Psychosocial stressors at work and musculoskeletal problems

OBJECTIVES - This paper examines the relationship between work stressors and the following health indicators: psychosomatic complaints, health behavior, and musculoskeletal problems. METHODS - Secondary analyses were performed on data from the National Work and Living Condi tion Survey, which provides a representative sample of the working population in The Netherlands. The survey was made in 1977, 1983, and 1986 by The Netherlands Central Bureau of Statistics. By means of factor analysis the following three risk dimensions were identified in the survey: work pace, intellectual discretion, and physical stressors. RESULTS - High work pace, low intellectual discretion, and physical stressors were associated with increased health complaints (both psychosomatic and musculoskeletal) and musculoskeletal disorders after adjustment for gender, age, education, and sports participation. Low intellectual discretion, but not high work pace, was associated with poor general health and health behavior indicative of poor health. Physical stressors were associated with general health as well, but not with health behavior, except for reported absenteeism. CONCLUSIONS - Psychosocial stressors are not only associated with psychosomatic complaints and health indicators, but also with musculoskeletal problems, both acute and chronic. Especially the re lation between intellectual discretion and musculoskeletal problems can be partly attributed to physi cal load. Even after adjustment for physical stressors and moderating personal characteristics, the re lationships between the psychosocial stressors and musculoskeletal problems remained significant and comparable in strength to the relationship between psychosocial stressors and several other health out comes, such as psychosomatic complaints.

problems, especially back pain, are considered to be important problem s assoc iated with poor well-being, the use of medical c are , and re striction of acti vities or ab senteeism at work in oth er countrie s as well ( 13). Although the relationship between psychosocial stressor s and psychological (dys )func tio ning ha s been studied exten siv ely and mod els of stress and indicators of psychological health have been construc ted (I), re search on the rel ationship between psychosocial stressors and musculoskeletal problems ha s been ver y lim ited. Positive evidence of a relationship bet ween psychosocial stressors at work and musculoskeleta l problems wi ll, ho we ver, provide interes ting options for p reventive actions aimed at saving co st s through a reduction in complaints, absenteeism , and, perhaps, ev en disab ility du e not onl y to psychologic al dysfunctioning but al so to mu sculoskeletal problems.
A recent review shows that studies on the relationship bet ween psychosocial stressors and mu sculoskeletal problems are het erogeneou s, both with respect to the inde pende nt variable and the outcome variable (II). No clear picture exi sts as to whi ch psychosocial stressors are the most strongly rel ated to mu sculoskelet al problems, if th is rel ationsh ip is a ca usal one, as to ho w speci fic the relationship is with respect to the type of musculoskeletal problem (back pain or other musculoskeletal problems, chronic or acute effects) and if nonwork activities like sports confound the relationship. Several mechanisms have been proposed. Psychological and physiological explanations have been put forward that suggest that psychosocial loads might either change the perception of pain (14) or induce physiological changes that might result in musculoskeletal problems (11). A methodological explanation might, however, apply as well (II). Although there might be a direct relationship between psychosocial stressors (eg, work pace and monotony) and musculoskeletal problems, the relationship might well be confounded by physical stressors such as static load or repetitive work.
In the present study, an analysis was performed on a representative sample of a cross-section of the Dutch working population. The study was descriptive in nature and aimed at determining the relationship between psychosocial stressors at work and sev- eral indicators of health, including psychosomatic complaints, health behavior, and acute and more chronic musculoskeletal problems. The study attempted to examine both the specificity of the relationship between psychosocial stressors at work and health outcomes and to test the possible confounding effect of physical load on the relationship between psychosocial stressors and musculoskeletal problems, while adjusting for several confounding personal characteristics.

Subjects and methods
The National Work and Living Condition Survey was administered by the Central Bureau of Statistics (CBS) to a representative sample of the working population in The Netherlands once every three years in 1977-1986 (15). The total sample consisted of 5865 employees aged 18 to 65 years. The representativeness of each sample has been corroborated in a comparison of the distribution of employees across occupations and company types with that of another CBS survey, directed at estimating the magnitude of the Dutch labor force (16). In an attempt to improve the reliability of the results, the different samples across time were pooled. The pooling was considered to be justified since the interrelations of the different variables were found to be stable across the years (17,18). Table 1 shows the variables used in the study, including psychosocial and physical stressors at work (independent variables), health outcomes (dependent variables), and personal characteristics (confounding variables). All questions on stressors at work in the survey were used in the analyses for this study, except the question on shift work. With respect to the personal characteristics, gender, age, and education are known to be important potential moderators in the relationship between psychosocial and physical stressors at work and health outcomes. These three characteristics were used, including sports participation as well. Sports participation was considered especially interesting since it has not only been found to be beneficial for well-being (19), but it might be positively related to musculoskeletal problems, especially of the limbs.
