There is evidence that shift work negatively affects employee overall health. It has been shown to increase risk for cardiovascular diseases [1-3], metabolic disease, diabetes, overweight and obesity [4-7] and also depressive symptoms [8,9]. Sleep disorders, such as insomnia, poor sleep quality and daytime sleepiness is more common among shift workers than non-shift workers [10]. Shift and irregular work has acute effects on sleep and alertness, especially if work is done at night and/or the morning. The effects seem to linger and also extend to days off [11]. Drake et al. (2004) reported that employees who met the criteria for shift work sleep disorder had significantly higher rates of ulcers, sleepiness-related accidents, absenteeism and depression than shift workers who did not meet these criteria [12]. In women, shift work, especially at nights, was associated with risk for breast cancer [13]. In a review study, Saksvik et al. (2011) showed that adaptation to shift work was related to characteristics of the individual; for example, young age, male gender, high scores for flexibility and extraversion were association with higher shift work tolerance [14]. Prospective cohort and cross-sectional studies have shown that shift work is also associated with sickness absence and work ability [12,15-18]. Moreover, the results of these studies are conflicting. The research results on the links between different shift work schedules and sickness absences ranging from strong connection [12,15,18-20] and or no links [16,17]. In addition, previous studies have focused more on health care workers [16-17,21-22] to the relative neglect of the industrial sector [ 8,18].

Working time planning is important supporting the wellbeing of employees [20,23-26]. For example Hornberger & Knauth [1995,1998] studies and Neil-Sztramko et al. (2014) review results showed that fast-forward shift rotation has positive effect employee’s health [24-26]. Employees asked about working a fast-forward rotating shift versus a long rotating shift reported that the former work shift pattern had a better effect on sleep, alertness, wellbeing at work and general health [23]. However, studies focusing on the effects of shift rotation on sickness absence are lacking, despite the fact that about 22% of the Finnish workforce are employed on shift work [19].

Shift work in the forest industry has been little investigated, although forest industry work is physically and psycho-socially demanding, as it includes a lot of repetitive loading and weight-bearing, carrying to wood products and workers possibilities to influence the tasks they perform or how they perform them are low. Pahkini et al. (2010) found that varying the work shift pattern was a good way to support employee wellbeing at work in the paper industry. Employees asked about working a fast-forward rotating shift versus a long rotating shift reported that the former work shift pattern had a better effect on sleep, alertness, wellbeing at work and general health [23]. The aim of this study was to investigate sickness absence according to different work shifts models in the forest industry in Finland.

This study was a descriptive cross-sectional investigation of a group of forest industry workers at two locations in Finland. The source population consisted of 636 forest industry employees [timber and wood products workers (n=600) and office workers (n=36)]. The study group comprised 280 employees, of whom 90 were women (32%) and 189 men 67%, (Table 1). Average work experience was 18.1 (SD 2.3) years. The research project of which this study forms part has been described in an earlier article on the educational background and workload of forest industry employees. The educational level of the present participants was mainly vocational school or basic education [27]. Workload was assessed subjectively by an occupational physical activity questionnaire [28]. The three most common workload levels were “medium heavy work” (5.0 MET=Metabolic Equivalent), “physically light standing work or light work involving movement” (3.5 MET) and “other sedentary work” (2.5 MET) [28].

Table 1: Distribution for each variable, N (% or min-max).

During the of 6.5-year study period January 2006–June 2012sickness absence was ≤1.5% (of working time) for half of the employee sample (n=140; cases), and was >1.5% for the other half (n=140; controls). The sample (N=280) was randomly selected from the source population [Figure 1]. The study was approved by the ethical committee of the University of Jyväskylä, Finland (12.11.2012).

Figure 1: Study design and study subjects.

Data were gathered from sick-leave registers. The employer`s registers gave detailed information on hours worked, overtime hours, sickness absence, accidents at work and other reasons for absence. There were five different work shift schedules [Table 2]. The tasks (forest industry factory work) were similar across the different shift schedules, except for the day shift (TAM15), which also included administrative tasks.

