Objective To determine the effect of consultant surgeon sleep hours on patient outcomes in cardiac surgery.
Design Prospective observational cohort study.
Subjects Between January 2004 and December 2009, we prospectively collected sleep hours of 6 consultant surgeons, ranging in age from 32 to 55 years, working in a tertiary care academic institution. The prospective study cohort included all patients undergoing coronary artery bypass, valve, combined valve–coronary artery bypass, and aortic surgery. The predicted risk of death and/or any of 10 major complications was calculated using our institutional multivariable model, which was then compared with observed values. Additional prespecified analyses examined the interaction between surgeon age, sleep hours, and postoperative outcomes. This study had more than 90% power to detect a 4% (clinically important) difference in overall complication rates among groups.
Main Outcome Measures Complication and mortality rates in operations performed by surgeons with 0 to 3, 3 to 6, or more than 6 hours' sleep the evening prior to surgery.
Results Of 4047 consecutive surgical procedures, 83 were performed by a consultant with 0 to 3 hours, 1595 with 3 to 6 hours, and 2369 with more than 6 hours of sleep. Rates of mortality (3 [3.6%], 44 [2.8%], and 80 [3.4%], respectively; P = .53) were similar in the 3 groups, as were the observed vs expected ratios of major complications (1.20, 0.95, and 1.07, respectively; P = .25). There was no significant interaction between surgeon age, hours of sleep, and occurrence of death or any of 10 major complications (P = .09).
Conclusion This well-powered prospective study showed no evidence that consultant surgeon sleep hours had an effect on postoperative outcomes.
Sleep deprivation has been shown to have negative effects on mood and cognitive and psychomotor function in recent meta-analyses,1,2 but the clinical relevance of these findings in the high-hazard domain of surgical practice remains controversial. Controlled studies of surgical trainees using laparoscopic or virtual reality stimulators while acutely sleep deprived have shown no performance or cognitive impairment,3 significant decrements in cognitive and psychomotor skills,4 or mixed results.5 Two retrospective studies by the University of Virginia cardiac surgical group demonstrated that sleep deprivation of consultant6 and resident7 surgeons had no effect on the morbidity or mortality of patients after cardiac surgical operations. However, the definition of sleep deprivation in these 2 clinical studies was imprecise and was assigned retrospectively if a surgeon was on record as having performed an operative case that started after 10 PM the night before or began prior to 7:30 AM on the morning of surgery. Furthermore, this surrogate definition of sleep deprivation did not account for the fact that cardiac surgeons are frequently interrupted at night by patient care issues from the hospital, referring physicians, transplant donor coordinators, and others. We therefore designed a well-powered, observational study in which surgeon sleep hours were recorded prospectively by the consultant surgeon involved. Our primary research question was as follows: did moderate (3-6 hours) or severe (<3 hours) acute sleep deprivation in consultant surgeons result in an adverse effect on postoperative cardiac surgical outcomes? We also decided a priori to analyze the interaction between consultant surgeon age and hours of sleep on the outcomes of cardiac surgical procedures because practitioner age and experience may be potential modifiers of the effect of sleep deprivation on performance.
This study was performed at the London Health Sciences Centre, University Hospital, an academic health sciences center affiliated with the University of Western Ontario that currently performs 1300 to 1400 cardiac surgical procedures per year. Approval for data collection for this study was granted by the research ethics board of the University of Western Ontario, which waived the requirement for informed consent from individual patients. Six of our 8 consulting surgeons consented to participate in this study; 4 were younger than 45 years and 2 were 50 to 55 years of age during the study.
The study cohort included all 4047 consecutive patients undergoing coronary artery bypass grafting (CABG), valve, combined valve-CABG, and aortic surgery by the 6 participating surgeons between January 2004 and December 2009. Patients undergoing heart transplantation, primary arrhythmia surgery, and operations for congenital heart disease were excluded. Our institutional database has prospectively recorded data on all cardiac surgical patients for the past 10 years, including the rates of hospital mortality and 10 major complications8-10; in 2004 we added the variable “consultant surgeon sleep hours” to the database. This variable was prospectively recorded on the day of surgery by participating surgeons who specified the total number of sleep hours, exclusive of interruptions, that they had the night prior to performing surgery. The baseline incidence of a patient developing 1 or more of the 10 major complications in our database is approximately 16%,10 and an a priori sample size calculation indicated that a 4000-patient study would have more than 90% power to detect a clinically important 4% difference in overall complication rates among groups.
