JCQ Version 2.0

Part II. Outline: Detailed Discussions of Steps in Development of a New Design

Different measurement goals have different implications for instrument design. The typically mentioned JCQ goals are:

a. A scientifically precise tool for assessment of work/illness associations. This requires extended scales, supporting multiple new hypothesis testing, and multiple methodologies

b. A social policy tool to assess the well-being and illness costs of poor work organization in the global economy: This requires a good quality , small core of questions with broad coverage, valid worldwide

c. A job redesign diagnostic tool for problem assessment and problem solving. This requires tailor-made assessment packages for different work groups and industry branches

The criteria that were used to develop the current JCQ (from 1985) are the following, in rough order of importance (see also discussion at beginning of section II-4). Criteria "c," "d," "e," and "f," are of roughly equal weight.

d. Broad range of working life breadth / spanned with a small set of macro dimensions

1. The JCQ Center decided in March 2002 (Dusseldorf 3rd ICOH) to drop the "cognitive load," (and "stressed" question) subset of questions and retain only the 5-question quantitative demand questions because it was inconsistent in theoretical interpretation and an inconsistently performing subset of psychological demand questions.

2. Single questions in scales which are not performing well, or are imprecise in less-used scales (where perfect comparability less crucial), could be dropped:

b. "conflicting demands," (see JCQ Workshop results Dusseldorf and ICBM Brisbane)

c. "job requires physical exertion," much less precise than other four questions - allows comparability to original QES.

The most currently cited difficulties are low scale correlations of questions: "repetitive work" and "conflicting demands." However, these both are components of standard scores in the important scales of D/C quadrant modeling. Predictive validity tests need to be made to determine if they are inconsistent and should be dropped. Repetitive work may extend the skill discretion variance in low end work situations. Conflicting demands has been defended by Hurrell as a "white collar" stressor needed for scale balance, however perhaps a separate new emotional-interpersonal-cognitive load scale would be better.

a. Decision latitude (skill discretion and Decision Authority) have no consistently reported major scales problems: but the two sub-dimensions need to remain separate for theoretical and empirical reasons, and some times will be separately used by researchers.

b. The macro-decision latitude questions that have rarely been used (1985 to 2004) by researcher, and should now be dropped. These issue are far better covered with the proposed dimensions.

Significantly, Psychological Demands 5 question scale - now the quantitatively-focused, "core" - still has inconsistent relations with other scales (also class-related), and interpretability problems relating to some overlap of mental and physical demands, particularly in blue collar workers. There are other critiques relating to low/moderate levels of Cronbach alpha reliability for psychological demands. However, no implications of any of the above problems for predictive validity, ultimately the most important criteria, have yet been demonstrated, and the scale should be improved - not dropped.

a. Low scale reliabilities have been noted since the inception but this may not be a critical problem. This reflect the inevitable consequence the JCQ instrument design decision to include diverse forms of measurement and to span somewhat diverse sub-dimensions within a still theoretically coherent broad dimension (reducing numerical correlations): both individual 'objective" experience (actual layoffs - very skewed distribution) and "subjective impressions ("insecurity" - normal distribution) are included. Perhaps a different statistical technique to assess the psychometric reliability should be employed here.

b. However, the current job insecurity scale has some redundancies: q. 34 "security" vs. q. 36 "loose job" and q. 33 "steady work" vs. q. 35: job loss experience". At least one of these questions should be dropped. The scale scores are not consistently calculated across studies, and this removal would not hurt future JCQ comparability.

In the case of social support, eleven questions are included in the recommended version JCQ 1.0. However, two questions of hostile supervisor or coworker relations have either been rejected in some large studies or found to represent a different dimension in factor loadings (Warren, personal communication, 2002 ?). One "conducive production question has rarely been used (an orphan). The remaining eight questions (two each for supervisor and coworker instrumental and socio-emotional support), have formed two high reliability (supervisor and coworker) scales, or one combined high reliability scale.

