Influence of personal characteristics, job-related factors and psychosocial factors on the sick building syndrome. Danish Indoor Climate Study Group.

The influence of personal characteristics, life-style, job-related factors, and psychosocial work factors on symptoms of the sick building syndrome was investigated in Greater Copenhagen, Denmark. The buildings were not characterized beforehand as "sick" or "healthy." Of the 4369 employees sent a questionnaire, 3507 returned them. Multivariate logistic regression analyses of the multifactorial effects on the prevalence of work-related mucosal irritation and work-related general symptoms among the office workers showed that sex, job category, work functions (handling of carbonless paper, photocopying, work at video display terminals), psychosocial factors of work (dissatisfaction with superiors or colleagues and quantity of work inhibiting job satisfaction) were associated with work-related mucosal irritation and work-related general symptoms, but these factors could not account for the differences between the buildings as to the prevalence of the symptoms. The building factor (i.e., the indoor climate) was strongly associated with the prevalence of the symptoms.

A n increasing number of co mplai nts a bo ut discomfort and health effects in relat ion to time spent in residenc es and nonindu strial workplaces ha s been reported during the last 10 to 15 yea rs. In 1983 the World Health Organization defined the concept of the sick building synd rome (l) as being characterized by an increased frequenc y o f a number of irritative sym pto ms of the eyes, no se, thro at , and low er airways; skin reac tio ns; non specifi c hypersensitivity; mental fatigue; headache; and nausea and dizziness among persons staying in the building in question . Typically, the sym ptom s grow wo rse during the workday in the building and disappear or diminish after the workers leave th e building. They are thus experienced as work-related. The etiolog y o f this synd rome is cur rently not fully understood .
Various factors, some ch emi cal (2), 'physica l (3) , biological (4), and psychoso cial (5), have been related to th e synd ro me . The relative importan ce of these fa ctors for th e sick building synd ro me has been inve stigated very little thus far. In addition the epidemiolog y of the sick building syndrome has onl y been an alyzed in a few studies (6)(7)(8)(9)(10) . Complaints about wo rkrelated irri tation of the mu cous membranes and workrelat ed general symptoms, such as headache, fa tigue, and malais e, have been found to be prevalent , and they differed significantly from one building to another witho ut bein g explai na ble by the results of a determina tion o f several co m po nents of indoo r pollution (7)(8)(9) . Ho wever no analysis has been made of the influ ence of potential confounders such as wo r k fun ctions, psychosocial factors, and life-style.
The purpose of our Danish town hall study was to examine the influen ce o f va rio us indoor climate factors on th e prevalen ce of the sym pto ms of the sick building synd ro me. The study included a large populati on, many different buildings, and measurements of many indoor climate factors. We have previously reported that sex, job category, photocopyin g, wo rk at video display terminals, a nd the handling of carbonless paper were significa ntly associated with workrelat ed mucosal irritation and wo rk-rela ted gen eral symptoms, such as headache, fatigue, and malaise (II). In thi s paper we have further analyzed the se associations, together with the po ssibility of additio nal asso ciat ions with certain personal characteristics and ps ychosocial factors, and have co nsidered wh eth er these factors can acco unt for th e differences between the pr evalence of symptoms among th e office workers in the buildings studied. This anal ysis deal s only with the office wor kers. The multi variate analysis comprised 2829 office wo rkers from the 19 buildings.

Questionnaire study
The questionna ire included que stions on the work , its type, pre vious and present diseases, presence of symptoms from the mucous membranes , presence of general symptoms , and the frequ ency and time variat ions of such symptoms. Qu estions about var iou s other symptom s, compl aints about the indoor climate (draft, low temperature , etc), family and housing conditions, exercise habits, and consumption of tobacco, alcohol and other beverages were also included.
The po ssible categories of an swers to th e que stions about indi vidual symptoms were: " no, " " yes, some times a year," "yes, some times a month ," "yes, some times a week," and "yes, daily. " We also asked about variation in the degree of symptoms , for which we employed the following categories of answers: "yes, it improves at work ," " yes, it improves on da ys off/during weekend s or vacations ," " no , it is alwa ys the same ," and " don't know." If a pa rticular symptom occurred some time s a week or more and improved on da ys off, the sympto m was classified as work-related . Besides single symptoms, we employed two gro ups of work-related symptoms, ie, work-related irritation of the mucous membranes (at least one of the workrelat ed symptoms of irritation or dr yness of the mucous membrane of th e eye, no se, or throat) and work-related genera l symptoms (at least one of the Women % N %

