Safety performance measurement significance

In the second edition of this book, I wrote that a renewed interest had developed in having measurement systems that effectively assess occupational safety performance, measurements that are universally applicable. Preferably, those measures would not only be historical but also predictive and serve as a base from which to prioritize future safety efforts. A significant goal was to have those measures communicate well in terms that managements understand. This renewed interest in performance measures arose out of the increased desire of some environmental, safety, and health professionals to move the profession forward by being able to establish more definitively the value of their work in relation to organizational goals. 

This figure provides an indication of the significance of the safety performance measure. Obviously, if the percentage unseen is very high, less reliance can be placed on the safety performance measure than when the percentage unseen is very low. 

But using time as the measure does of course become a big problem should the year be reached and the target not be achieved. Despite making significant improvements in overall safety performance, Balfour Beatty did not achieve Zero Harm by 2012, and indeed have since revised its programme, omitting any timescale from its strapline. 

Even for the large organization with significant annual hours worked, in addition to historical data, hazard-specific and qualitative performance measures (safety audits, perception surveys, the incident... 

From 1933 to 1999, the death rate dropped 90%. It is recognized that fatality rates are not the only measure of safety performance. Nevertheless, these numbers are significant. 

Moreover, analyses of new entrants are subject to significant caveats. First, they are based on very few accidents—e.g., three fatal accidents among new entrants in the Barnett and Higgins database, five total accidents among new entrants in the data used by Oster et al. In addition, as noted by Rose (1992, pp. 84-85), the results are sensitive to the measures of safety performance employed in the study, and also may depend on the definition of entrant carriers... 

The team that developed Process Safety Research for the 2Century expressed the elusive nature of understanding process safety culture in these words. "The safety culture of an organization has a significant impact on its safety performance. Although the level of safety culture of a work community can be subjectively observed immediately after entering that community, an objective measure is difficult to define. Moreover, establishing effective improvement, if needed is not an easy task... [6]. ... 

Develop objectives to include specific areas of performance that can be measured and/or verified. For example, Improve safety performance in my department next month is too general as an objective. A better objective may be, for example, Reduce first aid cases by 10 percent next month. Even more measurable goals are those objectives where managers and/or supervisors have complete control— for example, Hold 30-minute safety meetings for all employees in my area every Monday morning at 7 00 a.m. As noted previously, objectives must be realistic and attainable and should represent a significant challenge. 

It is probably fair to say that for measures of safety performance that occur less than fifty times a year, the size of the year-to-year variation is so large in percentage terms that it may be difficult in practice to draw meaningful statistical conclusions. Fortunately, the situation is less discouraging if a railroad is observed over multiple years. Consider the twenty-five-percent significance level. A railroad would have an occurrence count above this level by pure chance one year in every four. However, if the railroad falls above the critical value for two years in a row, then the probability that this event will occur purely by chance, and not due to poor safety precautions by the railroad will be 14 or one chance in sixteen. The probability that the railroad would fall above the critical value by pure chance for three years in a row is or year in sixty-four, a very small probability. On this... 

---