We began our analysis by determining the predicted risk of death and/or any of the 10 major complications using a previously developed institutional multivariable model that had a C statistic of 0.74 and a Hosmer-Lemeshow goodness of fit P value of .12.10 Variables used in this model included patient age, sex, body mass index, preoperative ejection fraction, urgency of surgery, primary or redo surgery, chronic obstructive pulmonary disease, recent myocardial infarction (within 30 days), peripheral vascular disease, preoperative Canadian Cardiovascular Society–New York Heart Association functional class, diabetes, cerebrovascular disease, coronary anatomy, preoperative creatinine level, congestive heart failure, ascending aortic atherosclerosis, planned on- vs off-pump CABG surgery, the presence or absence of a preoperative intra-aortic balloon pump, and consultant surgeon. The predicted values were then compared with observed values in the 0 to 3 hours’, 3 to 6 hours’, and more than 6 hours' sleep groups using χ2 tests. Given the relatively small number of patients in the 0 to 3 hours’ sleep group, the analyses were also repeated for the combined 0 to 6 hours vs more than 6 hours’ sleep groups. Furthermore, the prespecified interaction between consultant surgeon age, hours of sleep, and risk of mortality and/or any of 10 major complications was investigated via logistic regression. In addition, post hoc analyses included the effect of surgeon sleep hours (0-6 hours vs >6 hours) on the outcomes of high-risk patient cohorts with predicted complication rates of greater than 30%. Categorical data were analyzed using a χ2 test, whereas median lengths of intensive care unit and hospital stay were compared using the Wilcoxon rank sum test. P < .05 was considered significant.
From January 2004 to December 2009, 4047 patients underwent CABG, valve, combined valve-CABG, and thoracic aortic procedures by the 6 participating surgeons. Eighty-three of the operations (2.1%) were performed by severely sleep-deprived surgeons who had fewer than 3 hours' sleep, 1595 (39.4%) by moderately sleep-deprived surgeons with 3 to 6 hours' sleep, and 2369 (58.5%) by surgeons who had more than 6 hours' sleep on the night before the operations. The different types of procedures and the predicted risks of mortality and/or any of the 10 major complications related to surgeon sleep hours are shown in Table 1. It is interesting that the predicted risk of patient mortality or major complications in the 3 groups was almost identical; the slightly higher percentage of aortic cases in the 0 to 3 hours’ sleep group was balanced out by the higher frequency of combined valve-CABG cases in the more than 6 hours’ sleep group.
Table 2 shows the observed incidence of mortality and each of the 10 major complications whose operational definitions have remained unchanged in our database since its inception 10 years ago.8-10 There was a statistically significantly higher incidence of septicemia in the 0 to 3 hours’ sleep group but no significant difference among groups in any of the other individual or composite outcomes. The median intensive care unit stay did not differ among groups, although the median postoperative hospital stay was 1 day shorter in the 3 to 6 hours’ sleep group than in the other 2 groups, a difference that was statistically significant. When data in the 0 to 3 hours’ and 3 to 6 hours’ sleep groups were combined, the major complication rate was 259 of 1678 (15.4%) in the 0 to 6 hours’ sleep group vs 403 of 2369 (17.1%) in the more than 6 hours’ sleep group (P = .17). Furthermore, the median hospital stay was 6.0 vs 7.0 days in the 0 to 6 hours’ vs more than 6 hours’ sleep groups, respectively (P < .001).
Table 3 shows the observed vs expected ratios for the occurrence of 1 or more of the 10 major complications according to consultant sleep hours; these ratios were not significantly different. If the 2 sleep deprivation groups (ie, 0-3 and 3-6 hours’ sleep) groups were combined and compared with the more than 6 hours’ sleep group, the respective observed vs expected ratios were 0.96 and 1.06 (P = .16).
Table 4 shows the interaction between sleep hours, consultant age, and the occurrence of mortality and/or any of the 10 major complications. This table shows that, although the surgeons older than 50 years exhibited slightly higher incidence of complications when sleep deprived compared with younger surgeons, there was no significant interaction between sleep hours and consultant age on adverse clinical outcomes (P = .09).