Materials and methods
The buildings Participation in the study was requ ested of the municipalit ies of Copenhagen, Frederiksberg, and the County o f Co penhagen, the tot al being 20 municipaliti es. Thirteen of them entered the study, three were excluded because comprehensive indoor climat e studies had been recentl y carr ied out or were in progress, one an swered too lat e for it to be included , and three did not want to participate. The to wn hall in each mu nicip ality was selected as the study building s. In one municipality two buildings (number 120 and 121) were included . Three of the study buildings had minor building con struction differences (eg, type of ventilation) and were therefore divided into two parts on the ba sis of these differences in the multi variate analysis. Am ong the 14 buildings affilia ted with the town halls , two had a population of office workers (168 and 110 for building s number 61 and 13I, respectively) of a size suitable for them to be incor porated into the multivariat e analysis, giving a to ta l of 19 buildings in the analysis. The age of the bu ildin gs ran ged from I to 80 (mean 18.8) years. Six buildings were situated in rural are as, six in urban communities, and seven in residential areas. Seven build ings had natural vent ilation , and 12 had mechanical ventilation. Of these 12, eight had air intake and four only air exhaust. Of the eight buildings with air intake , two systems provided recirculation and two had humidifiers, of which , howe ver , onl y on e was in use. The floor area ran ged from 1250 to 12 900 (mean 4418) m' , and the number of employees in the buildings ranged from 84 to 419 (mean 205). Th e measurements of the indoor climate ha ve already been described previously (11). The many indoor climate factors determined resulted in values mainly at the level normally con sidered acceptable or in values in accor dance with levels previou sly reported .

Statistical methods
The significance of single factors in the total population of office workers was evaluated by the chi-square test.
The multifactorial effects on the prevalences were analyzed in a logistic regression model according to Breslow & Day (12), in which the probability of response is related to a set of regression variables x=(x!, ... , x p ) via the following equation: In our case, the probability for each individual to develop the symptom was expressed as a function of the characteristics of that same person. The betas are parameters to be estimated. These parameters were interpreted via the odds ratio. For individuals with characteristics x and x*, respectively, the odds ratio was The 95 010 confidence interval of the odds ratios was estimated from the corresponding confidence interval of~[~± 1.96 SD (~)].
Because of the very large number of factors investigated, we used a modified backward elimination procedure. First, one set of important variables was analyzed, ie, building, sex, age, job category, seniority, and weekly workhours. The relative importance of these variables was judged from the single factor analysis and from the results of other studies. The model was reduced with a standard backward elimination procedure until only factors jointly significant at the 5 % level were retained. These factors were considered together with the variables on work function, and a similar stepwise reduction was performed. The resulting model was analyzed, together with two different groups of variables, for psychosocial aspects of the work and personal characteristics. Again the same stepwise elimination was carried out. Finally, all independent variables retained in the foregoing analysis at a 5 % significance level were included in the final model, and a last backward elimination was performed. In this last model, only variables with a significant influence at the I 010 level were retained.
For several independent variables, different scorings were tried, but in the final analysis many were dichotomized. Variables such as age, seniority.workhours, and number of clients served per day (and others) were divided into three or five groups and analyzed as continuous variables as well. The job categories, although not all significantly different, were kept separate, as there were no natural combinations and as a combination directly based on prevalences would have been different for the mucosal irritation symptoms and the general symptoms.
Apart from the variables included in this report, many others were investigated, and the interactions between some of the significant variables were tested. For example, it was determined whether the differences between the sexes or between the job categories remained constant across the buildings. There were no marked indications of interactions between the buildings and the other significant effects.
In the analyses, the buildings were separated by a categorical variable (highly significant). In a subsequent paper this variable will be replaced by the measurements of the indoor environment in the individual town halls.

