Effect of a protected sleep period on hours slept during extended overnight in-hospital duty hours among medical interns: a randomized trial

Abstract

Context: A 2009 Institute of Medicine report recommended protected sleep periods for medicine trainees on extended overnight shifts, a position reinforced by new Accreditation Council for Graduate Medical Education requirements.

Objective: To evaluate the feasibility and consequences of protected sleep periods during extended duty.

Design, setting, and participants: Randomized controlled trial conducted at the Philadelphia VA Medical Center medical service and Oncology Unit of the Hospital of the University of Pennsylvania (2009-2010). Of the 106 interns and senior medical students who consented, 3 were not scheduled on any study rotations. Among the others, 44 worked at the VA center, 16 at the university hospital, and 43 at both.

Intervention: Twelve 4-week blocks were randomly assigned to either a standard intern schedule (extended duty overnight shifts of up to 30 hours; equivalent to 1200 overnight intern shifts at each site), or a protected sleep period (protected time from 12:30 AM to 5:30 AM with handover of work cell phone; equivalent to 1200 overnight intern shifts at each site). Participants were asked to wear wrist actigraphs and complete sleep diaries.

Main outcome measures: Primary outcome was hours slept during the protected period on extended duty overnight shifts. Secondary outcome measures included hours slept during a 24-hour period (noon to noon) by day of call cycle and Karolinska sleepiness scale.

Results: For 98.3% of on-call nights, cell phones were signed out as designed. At the VA center, participants with protected sleep had a mean 2.86 hours (95% CI, 2.57-3.10 hours) of sleep vs 1.98 hours (95% CI, 1.68-2.28 hours) among those who did not have protected hours of sleep (P < .001). At the university hospital, participants with protected sleep had a mean 3.04 hours (95% CI, 2.77-3.45 hours) of sleep vs 2.04 hours (95% CI, 1.79-2.24) among those who did not have protected sleep (P < .001). Participants with protected sleep were significantly less likely to have call nights with no sleep: 5.8% (95% CI, 3.0%-8.5%) vs 18.6% (95% CI, 13.9%-23.2%) at the VA center (P < .001) and 5.9% (95% CI, 3.1%-8.7%) vs 14.2% (95% CI, 9.9%-18.4%) at the university hospital (P = .001). Participants felt less sleepy after on-call nights in the intervention group, with Karolinska sleepiness scale scores of 6.65 (95% CI, 6.35-6.97) vs 7.10 (95% CI, 6.85-7.33; P = .01) at the VA center and 5.91 (95% CI, 5.64-6.16) vs 6.79 (95% CI, 6.57-7.04; P < .001) at the university hospital.

Conclusions: For internal medicine services at 2 hospitals, implementation of a protected sleep period while on call resulted in an increase in overnight sleep duration and improved alertness the next morning.

Trial registration: clinicaltrials.gov Identifier: NCT00874510.

Figure 1

Flow Diagram at Participating Institutions Interns may have been randomized to either study group from one rotation to another. Subinterns represent fourth-year medical students.

Figure 2

Total Continuous Time Awake Among Participants Hours awake include the last wake-up before an on-call shift to the first sleep during or after an on-call shift. The unit of this analysis is the awakening period that starts before the precall and ends during or after the postcall. The measurement is duration in hours of these awakening periods. Error bars indicate 95% CIs.

Figure 3

Effects of Intervention on Hours Slept During On-call Hours slept during on-call shifts are measured as the sum of sleep time in hours during a 24-hour period from midnight to midnight. We report the robust CIs that accounted for the repeated measures using generalized estimating equations. The figure shows the mean effect of the intervention on sleep for each of the subgroups for which a positive number indicates that participants slept more during intervention months than during the control months. The CIs for number of patients admitted and for the number of patients for whom the interns and medical students (subinterns; fourth-year students) were responsible were calculated using the method suggested by Agresti and Caffo of adding 4 pseudo observations to the intervention and control groups, half with no sleep and half with sleep. The Agresti and Caffo study show that the 95% CIs resulting from inverting large sample Wald tests often have lower coverage probability than the intended 95% and that adding the 4 pseudo observations before calculating the CIs generally brings the coverage probability close to 95%.

Figure 4

Percentage of On-call Days Without Sleep Hours slept during on-call shifts are measured as the sum of sleep time in hours during a 24-hour period from midnight to midnight. We report the robust CIs that accounted for the repeated measures using generalized estimating equations. The proportion of on-call days without sleep for which a reduction in this proportion is seen as beneficial. For calculating CIs, see the Figure 3 legend. Subinterns represent fourth-year medical students.