The relationships between stressors and health indicators were studied with multiple logistic regression analyses. Since the mutual independence of the stressors at work were not empirically established, a factor analysis (Varimax rotation) was first performed on these stressors to identify possible underlying stressor dimensions. It was expected that a factor analysis would identify dimensions which would have more theoretical relevance than the individual stressors. Although the regressions of the dimensions to be extracted from the health indicators might be interesting from a theoretical perspective, the dimensions to be extracted would probably explain only Table 2. Mean percentages of confirmatory answers to questions about the presence of individual risk factors for work stress and health, across 1977, 1983, and 1986. Only the psychosomati c com plaint score is a summary sco re.
part of the variance explained by the individu al stressors in the survey. The refore , the regressi on analyses were performed on both indi vidu al stressors and underl ying dimen sions of stress risk.
Mult iple logistic regression analyses performed with the individu al stressors (all dichotomous variables) as independent variables result in an odds ratio for each stresso r, indicatin g the relati ve chance that the health indic ator is present when the stressor is present, when the other independent and confounding variables are adjusted for in the regress ion model. When these logistic regressions are applied to the dim ension s extrac ted from the factor analysis, the independent parameter is no longer dichotomous but has a scoring range from zero to the number of stressors that constitute the dimen sion. The odd s ratio of a stressor dimension indicate s the chance that a particular health ind icator is present when one, two, three, or more (up to the maximum number of stressors in the dimen sion) stressor s of the dimension are present relative to the situation in which all stressors of that dimension are absent.

Results
Th e factor analysis of the stressors resulted in the following three independent factors, explaining 45% of the variance of the individual stressors : (i) work pace (Eigenvalue = 1.02), (ii) intellec tual discreti on (co nsisti ng of monotonous work, poor possibilities for personal development , poor fit between the actual work and education or experience level, and poor promotional prospects) (Eigenvalue = 1.62, Cronbach a.=0.55, factor loading s ranging from 0.56 to 0.77, indicating an acceptable homogeneity and rather equivalent contribution of the individual stressors), and (iii) physical stressors (heavy physical work, bad smell at work , dirty work , noise at work, dangerous work) (Eigenvalue = 2.27, Cronbach a. = 0.63, factor loadings ranging from 0.59 to 0.71 , indicating an acceptable homogeneity and rather equi valent contribut ion of the ind ividual stressors ). Tabl e 2 shows the percentage of employees said to have been co nfro nted with a specific stressor, as well as the prevalence of the health indicators. A summary score was calculated only for the psychosomatic complaints. The stress dim ension s are also shown in th is table.
The odds ratios and confidence interv als for the individual stressors for each of the health indicators, estimated on the basis of the mult iple logistic regression analyse s, are presented in table 3. In table 4, the estimated odds ratios and co nfide nce intervals are presented for the three risk dimensions which resulted from the factor analysis . Tables 3 and 4 indicate that a high work pace was positivel y rela ted to psychos omatic complaints, as well as to complaints of back pain and muscle and joint co mplaints. A high work pace was, however, unrelated to a feeling of health in general and to the health behavior factors.
Poor intellectual discretion, especially monotonou s work, was related to all of the indicators of health , including all of the indica tors of musculoskeletal problems. Table 4 shows that the health probl em s steadily increase d with an increasi ng number of indicato rs for poor intellec tual discret ion. Physical stressors, especiall y heavy physical work, were positively related to most of the health indicators , especiall y back problem s. The odds ratios for the back problems were relatively high, mainly due to the explanatory power of physically heavy work.
The estimated odds ratios for the personal characteristics are show n in table 3 but, for efficiency, not in table 4. The odds ratios for the individual characteristics were found to be about the same when they were entered into the model either with the individual stressors (ta ble 3) or with the risk dimensions (table 4) .
The women reported more health complai nts and more doctor' s visit s and use of medication than the men, as did more of the older employees than the younger ones. Older employees , on the other hand , less frequently reported having stayed home ill during the last three month s, Sport s particip ation was positivel y associated with a feeling of good health in general but not with other indicators of health.
