|
|
|
|
|
|
|
|
NIOSH Safety and Health Topic:Work Schedules: Shift Work and Long Work Hours |
|
|
|
Extended Abstracts from Conference: Simon Folkard and David A. Lombardi * * Liberty Mutual Research Institute for Safety IntroductionThis paper proposes a modelling approach to predicting the effect of prolonged work hours on the incidence of injuries and “accidents”. We have chosen this approach in view of the various methodological difficulties in attempting to examine the impact of long work hours on health measures. These problems are illustrated with reference to a survey we conducted of over 2000 UK aircraft maintenance engineers in which measures of both the “normal” hours worked per week (including overtime) and various health problems were obtained (Folkard 2003). For example, as hypothesized the frequency of reports of minor infections increased with increasing work hours. However, the frequency of cardiovascular symptoms decreased which suggests a possible self-selection bias of the fittest workers into longer work hours, i.e. a healthy worker effect. In addition, our data indicated that age is negatively correlated with normal work hours, such that on average, older workers work fewer hours per week. Thus any relationship between work hours and health measures is likely confounded by age which would have to be controlled for. Finally, many health outcomes are “chronic” requiring in some cases long induction and latency times . Thus ideally, any examination of the impact of long hours on health needs to control for both age and years of experience of long work hours, which typically are highly correlated. In the light of these potential problems with chronic health measures, we concentrate in this paper on acute “accident” and injury measures (described further as incidents) that have a clear time of onset that can be related to specific features of work schedules. We describe our initial attempt to develop a statistical model based directly on the trends in the relative risk of incidents. In the vast majority of cases the incidents on which these trends are based were not severe, but it is likely that they may represent a relatively direct measure of the occurrence of mistakes and omissions. Unfortunately, however, many available studies of industrial accidents and injuries do not allow for an unbiased calculation of relative risk estimates due to non-homogeneous a priori risk (see Folkard and Tucker 2003, Folkard and Lombardi 2004 for further details). For example, the number of individuals at work or level of supervision may not be constant over the 24-hour day or the nature of the job and the associated tasks being performed may vary considerably across the 24-hour day. Additionally, longer, and possibly safer production runs may often be performed on the night shift. However, many studies have either avoided or attempted to correct for this problem, such that any residual variation in risk could reasonably be assumed to reflect on the state of the individuals concerned. In earlier papers we have established that there are four reasonably consistent trends in accidents and injuries associated with features of work schedules ( Folkard and Tucker 2003, Folkard and Åkerstedt 2004). Summary of the trends in riskThe first trend concerns the impact of different shift lengths on relative risk. Studies that have interpolated performance measures have typically found a deterioration in performance and alertness on 12-hour shifts compared to that on 8-hour ones (e.g. Rosa 1991). Four studies have recently reported a trend in the risk of incidents over successive hours on shift and have managed to correct for exposure in some manner. These studies were reviewed in detail by Nachreiner (2000), and Folkard and Lombardi (2004). However one of the trends, namely that of Folkard (1997), statistically combined several relatively small studies and made various assumptions in deriving an overall trend. Since the remaining three studies were based on substantial numbers of incidents and report fairly similar trends to that derived by Folkard (1997) the latter was omitted from consideration in deriving an average trend. The three remaining studies examined trends in national accident statistics and corrected for “exposure” on the basis of time budget studies of relatively large samples of the population under consideration. In order to combine these three trends into a single averaged trend, the mean risk for the first eight hours was set to one and then hourly relative risk values were calculated for each study. These hourly relative risk values were averaged to produce a combined trend. Apart from a slightly heightened relative risk from the second to fifth hour (see Folkard 1997; Tucker et al 2000 for a discussion of this), risk increased in an approximately exponential fashion with time on shift such that it was more than doubled in the twelfth hour relative to the average for the first eight hours. Using this combined trend, the relative risk of shifts of different lengths was estimated by differentiating it. Note that the risk of an eight-hour shift was set to one as described above. This differentiated combined curve estimated that relative to eight hour shifts, ten hour shifts were associated with a 13.0% increased