Desktop version

Home arrow Management

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Fatigue and Safety – Cause or Risk Factor?

This chapter has looked at factors that both shape and impair performance. While individual differences can undoubtedly affect the efficacy of our interventions, the question we want to answer, of course, is what effect does stress and fatigue have on safety. Fatigue as a causal factor has been demonstrated in some contexts, such as road traffic accidents, but the same relationship in aviation has been more difficult to establish (Williamson et al., 2011). The authors of the CASA review of aviation fatigue rules (CASA, 2018) make the point that there are simply too few accidents at the extremes of duty periods to draw sound conclusions and, in any case, fatigue is likely to be just one contributory factor. Thomas and Ferguson (2010) suggest that fatigue effects are masked because the crew protect the safety of the operation by modifying their behaviour.

We saw earlier that relationships can be found between hours of sleep and wakefulness before duty and subjective measures such as the S-PS and KSS. Working with LOSA data, Thomas and Ferguson (2010) looked at hours awake before a duty, hours of sleep, error rates and error management. They identified a threshold value of 6 hours of sleep before a duty below which error rates increased, and error management was degraded. No effect was found for hours of wakefulness. However, a second LOSA study (Clockwork, 2014) found a significant relationship for both sleep and wakefulness. If both the crews had been awake for more than 8 hours before the start of the duty then error rates were higher, and if one or more crew members had less than 6 hours of sleep before a duty, error management was impaired. Unfortunately, increased error rates do not necessarily mean reduced levels of safety. It is not possible to predict exactly what errors will be made or how significant they might be to the operation. In another LOSA study, the crews with the highest S-PS scores committed least errors, the difference being that these were daytime operations. The relationship between fatigue and error is complicated.

The effect of hours of wakefulness was evident in a 2017 sample of cargo flights. The aim of the study was to investigate fatigue effects in two-sector operations through the night. The crew would be operating in the WOCL, which is the phase of the circadian cycle when the body is really supposed to be asleep. This circadian effect has been observed in aircraft maintenance where Hobbs et al. (2010) found that skill-based errors (using Rasmussen’s SRK taxonomy) were significantly more frequent during the early hours of the morning. In a study of Brazilian pilots (de Mello et al., 2008), error rates were found to increase by 50% for the early morning period.

The bulk of the observations of the cargo crews were done on triangular patterns. The first sector was operated by a crew who took the aircraft from home base to a major cargo airport. They then handed the aircraft over to a second crew who operated two sectors through the night, back to home base. Data were collected on all three sectors. In addition to the normal LOSA data, the crew were asked to provide information about sleep, hours awake before the duty and to complete the S-PS and KSS. Using the LOSA data from a previous, mainly daytime, study as a benchmark, I compared the error rates for the two-sector WOCL flights. I need to add a caveat here: there is no such thing as a ‘normal’ error rate, and we will explore error in more detail in Chapter 6. The benchmark error rate across all aircraft types was 6.25/sector. The error rate for the cargo crews operating through the night was 10.6/ sector. However, the error rate for the crew on the initial sector was also above 10 and yet this sector, technically, was not through the WOCL. The explanation seems to lie in hours of wakefulness. The WOCL sector crews had, on an average, been awake for 2.5 hours before starting the duty. The crews operating the first sector had been awake, on an average, for 7.5 hours before the duty: a value very close to that found in the 2013 study to be associated with increased error rates. Hours awake before a duty, then, is a predictor of performance, and its effect is comparable to that of operating through the WOCL.

It seems, then, that there is a relationship between the two sleep parameters (prior sleep and wakefulness) and performance (error rate and error response), but that is still not an incontrovertible proof of a threat to safety. In Chapter 2,1 discussed the concept of risk, and so it might be useful to view fatigue through a safety management system (SMS) prism. An SMS risk matrix compares the probability of an event against its severity to arrive at a ‘risk level’. De Matos Alceu (2015) looked at the distribution of fatigue risk using S-PS data. The S-PS intervals were mapped onto the ‘severity’ scale of a risk matrix while the frequency of S-PS scores in a sample of flights was mapped onto the ‘likelihood’ scale. A problem with this approach is that there is no universal agreement on the grading of risk and severity. However, a risk assessment of night-time operations into Queenstown Airport in New Zealand’s South Island recommended that crews with an S-PS of 4 or more should not be permitted to fly the approach (Navigatus Consulting, 2017). The risk was too high. Figure 4.10 illustrates the risk distribution of SPS scores for cargo flights. The crew were sampled before flight (BF1 and BF2) and after landing (AF1 and AF2) across their two-sector pattern. The rate of scores of ‘4’ was normalised, and then the percentage of scores above and below ‘4’ were calculated. The data for the daytime B-777 regional flights (SI and S2) are included for reference. We can see that the fatigue scores for the daytime regional flights suggest a low-risk profile. However, for the two-sector operation through the night, the risk profile shifts progressively through medium towards high risk. In an SMS, the output from the matrix is an instruction to management to take appropriate action in response to the specific risk.

