California Focus Study Group Session I: Summary January 21, 1998; Irvine California FOCUS ON JOB STRAIN RESEARCH
Summary Prepared by: Karen Belkic
Overview of Research on Job Strain
Dean Baker
The Job Strain Construct was originally conceived much more broadly and complexly than is reflected in the instruments under current use. Due to the need to rely upon existing data bases, a series of questions has been used repeatedly, and these have yielded a substantial number of positive results. However, there is clearly a need to refine both the models and techniques, with the basic aim of applying these successfully in intervention research.
June Fisher and David Ragland
There is found to be limited variance when a single occupational cohort is assessed using the current Job Strain methodology. Since in-depth investigation and intervention often needs to be focused upon a single occupation, the dilemma emerges: Can the Job Strain Model help answer the relevant questions, many of which are raised by the Trade Unions, e.g. "how long should a rest break be" for a given occupational group? There is a need for job-specific instruments. On the other hand, the conceptual framework of the Job Strain Model is has been quite enthusiastically adopted by Trade Unions. Many of the worksite changes endorsed by Trade Unions de facto serve to ameliorate burdensome demands and to promote worker control on the job. This is seen particularly in efforts to increase worker input into scheduling and attenuation of time pressure.
Paul Landsbergis
There is an ever-increasing body of job strain research as it pertains to CVD. Some of the most recent European studies are still unpublished. Overall, approximately half of these studies are concerned with heart disease, and half on CVD risk factors. The majority do show a positive relation between job strain and health outcomes. Furthermore, the cohort studies are positive, by-and-large. On the other hand, investigations using casual blood pressure as an endpoint are often non-significant, whereas there is a more consistent, significant positive association between job strain and ambulatory blood pressure.
Based on the existing epidemiologic data, the Job strain model seems to be operative both in men and in women. Blood pressure may be a major mediating mechanism of job strain-related heart disease. The epidemiologic evidence is not as consistent for obesity nor for hypercholesterolemia, while smoking as well as elevated fibrinogen seem to have more confirmatory data.
With regard to the two major dimensions of the Job Strain Model in relation to CVD outcomes, the findings are most consistent for low decision latitude, while there seems to be less data supporting the role of psychological demands. There are even some recent empirical studies suggesting that low demands may also be related to CVD.
Blue-collar workers appear to be the most vulnerable to the cardio-noxious effects of job strain. With increasing age, the effects of job strain are also augmented. Low social support coupled with job strain ("isostrain") is a particularly deleterious combination.
Epidemiologic studies using the imputation method have yielded predominantly positive results. This method assigns national average Job Strain scores to a given occupation, thereby obviating self-report bias. As of the time of the conference, there had been no published intervention studies of Job Strain in relation to CHD or blood pressure outcomes.
Peter Schnall
This study has a nested case-control design, and was extended to become a cohort investigation. Originally there were 264, predominantly white, male participants from seven different New York City Worksites. At three years, there were 195 men who had participated in both waves. A third wave of ambulatory blood pressure analysis was performed after 6 years, with 194 of the original 264 men participating. At Wave 3, 100 female nurses and aides were included, in addition. Job strain in the first wave was defined using median cutpoints for psychological demands and decision latitude, and was present in 26% of the participants. These same cutpoints were used in the subsequent waves, yielding 16% job strain exposure at Wave 2, and 11% at Wave 3. Cross sectionally, there was an approximately + 6-7 mmHg work systolic blood press (BP) and + 3-5mm diastolic BP effect of job strain. These findings persist into leisure and even sleep periods. Among the women, a similar, significant although slightly lower job strain effect was found.
Longitudinally, those repeatedly exposed to job strain at Waves 1 and 2 had a mean level of ambulatory BP +12/+7 mmHg greater than those without exposure at either wave. A similar but slightly smaller effect was seen for repeated exposure at Waves 1 and 3.
