X hits on this document





116 / 156

with Table 8.1 giving guidance on choice of test (though these aren’t the only possibilities) and Appendix B illustrating some of those tests. Some cautions appear in the text:

      • o

        When numbers of injuries are less than five, “exact” methods should be used (p. 103)

        • o

          When comparing the change in rates between an intervention and control group, statistical tests should not be carried out on each group separately. Instead, the test should be carried out on the difference in rate changes (pp. 105-107)

        • o

          If groups are being compared and initial injury rates are quite different, adjustment needs to be made, e.g. through multiple regression (p. 107)

    • If no statistical tests were used, and you think they were not needed, select the answer corresponding to this. If no statistical tests were used and you think they should have been, then use one of the following three options: minor deficiencies; inappropriate statistical test, which might modify the conclusions; inappropriate statistical tests, which could substantially affect conclusions.

    • Consider issues of statistical power in Q43 not here.

  • 17.

    Are you confident that measurement methods did not introduce bias to the corresponding financial benefits findings?

  • In the comment box following your response, note the rationale for your selection.

  • This question addresses measurement bias; i.e., systematic error arising from inaccurate measurement (or classification) of subjects on study variable(s).

  • Consider the validity of the measurement method; consider reliability to the extent that could bias findings. Examples:

    • o

      Consider whether blinding was used in the assessment of final outcome(s) and whether it was necessary

      • o

        The introduction of an OHSMS could alter reporting practices, so this potential source of bias should be eliminated by some means

  • Consider the measurement methods independent of the sample issues. For example, in the Nytro et al. study, there was concern that the methods used for implementation might have created a bias in the direction of better compliance because company managers would have an incentive to appear as though they are complying with the law. If we had been told that the self-report method had been verified in a sub-sample through an external audit, we could have given a high rating for the measurement methods, even though the means of selecting the sample (quota sampling; no indication of refusals) could nevertheless have biased the results towards more compliance.

Effectiveness of Occupational Health & Safety Management Systems: A Systematic Review


Document info
Document views690
Page views694
Page last viewedMon Jan 23 21:20:29 UTC 2017