Influence of job strain and emotion on blood pressure in female hospital personnel during workhours.

THEORELL T, AHLBERG-HULTEN la TORRE B. Influence of job strain and emotion on blood pressure in female hospital personnel during workhours. Sand J Work Environ Health 1993;19:313- 8. A homogeneous sample of 56 women who were between the ages of 20 and 59 years and worked in acute emergency care, child psychiatry, or a pediatric outpatient clinic comprised the subjects of this study to determine the relationship between job strain and blood pressure. Job strain was measured with a standardized questionnaire, and blood pressure during workhours with self-triggered equipment. Endocrine factors (morning concentration of plasma pro lactin, cortisol, and dehydroepiandrosterone) and emotional states recorded in diaries werealso studied. Significant interrelationships occurredamong perceived job strain, plasma prolactin,and diastolic blood pressure during workhours even when body mass index, age, family history of hypertension, level of education, and moodstate were adjusted for in a multiple regression analysis. Thusjob strainof female care givers was associated with systolic and diastolic blood pressure during workhours and also with diastolic blood pressure at rest, but not with blood pressure during leisure time.

Since the introductio n of the co ncept of job stra in (the result of a combinat ion of high psychol ogical demands and low decision latitude), several studies have been publi shed regarding its relationship to the risk of cardiovascular disease ( I) . One of the reasons for this rel ati on ship may be th at job strain indu ces a state of sympa thoa drenal act ivati on. If this activation is long-la stin g, it might influence several physiol ogical processes which incre ase the risk of cardiov ascular disea se. Blood pressur e elevation is suc h a process. Accordingly, blood pressure should be studied in relation to job strain. Existing data on this relation have been co ntrove rsial (2)(3)(4)(5)(6)(7). Th e discrepancies may be due, at lea st in part , to the way in which blood pre ssure has been measured -at rest, during workhours, or during leisure time (8). The conflicting results may also be explained by the varying degree of aw areness in studied subjects about their own blood pressure -subjects who are unawa re of their ow n hyperten sion are likel y to und erreport life probl ems (9), whereas the opp osite may be true of tho se who know that the y are hypertensive.
Gene ral knowledge exists regardin g the role of the pitu itary and the adrenal medulla and cortex, and spe- ci fica lly cortisol and catec holami ne conce ntrations in pla sma and urine ha ve been studied e xten sively in relation to blood pressure ele vation. (For a review, see reference 10.) H owever, there is also a need for more specific knowledge regarding other endocrine facto rs. For instance, it has been pointed out th at plasm a prolactin , which mirrors the dopaminergi c act ivity in the brain (the lower the dop aminergic activity, the higher the pla sma prolactin), seems to react to real life stressors in a sys tematic way . When subjec ts fee l over whelmed and pow erl ess in a crisis situation, pla sma pr olactin tend s to rise, whereas no elevation or even decrea sed levels may be seen during a cris is situatio n if the subject is active and feels that he or she is in co mmand of the situation. (For a review, see reference II.) This phenomenon could be of importance to blood pressure although the relati on ship between prolactin and blo od pressure has not been studied extensively in nonpregn ant study gro ups. Another hormone of potenti al rele vance that has been studied during recent yea rs is dehydroepiandrosterone (DHE A). It has been found that DH EA plasm a levels are low in subj ects who are likel y to develop heart disease or to have aggrav ated heart disea se in the near futu re ( 12).
It has been pointed out by several authors th at there may be spec ific physiological reactions to certain co mponents of work stress and that, accordingly, it may be meaningful to study the psychoendocr inologic aspects of work stress spec ific ally in defined occ upational branches. One such ex ample is health ca re, which always entail s co nfro ntation wi th and closene ss to suffe ring (13 ). Thi s confrontation

