Interns met with study staff on the first day of each four-week rotation for orientation, including to receive instructions on use of an Actiwatch Spectrum, a watchlike device that contains a highly sensitive accelerometer to measure physical motion. Data from the Actiwatches were collected in one-minute epochs and stored in the watch until being downloaded at the end of each rotation.6,8 We assessed the behavioral alertness of study participants by two techniques. They completed a validated three-minute Psychomotor Vigilance Test (PVT)9 each morning they were at the hospital and every night they were on call. The PVT measures alertness on the basis of reaction time to stimuli that are presented at random interstimulus intervals (in this case, two to five seconds).10 We used response speed (reciprocal response time) and the number of lapses (response times ≥ 355 milliseconds) as our primary PVT outcome variables.11 Participants also filled out an electronic sleep log, which contained the Karolinska Sleepiness Scale (KSS),12 a nine-point verbally anchored scale ranging from “extremely alert” (score = 1) to “extremely sleepy–fighting sleep” (score = 9), and questions about patient load and sleep disruptions.
In the last three days of the rotation, research assistants met with the interns to collect the Actiwatch and distribute $25 gift cards for every week of adherence to study measures.
Monthly schedule and randomization
During the study year there were 12 four-week blocks in the schedule. Within six consecutive pairs of blocks, one block was randomly assigned to an intervention schedule or the standard schedule. Participants underwent the assigned condition for all four weeks they were on rotation, with data collection planned from Monday of the first week through Friday of the final week (25 days). Assignment to rotation blocks was independent of this project, and randomization allocations of each four-week period were concealed from participants at the time of con sent. Because of the nature of the inter vention, blinding of the staff could not be maintained.
In the standard schedule months, the Internal Medicine Service at PVAMC comprised four teams, each with either one resident and two interns or one resident, one intern, and one senior student in an intern role. Pairs of interns on a team were on call with a night-float resident every fourth night. On-call interns admitted patients throughout the night, were responsible for cross-coverage until the primary team returned around 7 AM, and generally worked until approximately 1 PM the next day (a 30-hour duty period). The HUP Oncology Service was staffed with the same model, but there were no medical student substitutions for interns. At both sites, the night-float residents were both second- and third-year residents.
During intervention months, the protected nap period was between 00:00 (midnight) and 3:00 AM for one intern and 3:00 and 6:00 AM for the second intern. Interns were assigned to alternate between early and late protected periods over the course of the rotation. They were instructed to give their cell phone/beepers to the covering night-float resident, and encouraged to sleep in a bed in a dark quiet room. Night-float residents were instructed to wake the interns if urgent assistance was required.
To address feasibility, night-float residents were called each morning by study staff to report time of cell phone/pager handoff and retrieval. The primary outcome was sleep time during the protected period as measured by Actiwatch Spectrum wrist activity monitors and supported by self-reported sleep diaries, completed each morning. Secondary outcomes were percentage of on-call nights without sleep, mean sleep time overall when on call, mean sleep amount within the four-day call cycle, number of attentional lapses and mean response speed on the PVT, subjective reports of sleepiness (KSS score), and whether sleep during on-call nights was disturbed. We also examined patient outcomes related to length of stay (LOS), discharge to the medical intensive care unit (MICU), death, and 30-day readmission.
The primary analyses were unadjusted intent-to-treat analyses testing for differences between the intervention and control groups at each of the two sites, analyzing each site as a separate trial and comparing early and late protected sleep shifts with the control group. The unit of analysis was the participant (intern) day. To account for the correlation among a participant’s multiple observations (days), we used Huber–White robust standard errors with the participants as the clusters.13 The similarity of the treatment groups with respect to baseline covariates was assessed by chi-square tests for categorical variables and Student t tests for continuous variables, as appropriate.
Participants had missing Actiwatch data 6.4% (1,736 hours/27,000 hours) of total study time on intervention months and 8.2% (1,835 hours/22,464 hours) on control months at HUP, and 8.0% (1,896 hours/23,664 hours) on intervention months and 7.3% (1,775 hours/24,264 hours) on control months at PVAMC. When Actiwatch data were missing, we used self-reported sleep over the missing time period if available. Self-reported sleep time was highly correlated with Actiwatch-recorded sleep time (Pearson correlation = 0.82), and the mean Actiwatch-recorded sleep time was similar to self-reported sleep time (5.3 hours versus 5.6 hours, respectively). When both Actiwatch and self-report data were missing (<4.3% in any group), we used multiple imputation in a manner that provides valid inferences under the assumption that the data were missing at random.14 We did two sensitivity analyses: using only days that contained no missing data; and assuming that all missing data were sleep or all missing data were no sleep, which provides bounds on the effect of the intervention.