However, strong reports of predictive validity are infrequent, and in has been well known for some time that objective validation is an very difficult goal for these questions (Hackman and Lawler, 1971, self-report/expert congruence; and Karasek and Theorell, 1990, appendix showing very low between-occupation variance). This leaves a great potential subjective component to interpretation of findings. Additionally, there have never been any reports of low scale reliability. Because of the low risk of unreliable scales, it is possible that the remaining eight questions could be cut to as few as four (one in each of the four categories above).

Cutting some social support questions is also consistent with the limited predictive payoff of this section of the questionnaire. Other components, specifically macro-level social support, and even socially creative components (under the "conducivity" construct) need questionnaire space. Furthermore, scale scores are rarely used (and can't really be used between-occupation) and this not a major component of future JCQ comparability.

Social policy goals, and epidemiological goals - i.e. long-term follow-up studies, both require consistency in questionnaire structure over long time periods - twenty years and beyond. This is comparability across "space" and time" as Per-Olof Ostergren noted at ICBM 2004. One of the major strengths of the "JCQ System" is the huge accumulation of investments made by users over two decades in collecting data bases with the most consistently applied instrument in the world in this research areas (JCQ User's database with over 500 e-mail addresses, translations in 22 languages, researchers from 14 countries in each of the last two Workshops). This is also an investment in time spent in understanding the scales, their interpretative frameworks, and the critiques and advantages of these sets of questions. It is a serious step to undermine the utility of this investment by changing the scales - even for a marginal improvement in scientific validity.

The JCQ was originally designed for a US occupational health and safety activist who wanted an easy-to-apply, easily-to-interpret package to introduce the "job stress risks" concepts in-the-job OHS hazard monitoring. The instrument was short to encourage wide use. Criteria (described in JOHP, 1998, p. 328) were applied to get the best "overall" instrument, even if scale parameters could not be optimal for all criteria. Surprisingly, over the next decades it was primarily epidemiologists, many outside the field, who used the instruments The two primary reasons for adoption were consistency with the published demand/control research, and compact scales (alternative, much longer "researcher" instruments, such as US NIOSH's own [quite good in many respects] were less used).

No large study in JCQ history has failed to bargain strongly for shorter-than-recommended scales, and then settled reluctantly for the full recommended scales - even if some areas had to be dropped. Specific implementations by large JCQ Users have modified the standard (also called 1977-QES-1985-Framingham Core, see JCQ User's Guide) by default, for example: Core 1993-JACE (36 questions, dropping eight macro DL, three social support, and two job insecurity questions). These large studies such as the JACE have built the predictive validity platform for the Demand/Control model. The JCQ Center feels is unlikely that the JCQ can be made longer than the current recommended 49 questions. Thus if new areas are to be added, current areas will have to be cut back.

However, a "modular approach" successfully used by Tage Kristiansen's Copenhagen instrument has a "mid-range" version of about the JCQ length, as well as a shorter version and a longer researcher version. In the JCQ 2.0 case something similar could be done. We would probably recommend that the "reduced core" (see below: 26 questions in six scales) be the "short form". A longer "researcher version" could now be discussed for the JCQ 2.0 Thus, some scales of lower priority, but with scientific importance should also be included in the design. Possibly it could be suggested to users that a set international researchers has agreed to pilot test this version - to insure the user will not be the only person to ever use this set of questions, without the power of the "all previous users have adopted this recommended scale" argument.

A. Recommended Reductions to Define an "Original JCQ Core" (Note: - Proposal Only)

1. Drop four questions: Drop "cognitive load" psychological demand questions [the JCQ Center decision 2002]]

2. Drop eight questions: Macro-decision latitude questions that have rarely been used.

3. Drop four questions: Single questions in scales which are not performing well, or are imprecise in less-used scales could possibly be dropped.

4. Drop seven questions: Many social support questions could be dropped with little loss.

Total reductions = Drop 23 of the 49 recommended question from JCQ 1.0; leaving a core of 26 questions.

The calculations above suggest that with an Original Core JCQ 2.0 version could be developed with just 26 questions. Retaining the basic JCQ instrument size of 49 questions would allow the addition of 23 new questions in the New Scale Areas above. Furthermore, an initial pilot instrument would have to anticipate that some questions would not work as well as other: perhaps 33% more than this number (32) should be included, yielding an initial Recommended Version JCQ 2.0 - Pilot with 57 questions. For the Researcher Version JCQ 2.0, additional questions, or additional modules could be included.