Department
Number Age (years)  of  of  employment  subjects  Mean  Range   Child psychiatry   17   35  23-58   Emergency   29  37  20-56   Pediatric outpatient   10   43  33-59   Total   56  38  20-59   Table 1. Age of the participants, none of whom were pregnant, birth control medication (N = 8) were excluded from lactating or on birth control medication. the present analysis. Thu s 56 women remained. Of these subjects, 27 were registered nurses and 29 were nurses and aides with a lower educational level. The numbers of partic ipants (blood samples) in the different wards, as well as their age distribution , are presented in tabl e I.
is a special characteristic of caring which may have profound effects on endocrine factors. Other features of health care are the frequent risks of intense emotional conflicts and risks of error making which may threaten the lives of others. The specific relationships between work conditions in health care and psychophysiological activation are of great interest in the stud y of blood pressure in relation to work conditions. In the present investigation an effort was made to combine a study of a psychoph ysiologically interestin g homogeneous group (female care givers) with regard to blood pressure (both diastolic and systolic blood pressure) at rest and during activity at work and during leisure time. Since a criticism in previous studies has been that researcher s have not taken into acco unt social class or education, which may determine blood pressure leve ls independentl y of work conditions (6), level of education was used as a confounding variable. Finally emotional states were studied since it has been remarked that the mechanism underlying possible relationships between job strain and blood pressure may be emotion al. Body mass index and family history of hypert ension were also taken into account as potential confounders in some of the analyses .

Study sample
This study is part of a larger one concerned with improving the psychosoci al work environm ent in health care . As part of this study, the personnel of three different departments (an acute eme rgency ward, a departm ent of child and youth psychi atry, and finally a pedi atri c outpati ent clini c in Stockholm county) were asked to parti cipate. All fem ale personnel who had been working for more than half a year at the ward were eligible. Out of those eligi ble, 90% particip ated in the questionnaire part, 9 1% agreed to have a blood sampl e taken in the morning for the endocrine part, and 82% made a self-tri ggered blood pressure measurement every hour during an ordinary workd ay. The personnel in the acute emergency unit were working on a co nstantly rotat ing day and night shift. The measurements took place during daytim e work, with an ordin ary night' s sleep preceding the work. The resultin g number of fem ale participants was 64. Women who were pregnant, lactating, or on The participants were asked to arrive at work in the morning after overnight fasting. Between 0700 and 0800 a blood sample was drawn from an antecubital vein by a licensed nurse. Thereafter, the subjects were instructed to measure their own blood pressure by means of a Cardiocar e 2000 apparatus once every hour dur ing waking time (of a day with no evening or night work) from 0800 to 0800 of the next day, both at work and during leisure time and to note activities and moods at the time of each measurement. The protocol s used for this proced ure have been described earlier (14).
The average systolic and diastolic blood pressures were calculated separately from all of the ava ilable measurem ent s for work and leisure time . Measurement s at rest were made immediately afte r the subjec ts woke up in the morning, before they started any activities. The comparability of these measurements to both con ventional and full y automated equipment measurements has been reported as good (14,15).
Emotional states were monit ored by means of a diary, which was fill ed out on every blood pressure measurement occasi on. On the basis of the ratio of report ed " happy" occasion s to total numb er of observations, percenta ges of specific moods were calculated for each subj ect. For the prese nt analysis, the specifi c moods "perce nt happ y" and "perce nt worried" were used as indi ces of emotional states . "Percent happy" showed a distribu tion that was reasonably close to a norm al distribution (skewness -0.28, kurtosis -1.1 4), whereas the variable "percent worried" was less normally distributed (skewness 2.33, kurtosis 5.28). The logarithmic transformation of the "percent worried" variable was tested in the statistical analyses, but it made no difference to the result s -all of the relationship s were far from statistica lly significant.
For the description of the psychosocial work environment a Swedi sh version of Karasek' s demanddecision latitude questionnaire ( 16) was used . In the present stud y, howe ver, the modification fro m the ordinary Swedish version was made in that the participants were asked to categ orize intensity of conditions rather than frequency . A Swedish work social support scale (positive factor s) was also used ( 17). The Cronb ach alpha was 0.5 1 for psychol ogical demands, 0.56 for decision latitude, and 0.77 for positive factors. For the present report only positive factors and the ratio between psychological demands and decision latitude have been used. The job strain ratio of this sample showed a distribution which was close to normal (skewness 0.21, kurtosis -0.73).
Radio immun oassay was used for the analysis of the cortisol, prolactin , and serum DHEA concentrations ( 18). Logarithmic transformations were used in all of the statistical analyses of plasma prolactin since its distribution was skewed.
Body mass index was calculated on the basis of reported weight and height (kg ' m'). The subjec ts were specificalIy asked about famil y history of hypertension in their father, mother, and siblings. Those who had respond ed affirmatively to any of these questions were operationally defined as subjec ts with a family history of hyperten sion. Conventi onal statistical tests were used, including a one-way analysis of variance with post hoc testing according to Scheffe and a linear multiple regression. Table 2 shows the mean systolic and diastolic blood pressures and their 95% confidence interv als during rest, work, and leisure time.