Personal characteristics
In table 2 the association between certain personal characteristics of the subjects and the prevalence of work-related symptoms is shown, together with the estimated odds ratios from the final model of the logistic regression analyses. The factors shown are some basic characteristics and information given by the subjects on diseases relevant to the context of the sick building syndrome (allergy in the family, hay fever, and migraine).
The women had a substantially higher symptom prevalence than the men. Only in the first multivariate model containing the variables of building, sex, job category, and number of workhours weekly or number of hours in one's office per day was the odds ratio for women somewhat reduced compared with the crude odds ratio. In the other models of the multivariate analysis, the odds ratio did not change much during the steps. Other than sex, only hay fever and migraine were significantly associated with the symptoms in the multivariate analysis. Hay fever had an (expected) effect on work-related mucosal irritation, and migraine was associated with the work-related general symptoms (also expected).

Life-style and residential factors
The life-style factors, ie, smoking, alcohol and coffee consumption, and exercise, were only weakly associated with the symptoms. However, there was a somewhat increased frequency of work-related general symptoms among the smokers, whereas coffee drinkers tended to have a lower frequency of work-related mucosal irritation (table 3). People wearing contact lenses had a higher frequency of work-related mucosal irritation than the others.
Housing conditions seemed to have an effect on both work-related mucosal irritation and work-related general symptoms. People living in apartments tended to have a slightly higher prevalence of both workrelated mucosal and work-related general symptoms than those living in other kinds of dwellings. Persons with small children or with indoor climate problems Table 2. Prevalen ce of work-related sy mptoms (%) amo ng the office workers by personal chara ct eri sti cs . (N =number of perso ns in th e cat ego ry, % =sympto m prevalence, P =P-value of th e chi -squ are tes t , OR =odds rat io est imate d fro m t he fin al mod el of th e logi st ic regres sio n analysis , 95 % CI =95 % co nf ide nce inte rval, NS =nonsignificant in the multivari at e ana ly si s)  in th eir residence seemed to suffer from more workrelated genera l symptoms than other persons. However, the effect of these housing con ditions was insignificant in the multivariate analysis, except for indoor climate problems in the residence (table 3).
Job category Ta ble 4 shows that work-related mucosal irritati on and work-re lated general symptoms were frequent in the clerk categories. The effect of job category gradually diminished during the steps of the multi variate anal ysis when information on the type of work, etc, was included in the model, especially for the clerk categories and social workers . For work-related mucosa l irri-ration , the reduction was the most marked when the covariates for type of work and workhours in the office were added to the model. For work-related general symptoms , the most marked reducti on was seen when the psychosocial factors of work were added. The odds ratio for work-related mucosal irritation among the clerks was, for example, reduced from 5.2 to 3.1 and that of work-related general symptoms among the social workers from 3.8 to 2. 1.

Seniority, type of office and workhours
The small and weak effect of length of time in office wor k and number of occupa nts in the office was insignificant in the multivariate analysis (table 5). Num- ber of workhours per week, when anal yzed as a continuous variable, and hours in one's office during the day had a significant effect on the prevalence of workrelated mucosal irritation and work-related general symptoms. These two covariates were significant in the multivariate anal ysis when added separ ately to the model. When they were added together, only number of workhours per week had an effect on work-related general symptoms, whereas hours in one's office per day had an effect on work-related mucosal irritation (table 5). The effect of number of hours worked outside the town hall building became insignificant in a model containing workhours per week and hours in one's office per day.

Type of work
Various common types of work in a town hall office, other than writing and typing (which were not asked about) were related to the prevalence of symptoms (table 6). It was found that the handling of carbonless paper and the handling of carbon paper were significantly associated with the occurrence of work-related mucosal irritation and work-related general symptoms, as were photocopying and work at video display terminals. The number of clients was clearly associated with the occurrence of work-related general symptoms, but not with work-related mucosal irritation.
In the multivariate analysis, the handling of carbonless paper was the only covariate with a significant effect on work-related general symptoms.
Handling carbonless paper weekly or daily, photocopying more than 25 sheets weekly, and work ing at a video displa y terminal more than I h some times a week or daily had a significant effect on work-related mucosal irritation.