risk and twelve hour shifts exhibited a 27.5% increased risk. The second trend relates to the relative risk of incidents on the morning, afternoon and night shifts on 8-hour shift systems. There are five studies that are based on relatively large numbers of incidents that appear to have overcome the potential confounders. In these studies, incidence rates were reported separately for the morning, afternoon and night shifts (see Folkard and Åkerstedt 2004 for details of these studies). In addition, three of these studies report two separate sets of data, for different areas or types of incident, giving a total of eight data sets across the three shifts. While some of the studies give no precise details of the shift system in use, many of them involved a total of only four days on each shift before a span of rest days. Based on the pooled frequencies across these eight data sets, risk increased in an approximately linear fashion, from morning to night, with an increased risk of 18.3% on the afternoon shift, and of 30.4% on the night shift, relative to that on the morning shift (see Folkard and Lombardi 2004 for more details of this analysis). This finding suggests that when the a priori risk appears to be homogeneous across the three shifts, there is a tendency for the relative risk of incidents to be higher on the afternoon shift than on the morning shift, and highest on the night shift. The third trend relates to relative risk of incidents over successive night shifts. The authors are aware of seven published studies that have reported incident frequencies separately for each night over a span of at least four successive night shifts (see Folkard and Åkerstedt 2004 for details of these studies). Based on the pooled frequencies across these seven data sets, incident risk was about 6% higher on the second night, 17% higher on the third night, and 36% higher on the fourth night than on the first night shift. Two important questions arise over this substantial increase in risk over four successive night shifts. The first question is what happens to risk over longer spans of successive night shifts, however there is a paucity of data relating to this (see Folkard and Lombardi 2004). While it is possible that over longer spans of night shifts risk may begin to decrease, there is currently no evidence to indicate that this is actually the case. The second important question is whether the increase in risk over successive shifts is confined to the night shift, or whether it might be general to all shifts and represent an accumulation of fatigue over successive workdays. Of the seven studies that provide evidence on the trend over successive night shifts, five studies also reported the risk over successive morning or day shifts. Based on the pooled frequencies across these five data sets , the risk of an incident was about 2% higher on the second morning/day, 7% higher on the third morning/day, and 17% higher on the fourth morning/day shift than on the first shift. There is clear evidence of an increased risk over successive morning/day shifts, but this increase in risk was substantially smaller than that over successive night shifts. Thus, the increase in risk over successive workdays is substantially larger on the night shift than on the morning/day shift. Finally, it should be noted that the trend for hours on duty described above does not control for the influence of breaks during a duty period, and indeed one possible explanation for the decrease in risk after the fifth hour may be that it reflects the influence of rest breaks. A number of laboratory studies on the effects of breaks have been conducted (e.g. Dababneh et al., 2001), but there appears to be only a single, recent study that has examined the impact of rest breaks on the risk of incidents (Tucker et al., 2003). This study examined industrial injuries in an engineering plant in which a 15-minute break was given after each period of two hours of continuous work. The number of injuries within each of the four 30-minute periods between breaks was calculated, and the risk in each 30-minute period was expressed relative to that in the first 30-minute period immediately following the break. The results indicated that risk rose substantially, and approximately linearly, between successive breaks, such that risk had doubled by the last 30-minute period before the next break. There was no evidence in this study that this trend differed for the day and night shifts, or for the three successive periods of two hours of continuous work within a shift. Combining the relative risk estimatesGiven that the various trends discussed above are based upon estimates of the relative risk of incidents, the combined effects of the type of shift, shift length, the number of successive shifts and the interval between breaks can be estimated in a relatively straightforward manner. For simplicity, this is illustrated using a model in which the single effects are assumed to combine in a simple additive manner. However, the use of a multiplicative model would likely result in an essentially similar pattern of results for normal ranges of shifts. The additive model can be expressed simply as: [formula 1] RR S = RR T + CR N + CR L + CR B