Results
The mean job strain ratios in the three departments were 0.80 (SE 0.04 ) (emergency), 0.55 (SE 0.03) (child psychiatry), and 0.53 (SE 0.07) (pediatric outp atient). The differences between the emergenc y department and the other departm ents were significant, job strain being perceived as higher in the emergency departm ent than in the other two department s.
A linear regression was performed between job strain and the systolic and diastolic blood pressure measurements and durin g rest, work, and leisure time, as welI as between job strain and body mass Scand J Work Environ Health 1993, vol 19, no 5 index, plasma levels of (log) plasma prolactin , plasma cortisol, plasma DHEA , age, education, percent happy, and percent worried. Significant relationships on the 5% level were found between j ob strain and diastolic blood pressure durin g rest (standardized beta = 0.30) and during work (standardized beta = 0.29). There was also a significant relati on between job strain and (i) body mass index (standardized beta. -0.30) and (ii) log plasma prolactin (standardized beta = 0.31). Log prolactin was tested in relation to the blood pressure measures. Significant relationships were found with diastolic blood pressure during work and sys tolic blood pressure during leisure time (standardized beta 0.27 and 0.31, respecti vely).
The percentage of time feeling "happy" was also tested in relatio n to the six different blood pressure measures. Significant inverse relation ships were found in all instances, the standardized regression coeffici ents ranging from 0.35 (diastolic blood pressure during work) to 0.49 (systolic blood pressure during work). The relationships between j ob strain and diastolic blood pressure and between j ob strain and log plasma prola ctin are displayed in figures I and 2, respectively.

15
Scand J Work Environ Health 1993, vol 19, no 5 A multiple linear regression analysis was then performed with each of the six blood pressure measures as dependent variables and the following explanatory variables in each of these equations: age, family history of hypertension, body mass index, education, job strain, positive factors, percent happy, and percent worried. The factors which were significant in any of these analyses are presented in table 3. Due to the fact that some subjects were lost in some measurements, the numbers are smaller than in the full sample.
In the analyses of blood pressure during workhours the analyzed sample of 41 women corresponded to 66% of the total sample invited to participate. The standardized regression coefficient between job strain and diastolic blood pressure during rest was 0.34, which was the only nonsignificant coefficient that was close to statistical significance (P =0.08) The magnitude of the regressions was also estimated by means of nonstandardized regression coefficients, with their 95% confidence intervals (95% CI). The nonstandardized regression coefficients for job strain in relation to systolic and diastolic blood pressure during work was 12.8 (95% CI -4.2-29.8) and 13.9 (95% CI 0.8-27.0), respectively, in the single regression analysis (N = 50 for both systolic and diastolic blood pressure). In the corresponding multiple regressions the nonstandardized regression coefficients were 29.2 (95% CI 10.7-47.6) and 24.6 (95% CI 7.6-41.1), respectively.
To estimate the importance of the endocrine variables in relation to blood pressure during work, a multiple regression analysis was also performed 316 using age and log plasma prolactin, plasma cortisol, and DHEA in serum as the explanatory variables. Table 4 shows the results of this analysis.
Since standardized beta coefficients have been criticized, we also calculated the percentage of possible change in systolic and diastolic blood pressure due to various explanatory factors. These percentages were based upon the nonstandardized multiple regression coefficients. The difference in blood pressure corresponding respectively to the minimum and maximum age, body mass index, job strain score, and percent happy was calculated. These differences were divided by the mean systolic and diastolic blood pressure. The percentages obtained for systolic blood pressure during work were 19 (age), 23 (body mass index), 22 (strain), and 17 (percent happy). For diastolic blood pressure during work the corresponding percentages were 24 (age), 32 (body mass index), 28 (strain), and 14 (percent happy). For systolic blood pressure during rest the numbers were 19 (age), 20 (body mass index), 21 (strain), and 14 (percent happy). The contribution of job strain to diastolic blood pressure during rest was not significant. (See the preceding results.) In this analysis body mass index, closely followed by job strain, was the most important determinant of systolic and diastolic blood pressure during work and of systolic blood pressure during rest.
The analysis showed that high prolactin levels were associated with high systolic and diastolic blood pressure. Low cortisol levels were associated with high blood pressure as well. Plasma cortisol was not significantly associated with job strain.