To ascertain whether a lack of consultant surgeon sleep hours had an effect on the results of patients with the highest predicted risk, we examined a cohort of 399 patients with a predicted risk of death or major complications in excess of 30%. The incidence of mortality and/or any of the 10 major complications was 52 of 141 (36.9%) if the surgeon had 0 to 6 hours of sleep vs 113 of 258 (43.8%) if the surgeon had more than 6 hours of sleep (P =.18).
The results of this well-powered, 6-year prospective study showed no important adverse effect of consultant surgeon sleep deprivation on the outcomes of cardiac surgical procedures. This finding was robust and held up even in a high-risk subgroup of 399 patients with a predicted risk of mortality and/or complication rate greater than 30%. Furthermore, there was no evidence that the effect of sleep hours on mortality or any of 10 major postoperative complications was affected by surgeon age. In this current era, when health care delivery is often scrutinized against benchmarks and vigilant outcomes analyses are performed, this study provides important reassurance to the general public that consultant surgeon sleep deprivation does not affect the quality of clinical outcomes or patient safety following cardiac surgery.
It has been hypothesized that sleep deprivation reduces surgical performance by affecting technical skills5,11-13 and cognitive abilities.1,2,4,14 Many studies have reported impaired fine motor skills following acute sleep deprivation,5,11-13 which may be particularly important in a technically demanding field such as cardiac surgery where small technical errors can contribute to significant complications. However, the generalizability of these studies to cardiac surgeons may be limited by the fact that most of these studies evaluated technical performance in laparoscopic simulators,5,11-13 assessing skills that may not be transferable to cardiac operations. Unfortunately, there are no good studies evaluating the effects of sleep deprivation on the fine motor skills required to perform CABG, valve, or aortic operations. More important may be the effects of sleep loss on cognitive abilities such as problem solving and decision making.1,2,4,14 Disruption of normal circadian physiology may reduce alertness, wakefulness, and intellectual acuity.15 Fatigue has been demonstrated to impair performance and compromise safety in commercial airline pilots and professional truck drivers16; however, similar evidence has yet to be proven in real-world surgery. Surgical simulator studies have demonstrated mixed findings of the effects of fatigue on cognitive abilities.3,4,14 It is certainly plausible that these effects could have impaired preoperative decision making or patient selection in our study. However, our results might suggest that this was not the case because all 3 groups had similar preoperative predicted risk profiles, perhaps indicating that surgeon decision making did not elicit any detectable change in preoperative patient optimization and, ultimately, the postoperative outcomes did not show any significant differences either.
Overall, the literature evaluating the effect of work hours and sleep deprivation on consultant physicians has mixed results.15,17 Some evidence in the trauma literature suggests that sleep-deprived consultant surgeons make fewer cognitive errors than postgraduate trainees, although the psychomotor skills of both groups are equally affected.14 Unfortunately, the only other article examining the clinical effect of sleep deprivation in consultant cardiac surgeons used an imprecise, surrogate measure of sleep deprivation6 with a resultant high risk of misclassification bias. Nonetheless, the investigators did not find any association between sleep loss and cardiac surgical outcomes.6 Recently, a matched retrospective cohort study of procedures performed by 86 consultant surgeons and 134 obstetricians and/or gynecologists showed no significant increase in complications in postnighttime procedures (5.4%) vs control procedures (4.9%) (odds ratio, 1.09; 95% confidence interval, 0.84-1.41).18 However, among surgical, but not obstetrical/gynecologic, postnighttime cases with 6 hours or less of sleep opportunity, there was a substantially elevated rate of complications (6.2% vs 3.4%; odds ratio, 1.72; 95% confidence interval, 1.02-2.89) compared with cases in which the consultant surgeon's sleep opportunity exceeded 6 hours.18 This study, although well powered and matched, still had a high likelihood of misclassification bias because surgeon sleep hours were assessed retrospectively.