Most striking was a significant drop in blood pressure among those who were exposed to job strain at Wave 1 but not at Wave 2. This effect was mainly due to an +11/+6 mmHg drop in blood pressure among the hypertensive participants. These findings illustrate the importance of assessing exposure at repeated intervals, and that changes in exposure can have noteworthy effects upon ambulatory (A) blood pressure. Furthermore, the fact that a change in exposure to job strain leads to a change in ABP provides yet another argument for a causal relationship. These findings, if replicated, could provide the basis for a clinical intervention trial aimed at attenuating job strain among hypertensive working people. Do the negative and non-confirmatory studies provide serious evidence against the job strain hypothesis with respect to Cardiovascular Disease:
Clinical Implications
Karen Belkic
Although the majority of studies concerning job strain and cardiovascular disease outcomes have shown a significant positive association, there have also been a number of publications which have reported non-significant, and in a few cases, seemingly negative results. In contradistinction to the general lack of attention which the medical community has given to the role of the job strain or the workplace in general with respect to cardiovascular disease, these latter studies have been often cited in cardiologic texts and other publications as evidence against the job strain hypothesis. Upon closer scrutiny however, it becomes clear that many of these studies are fraught with major flaws which undermine their internal validity. Among the most serious of these flaws is selection bias, whereby a sample of convenience such as patients undergoing coronary angiography, fundamentally and inextricably distorts the etiologic relation between the exposure and outcome variable. Other flaws common to a number of these studies include, inter alia, failure to adjust for major confounders such as socioeconomic status, lack of follow-up with respect to job strain or even employment status in long-term cohort studies, use of median cutpoints to define job strain exposure among a single-occupation cohort, failure to stratify analysis by gender, as well as non-differential misclassification which occurs particularly when the imputation method is used. Since current level of exposure to job strain can be strongly influenced by selection factors, especially seniority, the need for assessment of cumulative exposure to job strain was underscored.
On the other hand, these null studies also raise several important issues with respect to the job strain model itself. In particular, the psychological demand dimension shows low internal consistency and is most vulnerable to bi-directional report bias (i.e. over-report as well as denial). The psychological demand dimension, as now operationalized in the Job Strain questionnaires, is quite subjective (most notably, the "working fast" and "working hard" items). A numerical example was given with respect to professional drivers, in which the psychological demand dimension was theoretically formulated so as to reflect the actual brain processes involved in human information transmission and then operationalized using concrete items specifically relevant to professional drivers. Whereas the Job Strain questionnaire demand dimension showed no significant difference between the drivers and a control group of subway guard attendants, the latter approach not only demonstrated a significantly augmented demand dimension among the drivers, but also pinpointed the specific nature of those demands. Most notably, the drivers had a very high burden at the level of receiving incoming signals. Furthermore, a number of the potentially modifiable workplace demands were identified using this approach. The overly parsimonious framework of the currently used methods for assessing Job Strain was also emphasized.
Peter Schnall
Using existing job strain questionnaires, the definition of job strain is still a subject of debate. Arbitrarily delineating an approximate 25% exposure level based on median cutpoints inevitably leads to non-differential misclassification. The quotient term has been shown to yield a stronger effect size than that does the quadrant term. Furthermore, a "J" shaped relationship is found when the quotient term is broken up into segments. In other words, exposure to low demands which would imply understimulation, also appears to be detrimental to the cardiovascular system. This finding is reminiscent of the physiologic model of underload and overload as described by Frankenhauser and colleagues. On the other end of the curve, those with exposure to very high levels of demands together with low control show a strong vulnerability to CVD. When job strain exposure is assessed using a 9-cell model (or tertile term), the 9th cell shows a very large effect size with respect to CVD. The finding of a dose-response relationship also supports the contention that Job Strain plays a causal role in the development of cardiovascular disease. The question was raised as to whether cutpoints should be routinely adapted to the population under study, i.e. with respect to age, gender, class etc. Dean Baker, elaborated this idea in relation to developing population norms, against which individually-collected data could be compared. He also emphasized the need for caution particularly in the transcultural setting with respect to how the individual items are worded. An example was cited in which even slight changes in the wording of an item completely altered its meaning so that rather than reflecting demands, decision-latitude was in fact being assessed. When meta-analysis is to be performed, the aggregability of the data becomes a key question, in which the measurement issue looms large.