Table 2: Five different shift work schedules (in Finnish “TyöAikaMuoto” TAM).

The register shows that many employees had worked several of these different schedules during the study period (6.5 years). Tables 3a and 3b presents the numbers of employees working the different shift work schedules.

Table 3: a) Shift work schedules number and percentage of employeesworking those schedules; b) Number and percentage of shift workschedules worked and by how many employees.

On every shift at least one person had no sickness absence hours while others had sickness absence that lasted almost the whole shift. Owing to these exceptional values, it was considered more useful to compare the sickness absence rates between the different shift work schedules using median rather than average values. Half of the employees were below and half above the median value. The median was highest in the three-shift schedule (TAM35) and lowest in the two-shift schedule (TAM25) [Table 4]. In the case group, in both the day-shift (TAM15) and three-shift six days a week (TAM36) schedules, the median was zero or almost zero.

Table 4: Statistics for sickness absence rate (%) in different work shiftschedules separately for cases and controls and for all (cases andcontrols together).

Statistical Analysis

The sickness absence rate was calculated separately for each shift work schedule and schedules were compared using independent and paired samples t-test. Statistical analyses were performed using SPSS 22.00.

The total number of working hours differed between the different shift work schedules, which mean that simple comparison is not possible. Therefore, sickness absence rates were constructed.

Sickness

100* Sickness absence hours

Absence rate

Sickness absence+workaccident+anything other reason for sickness absence+hours worked

In this study, some employees had worked all five different shift work schedules (Table 2). Table 3 shows the number of employees in each schedule (left) and the number of different schedules worked by employee over the 6.5 year period. The most frequently worked schedule was the three-shift rotation, five days a week (TAM35) n=178 (24.6%). Many employees had worked in several different shifts over the study period; only 45 had worked the same shift and nine had worked in all five schedules.

As only nine employees had worked all five schedules, analysis of variance for repeated measures cannot be applied, and as most of the employees had worked several different schedules shifts, the observations are not independent and therefore analysis of variance cannot be used in comparing the sickness rates across schedules. Instead t-tests are used. We compared every ten pairs of schedules using t-tests. In each pair of schedules, i.e., shift1 and shift2, we separate three groups of people: G1.2 comprises those who have worked both shift1 and shift2; group G1 comprises those who have worked shift1 but not 2; and group G2 comprises those who have worked shift2 but not in shift1. We can then compare the sickness rates in shift1 and shift2 in group G1.2 using paired samples t-test and groups G1 and G2 using independent samples t-test. We also compared group G1.2 with groups G1 and G2 to see if these employees differed in their sickness absence rates when working both two schedules and when working only one of them.

Because we used a series of tests, we used the Bonferroni method and a significance level of P=0.005, but results are also reported for P<0.05.

The results are presented for three different groups. First, the results are reported for the whole study group (cases and controls together); second, the results are reported by gender (women and men); and third, the results are reported by age group (≤ 45 and ≥ 46).

Sickness absence in the whole study group

The variation in sickness absence rates is highest for the day-shift (TAM15) in both groups, but especially in the control group (see standard deviations table 4 and Figures 2a and 2b).

Figure 2a: Confidence intervals for the mean sickness absence rate in the different work shift schedules for cases and controls combined.

Figure 2b: Confidence intervals for the median sickness absence rate in the different work shift schedules for cases and controls combined.

Among all the employees (N=280) working in the different schedules, those working the three-shift rotation five days a week (TAM35) show a statistically significantly higher sickness absence rate. For example, of all employees who worked the two-shift (TAM25) and three-shift (TAM35) rotations, those who worked the two-shift (TAM25) but not the three-shift (TAM35) rotation had a lower average sickness absence rate than those who worked the three-shift (TAM35) but not the two-shift (TAM25) rotation (independent samples t-test, P=0.001). Those who worked only the two-shift TAM25 rotation had lower sickness absence rate than those who worked both shifts (P=0.002). The statistically significant differences between the different pairs of shifts are presented in more detail in table 5.

Table 5: Comparisons of the work shift schedules for all employees combined, for women and men separately, and for the two age groups separately.