When physical stressors and personal characteristics were entered stepwise into the logistic model, a stepwise reduction in the estim ated odds ratios of the intellectu al discretion dimension was observed (table 5). As has been shown in tabl es 3 and 4, in which the results fro m the most extensive regression  23 1.00-1.52 1.14 0.99-1 .32 1.05 0.91-1 .21 1. 04 0.90-1.18 1.12 0.97-1.29 1.05 0.88-1.23 1.04 0.89-1 .22 1.23 Table 4. Estimated odds ratios (OR) of the psychosocial risk dimens ions for the different health indicators. The odds rat ios indicate the health risk associated with the level of the independent variabl e as opposed to the situation in which th is independent variable was absent , all other variable s being adjusted for in th e log is ti c model (inc luding personal characteristics; the risks assoc iated with these charact eris tic s are not show n sinc e they do not differ from those in models were shown, the health effects of (poor) intellectual discretion did, however, remain significant. The estimated odds ratios of the physical stressors did not diminish but, instead, increased when the personal characteri stics were entered into the model.

Discussion
One primary conclusion of this study is that a high work pace and poor intellectual discretion, two risk dimensions for work stress which show resemblance to the two main risk dimen sions for work stress in Table 5. Changes in the estimated odds ratios due to the addition of physical stressors and personal characterist ics to the model. (model 1 =all three risk dimensions and the personal characteristics in the model , model 2 =only work pace, intellectual discretion and physical stressors in the model , model 3 =only work pace and intellect ual discretion in the model)  1  2  3  1  2  3  1  2  3  1  2  3  1  2  3  1  2  3  1  2  the Karasek model of "j ob dem and s-deci sion latitude" ( I, 20), were associated with several indic ators of health in a cross-section of the Dutch working populati on. The se relationships remained significant even when physical stressors like heavy work load and moderating personal characteristics were adjusted for. A high work pace was an important aspect of job demands, whereas intell ectual (or skill) discretion was an aspect of decision latitude. A high work pace was found to be associated with psychosomatic complaints and musculoskelet al problems, including complaints of back pain , jo int and muscle problems, and the more chronic back problems. Poor intellectual discretion, especially monotony on the job, was related to a feel ing of poor health in general and to seve ral indic ators of (ill-) health beha vior as well.
The strength of the associations found for both the psychosocia l stressors and musculoskeletal problem s was co mparable with the strength of those found between psycho social stress ors and psychosomatic complaints. This finding sugges ts that the relationship betwe en the stressors and the health outcomes was either rather nonspeci fic or might have been the result of a strong interrel ation ship between the indicators of health in the present study. Thi s last suggestion does not, howe ver, apply since the correlation between the health indicators was rath er low; it ranged from 0.17 to 0.47, the highest correlati on being found for the subject's opinion of his or her health in general and psychos omatic co mplaints.
Other meth odol ogical problems, like a lack of reliability of the measurem ent instruments or fluctu ations in time, might have influenced the associations between the psychosocial stressors and health ind icators , as well as their (non)specificity. Apart from the fact that the interrelations between the variables in the study remained stable across the years, logistic regress ions were also performed on the data of each year in which the survey was admini stered. AI-though, for some subgroups (eg, employees indicating five physical stressors and having chron ic back problems in 1983), the number of cases was extremely small, the results of these analyse s indicated that the relationsh ips found for each year were in fairly good agreement with the results of the overall analyses presented in thi s articl e.
A criticism that might apply to all of the findings of our study is the fact that the operationalization of both the independent and dependent variables was by way of a self-report instrument. This limitation may have res ulted in infl ated associations due to "common measure variance." Th is explan ation is, however, not very likely since studies in which relations are reported between survey-based independent variables and more objective indicators of health , like morbidity and mortality figure s, show odds ratios or relative risks of comparable magnitude as, or even larger than, those of our study. (See, eg, references 4 and 7.) A recent study in which psych osocial stressors, operati onali zed by means of a que stionnaire , were found to be associ ated with the results of musculoskeletal function testing by medic al doctors has reported risks of comparable magnitude as well (2 I). The fact that report ed stress and health problem s in themsel ves do have significant prognostic value is shown in longitudinal studies in which these variables were found to predi ct mortality (22,23).
