frequent interruptions) are sig- nificantly more prevalent among hypertensives, as well as one indicator of low occupational reward forced job change). Almost three times as many hypertensives as compared to the remaining group are characterized by a co-manifestation of high effort and low reward (12% versus 4.2%). This finding is substantiated by multivariate logistic re- gression analysis whose results are given in Table 4.
Both types of logistic regression models, the three analyses related to sickness absence (see Table 5) and the analysis related to hypertension (see Table 4) support the hypotheses of differential health outcomes according to the components of active versus passive coping within the flame of effort-reward imbalance at work.
DISCUSSION
This study shows preliminary evidence of differ- ential statistical prediction of work stress-related
health outcomes according to the criterion of active versus passive coping. Active coping was assumed to operate if both conditions postulated by the theoretical model of effort-reward imbalance at work were simultaneously present whereas passive coping was assumed to exist if low reward in the absence of high extrinsic or intrinsic effort was ex- perienced. Middle managers suffering simul- taneously from high effort and low reward (active coping) were at substantially elevated risk of exhi- biting manifest hypertension (multivariate odds ratio: 5.77). On the other hand, middle managers characterized by passive coping (low occupational rewards only) were more likely to experience sick- ness absence (multivariate odds ratios ranging from 2.67 to 9.15). Both findings are in accordance with stress-theoretical considerations on different health outcomes of active versus passive coping.
To our knowledge, this is the first report on differential statistical prediction of work stress-re- lated health outcomes based on a theoretical model.
Table 5. Logistic regression analyses: psychosocial factors associated with sickness absence in middle managers
Variable Multivariate OR 95°/,, CI Outcome (sample)
Model 1 Self-reported health (rather poor/poor) 8.23** Sustained anger (often. very often) 4.90* Low reward:
Status incongruence (efforts and achievements) 3.30(*) Status discrepancy (education and training) 6.00** Forced job change 9.15**
Model 11 Self-rated health (rather poor/poor) 2,49" Sustained anger (often. very often) 2.44*
Low reward: Status incongruence (efforts and achievements) 2.67*
2.19-30.99 1.30-18.47
0.86 12.63 1.65 21.82 1.96 42.68
1.18-5.27 1.14 5.22
1.24 5.78
Model IlI Self-reported health (rather poor/poor) 14.40"** 3.76-.55.40
Low reward: Status incongruence (efforts and achievements) 4.05* 1.25 13.13 Forced job change 6.33"* 1.81-22.07
Short-term absence (n = 85)
Long-term absence (n = 132)
No. of absence episodes >.>_ 2 (n = 86)
***P:~ 0.001; * * P < 0.01: *Zg P = 0.05; (*)P_< 0.10.
1118 Richard Peter and Johannes Siegrist
Evidence on associations between chronic work stress in terms of active coping and cardiovascular risk (especially manifest hypertension) was demon- strated by Schnall et al. (1990) using the demand control model of job strain. Workers scoring high on task demands and low on task control showed a relative risk of 3.09 of being hypertensive if com- pared to low strain workers. However, the compar- ability of the two studies is somewhat limited due to different definition criteria of manifest hyperten- sion. In two large-scale studies the model of job strain was used in predicting sickness absence (Kristensen, 1991; North et al., 1993, 1996). Interestingly, in both studies, the control dimension was more important than the demand component in predicting sickness absence. In the British study, the demand component produced some significant effects in the opposite direction (North et al., 1996). Again, these findings suggest that workers involved in highly demanding job conditions are not likely to exhibit passive coping behavior in terms of sickness absence. An additional body of empirical evidence indicates that distinct work-related and personal behaviors associated with sickness absence are inter- preted as an expression of passive coping or with- drawal behavior, thus giving indirect support to the conceptual approach suggested here (Dimberg et
al., 1989; Erlam, 1982; Lazarus, 1989; Malmgren and Andersson, 1984; Melamed et al., 1992). A clo- ser inspection of the findings reported in Tables 2 and 5 reveals that one indicator of low occupational reward exerts the most consistent effects on sickness absence: status incongruence. This indicator measures the awareness of a particularly frustrating job situation where high efforts are not compen- sated by any progress in occupational achievements. It is likely that this experience may increase the pro- pensity towards withdrawal and "giving up" beha- vior.
On the other hand, reported findings suffer from several restrictions. Although the sample under study is very homogenous in terms of age, sex, oc- cupation and socioeconomic status and although the reduced sample was not biased to any relevant degree results are restricted to a rather small obser- vational basis. Large confidence intervals in logistic regression analysis and the restriction of statistical model testing to a "weaker" strategy may be inter- preted as signs of limited robustness. However, a clear risk of falsifying the hypothesis was given, and additional support in favor of predictive validity of the concept of effort-reward imbalance with respect to cardiovascular risk was found in this sample: middle managers suffering from frequent interrup- tions in combination with lack of reciprocal support had significantly higher levels of atherogenic lipids (F = 4.3; P < 0.05). Moreover, the relative risk of suffering from a co-manifestation of high athero- genic lipids and of hypertension was consistently el-
evated under conditions of high effort and low reward (Siegrist and Peter, 1996).