What are the priorities for new questions? This issue will have to be discussed in upcoming JCQ Workshop (at the 4th ICOH Newport Beach, March, 2005, with a possible second discussion at the 2nd ICOH WOPS in Okayama in August, 2005).

However, we can gain some idea of the priority from examination of the voting record for the topic area to be discussed in the March 19, 2003 meeting (see Minutes, Karasek, Table 1). After "instrument goals," the highest single topic priority was for psychological demand modifications. Then came macro decision latitude, but if the several related topic's votes are included (organizational trust and justice, downsizing, globalism and job insecurity, etc), it had the largest total. Social support came next (but its topics overlapped the macro decision latitude).

We can generate one speculative idea of possible scale question allocations, since there are roughly 30 questions to allocate among a half dozen new scale areas. Marking the highest priority with three stars, and significant interest with two stars, we might give the following allocation of 31 questions for a new instrument [Readers will note that there are many ways to do this, making for a rich future discussion].

25% of these would be dropped after the pilot, for a final 49 question Recommended JCQ 2.0 (this priority was used as a very rough guide for the appended Possible JCQ 2.0 Design Example).

A major reason for using expert interviews is to reduce the numbers of questions that need be asked of each individual respondent in a workplace. Use of company-wide sources of information can be appended to the worker response information gathered independently of the questionnaire - this could reduce questionnaire size by 30 to 40%. An attempt has been made to maximize the number of new areas that could be assessed in this manner - using a set of diverse experts from the company, reflecting different viewpoints and areas of expertese. The experts question would be posed by JCQ researchers in short interview sessions. Expert respondents would be specifically asked to describe the conditions that pertain to the workers who will answer the questionnaire.

Three experts would be selected for each workplace. A production manager or general manger would be needed to answer some questions, but not many ("fewer than 10 questions - to facilitate agreement top participate). A human resource manager and a union representative - or informally identified worker spokesperson - would also be interviewed and asked a longer list of questions (but still fewer than 20). Perhaps with very large work sites several triads of experts could be used - to insure familiarity with specific worker conditions.

Thus, the JCQ 2.0 has been designed to allow corroboration of new areas of job hazard ( and some old ones) with company-based experts who will give a separate assessment of hazard level information. One important issue is the match-up of specific worker questions with particular expert information (response scales would have to be specially designed to facilitate comparison - given different forms of information and response).

A primary comment in much job stress research, and JCQ research, is that objective verification of worker self-report hazard assessment would improve the predictive validity - and the interpretive power of the research.

Data from social registers will often not to available to corroborate worker self-reports in the area of JCQ. However, there are two areas where commonly collected data might be useful - although w it would have to reflect quite local conditions, not always available.

a. Industry/regional competitive pressures, and industry profitability, growth levels.

7. Demand Assessment Packages for Specialized Occupational Groups: Precision and Job Design Utility

One suggestion is to develop specific "job demand" packages for specific occupation or industry groups (less discussed for other dimensions). Jeff Johnson suggested that to avoid an overwhelming task here and to retain the generalizability hoped for in the JCQ, a small set of "ideal-typical" jobs could be identified, and a special set of demand questions developed for each of these. The JCQ Center recommended strategy here would be to include one/two key questions from each of the relevant sub-dimensions within the Core JCQ to provide a "scaling reference" for the other new questions (which allow much greater precision to be attained with in each sub-area by specifically design questions within the special occupation). This would allow both the overall "demand levels" to be interpreted across all occupations (as is currently possible) while addressing the goal of greater monitoring precision in specific areas.

