X hits on this document





55 / 156

Table 4.4.1. Summary of Evidence in Mandatory OHSMS Studies

First Author (Year) Dufour (1998)




All Quebec workplaces in manufacturing sectors subject to LSST regulations 1985-88.

LSST (1979)

Economic outcomes: Prevention program (PREVENT) and INFRACTION variables had statistically significant and positive regression coefficients, so appeared to have a positive effect on productivity growth. Implied contributions of both variables on productivity growth 0.007 (at sample means).

Lewchuk (1996)

637 Ontario workplaces (497 who responded to 1991 IAPA survey; 140 others)

Bill 70 (1979)

Final outcomes: Regression coefficient, β = -0.015 for LTI frequency regressed on legislation change. In other words, 18% decrease in LTI (10 yrs post-intervention vs. 4 yrs pre-intervention)

Nytro (1998)

Randomized quota sample, representative of Norwegian firms re: size, industry, location. 1993: n=2092 1996: n=1182

Internal Control (IC) regulation 1992


Implementation: Increase in % of companies that reported having fully implemented IC, 1993-96: 37%. 1993: 8% implemented; 25% in progress; 67% not yet started. 1996: 45% implemented; 36% in progress; 19% not yet started.

Authors: Prevention programs and penalties have reduced the incidence of workplace injuries in manufacturing sector, leading to reduction of direct and indirect costs sufficient to have an enhancing effect on productivity growth. Reviewers: The question remains as to the nature of the OHSMS - IC regulations have the intention of moving firms to OHSMS and the PREVENT variable may be a good indicator of companies actually implementing them. Authors: Bill 70 had a significant effect in the direction of lower frequencies for manufacturing but was not significant for retail sector. The reduction in lost-time accident frequencies was around 18% for manufacturing workplaces. Time variable was significant and positive in both regression analyses, indicating growing accident/illness rates in both sectors. Reviewers: A major confounder is the change in workers’ compensation administration and shift to New Experimental Experience Rating (NEER). Weaknesses in the statistical analysis: potential for selection bias; confounders. Authors: No specific conclusions on implementation (study’s data on other outcomes not reported here due to exclusion at QA stage) Reviewers: Social desirability bias in reporting is possible, but is unlikely to explain all of the change seen here. There do not seem to be alternative explanations for the observed change.

Saksvik (1996)

Torp (2000)

Randomized quota sample, representative of Norwegian firms re: size, industry, location. N=2092

1567 managers and garage workers in 237 garages in 1996

Internal Control (IC) regulation 1992

Internal Control (IC) regulation 1992

Implementation (% respondents report category of change)13: More clear lines of responsibility 25% More/better risk assessment 21% Better documentation 25% New strategic plans 18% Intermediate outcomes (% change)13: HES awareness 39% Final outcomes: Regression results for absenteeism development: T values: IC status 2.94 (p<0.01); Regression results for accident development: T values: IC status 1.66 (p >0.05);

Intermediate outcomes: IC status significantly correlated with 7 of 9 outcomes, indicating IC had positive effects. Final outcomes: Employees at garages with higher IC states reported significantly fewer musculoskeletal symptoms (p<0.01). No significant relationship between IC measures and sick leave. Standardized regression coefficients for (1) internal control index; and (2) internal control status rated by manager: Musculoskeletal symptoms: -0.026;

  • -

    0.076**. Sick leave in last 30 days:

  • -

    0.048; -0.013.

Authors: The most frequently reported changes in HES practices after introduction of IC regulations were as follows: increased HES awareness, clearer lines of responsibility, improved risk assessment and documentation, and new strategic plans. IC status contributed significantly to explaining the variance in development of absenteeism 1990- 1992, but not for accidents. Two regression models were able to explain only a small part of the total variance in absenteeism and accidents. The variables that contributed most to the models were level of absenteeism and accidents prior to implementation. Reviewers: Study examined a stratified random quota sample, but response rate not reported. Possible recall bias for final outcomes (absenteeism and accident rates). These were based for the most part on administrative archival data, but some workplaces presented recalled self- report data. Authors: Systematic HES activities positively and significantly correlated with following outcomes, measured at individual level: satisfaction with HES activities, physical working environment, social support, HES-related management support, health-related support, and control and workers’ participation in activities related to occupational health. Reviewers: Some sources of potential bias noted, e.g., selection bias, (low response rate); social desirability bias (exclusive use of self-report measures). Conclusions about causality not possible because of cross sectional design.

13 These values are derived from the values given in the report for the sub-sample that answered these questions (n=911) by accounting for the size of the sub-sample that responded divided by the size of the entire sample (the non-respondents were those who hadn’t started implementing IC at all)

Institute for Work & Health


Document info
Document views578
Page views582
Page last viewedTue Jan 17 16:15:00 UTC 2017