The results of our study may have important ramifications in restricting current trainee work hours when future practice may demand optimal performance during sleep-deprived conditions. It is now almost 7 years since the Accreditation Council for Graduate Medical Education (ACGME) in the United States imposed national work-hour limits for postgraduate medical trainees. However, the effects of the ACGME work-hours mandate on patient mortality and educational opportunities have been equivocal.17 Similarly in Europe, more aggressive work-hour restrictions have been implemented that have improved trainee lifestyle satisfaction; however, again, there are few data demonstrating any reduction in medical errors or improved patient safety.19,20 An analysis of the sleep characteristics of 180 attending physicians from varied specialties found that reduced sleep, not hours worked, was associated with increased sleepiness,21 highlighting the weak link between work hours and quantity of sleep. This finding undermines the intuitive belief that reduced work hours improves daytime fatigue, trainee performance, and, hence, patient safety, which was a common tenet used to justify maximum-work-hour legislation. Regardless, the findings of our study might suggest reassessment of this belief when contemplating trainee work-hour limitations.
Given the results of our study, we hypothesize that well-developed compensatory mechanisms must exist to combat the effects of sleep deprivation. Most attending cardiac surgeons could have developed physiologic adaptation to chronic sleep deprivation and, through conditional learning, practiced enough to reduce errors under these conditions. With development of expertise, the complex decision making and manual dexterity required for cardiac surgical procedures has likely become more routine for attending surgeons, thus likely requiring less of the sleep-dependent, higher executive cognitive functions.22 Most importantly, cardiac surgery is a “team sport” that involves multiple levels of trainee, surgical, and nursing staff who collectively could potentially compensate for a consultant surgeon's fatigue-related impairment. Simple countermeasures that have been recommended,17 such as continuous social interaction and brightly lit operating rooms, can also protect against fatigue-related error. Caffeine may mitigate the effects of sleep deprivation by improving cognitive function23-25 and has been the standard “postcall meal” for generations of surgeons.
There is a burgeoning body of literature on the potential influence of personality traits on the manifestations of sleep deprivation, especially the effect of introversion-extraversion traits26 and how individual differences in emotional intelligence determine the response to frustration following sleep loss.27 There is also emerging evidence that differential vulnerability to the effects of acute sleep deprivation is a stable, traitlike characteristic28 that may be mediated by specific polymorphisms in the PER gene.29 Future studies should consider administering validated personality inventories to participating health care practitioners to explore a possible interaction between personality traits, observed behaviors during acute sleep deprivation, and patient outcomes.
The strengths of our study include the accurate, prospective ascertainment of surgeon sleep hours in a relatively large patient cohort and the fact that our group of participating consultant surgeons were of different sexes and ages. Limitations of our study included the relatively small number of patients (ie, 83 [2.1%]) who were operated on by consultant surgeons with severe sleep deprivation (<3 hours’ sleep). Nonetheless, 1595 patients in this cohort (39.4%) were operated on by surgeons who had at least moderate acute sleep deprivation (3-6 hours’ sleep). Another shortcoming of our study may be the limited generalizability of the results of 6 consultant surgeons (4 aged <45 years and 2 aged >50 years) in a single academic center, with none older than 55 years and in their last decade of practice. Conclusions of the interaction between surgeon age, sleep hours, and postoperative outcomes must therefore be made with caution. Furthermore, the generalizability of our study findings is limited by the fact that the clinical setting was a busy academic cardiac surgical unit with a strong cardiac surgery residency training program. Whether our study results and conclusions would be the same in a nonacademic setting without postgraduate trainees as potential mitigating agents remains uncertain.
An additional methodological issue in this and most other sleep-deprivation studies in the literature is the failure to control for potential confounding variables that may have had an effect on the observed clinical results. These include the possible concomitant influence of chronic partial sleep restriction,15,17,30 the effect of differing personality traits on vulnerability to sleep deprivation,26-28 and the failure to control for stimulant use.23-25 Although surgeons may subjectively adapt to restricted sleep,17 objective measures of performance continued to worsen during conditions of chronically reduced sleep hours.17,18,31,32 The accurate measurement of chronic partial sleep deprivation could not be accomplished in our study given the variability in operating room and vacation schedules of participating surgeons during the 6 years that patient outcome data were recorded.
In summary, our 6-year prospective study showed no evidence that consultant surgeon sleep hours had an adverse effect on postoperative outcomes, even in patients with a high predicted risk. Future studies should explore the compensatory mechanisms that individual health care practitioners and surgical teams use to maintain good patient outcomes when key members of the team are sleep deprived.
Correspondence: Michael W. A. Chu, MD, FRCSC, B6-106 University Hospital, London Health Sciences Centre, 339 Windermere Rd, PO Box 5339, London, ON N6A 5A5, Canada (email@example.com).