Sickness absence among female and male

In this study, women had more sickness absence than men [Figure 3a and 3b].

Figure 3a: Confidence intervals for the mean sickness absence rates of men and women in the different work shift schedules for cases and controls combined.

Figure 3b: Confidence intervals for the median sickness absence rates of men and women in the different work shift schedules for cases and controls combined.

All the comparisons were made also for women and men separately. For women, sickness absence was higher among those women who only worked the three-shift (TAM35) rotation than for those who worked the day shift (TAM15, P<0.001) or two-shift (TAM25, P<0.001) but three-shift (TAM35) rotation. Women who only worked the two-shift (TAM25) rotation had a lower sickness absence rate than those who worked both the two-shift (TAM25) and three-shift (TAM35, P=0.001) rotations. Women who worked both the three-shift rotation five days a week (TAM35) and three-shift rotation seven days a week (TAM37) rotations had a lower sickness absence rate in the three-shift rotation seven days a week (TAM37, P=0.002). Women who only worked the three-shift rotation seven days a week (TAM37) also had a lower sickness absence rate than those who worked in both the three-shift rotation six days a week (TAM36) and three-shift rotation seven days a week (TAM37, P=0.003). It seems that working a three-shift rotation five days a week (TAM35) increases the sickness absence rate, especially among women.

For men the only (almost) significant difference was that those who did not work the day-shift (TAM15) but only the two-shift (TAM25) rotation had a lower sickness absence rate than those who worked the day-shift (TAM15) but not the two-shift (TAM25, P=0.006) rotation.

Sickness absence in the younger (≤ 45) and older (≥ 46) age groups

Of comparison by age group are shown in Figure 4a and 4b). For the age group ≤ 45, the only difference was that those who worked the two-shift (TAM25) rotation but not three-shift rotation five days a week (TAM35) had lower sickness absence rate than those who worked the three-shift (TAM35) but not two-shift (TAM25, P=0.001) rotations. For the age group ≥ 46 the only significant difference was that when working both the three-shift rotation five days a week (TAM35) and the three-shift rotation seven days a week (TAM37), the sickness absence rate was higher for the three-shift rotation five days a week (TAM35, P=0.002). Some almost significant differences were observed: for example, those who worked the day-shift (TAM15) but not the three-shift rotation five days a week (TAM35) had a lower sickness absence rate than those who worked the three-shift (TAM35) rotation but not the day shift (TAM15, P=0.006). Again, working the three-shift rotation five days a week (TAM35) was associated with a higher sickness absence rate.

Figure 4a: Confidence intervals for the mean sickness absence rates by age group and work shift schedule, cases and controls together.

Figure 4b: Confidence intervals for the median sickness absence rates by age group and work shift schedule, cases and controls together.

Table 5 presents the results of the comparisons of the different age groups for women and men separately. The results are consistent with those reported above. The sample sizes of the groups are rather small so they are not reported here in any further detail.

Women over age 46 had more sickness absence than men in the same age group. A high frequency of absences was observed among women working the three-shift rotation five days a week (TAM35, independent samples t-test, P0.001). Men in the same age group also showed sickness absence when working the day shift (TAM15, independent samples t-test, P0.017). Men under age 45 had more sickness absence than women in the same age group.

Table 5 presents the results of the various comparisons between the shift work schedules in more detail. Comparisons were also made between women and men and by age group (under 45 years, over 46 years).

The main findings of this study were that the work shift schedule associated with the highest sickness absence rate was the three-shift rotation five days week (TAM 35). This result was found for both sexes in both age-groups, in particular women over age 46, and men in both age groups. For both day shift (TAM15) and, in particular, two-shift (TAM25) work, the sickness absence rate was lower. It appeared that, when measured by the sickness absence rate, the day-shift (TAM15) and two-shift (TAM25) schedules best supported the present employees’ health. According to the information received from the employer, employees who found shift work challenging were assigned to the day shift (TAM15).