Theoretical congruence with a useful underlying theoretical model is probably the major reason for broad use of the JCQ. The current expansion, with some of its "beyond-model" dimensions, in the JCQ 2.0 thus raises the question of whether a loss of theoretical coherence could occur, and with that a perceived lack of rigor in the JCQ (this was mentioned for "fairness" (E/R); social capital; and emotional coping styles) . What would be the theoretically valid supports for model testing with the new dimensions (and even some of the old ones) in a clearly "broadened" demand/control analytic framework

A. Need for the Stress-Disequilibrium Theory (the Demand/Contol model underpining)?

It may well be true that a broadened statement of the Demand/Control model - such as that provided by the Stress-Disequiibrium Model - will be needed to encompass the new JCQ scales arising from the breadth and dynamism of changes in working life in the global economic context. The question arises: what extensions beyond task-level measurement are supported by this new framework (which underlies the D/C model, but is only now being articulated)?

The demand control model has always been a far more general model that the task level measurements in the JCQ scales, however, the precise operation of the mechanisms at other levels, while addressed in Karasek and Theorell 1990, has not been formulated in the most general terms. Multiple levels of organization are clearly supported in the Stress-Disequilibrium framework - which describes energy-into order transformations in terms which are compatible with "demand and control" concepts, but far more general (and can be seen as the general form for that model, ultimately linked to basic physical principles: thermodynamics). This allows, on the one hand, the physiological hypotheses about chronic disease processes to be formulated, and at the same time, at higher levels, provides hypotheses about organizational structure and behavior pattern integration which have implications at psychological and sociological levels. This full formulation cannot be articulated here (and is not yet complete, but draft materials and a chapter are available), but some examples can be given. (footnote).[1]

Can the simple four cell D/C model version be maintained in the context of addition of the new macro dimensions of job insecurity? An early empirical validation in the QES shows that job insecurity increases the job stain-mental distress associations clearly when added to psychological demands (at the occupational level).

2. Macro-level Control-Insecurity-Support: An Empowerment Indicator, a Demand, or a Support Buffer

Is this new dimension an empowerment indicator (it might be added to decision latitiude), otherwise used in a buffering (interaction) test of D/C moderation.

3. Along the Active Passive Diagonal: Conducive Coordination vs. Ideological Value Conflict

Conducive Coordination (Karasek, BSTS, 2004a, Karasek, BSTS, 2004,b), in terms of collaborative social relationships, may describe the necessary social context of "active job" situations. Also, in other circumstances, ideological conflicts - wherein a worker is forced to work against personal values - could perhaps potentiate passive withdrawl.

(This sectionof the memo will be completed after JCQ Board and JCQ User response)

New "voter-education" summary packages of materials could be developed by scale subgroups to support discussion platform for JCQ users everywhere (including web-based) - to reach beyond the 50 Workshop participants of the last two years. This would prepare for decision-making discussions to be undertaken at an upcoming international meeting [now scheduled JCQ Workshop organized in combination with the 4^th ICOH in Los Angeles in March 2005]. Each packet would included the following components:

Decision trees could serve as a tool for combining several issues involved in a scale area, into an ordered sequence of decisions for JCQ User Group Discussions.

This is a speculative example, and incomplete. Many other alternatives could be constructed. For example, almost none of the questions are listed in the new Control-Insecurity-Support Dimension area (DL-IIb). It does graphically illustrate, however, the nature of the design challenges that must be resolved.

[1] (footnote)

On the "demands side:" stressors create requirements for the organism to utilize its "control capacity" to integrate internal and external adaptive actions. Typical S-D theoretical issue's relating to "control" are: maintaining degrees of freedom of action for one's repetoire of capabilities; of limiting external threats to own control, of utilizing external possibilities of extending one's own control; and finally expansion of long term control capacity (and in the short term daily physiological balance), and finally of growth of new capabilities. For example, themes relating to emotional work demands of complex situations can be addressed, because they require organization of both internal motivations and learned constraints on action at multiple levels: requiring high levels of control capacity. Job insecurity can be addressed because the uncertainly of the worker's external environment creates multiple adaptive demands at both behavioral, organizational and physiological levels (stressors). Furthermore, the concept of maintaining a high-level equilibrium of functional capacity is crucial in the model - and threats to such an stable equilibrium for the worker are obviously a central problem addressed by the new "macro decision latitude-insecurity-support dimension".