Distribution of S-PS scores

FIGURE 4.10 Distribution of S-PS scores.

Whereas hours of sleep and wakefulness are direct measures, the other fatigue scales we have looked at in this chapter are indirect measures. A pilot’s selfperception of fatigue and its subsequent manifestation in terms of performance is complicated. It has been suggested that fatigued pilots fly more conservatively, but commercial aircraft are usually operated by a crew. Therefore, the fatigue status of the individual crew members might not be an adequate measure of the risk given that the performance reflects the inputs of two or more crew members. This final example illustrates the problem.

In November 2018, the crew of a B-747 made two attempts to get airborne for a night-time departure without first selecting take-off flap (personal communication). After their second attempt, they realised their error and finally got safely away. The story starts at pushback. The captain was waiting for the ground engineer to indicate that pushback was complete, that he had disconnected his headset and that the crew were clear to taxi. There had been some discussion, generally, among pilots that this process seemed to be taking longer and longer. Once the engineer had disconnected, the next step in the process was for the take-off flap to be selected. As usual, on this evening, the engineer was taking his time, but once he had he finished, the captain turned back to the task in hand. Under normal circumstances, the pilot monitoring, in this case the FO, would have had his hand on the flap selector waiting for the instruction to select the flap. On this occasion, for whatever reason, the FO had moved his hand and a visual cue that would normally act as a place marker in the procedure, triggering the captain’s next call, was missing. The checklist was continued but without that crucial step. There was one more opportunity to check whether the flaps had been set, which was when the crew verified that the take-off checklist had been completed. This step has some discretion as to when it is undertaken. A new requirement had been introduced to verify the correct runway for take-off, and so the task was usually delayed until the crew had arrived at the nominated runway entry point. On this occasion, because traffic was light, АТС was gently pushing the crew along and had already given them clearance to take-off The crew turned onto the runway and brought up the power, having omitted to finish the take-off checklist. When they pushed the TOGA buttons, they got a warning. They did a slow-speed rejection of the take-off, turned off at the next intersection and come back to start again. As they were turning off the runway, while investigating the reason for the warning, the crew did a quick check that autothrottles had been selected. As the relevant switch was in the correct position, they assumed that it must be a TOGA switch malfunction. The crew knew' that manual selection of thrust for take-off was permissible so the apparent malfunction w'as no impediment, and they did not need to call an engineer. Cleared for take-off a second time, they brought up the power and immediately got a ‘CONFIG FLAP’ warning. The problem was now' clear. We w'ill look at error in a later chapter but, technically, the crew made two lapses - or forgetting to do something - and one mistake. They forgot to select flap in the first place, and then they forgot to complete the checklist. They then misdiagnosed the reason w'hy the TOGA buttons did not work: that w'as the mistake. Both crews had positioned to operate the day before so neither was fully acclimatised. The time zones crossed by each pilot prior to their duty differed, one travelling five zones w'est (supposedly the more beneficial direction) while the other travelled eight zones in an easterly direction.

One pilot had managed to sleep until about 3 hours before the reporting time while the other pilot had woken at his normal time as though he were at home. He had been awake for more than 14 hours when the event happened.

We cannot say that fatigue ‘caused’ this crew to perform well below their usual, professional standards. What we can argue is that fatigue is a risk factor: the presence of a fatigued state - and we have seen more than once in this chapter that hours awake before a duty correlates with error - increases the risk that something contrary to expectations will happen. We cannot predict what error will be committed, just that the performance of both individuals and crews as a unit is likely to be more erroneous when fatigued.

 
<<   CONTENTS   >>

Related topics