Our findings have theoretical significance. The risk dimensions, found to be associated with the health indicator, did not, however, cover the "j ob demands-decision latitude" model completel y, and ge neralization of all findin gs to the enti re model is therefore limited. The other aspect of deci sion latitude, autonomy, was absent in the National Work and Living Condition Survey, as well as another stressor which is considere d to be an important risk factor for work stress: social support (24)(25)(26). In the "jo b demands-deci sion latitude" model, the risk di-mensions are assumed to interact and result in a combined health outc ome which cannot be solely predicted from a summing of the effects of the two main risk dimensions. A multiplicative interaction of work pace and intellectual discretion was not, however, found to increase the variance in health outcomes in our study. In the literature on the "jo b demands-decision latitud e" model, the interactive effect of the risk dimensions is also under debate. (See, eg, references 27 and 28 for a discussio n.) One of the criticis ms is that the co mbined effect of the two risk dimensio ns for work stress is not best described as multiplicative but as addi tive , whereas in the latter case the relations may very well be curvi linear. Curvilinearity cannot, however, pro perly be tested for dichotomous health indicators. The absence of a significant contribution of the multiplicative interaction term of work pace and intellec tual discretion does, however , fail to support the assum ption of interaction between two main components of the "job demands-dec ision latitude" model whe n added to the logistic mode l.
Although the rela tionship of work pace with intellectual discretion remained significant after adjustment for physical stressors , adding physical stresso rs to the logistic model resulted in a reduction of the magni tude of the esti mated odds ratio s of (poor) intellect ual discretion. A similar findi ng was reported by Theorell et al (29) . Th is finding may be due to the fact that poor intellect ual discretion, especially of monotono us work, is often associated with work which is either short-cycled or involves a high static (post ural) load. This hypothesis should, howeve r, be teste d in a study in which the physical load is more elaborate ly operationalized .
Not only physical stressors but also personal characteristics reduced the odds ratios for (poor) intellectual discretion in the regression models. Thi s reduction was probabl y due to the fact that women and older employees, and employees with little education, more often have work with poor intellectual discretion. The increase in the odds ratios for physical load when these characteristics are entere d next into the model is, on the other hand, pro bab ly due to the fact that men and younger employees more often perform heavy work in a work enviro nme nt which is physica lly loading ( 16,17). Even after adjustment for the moderating effec ts of gender, age , and education level, the odds ratios of both the indiv idual stresso rs, especially monotony , and the main risk dime nsions for work stress generally remained significant.
Even afte r adjustment for physica l load and personal characteristics, the odds ratios estimated for (poor) intellectual discretion were consistently found to be higher than those found for work pace. This finding sugges ts that the impac t of inte llectual discretion was larger than that of jo b dema nds. Two alterna tive ex planatio ns can be put forward . The first one is purely methodologica l in nature and concerns the number of questions asked to measure the con-144 cepts. The "demands" dimension was operationalized with one question only, whereas the "intellectual discretion" dime nsion was operat ionalized more exte nsively. The more questions aske d, the more reliable the score and the more variance to explain (anot her concept) or to be explained. A seco nd exp lanation can be found in recent findings of Theorell et al (30), which indica te that the perception of pain might be different under situatio ns that differ in job demands and decision latitude. These researchers found that high demands are assoc iated with an increase in pain threshold. Sympathetic activation might be the underlyi ng physio logical mechanism. An increase in pain threshold may result in the underreporting of health symptoms. Low intellectual disc retion, on the other hand, was found to be associated with a lowering of the pain threshold, which, instead, might result in an increas ed tendency to report symptoms.
Before the conclusions of the present study are summarize d, it must be kept in mind that this study was cross-sectional in nature. All significant relations do not necessarily indicate a causal relationship. As stated in the introduc tion, longitudinal studies which provide support for causal proce sses linking psychosoc ial stressors to musculoskeletal problems are, to our knowledge, unavailable. In addition, the streng th of the relationships found in the present crosssectio nal study may, on one ha nd, be infla ted by methodo logica l problems such as recall bias . On the other hand, however, the strength of the relationships found might be reduced because of selection proce sses both into and out of professions, resulting in a selection of "fit" employee s into specific jobs and a loss of those emp loyees who leave jobs high in stress risk beca use of health prob lems they have deve loped over time.
To summarize, it ca n be stated that psyc hosocial stres sors are not only assoc iated with psychosomatic comp laints and hea lth indica tors, but with musculoskeletal probl ems as well. Especially the relation between intelle ctual discretion and musculoskeletal probl ems can be partly attributed to physical load, although the exact nature of this confounding effect is not clear . Even after adjustment for physical stressors and moderating personal charac teristic s, the relationships between the psychosocial stressors and musc uloskeletal prob lems remained sig nifican t and comparable in strength to the relatio nships betwee n psycho socia l stressors and several other healt h outcome s, like psychoso matic complai nts and seve ral health behavior s.