MSF E Is the Best Choice for Assessing Chronotype in Shift Workers

Although small in its effect, mid-sleep on free days did significantly differ between the three shifts. As the timing of sleep on evening shifts was very similar to that on free days (see Figure 3.9.), mid-sleep on free days following the evening shift (MSFE) represents the best-suited parameter for the quantification of chronotype in rotating shift workers, as it should, if at all, be only minimally affected by the sleep-wake behaviour on evening shifts. The slight delay in sleep onset and the slight reduction in sleep duration on evening shifts, compared to free days, suggest that some shift workers do recover from sleep debt on free days following evening shifts. MSFE should therefore be corrected for sleep debt, in order to control for inter-individual differences in sleep recovery on free days.

MSFscn (sleep corrected mid-sleep on free days) for shift workers was computed as indicated in Figure 3.5., in the same way as for day workers but by means of obtained parameters from the evening shift (MSFE_{; SD}E_{_w; SD}E_{_f; ØSD}E_{). After exclusion of those} subjects waking-up to an alarm clock on free days following an evening shift, MSFE values from 238 rotating shift workers remained. The sleep correction resulted in a mean MSFEscn of 4.23 (SD = 1.47), corresponding to a 20 minute advance compared to MSFE (see Figure 3.10. A). The distribution of MSFEscn is very similar to the distribution of MSFscn in day workers (M = 4.45, SD = 1.40, see Figure 3.10. A.). To test the hypothesis that shift workers do not differ from day workers, an independent t-test was computed, whereby a more stringent decision criterion was set, α = 0.2. Results from the Kolmogorov-Smirnov showed a violation of normality in MSFEscn. Results from the t-test showed a significant difference between MSFscn in shift and day workers: t (70294) = 2.51, p ≤ .2, g = -.11, 1-β = .80. Results from the Mann-Whitney test confirmed these results: z = - 2.17, p ≤ .2, g = - .11, 1 - β = .80.

To explore the possibility that the MSF differences between shift workers and day workers were due to a systematic difference in the local time of sunrise-- the longitude of the study sites (Roenneberg et al., 2007)-- the largest shift work sample was examined for further analysis. This was a sample of 106 shift workers from Cham (12.66 degrees of longitude), with an average MSFE_{scn of 4.18, (SD = 1.57). MSF}E_{scn values of this sample were then} transformed to match the average degree of longitude of the large MCTQ database of 82,000 day workers, being 9.18 degrees of longitude (see Figure 3.10. C and D). This transformation was achieved by adding 13.92 minutes to each MSFEscn score. After matching the two samples, the shift worker distribution delayed to an average MSFEscn of 4.41 (SD = 1.57). Results from the independent t-test showed no difference in MSFscn between shift and day workers: t (105.35) = - .11, p >.2, g = .03, 1-β = .30. Results from the Whitney Mann test confirmed these results: z = - .20, p >.2, g = .03, 1-β = .30. As such, when corrected for sleep debt, the sleep phase of shift workers on free days following the evening shift is comparable to the sleep phase of day workers on free days. Individual differences in MSFEscn can therefore be applied for the assessment of chronotype in rotating shift workers.

Figure 3.10. A. Frequency distribution of MSFE_{scn in 238 rotating shift workers (green), plotted} against the distribution of MSFscn in 70,058 day workers (grey). B. MSF scores are highly associated to the degree of longitude in people living in areas up to 300,000 inhabitants in Germany. C. In order to match the shift work population from Cham (12.66 degrees of longitude) to the large MCTQ databank of >82,000 day workers, MSFEscn scores were transformed so as to match the average degrees of longitude in day workers (9.18 degrees of longitude). D. After standardizing the degree of longitude, the distribution of MSFEscn is indistinguishable from that of MSFscn in the day worker population (for a statistical analysis, see text). Source for panel B: Roenneberg et al. 2007. Source for panel C: Kantermann et al. 2007.

3.5.4.1 What about shift workers without evening shifts?

For the assessment of chronotype in shift workers without evening shifts, different algorithms will need to be applied. Based on the data obtained by rotating shift workers from the three-shift-model, expected MSFEscn values can be deduced for shift workers without evening shifts. Figure 3.11. presents the correlations between MSFEscn and MSF following the morning shift (MSFM), and between MSFEscn and MSF following the night

shift (MSFN) in shift workers from the three-shift-model that reported waking-up without the use of an alarm clock on all free days. For the latter correlation, standardized residuals greater than 3 were removed to exclude values from the second peak obtained by individuals with an inversed sleep cycle. A total of 169 shift workers remained, on the basis of which equations of a straight line were computed by means of the obtained slope and intercept. MSFM and MSFN can be corrected to a perfect fit with MSFEscn by dividing MSF through the respective slope and intercept (see Figure 3.11.). The observed values of the slope and intercept can then be applied for the computation of expected MSFEscn values in shift workers who do not work evening shifts (e.g. permanent morning and night workers). In the case of a shift rotation between the morning and the night shift, MSFM _{appears to be} a better choice, as free days appear to be less influenced by morning shifts than night shifts (see Figure 3.7. B). Expected MSFEscn values can be obtained by means of the following equations:

For Permanent Morning Workers: MSFM_{/ (0.7741- 0.6032)} For Permanent Night Workers: MSFN/ (1.0565 – 0.2984)

Figure 3.11. Correlations between MSFEscn and A. MSF following the morning shift (MSFM) and B. MSF following the night shift (MSFN) in 169 rotating shift workers from the three-shift-model, with “best fit” straight line, governed by the equation (y=mx+b), containing the slope (m) and intercept (b) of the association between variables. MSFM _{and MSF}N _{are adjusted to a perfect fit with} MSFscn, by means of the method of least square MSFM / m-b for the morning shift and MSFN /m-b for the night shift.