Estimating hourly work schedule risk in railway traffic controllers. (July 2022)

• Record Type: Journal Article
• Authors: Roets, Bart; Folkard, Simon
• Is Part Of: Safety science. Volume 151(2022)
• Publication Date: 2022-07
• Subjects: Fatigue -- Risk assessment -- Accidents -- Work schedule -- Workload -- Railway control; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2022.105757
• Languages: English
• ISSNs: 0925-7535

Abstract

Highlights: We introduce an "Hourly Risk Index" to assess fatigue-related work schedule risk. We consider the relative risk of cumulative, time-of-day, and time-on-duty effects. We validate with a large-scale dataset from Belgian railway traffic control centers. The Hourly Risk Index outperforms its predecessor, the Risk Index. Error severity is significantly related to a further increase in risk. Abstract: Around-the-clock staffing in safety-critical environments calls for an accurate risk assessment of work schedules. In this paper we introduce and evaluate an hourly version of the "Risk Index", an established risk model in the UK rail industry. This "Hourly Risk Index" is the first risk-based model that allows the prediction of risk at the hourly level, and is founded on the most recent scientific insights in fatigue-related accident risk. It considers the build-up of risk over consecutive shifts, recovery during days off, quick returns, and time-of-day and time-on-duty effects. Improved understanding of hourly risk is highly relevant for both ex-ante (staff rostering) and ex-post (accident and incident analysis) evaluations. We evaluated the performance of the Hourly Risk Index by analyzing a purpose-built real-world dataset from Belgium's railway traffic control centers (containing close to 8 million round-the-clock working hours in the period 2013–2018). Results not only show that the occurrence of safety-critical errors is significantly associated with higher Hourly Risk Index values, but also that error severity is significantly related to a further increase in risk. In addition, the Hourly Risk Index outperformed the original Risk Index. Our empirical results also emphasize the importance of job demands, by indicating a higher performance of the Hourly Risk Index under conditions of low or high workload. Given these outcomes, it is currently being deployed and tested at Belgian national railways. This should allow the further refinement of the Hourly Risk Index in an iterative manner.