The PVT outcomes (response speed and number of lapses) were analyzed in the mornings on the first day post call. In PVAMC, the nonadherence rate was 36.1% in intervention and 43.6% in control. In HUP, the nonadherence rate was 39.4% in intervention and 47.2% in control. Multiple imputation was used for missing PVT outcomes. We used a variance-stabilizing transformation of the count of number of lapses to the square root of the number of lapses.15
Patient care data came from the HUP Patient File and the VA Patient Treatment File. We selected patients who were newly admitted to the studied services at both VA and HUP. We examined mean LOS; rate of transfer to the MICU; mortality rates and length of time to death on the study floors, in the MICU, and anywhere in the hospital; and readmissions within 30 days of hospital discharge to the same hospitals.
Readmissions were tracked only for patients admitted back to the study services of PVAMC or HUP. We conducted t tests or chi-square tests on the differences in outcomes between control months and intervention months.
On the basis of a conservative estimate of an intern’s standard deviation of on-call sleep per month, we calculated that there was 90% power to detect a 30-minute difference in the average on-call sleep between the intervention and control at each site—the primary study outcome. No interim analyses were planned or conducted. All reported P values are two sided and were not adjusted for multiple comparisons.
One hundred percent of the invitees agreed to participate, for a total of 102, but 8 were not scheduled on the study rotation. Among 36 interns scheduled to rotate at PVAMC only, 92% rotated once (all subinterns only rotated once) and the rest rotated twice. Among 7 interns scheduled to rotate at HUP only, 4 rotated once and the rest rotated twice. Among 49 interns who rotated at both sites, 80% rotated three times or more. Of the total 94 participants, the mean age was 27.8, and 43 (46.7%) were male.
At both PVAMC and HUP, there were no significant differences between the control and intervention arms in terms of baseline characteristics related to interns, the number of patients admitted on call, or number of patients for which interns were primarily responsible (Table 1).
On 97.4% (561 nights/576 nights) of intern on-call nights with handover data, cell phones were signed out to residents as designed. Mean sleep time during the protected period for those assigned to the early nap shift at PVAMC was 2.33 hours in the intervention months compared with 1.90 in the control months (P= .099) (Table 2). Mean nap sleep periods were significantly longer among interns assigned to the intervention arms in the other shifts: PVAMC late shift: 2.40 versus 1.90 hours, P = .036; HUP early shift: 2.40 versus 1.55 hours, P < .0001; HUP late shift: 2.44 versus 1.55 hours, P < .0001. Interns with protected sleep periods were less likely to have on-call nights with no sleep during both shifts at both sites: PVAMC early shift: 6% (7/120) versus 21% (53/257), P < .0001; PVAMC late shift: 10% (13/124) versus 21% (53/257), P = .014; HUP early shift: 12% (17/141) versus 23% (53/232), P = .01; HUP late shift: 11% (16/142) versus 23% (53/232), P = .01. The proportion of interns who reported sleep disturbances was significantly lower in each of the protected sleep periods: PVAMC early shift: 57% versus 89%, P < .0001; PVAMC late shift: 56% versus 89%, P < .0001; HUP early shift: 56% versus 89%, P <.0001; HUP late shift: 54% versus 89%, P < .0001. Mean amount slept across all four days of the call cycle were similar: PVAMC early shift: 6.76 versus 6.47 hours, P = .12; PVAMC late shift: 6.73 versus 6.47 hours, P = .275; HUP early shift: 6.73 versus 6.47 hours, P = .12; HUP late shift: 6.51 versus 6.47 hours, P = .81 (data not shown). The results were qualitatively similar in sensitivity analyses in which we used only days that contained no missing data and assumed all missing data were sleep or all missing data were no sleep.
At the PVAMC, response speed on the PVT was significantly faster after on-call nights in the intervention relative to the control group: PVAMC early shift: 4.15 versus 3.81 per second (s−1), P = .015; PVAMC late shift: 4.11 versus 3.81 s−1, P = .002. The number of lapses of attention was lower, albeit not significantly for the early shift at PVAMC: PVAMC early shift: 3.35 versus 6.43, P= .059; PVAMC late shift: 3.14 versus 6.43, P = .002. Response speeds and lapses did not differ at HUP. There were no differences in subjective sleepiness on the KSS in the control relative to the intervention group (Table 2).