Psychosocial factors
All the psychosocial factors of work examined are shown in table 7. They were all strongly associated with either work-related mucosal irritation or work-related general symptoms in the multivariate analysis when added separatel y to the model. When all the psychosocial factors were included in the analysis, only dissatisfaction with one 's superior and the feeling that the quantity of work inhibits one's job satisfaction had a significant effect on work-related mucosal irritation. The possible mental strain of the combination of little influence on the organ ization of the daily work and a high workpace has been suggested by Karasek et al (20). In the present study it was self-assessed, and office workers considering their work pace too fast and feeling that they had only some or little influence on the organization of their daily work had a significantly higher odds rat io for general symptoms. The test probability for the effect of job category and the com- Table 6. Prevalence of work-related symptoms (%) among the office wOJi<ers by work function. (N =number of persons in the category, % =symptom prevalence, P =P-value of the cnl-square test , OR =odds ratio estimated from the final model of the logistic regression analyses, 95 % CI =95 % confidence interval, monthly or less =no, sometimes a year or sometimes a month, weekly or daily =some times a week or daily, NS =nonsignificant in the multivar iate analysis)  binat ion of little influence an d high wor k speed varied just about 0.01 in the last step of the stepwise procedure of the multi variate ana lysis (2572 person s) and in the final model (2597 persons). Thus both were retained in the mod el.

The buildings
In figure 1 the crude odds ratios for the occurrence of work-relate d mu cosal irr itation and work-related general symptoms in the 19 build ings included in the stud y are shown alon g with the adjusted odd s ratios. Th e adjusted odds ratios were der ived from the final model in the multivari ate ana lysis. Some buildings changed their ran k ord er, and in some cases the difference between th e crude and adj uste d odds rat io was sub stantial in comp ari son with the variation between the buildings. But the odds ratio between the buildings with the highest pre valence and those with the lowest prevalence was still about thre e.

Discussion
This study was not undertaken at th e request of either the management or the emp loyees, as most studies of the sick building syndrome have been (13,14). The buildin gs and their employees were chosen because we wanted to examine a uniform population with equ al socio economic status and type of work, but exposed to different indo or climate conditions.

292
We have prev iously rep ort ed that the many indo or clima te facto rs determin ed in this study resulted in values mainly at the levels normally considered acceptable or in values in accorda nce with levels previously reported. However, there was a considerab le variatio n between the buildings as to some of the indo or climat e param eters (11). Th e prevalence of work -relate d mucosal irritatio n and work-related general symptoms differed between the build ings too , and the symptom prevalence was significan tly associated with sex and some job-related factors.
Using mult ivariat e logistic regression analyses, we have studied these and other pot ential risk facto rs for the symptoms. An impo rtant find ing was that, although work-relate d mucosal irri tation and wor krelated general symptoms were significan tly associated with some of the per sonal characteristics, job-related facto rs, and psychosoci al facto rs of work , there was still a highly significant association between the symptoms and the building factor. Thu s the observed differences in prevalences amo ng the populatio ns in the bu ildings cannot be explained by different composition s of employee s.
Amo ng the man y personal characteristics, life-style facto rs, and resident ial factors analyzed, sex seemed to be the most import ant risk factor. In common with other studies, our investigatio n revealed that wom en ha d a higher prev alence of work-rel ated mucosal irr itat ion and work -related general symptoms than men (10,15), even when other important factors, such as job category, were taken into consideration.
Persons with a medical history of hay fever (14 070) had a higher risk for work-rel at ed mucosal irritation , probably owin g to the nonspecific hyper sensitivity of the mucous membrane connected with this disease.
Persons suffering from migraine had a higher risk for work-related general symptoms, a reasonable finding, since work-related headache is one of the symptom s in the group of work-related general symptoms. In an interview study of a random sample of the adult Danish population, smoking was significantly associat- ed with both work-related mucous membrane irritation and work-related headache (10). In this study, smoking more than 10 g of tobacco a day was significantly associated only with the occurrence of workrelated general symptoms, and the effect was weak (odds ratio 1.3).