Accepted for Publication: February 22, 2011.
Published Online: May 16, 2011. doi:10.1001/archsurg.2011.121
Author Contributions:Study concept and design: Chu, Fox, and Novick. Acquisition of data: Chu, Fox, Kiaii, Quantz, Myers, Hewitt, and Novick. Analysis and interpretation of data: Chu, Stitt, Fox, Guo, and Novick. Drafting of the manuscript: Chu and Novick. Critical revision of the manuscript for important intellectual content: Chu, Stitt, Fox, Kiaii, Quantz, Guo, Myers, Hewitt, and Novick. Statistical analysis: Stitt. Administrative, technical, and material support: Chu, Fox, Kiaii, Quantz, Guo, Myers, Hewitt, and Novick. Study supervision: Novick.
Financial Disclosure: None reported.
Funding/Support: This study was supported by internal research funds from the Division of Cardiac Surgery, University of Western Ontario.
Disclaimer: Internal research funds of the Division of Cardiac Surgery at the University of Western Ontario covered the salaries of the clinical research associates who run our clinical database as well as biostatistical consultation fees. All data collection, analysis, and interpretation as well as the writing of this report were performed by the study authors, without the involvement of any outside party.
Additional Contributions: Elizabeth Millar assisted in manuscript preparation and Vladimir Hachinski, CM, MD, FRCPC, DSc, reviewed an earlier version of the manuscript.
Pilcher JJ, Huffcutt AI. Effects of sleep deprivation on performance: a meta-analysis. Sleep
. 1996;19(4):318-3268776790PubMedGoogle Scholar
Philibert I. Sleep loss and performance in residents and nonphysicians: a meta-analytic examination. Sleep
. 2005;28(11):1392-140216335329PubMedGoogle Scholar
Lehmann KS, Martus P, Little-Elk S,
et al. Impact of sleep deprivation on medium-term psychomotor and cognitive performance of surgeons: prospective cross-over study with a virtual surgery simulator and psychometric tests. Surgery
. 2010;147(2):246-25420109624PubMedGoogle ScholarCrossref
Kahol K, Leyba MJ, Deka M,
et al. Effect of fatigue on psychomotor and cognitive skills. Am J Surg
. 2008;195(2):195-20418194679PubMedGoogle ScholarCrossref
Eastridge BJ, Hamilton EC, O’Keefe GE,
et al. Effect of sleep deprivation on the performance of simulated laparoscopic surgical skill. Am J Surg
. 2003;186(2):169-17412885613PubMedGoogle ScholarCrossref
Ellman PI, Law MG, Tache-Leon C,
et al. Sleep deprivation does not affect operative results in cardiac surgery. Ann Thorac Surg
. 2004;78(3):906-91115337018PubMedGoogle ScholarCrossref
Ellman PI, Kron IL, Alvis JS,
et al. Acute sleep deprivation in the thoracic surgical resident does not affect operative outcomes. Ann Thorac Surg
. 2005;80(1):60-6515975341PubMedGoogle ScholarCrossref
Novick RJ, Fox SA, Stitt LW,
et al. Cumulative sum failure analysis of a policy change from on-pump to off-pump coronary artery bypass grafting. Ann Thorac Surg
. 2001;72(3):S1016-S102111565718PubMedGoogle ScholarCrossref
Novick RJ, Fox SA, Stitt LW, Forbes TL, Steiner S. Direct comparison of risk-adjusted and non-risk-adjusted CUSUM analyses of coronary artery bypass surgery outcomes. J Thorac Cardiovasc Surg
. 2006;132(2):386-39116872967PubMedGoogle ScholarCrossref
Novick RJ, Fox SA, Stitt LW,
et al. Impact of the opening of a specialized cardiac surgery recovery unit on postoperative outcomes in an academic health sciences centre. Can J Anaesth
. 2007;54(9):737-74317766741PubMedGoogle ScholarCrossref
Grantcharov TP, Bardram L, Funch-Jensen P, Rosenberg J. Laparoscopic performance after one night on call in a surgical department: prospective study. BMJ
. 2001;323(7323):1222-122311719413PubMedGoogle ScholarCrossref