The relative risk of the first shift in a span of a given type must be estimated before this model can be applied. This value differs from the estimates given above since those values were based on various spans of shifts and we know that the trend over successive shifts differs depending on whether they are morning/day or night shifts. Five studies were selected that provided trends for both successive morning day and night shifts. These studies were used to estimate the risk on the first of a span of night shifts relative to that on the first of a span of day shifts. It should be noted that the trend in risk over the four successive night shifts from this sub-set of five studies was virtually identical to that described above. If we set the risk on the first day shift as 1.00 then we can calculate the relative risk on the first night shift as being 1.06. Further, rather than express the relative risk on different shifts, spans of shifts, and durations of shifts relative to a single 8-hour day shift it would seem appropriate to use a reference of a span of five successive 8-hour day shifts, i.e. the “normal working week”. Thus, the relative risk on a span of five successive 8-hour day shifts with a single, mid-shift break was set at 1.00, and the relative risks for all other combinations were expressed relative to this. Finally, a linear extrapolation of the trends over successive shifts and the interval between breaks was made in order to estimate the relative risk associated with longer spans of shifts or intervals between breaks. Estimating the risk of long work hoursThis simplistic model can be used to estimate the risk associated with various work schedules, including those that involve long work hours. For example, the European Union’s “Working Time Directive” limits the hours of work per week to 48 hours. If the 48 hours are comprised of six successive 8-hour day shifts, we estimate the associated risk to be only 3% higher than on the “standard” 40-hour week involving five successive 8-hour days. However, if the 48 hours are worked as four successive 12-hour day shifts, we then estimate that the risk increases by 25% over the “standard”. For night shifts, the estimated risk is increased by 25% for six successive 8-hour night shifts, but by 40% for four successive 12-hour night shifts, relative to the “standard” 40-hour week. Similarly, if we consider a 60-hour week, we can model this as six 10-hour day or night shifts, or five 12-hour day or night shifts. In this case the increased risk on six 10-hour shifts, relative to our “standard” 40-hour week involving five successive 8-hour days, is estimated to be 16% for day shifts and 38% for night shifts. In contrast, the increased risk for five 12-hour day shifts is estimated to be 28%, while that for five 12-hour night shifts is 46%. Thus, although it is clear that a single long shift is associated with an increased risk, when the accumulated hours per week are considered we need to take account not only of the length of the individual shifts, but also of the type of shift and the number of successive shifts worked. Finally, we will consider the impact of breaks and illustrate how, at least on the basis of our current estimates, these may mitigate the effects of long work hours. We conclude that the risk of an incident associated with long work hours may be influenced more by the precise work schedule, including the length and timing of the duty periods, and the provision of breaks, than by just the accumulated working hours. ReferencesDababneh AJ, Swanson N., Shell RL. (2001) Impact of added rest breaks on the productivity and well-being of workers. Ergonomics, 44, 164-174. Folkard, S. (1997) Black times: temporal determinants of transport safety. Accident Analyses and Prevention, 29, 417-430. Folkard, S. (2003) Work hours of Aircraft Maintenance Personnel. CAA Report No. 2002/6. http://www.caa.co.uk/publications/publicationdetails.asp?id=628 Folkard, S., Åkerstedt, T. (2004) Trends in the risk of accidents and injuries and their implications for models of fatigue and performance. Aviation, Space and Environmental Medicine. 75 (Section II), A161-A167. Folkard, S., Lombardi, D.A. (2004) Designing safer shift systems. In Designing safer shift systems. In P. Nickel, H. Grzech-Sukalo, K Haenecke & M. Schutte (Eds) Arbeits- und Organistionpsychologie – Impulse aus Oldenburg . Peter Lang: Frankfurt am Main ?? Folkard, S., Tucker, P (2003) Shiftwork, safety and productivity. Occupational Medicine. 53, 95-101. Nachreiner F . (2000) Extended work hours and accident risk. In Marek T, Oginska H, Pokorski J, Costa G, Folkard S, (Eds.) Shiftwork 2000 - implications for science, practice and business. Kraków: Institute of Management , Jagiellonian University , pp 29-44. Rosa , R. (1991) Performance, alertness and sleep after 3.5 years of 12h shifts: a follow-up study. Work and Stress, 5, 107-116. Tucker, P., Sytnik, N., Macdonald, I. , Folkard, S. (2000) Temporal determinants of accident risk: the “2-4 hour shift phenomenon”. In Hornberger S., Knauth P., Costa G. and Folkard S. (Eds) Shiftwork in the 21st Century. pp 99-105. Peter Lang, Frankfurt , Berlin , Bern , Bruxelles , New York , Oxford and Wien. Tucker, P., Folkard, S., Macdonald, I. (2003) Rest breaks reduce accident
risk. Lancet, 361, 680. |
|
|
|||||||||