It may be argued that the association between forced job change and sickness absence contradicts the evidence of an inverse relation between job instability and sickness absence. Yet, middle man- agers in this large-scale company experienced a safe employment status. Forced job change therefore was not directly related to job instability, but rather to an unfavorable change in task responsibilities which was out of control. Significant effects of negative mood such as sustained anger on sickness absence were observed in two out of three models of regression analysis. This finding is in accordance with the notion of a frustrating, anger-inducing job experience to which employees increasingly respond with passive coping. Clearly, we lack definitive in- formation on this latter argument as we did not explicitly measure the link between low reward and passive coping attitudes.
Furthermore, comparing the effects of short-term versus long-term spells, our findings are in accord- ance with those reported by North et al. (1996) and Marmot et al. (1995) indicating that psychosocial influences are more pronounced in short-term spells, even after adjusting for the effects of self-reported health, health-related behaviors, and age. However, in our data set, 14 middle managers were included into both analyses as they exhibited short-term and long-term spells simultaneously. When recalculating the logistic regression analyses without this sub- group findings remained more or less the same for long-term spells, but were less powerful for short- term sickness absence, due to the small remaining sample size. Furthermore, we also checked the poss- ible effect of co-manifestation of hypertension and sickness absence on the reported findings. When introducing hypertensive status as a confounding variable into the logistic regression analysis with sickness absence data as outcome criterion we did not find any substantial modification. In bivariate analysis no significant association was found between sickness absence and hypertensive status.
Finally, it must be stated again that this study is based on a cross-sectional design using one measurement point only. This limited design pre- cludes any test of causal relations. Yet, it is unlikely that in this sociodemographically and occupation- ally homogeneous group social selection factors op- erate to any significant degree. Moreover, reported findings are restricted to a middle-aged male white collar group in the car production industry of a country with advanced economic development. Obviously, it is not possible to extrapolate these findings to other occupational groups and employ- ment sectors.
Despite these limitations, this study gives prelimi- nary evidence on the potential benefit of a socio- logical model of chronic work stress which allows for differential prediction of health outcomes
Chronic work stress and sickness absence 1119
according to the core not ions of active versus pas- sive coping with the demands at work.
Acknowledgements--We thank Dr. Joachim Stork phys- ician at the Occupational Health Department of the inves- tigated study site for his helpful cooperation.
REFERENCES
Andress, H. J. (1986) GLIM-Verallgemeinerte Lineare Modelle (Generalized Linear Modeling). Vieweg, Braunschweig Wiesbaden.
Axelrod, J. and Reisine, T. D. (1984) Stress hormones. Science 224, 452-459.
Ballieux, R. E., Cullen, J., Fielding, J. F., L'Abbate, A., Siegrist, J. and Wegman, H. M. (eds) (1984) Breakdown in Human Adaptation to Stress." Towards a Multidisciplinary Approach. Martinus Nijhoff, Boston.
Beamish, R. E., Singal, P. K. and Dhalla, N. S. (eds) (1985) Stress and Heart Disease. Martinus Nijhoff, Boston.
Carver, C. S,, Scheier, M. F. and Weintraub, J. K. (1989) Assessing coping startegies: a theoretically based approach. Journal of Personality and Social Psychology 56, 267-283.
Cooper, C. L. and Payne, R. (eds) (1991) Personality and Stress." Individual Differences in the Stress Process. Wiley, Chichester.
Cremer, P., Nagel, D., Labrot, B., Muche, R., Elster, H., Mann, H. and Seidel, D. (1991) Gfttinger Risiko-, blzidenz- und Prdvalenzstudie (GRIPS). Springer, Berlin, Heidelberg, New York, London, Paris, Tokyo.
Dimberg, L. et al. (1989) Sickness absenteeism in an engeneering industry--an analysis with special refer- ence to absence for neck and upper extremity symp- toms. Scandinavian Journal of Social Medicine 17, 77 84.
Efron, B. (1975) The efficiency of logistic regression com- pared to normal discriminant analysis. Journal of the American Statistical Association 70, 892-898.
Erlam, A. R. (1982) Sickness absence in drivers of London taxis and buses. Journal of Occupational Medicine 32, 20 25.
French, J. R., Caplan, R. D. and Harrison, R. V. (1982) The Mechanisms o[" Job Stress and Strain. Wiley, Chichester.
Girl, A. ( 1981 ) Nonlinear Multivariate Analysis. Department of Data Theory, Leiden.
Henry, J. P. and Stephens, P. M. (1977) Stress, Health, and the Social Environment. Springer, Berlin.
House, J. S. (1981) Work Stress and Social Support. Addison Wesley, Reading, MA.
Johnsson, J. V. and Johansson, G. (eds) (1991) The Psychosoeial Work Environment and Health." Work Organisation, Democratization. and Health. Baywood, Amityville, NY.
Kahn, R. A. (1981) Work and Health. Wiley, New York. Karasek, R. A. (1979) Job demands, job decision latitude,
and mental strain: implications for job redesign. Administration Seience Quater O' 24, 285-307.