In earlier studies have focused on the risk factors associated with shift work and the negative links between health and shift work [1-4,7,9], whereas the association between shift work and sickness absence has not been systematically investigated and has remained unclear [8,16-17]. Our results are in line with those of Nakata et al., who reported that shift work was associated with higher sickness absence in the electronics sector among men [8], and with those of Tuchesen et al., who found that working evening may cause long-term sickness absence among employees in elderly care [17].

The present study supports earlier findings that the so called long work rotation (six or seven nights in a row) is a risk factor for employee health [23-26]. In order to get enough rest to return to work refreshed, a fast-forward rotating shift schedule has been recommended by researchers [26] and by the Finnish Occupational Health Institution [29]. Fast-forward rotation best supports the normal state of alertness and physiology in humans [23,29]. Hornberger and Knauth and Pahkini et al. also offer ergonomic recommendations for the design of shift systems. Regularity is also an important feature of an ergonomically designed work shift model [29].

This study showed that the work shift model that least supported physiological circadian rhythms (three-shift rotation five days a week, TAM35) was associated with a higher sickness absence rate (%). It is possible that changing the order of the evening and night shifts would have a positive effect on the sickness absence rate, and hence general health. The resulting shift model would then support normal human physiology [11,26] and the short work cycle (fast rotation of shifts: morning, morning, evening, evening, night, night, free, free) and fast shift model (only three or four night) be experienced positively [23-26]. It is known that work shift regularity is an important feature of an ergonomically designed work shift model [29].

Difficulties were experienced in this study in finding statistically appropriate and reliable methods for detecting the differences between sickness absences across the different shift work schedules, as many of the employees had worked several schedules but only nine employees had worked all five different shifts. Workers’ tasks on the evening and night shift schedules did similar tasks, whereas day shift workers also did some administrative work. Consequently, if we had used variance analysis for repeated measures, a lot of information would have been lost. On the other hand, the fact that some employees had worked many shifts meant that the observations were not independent, and thus analysis of variance would not have yielded correct results. The personal tendency to being absent would also have been ignored. Therefore, we performed a series of tests comparing all the shift work schedules in pairs separately, which allowed us to test for differences between all the different employee groups (those who worked both schedules in the pair and those who worked one but not the other schedules in the pair). In so doing, we were able to use all the information (i.e., all employees) we had. The strength of this study was that the sickness absence follow-up period of 6.5 years was long enough. In other studies [8,16] the follow-up periods has been much shorter, generally 1 or 1.5 years. Thus changes in the implementation of the workplace are easier to analyze, because the study design consists of the same employees in the organization.

Sickness absence covers a wide diversity of factors. The results of this study can partly be explained by reference to the mechanism of physiological circadian rhythms and how different shifts schedules support or impair general health [11,20,26,29]. Even if sickness absence behavior appears to be an individual behavior, it is often observed to reflect the functioning of the entire organization [23]. Therefore, supervisors and occupational health services should support healthy solutions to when work is done. This could mean better and more personalized work shift planning, as along with support of a healthy lifestyle [26]. In a review, Saksvik et al. review reported that young age, male gender, low scores on morningness, and high scores on flexibility and extraversion are related to higher shift work tolerance.

More high-quality research is needed that takes work schedules into account in the investigation of employee health and functioning and sickness absence, including, for example, personal factors such as whether the employee feels that s/he is a morning or an evening person. Shift work workload and the factors that support employee recovery both at work and during leisure time also merit further study.

In this study, the three-shift rotation five days a week (TAM35) showed the strongest association with the sickness absence rate, and can be concluded to least support the human physiological circadian rhythms. The day shift, five days a week (TAM15) and the two-shift rotation five morning shifts and five evening shifts; (TAM25) may better supports employee health, as measured by the sickness absence rate than the other shift models studied.

What is new? The shift work models used in the Finnish wood industry have been little investigated. This study confirms the view that a backward rotating three-shift schedule is associated with a higher sickness absence rate.

We would like to thanks Arja Piirainen, PhD (University of Jyväskylä, Department of Health Sciences) for assistance with the research design work and data collection. Also we thank Arja Piirainen and Ari Heinonen, PhD (University of Jyväskylä, Department of Health Sciences) for commenting on the manuscript.