We identified a total of 2,252 patients (2,577 new admissions) to PVAMC Medical Service and 827 patients (998 new admissions) to the Oncology Unit of HUP. As shown in Table 3 there were no differences between intervention and control on any of the patient-level outcomes except that at PVAMC patients cared for by the control group compared with the intervention group had shorter LOS on study floors (4.6 days versus 4.9 days, P = .014) and shorter overall hospital stays (6.7 days versus 7.9 days, P = .012).
In a recent study by our team, we found that giving residents a five-hour protected sleep period produced a significant increase in mean hours slept, a significant decrease in the proportion of interns with no sleep on extended duty overnight shifts, a significant reduction in sleep disturbances, and an increase in behavioral alertness on mornings after overnight shifts.7 Although suggestive that protected sleep periods during an extended shift could be a viable alternative to 16-hour shifts, instituting such a system at our university hospital proved infeasible because the manner in which it was tested required an extra resident and was thereby not personnel neutral. To our knowledge, this study represents the first examination of a personnel-neutral protected sleep period during extended work periods, though a recent pilot study did report beneficial impacts of a 20-minute daytime nap for interns.16 With three-hour nap sleep periods, interns with protected time do get more sleep (albeit only a mean increase of 0.5–0.9 hours compared with 0.9–1.0 hours in the previous study)7 and have similar reductions in the proportion of nights with no sleep or interrupted sleep regardless of whether they had the early or late sleep shift. However, we observe a smaller impact on postcall alertness levels. Notably in both the current and former study,7 interns left the hospital at the regular time. The sleep period did not require them to stay longer to complete their work.
These findings are in line with extensive laboratory and field research on the use of protected sleep periods for fatigue management in a variety of areas.17–20 Naps as brief as 10 minutes and extending up to four hours have been found to improve alertness and performance during prolonged periods of wakefulness relative to no sleep,21,22 if the sleep inertia immediately after the nap is dissipated with behavioral activity and caffeine before starting work.23–25 In this study, interns assigned to three-hour naps had fewer sleep interruptions and thus presumably less sleep inertia to deal with. However, if interns in either arm were awakened by the night-float resident in an emergent situation, one might imagine the activity would compensate for inertia. Although naps cannot prevent the accumulation of sleep debt over many days if sleep is chronically inadequate,26 they do offer an immediate countermeasure to fatigue during periodic cycles of long duty.27
As with the first iteration of this intervention, limitations of this study include its generalizability: to other institutions, other types of services, other specialties, and other fatigue management strategies. Most notably, we conducted the study using interns who were working 30-hour extended duty shifts. One might imagine that residents with more training are more efficient and confident and thus may make better use of potential rest times. Also, in our assessment of patient outcomes we were unable to link outcomes to a particular intern (or resident). Over the course of a hospital stay, patients would have been taken care of by multiple individuals; however, we can conclude that giving the on-call interns protected sleep periods did not worsen outcomes. Finally, we should restate that nearly 40% of the PVT data were missing, not only reinforcing the need for sound imputation, which we did, but also pointing to the difficulty of asking interns to do extra tasks during this very busy time of the day.
It is important to note that the design used in this study of three-hour protected sleep periods was cost- and personnel neutral and appears feasible to integrate into scheduling designs that have trainees working extended (24 + 4) shifts. Moreover, it worked in sites that traditionally have patients of very different levels of acuity. Compared with earlier nap studies,5,6 we also had much higher participation. This is likely because we made the protected sleep periods part of the standard intern schedule, which signaled to interns the important benefits of sleep and empowered interns to hand over pagers and phones to the covering residents. Our work contributes to a growing body of evidence that strategic napping that is based in sleep opportunities protected from nonessential interruptions is an effective and feasible fatigue risk management countermeasure. Strategic napping thereby provides an alternative to mandatory short shifts that create significant discontinuity in both care and education.
Comparative effectiveness research of alternative forms of fatigue management would be helpful in disentangling the debate as to the optimum way of reducing intern/resident fatigue while preserving and enhancing quality of care and education. Studies that test alternatives (including prophylactic napping,27 shorter periods of protected sleep, longer periods of protected sleep, or 16-hour shifts), that are powered to compare patient outcomes, and that also assess fatigue and cognitive alertness will help us to determine a path forward. It is critical that any solutions balance the many important competing considerations around cost, feasibility, and patient care and educational outcomes.
Acknowledgments: The authors wish to thank Ilene Rosen, MD, MSCE, and Karen Warburton, MD, for their assistance in conducting the study.