Karasek, R. A. and Theorell, T. (1990) Healthy Work: Stress, Productivity, and the Reconstruction o]' Working L([e. Basic Books, New York.
Kiecolt-Glaser. J. K. and Glaser, R. (1995) Psychoneuroimmunology and health consequences: date and shared mechanisms. Psyehosomatie Medicine 57, 269-274.
Kristensen, T. S. (1991) Sickness absence and work strain among Danish slaughterhouse workers: an analysis of absence from work regarded as coping behaviour. Social Science & Medicine 32, 15-27.
La Ferla, F. and Levi, L. (eds) (1993) A Healthier Work Environment. World Health Organization, Copenhagen.
Lazarus, R. S. (1989) Psychological stress at the work place. Journal of Occupational Environmental Health 1 I, 528-540.
Lazarus, R. S. (1991) Emotion and Adaptation. Oxford University Press, New York.
Lazarus, R. S. and Folkman, S. (1984) Stress, Appraisal. and Coping. Springer, New York.
Levine, S. P. (1987) Psychobiologic consequences of dis- ruption in mother-infant relationship. In Perinatal Development. A Psychobiological Perspective, eds N. A. Krasnegor, E. M. Blass, M. A. Hofer and W. P. Smotherman, pp. 359-376. Academic Press. Orlando.
Malmgren, S and Andersson, G. (1984) Corporate reported sick leave and its relationship with education, responsibility and blood pressure. Scandinavian Journal of Social Medicine 12, 171-176.
Manuck, S. B., Marsland, A. L., Kaplan, J. R. and Williams, J. K. (1995) The pathogenicity of behavior and ist neuroendocrine mediation: an example from cor- onary artery disease. Psychosomatic Medicine 57, 275 283.
Marmot, M. G., North, F., Feeney, A. and Head, J. (1993) Alcohol consumption and sickness absence: from the Whitehall II Study. Addiction 88, 369 382.
Marmot, M. G., Feeney, A., Shipley, M., North, F. and Syme, S. L. (1995) Sickness absence as a measure of health status and functioning: from the UK Whitehall I1 study. Journal o[" Epidemiology and Community Health 49, 124-130.
Mason, J. W. (1968) Organisation of psychoendocrine mechanisms. Psychosomatic Medicine 30, 565 588.
Melamed, S., Luz, J. and Green, M. S. (1992) Noise ex- posure, noise annoyance and their relation to psycho- logical distress, accident and sickness absence among blue-collar workers--the Cordis Study. Israel Jola'nal ~)l' Medical Science 28. 629--635.
North, F., Syme, S. L., Feeney, A., Head, J., Shipley, M. J. and Marmot, M. G. (1993) Explaining socioeconomic differences in sickness absence: the Whitehall II study. British Medical Journal 306, 361 366.
North, F., Syme, S. L., Feeny, A., Shipley, M. and Marmot, M. (1996) Psychosocial work environment and sickness absence among British civil servants: the Whitehall I1 Study. American Journal of Public fh'alth 86, 332-340.
Peter, R. ( 1991 ) Berufliehe Belastungen. Belastungsbewdltigung und koronares Risiko hei bldustriearheitern. LIT. Hamburg M~inster.
Sauter, S. C., Hurrell, J. J. and Cooper, C. L. (eds) (1989) Job Control and Worker Health. Wiley, Chichester.
Schnall, P. L., Piper, C., Schwartz, J. E., Karasek, R. A., Schlussel, Y., Devereux, R. P. and Pickering, T. G. (1990) The relationship between "'job strain", work- place diastolic blood pressure and left ventricular mass index--results of a case-control study. Journal Of the Ameriean Medical Association 14, 1929-1935.
Schnall, P. L., Landsbergis, P. A. and Baker, D. (1994) Job strain and cardiovascular disease. Annual Review O[ Publie Health 15, 381-411.
Seligman, M. E. P. (1975) Helplessness: On Depression. Developnwnt and Death. Freeman, San Fransisco.
Siegrist, J. (1996a) Adverse health effects of high-effort/ low-reward conditions. Journal q!" Oceapational Health P*Tchoh~gy 1, 27-41.
Siegrist, J. (1996b) So-iale Krisen und Gesundheit. Hogrefe, G6ttingen.
1120 Richard Peter and Johannes Siegrist
Siegrist, J. and Peter, R. (1996) Threat to occupational status control and cardiovascular risk. Israel Journal of Medical Science 32, 179-184.
Siegrist, J., Peter, R., Junge, A., Cremer, P. and Seidel, D. (1990) Low status control, high effort at work and ischemic heart disease: prospective evidence from blue-col- lar men. Social Science & Medicine 31, 1127-1134.
Theorell, T. and Karasek, R. (1996) Current methodologi- cal issues relating to psychosocial job strain and cardio-
vascular disease research. Journal of Occupational Health Psychology 1, 9-26.
Weiner, H. (1992) Perturbing the Organism: The Biology oJ Stressful Experience. University of Chicago Press, Chicago.
Wieland, H. and Seidel, D. (1983) Quantitative lipop- trotein electrophoresis. In Handbook of Electrophoresis III, ed. L. A. Lewis, pp. 83f. CRC Press, Boca Raton.
