The Mental Injuries Tool group was established out of a stakeholder sub-committee of worker representatives and the Occupational Health Clinics for Ontario Workers who were charged with “supporting worker representatives in taking action on prevention and workers’ compensation”. In February 2011 members of the working group and other interested people attended a workshop which reviewed the theory behind common psychosocial measurement tools. Participants were walked through the content of a number of surveys, filled them out, and reviewed the scored results. Based on many contacts and deliberations, the group decided to pilot test the COPSOQ survey at upcoming union events. We contacted Tage Kristensen, the author of the COPSOQ survey and received permission to use instrument (all the materials associated with the survey are freely available online at: http://www.arbejdsmiljoforskning.dk/en/publikationer/spoergeskemaer/psykisk-arbejdsmiljoe). No changes were made to the English language version of the COPSOQ questions.
Based on these successful pilot administrations of the survey, the feed-back we received from the pilot respondents, and discussions within the MIT group it was decided to adopt the COPSOQ survey as the basis for our assessment tool. For the symptoms however, we included extra questions from a longer version of the survey. Five symptom categories were included (burnout, stress, sleep troubles, cognitive and somatic symptoms). With respect to the questions about offensive behaviours, two questions concerning discrimination and vicarious offensive behaviours were added. We did not include any questions regarding an individuals’ history of mental illness or depressive symptoms since we were concerned the worker representatives using the survey might be able to trace an individual’s responses and “label” or “diagnose” the person (even though the surveys are anonymous).
In response to the feedback received during the union conferences and discussions during MIT meetings/calls, questions were considered about exposures to other health and safety workplace hazards. These questions address issues similar to the “Supportive Physical Environment”, which was added as a 13th Workplace Factor in the CSA Z1003 national standard on “Psychological Health and Safety in the Workplace”. Furthermore, various preliminary and demographic questions (often customized to the union or workplace) were also added. The decision to include the exposure questions was made by the MIT group whereas the decision to include various demographic and other questions (e.g. shift work), was left to the discretion of the parties using the survey for their particular workplace. Any additional questions (like the shift question) were usually taken from established sources (such as the Canadian Community Health Survey) so that the results will be comparable to published data/studies. The questions regarding behaviour based safety programs were taken from the Nordic Occupational Safety Climate Questionnaire (NOSACQ-50) (http://www.arbejdsmiljoforskning.dk/en/publikationer/spoergeskemaer/nosacq-50).
To test for possible associations between psychosocial risk factors and symptoms, a correlation matrix was constructed to identify those risk variables that have statistically significant associations with symptoms. From this matrix we select the top risk factors associated with the sum of all the symptoms (as measured by the coefficient of determination (r2)). These top risk factors are then presented as the main issues for the H&S reps to work on. The correlation matrix is also a part of the spreadsheet analysis tool. This list of risk factors for further attention is based on an internal comparison of only the respondents’ data and thus, does not rely on the comparison with the Danish reference data for this selection.
For large data sets we have performed additional multi-level regression analyses to check the performance of the spreadsheet in identifying the top three issues. So far the performance of the spreadsheet tool has been reasonable but not perfect. There are interactions between risk factors which are not accounted for by the bivariate statistical calculations in the spreadsheet which the more sophisticated multi-level regression analysis is able to detect and account for.