1. Asch DA, Parker RM. The Libby Zion case. One step forward or two steps backward? N Engl J Med. 1988;318:771–775
3. Ulmer C, Wolman DH, Johns MMH Resident Duty Hours: Enhancing Sleep, Supervision and Safety. 2009 Washington, DC National Academies Press
4. Nasca TJ, Day SH, Amis ES JrACGME Duty Hour Task Force. . The new recommendations on duty hours from the ACGME Task Force. N Engl J Med. 2010;363:e3
5. Richardson GS, Wyatt JK, Sullivan JP, et al. Objective assessment of sleep and alertness in medical house staff and the impact of protected time for sleep. Sleep. 1996;19:718–726
6. Arora V, Dunphy C, Chang VY, Ahmad F, Humphrey HJ, Meltzer D. The effects of on-duty napping on intern sleep time and fatigue. Ann Intern Med. 2006;144:792–798
7. Volpp KG, Shea JA, Small DS, et al. Effect of a protected sleep period on hours slept during extended overnight in-hospital duty hours among medical interns: A randomized trial. JAMA. 2012;308:2208–2217
8. Malmberg B, Kecklund G, Karlson B, Persson R, Flisberg P, Ørbaek P. Sleep and recovery in physicians on night call: A longitudinal field study. BMC Health Serv Res. 2010;10:239
9. Basner M, Mollicone D, Dinges DF. Validity and sensitivity of a brief psychomotor vigilance test (PVT-B) to total and partial sleep deprivation. Acta Astronaut. 2011;69:949–959
10. Basner M, Rubinstein J. Fitness for duty: A 3-minute version of the Psychomotor Vigilance Test predicts fatigue-related declines in luggage-screening performance. J Occup Environ Med. 2011;53:1146–1154
11. Basner M, Dinges DF. Maximizing sensitivity of the psychomotor vigilance test (PVT) to sleep loss. Sleep. 2011;34:581–591
12. Akerstedt T, Gillberg M. Subjective and objective sleepiness in the active individual. Int J Neurosci. 1990;52:29–37
13. Diggle PJ, Heagerty P, Liang K-Y, Zeger SL Analysis of Longitudinal Data. 20022nd ed New York, NY Oxford University Press
14. Rubin D. Multiple Imputation for Nonresponse in Surveys. 1987 New York, NY J. Wiley & Sons
15. Ramsey F, Schafer D The Statistical Sleuth: A Course in Methods of Data Analysis. 20012nd ed Pacific Grove, Calif Duxbury Resource Center
16. Amin MM, Graber M, Ahmad K, et al. The effect of a mid-day nap on the neurocognitive performance of first-year medical residents: A controlled interventional pilot study. Acad Med. 2012;87:1428–1433
17. Akerstedt T, Torsvall L. Napping in shift work. Sleep. 1985;8:105–109
18. Bonnet MH, Arand DL. The use of prophylactic naps and caffeine to maintain performance during a continuous operation. Ergonomics. 1994;37:1009–1020
19. Caldwell JA, Caldwell JL, Schmidt RM. Alertness management strategies for operational contexts. Sleep Med Rev. 2008;12:257–273
20. Rosekind MR, Neri DF, Dinges DF. From laboratory to flightdeck: Promoting operational alertness. Fatigue and Duty Time Limitations—An International Review. 1997 London, England Proceedings of The Royal Aeronautical Society:7.1–7.14
21. Dinges DF, Whitehouse WG, Orne EC, Orne MT. The benefits of a nap during prolonged work and wakefulness. Work Stress. 1988;2:139–153
22. Bonnet MH. The effect of varying prophylactic naps on performance, alertness and mood throughout a 52-hour continuous operation. Sleep. 1991;14:307–315
23. Mulrine HM, Signal TL, van den Berg MJ, Gander PH. Post-sleep inertia performance benefits of longer naps in simulated nightwork and extended operations. Chronobiol Int. 2012;29:1249–1257
24. Tassi P, Muzet A. Sleep inertia. Sleep Med Rev. 2000;4:341–353
25. Van Dongen HP, Price NJ, Mullington JM, Szuba MP, Kapoor SC, Dinges DF. Caffeine eliminates psychomotor vigilance deficits from sleep inertia. Sleep. 2001;24:813–819
26. Mollicone DJ, Van Dongen HP, Rogers NL, Dinges DF. Response surface mapping of neurobehavioral performance: Testing the feasibility of split sleep schedules for space operations. Acta Astronaut. 2008;63:833–840
© 2014 by the Association of American Medical Colleges
27. Dinges DF, Orne MT, Whitehouse WG, Orne EC. Temporal placement of a nap for alertness: Contributions of circadian phase and prior wakefulness. Sleep. 1987;10:313–329