MUCOSAL IRRITATION
Coffee is known to have a symptom-relieving effect on headache, but there was no association between coffee consumption and work-related general symptoms in our analyses. However, people drinking coffee had a lower risk for work-related mucosal irritation than those not drinking coffee. This is probably a casual finding; it might also be that those who do not drink coffee form a sensitive section of the population.
A minor portion (11 070) of the office workers reported a variety of indoor climate problems in their residence. Generally, they had a higher risk for workrelated general symptoms than those without such problems.
The job-related factors seem to be important risk factors for work-related mucosal irritation and workrelated general symptoms. Work hours per week and the average hours per day which the office worker stayed in the same office during a workweek somehow reflect the same job-related factor, since in the multivariate analyses they were mutually exchangeable. People staying 7 hid had more work-related symptoms than those staying less and also more work-related symptoms than those staying 8 h or more; this finding suggests that the problem was not merely a matter of the number of hours spent in the office.
Handling carbonless paper, photocopying, and working with video display terminals have previously been found to be associated with work-related mucosal irritation and work-related general symptoms (16,17). Although this population of office workers handled only minor quantities of carbonless paper, photocopied few sheets of paper, and worked few hours at a terminal, these factors were significantly associated with the work-related symptom prevalence. This finding emphasizes the importance of considering these factors in the study of 'indoor climates.
Some of the influence of job category on the prevalence of work-related mucosal irr itation and workrelated general symptoms could be ascribed to work function and psycho social factors of work, but, even when adjusted for these covariates, job category still showed a highly significant association with this prevalence. The effect of job category could either be due to uninvestigated work functions and psychosocial factors or to a general tendency to allocate a different indoor climate to the different job categories.
Social and psychological aspects of work have been related to complaints about indoor climate and workrelated symptoms (5), and in some cases employees have been accused by employers of being hysterical. The symptom pattern of the sick building syndrome is not very typical for mass psychogenic illness, and the course of the sick building syndrome is endemic 294 rather than epidemic in contrast to mass hysteria (18,15), even though the possibility of major psychogenic factors has to be considered.
Indeed, highly significant associations between the work-related symptom prevalence and the psychosocial factors were found, and, as expected, the impact on the occurrence of work-related general symptoms was more pronounced than that on work-related mucosal irritation. However, the psychosocial factors asked about in this study could not explain the overall variation between the buildings as to work-related symptom prevalence, but the multivariate analyses indicated that these factors had a substantial impact on the recorded differences between the job categories. It is interesting to note that the adjusted odds ratios for work-related general symptoms were higher for persons who felt that their workpace was high and that they had little influence on work organization, a combination which is considered to produce mental stress (20).
The symptoms and the independent variables were derived from the questionnaire, but information about age, sex, job category, and building was checked from the data of the clinical study and employment lists. The other factors were based on self-reported information, and people suffering from symptoms may be highly aware of what they are exposed to or those exposed to suspected factors may report symptoms more frequently. In both cases the associations between the occurrence of symptoms and the exposure to carbonless paper, photocopying, and video display terminals would be explained. However, the study was introduced as a general study of the indoor environment that especially emphasized the measurement of the indoor climate, and the results are in accordance with those of other such studies.
This study dealt with a selected population of office workers in a geographically selected population of buildings and thus was not intended to be a representative sample of all office workers or even civil servants . Nevertheless, the prevalence of work-related mucosal irritation reasonably agrees with that of a study of a representative sample of the adult Danish population (10), in which the estimated prevalence of work-related mucosal irritation at work for office workers was 44 % for women and 25 % for men, in comparison with 21 % for women and 12 % for men in the general population.
In conclusion, the building factor was strongly associated with the prevalence of work -related mucosal irritation and work-related general symptoms, and the result s supported the concept of using the term "sick building syndrome," but other factors such as sex, jobrelated factors, and psychosocial factors must be accounted for. Whether the building factor is related to special characteristics of the building or to the measurements of the indoor climate will be discussed in a subsequent paper.