Fatigue, Alertness and Risk Prediction for Shift Workers

This report describes the principal outcomes of an Impact Acceleration Account project (grant number EP/I000992/1) between the University of Surrey and Transport for London carried out between Oct. 2019 and Mar. 2020. The aim of the project was to compare the Health and Safety Executive (HSE) Fatigue Risk tool with SAFTE and other more recent models of fatigue, where fatigue here primarily means a reduced ability to function effectively and efficiently as a result of inadequate sleep. We have not sought to discuss the useability of the HSE Fatigue Risk tool or SAFTE since this has been discussed comprehensively elsewhere, we have instead focussed on the fundamental principles underlying the models. All current biomathematical models have limitations and make asumptions that are not always evident from the accompanying documentation. Since full details of the HSE Fatigue Risk tool and the SAFTE model are not publicly available, Sections 1 and 2 give a mathematical description of the equations that we believe underlie each of these models. A comparison of predictions made by our versions of the HSE and SAFTE equations for one particular shift schedule of relevance to the UK and global tunnelling and construction industries is shown in Section 3. In this comparison, we use data collected durings TfL9s Crossrail project by Dragados. Crossrail is the biggest railway infrastructure project in Europe and is one of the largest single investments undertaken in the UK. It is a joint venture between TfL and the UK Department of Transport. Construction started in 2009 and intensive operational testing is expected to take place in 2021. The project included digging 42 kilometres of new rail tunnel under London. Dragados was one of two engineering companies responsible for tunnelling. Essentially, both models give broadly the same message for the schedule we looked at, but the ability to display fatigue as it develops within a shift is a strength of SAFTE. A summary of the strengths and limitations of the use of these kind of scheduling tools is given in the final section, Section 4. Limitations include: (i) Models do not describe fatigue during times when people are not in shift (e.g. driving home). However, they could readily be extended to do so. (ii) Models assume people start well-rested. This is not always a good assumption and can lead to an under-estimate of fatigue. (iii) Most models are currently based on population averages, but there are large individual different. It would be possible to further develop models to include uncertainty in fatigue predictions associated with individual differences. (iv) Few models include the light environment, which is important both to promote short-term alertness and facilitate circadian alignment. (v) Models are not transparent, which makes them hard to independently validate. (vi) It is hard to relate the outputs of current models to measureable outcomes in the field. We also discuss briefly recent developments in mathematical modelling of fatigue and possible future directions. These include (i) Guidance on scheduling and education on sleep and fatigue should be considered at least as important as current biomathematical models. (ii) Only by analysing and integrating high quality individual data on sleep, fatigue, performance, near misses, accidents, actual shift patterns with models can we develop better models and management systems to reduce fatigue and associated risks. Wearables combined with apps present a great opportunity to collect data at scale but need to be used appropriately. (iii) The importance of making time for sleep is not always recognised. Education, early diagnosis of sleep disorders such as sleep apnea, and self-monitoring all have a role to play in reducing fatigue-related risk in the work-place. Section 3 and Section 4 may be understood without reading the intermediate more mathematical sections.