This paper quantitatively motivates the need for active monitoring of occupational safety incident data through the use of cumulative sum (CUSUM) control charts. The frequency of incidents within a subset of historical accident data is analysed. The performance of Poisson CUSUM and exponential CUSUM (time-between-events) charts is compared in an illustrative example to show that shorter periods of aggregation and time-between-events monitoring lead to more timely indications of increased accident frequency. An extension showing the anticipated performance of these charts with real-time data is given. Various adjustments to the monitoring system are also simulated to show that quick implementation of hazard controls can significantly impact safety performance. Decreases in the frequency of safety incidents as a result of implemented hazard controls can also be monitored.
Keywords: accident monitoring; control chart; cumulative sum; safety; surveillance; time-between-events.