DeMaria EJ, McBride CL, Broderick TJ, Kaplan BJ. Night call does not impair learning of laparoscopic skills. Surg Innov
. 2005;12(2):145-14916034504PubMedGoogle ScholarCrossref
Leff DR, Aggarwal R, Rana M,
et al. Laparoscopic skills suffer on the first shift of sequential night shifts: program directors beware and residents prepare. Ann Surg
. 2008;247(3):530-53918376200PubMedGoogle ScholarCrossref
Gerdes J, Kahol K, Smith M, Leyba MJ, Ferrara JJ. Jack Barney award: the effect of fatigue on cognitive and psychomotor skills of trauma residents and attending surgeons. Am J Surg
. 2008;196(6):813-82019095094PubMedGoogle ScholarCrossref
Olson EJ, Drage LA, Auger RR. Sleep deprivation, physician performance, and patient safety. Chest
. 2009;136(5):1389-139619892678PubMedGoogle ScholarCrossref
Rothschild JM, Keohane CA, Rogers S,
et al. Risks of complications by attending physicians after performing nighttime procedures. JAMA
. 2009;302(14):1565-157219826026PubMedGoogle ScholarCrossref
McIntyre HF, Winfield S, Te HS, Crook D. Implementation of the European Working Time Directive in an NHS trust: impact on patient care and junior doctor welfare. Clin Med
. 2010;10(2):134-13720437982PubMedGoogle Scholar
Maxwell AJ, Crocker M, Jones TL, Bhagawati D, Papadopoulos MC, Bell BA. Implementation of the European Working Time Directive in neurosurgery reduces continuity of care and training opportunities. Acta Neurochir (Wien)
. 2010;152(7):1207-121020373115PubMedGoogle ScholarCrossref
Chen I, Vorona R, Chiu R, Ware JC. A survey of subjective sleepiness and consequences in attending physicians. Behav Sleep Med
. 2008;6(1):1-1518412034PubMedGoogle ScholarCrossref
Bonnet MH, Balkin TJ, Dinges DF, Roehrs T, Rogers NL, Wesensten NJ.Sleep Deprivation and Stimulant Task Force of the American Academy of Sleep Medicine. The use of stimulants to modify performance during sleep loss: a review by the Sleep Deprivation and Stimulant Task Force of the American Academy of Sleep Medicine. Sleep
. 2005;28(9):1163-118716268386PubMedGoogle Scholar
Crochet P, Aggarwal R, Mishra A, Sirimanna P, Darzi A. Caffeine and taurine reverse the deterioration in laparoscopic and cognitive skill following sleep deprivation. Surg Forum
. 2009;209:S111Google Scholar
Wesensten NJ, Killgore WDS, Balkin TJ. Performance and alertness effects of caffeine, dextroamphetamine, and modafinil during sleep deprivation. J Sleep Res
. 2005;14(3):255-26616120100PubMedGoogle ScholarCrossref
Killgore WD, Richards JM, Killgore DB, Kamimori GH, Balkin TJ. The trait of introversion-extraversion predicts vulnerability to sleep deprivation. J Sleep Res
. 2007;16(4):354-36318036080PubMedGoogle ScholarCrossref
Kahn-Greene ET, Lipizzi EL, Conrad AK, Kamimori GH, Killgore WDS. Sleep deprivation adversely affects interpersonal responses to frustration. Pers Individ Dif
. 2006;41(8):1433-1443Google ScholarCrossref
Van Dongen HP, Baynard MD, Maislin G, Dinges DF. Systematic interindividual differences in neurobehavioral impairment from sleep loss: evidence of trait-like differential vulnerability. Sleep
. 2004;27(3):423-43315164894PubMedGoogle Scholar
Viola AU, Archer SN, James LM,
et al. PER3
polymorphism predicts sleep structure and waking performance. Curr Biol
. 2007;17(7):613-61817346965PubMedGoogle ScholarCrossref
Van Dongen HPA, Maislin G, Mullington JM, Dinges DF. The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and sleep physiology from chronic sleep restriction and total sleep deprivation. Sleep
. 2003;26(2):117-12612683469PubMedGoogle Scholar
Dinges DF, Pack F, Williams K,
et al. Cumulative sleepiness, mood disturbance, and psychomotor vigilance performance decrements during a week of sleep restricted to 4-5 hours per night. Sleep
. 1997;20(4):267-2779231952PubMedGoogle Scholar