Discussion
These results indicate that there are strong relationships between job strain and blood pressure during work done by female health care personnel. The most consistent relationships were found for diastolic blood pressure during work.
Mood was used as a potential confounder in the multiple regression analyses of job strain and blood pressure, since it could be argued that the reason for a relationship between job strain and blood pressure during work is that a bad or good mood could influ-ence the way in which the job situ ation is being described and that the important fact or could be mood independently of any objective work conditions. In this case there was no relationship between mood and reported job strain, and the introduction of the mood variables into the equations did not change the relationship between job strain and blood pressure. Intere stingly, the relation ship between jo b strain and diastolic blood pressure durin g work was considerably strengthened after adj ustment for body mass index. The partic ipants who had a higher body mass inde x experienced lower levels of job strain .
It should be noted that there were premenopausal, as well as menopausal and postm enopausal , women in the study sample. There was no relationsh ip whatsoever, however, between age and plasma prol actin, neither linear nor curvilinear.
The participants were in good health . Only two subj ects had (asymptomatic and previously undiagnosed) diastolic blood pressure equal to or above 90 mm Hg (11 .97 kPa) during rest, and neith er of these subj ects were on medication. In addition the body mass index was generally low. The attrition rate was acceptable. The sample studied may not be fully representative of health care personnel , but its members did have widely different work tasks. The acute emerg ency personn el were typically exposed to many dramati c situati ons requi ring quick and determi ned action s, whereas the other groups studied had closer contact with chronic suffering among children and their families. Other group s of health care personnel, for instance, those in geriatric care, are exposed to more physical demand s. It appears that the studied groups faced primarily psychosocial work strain. Therefore they could be appropriate for the study of the relationships between physiologic al factors and the psychosocial work environment. Health care workers in hospital s are, in general, aware of their own health, and they seem to be better able than subjects in other occupations to analyze their own jo b situation. Underrepo rting may therefore be a smaller problem in this group than in many other groups that have been studied, for instance, male bus drivers (6).
The magnitude of the observed relationships between j ob strain and blood pressure could be illustrated by the fact that an increase in job strain of 0.15 , which corresponds to an increase of 3.7 mm Hg (0.49 kPa) in diastolic blo od pressure durin g work in the multiple regression analysis, is the equivalent of the job strain difference between female physici ans (low) and waitr esses (high) in a previous study (15). The individual variation in j ob strain was considerable , and it was observe d that, in the multiple regression, job strain was of greater relative importancc to diastolic blood pressure during work than age, body mass index, family history, and emotional state during the day of measurement.
It should be no surpri se that job strain was more relevant to blood pressure durin g work than during Scand J Work Environ Health t 993, vol 19, no 5 leisure time. It is of interest, however , that significant, albeit weaker, relationships were found between blood pressure during rest and job strain. These findings are in general agreement with those of our previous study of ambulatory blood pressure recordings in relation to jo b strain (5). A general finding in both of our studies (5), as well as that of Schnall et al (3), was that neither psychological demands nor decision latitude alone explained blood pressure level. Instead, it was the ratio bet ween high demands and low decision latitude that was important.
Although the psych ophysiological finding s of the present study are difficult to interpret due to the cross-sect ional nature of the study, the finding of consistent relationships between jo b strain and plasma prolactin , as well as between plasma prolactin and blood pressure, merits attenti on. It is possibl e that plasma prolactin could mirror one physiological pathway between j ob strain and blood pressure. An interpretation of these find ings is that these women feel overburdened. Feelin gs of disaster potential in crisis situations may be associated with elevated prolactin levels (I I). For instance, it has been shown that subway drivers who are exposed to "person under train" episodes develop sleep disturb ances and elevated prolactin levels for some weeks after the episode (19). It is possible that health personnel who are exposed to many such emotionally trying episodes in their work may react in the same way.
The finding that serum DHEA was lower in subjects with elevated blood pressure during leisure time is in agreement with the findings of previous research (12).
In conclusion, the findings of the present study support the hypothe sis that job strain is associated with blood pressure elevation during work and possibly also during rest. The sampl e studied was homogeneous with regard to gender and occupational branch, showed a wide job strain variation, could be expected to describe j ob situations accurately, and included no subjec ts who have become " hypertensive patients." Adju stment for the potential confounders age, mood, family history, body mass index, and educational level incre ased the strength of the association. The results supp ort the assumption that plasma prola ctin may mirror a physiological pathway between j ob strain and blood pressure.