Applying Control Charts to Safety Data Detects Trends Earlier

Abstract

Performance monitoring and improvement are key elements of any safety management system. Companies record and review many different health, safety, and environmental (HSE) indicators on a monthly basis with the aim to continuously improve HSE performance. As a result, company management may initiate safety stand-downs and various campaigns based on month-to-month variations seen in key performance indicators (KPIs).Challenges involved with this include whether the changes seen in data represent an actual trend or whether they are just variations around a stable process. Other questions raised include whether the actions taken by management regarding an actual trend prove to be effective over time at countering a negative trend, and did an opportunity exist to identify the onset of a real trend earlier in time to enable a faster counteraction?To answer these challenges and questions, the authors investigated how control charts applied to one company's safety data detected statistically significant trends in a deterministic way. The paper describes how the authors incorporated control chart logic into a reporting engine to enable automated detection of trends and presents the authors' observations based on the results this method provided. The observations include how applying a control chart method reduced unnecessary or erratic safety initiatives due to otherwise incorrectly interpreted safety data.The paper also describes how the method provided a better indication of whether a safety campaign was having a statistically significant impact on safety performance over time and how, if it was not adding value, the campaign was stopped. Additionally, the paper examines how this method has a partially predictive capability as it enables identification of the earliest data point in a time series that could be the start of a new statistically significant trend and how this enables a company to take action before a negative trend can establish itself.