Author + information
- Received August 7, 2014
- Accepted August 28, 2014
- Published online January 1, 2015.
- Herbert D. Aronow, MD, MPH∗∗ (, )
- Hitinder S. Gurm, MD†,
- James C. Blankenship, MD‡,
- Charles A. Czeisler, MD§,
- Tracy Y. Wang, MD, MHS, MSc‖,
- Lisa A. McCoy, MS‖,
- Megan L. Neely, PhD‖ and
- John A. Spertus, MD¶
- ∗Michigan Heart, Ann Arbor, Michigan
- †Division of Cardiovascular Medicine, University of Michigan, Ann Arbor, Michigan
- ‡Department of Cardiology, Geisinger Medical Center, Danville, Pennsylvania
- §Division of Sleep Medicine, Brigham & Women’s Hospital, Boston, Massachusetts
- ‖Duke Clinical Research Institute, Durham, North Carolina
- ¶Section of Cardiology, Mid America Heart Institute, Kansas City, Missouri
- ↵∗Reprint requests and correspondence:
Dr. Herbert D. Aronow, Michigan Heart, 5325 Elliott Drive, Suite 202, Ypsilanti, Michigan 48197.
Objectives This study sought to compare in-hospital mortality and bleeding complications for procedures performed by sleep-deprived versus non–sleep-deprived operators.
Background To optimize the safety of percutaneous coronary intervention (PCI), it is essential to determine whether physicians performing emergent, middle-of-the-night procedures, and who may be sleep-deprived as a consequence, have equally safe outcomes when performing cases the following day.
Methods We used CathPCI registry data to compare in-hospital mortality and bleeding complications for procedures performed by sleep-deprived versus non–sleep-deprived operators using logistic regression with generalized estimating equations to account for within-operator clustering. Outcomes were risk-adjusted using previously validated models for in-hospital mortality and bleeding. Our cohort included 1,509,096 daytime PCI procedures performed by 5,014 operators between 7 am and midnight from July 1, 2009, through June 30, 2012. Operators were assumed to be acutely sleep-deprived if they began a middle-of-the-night PCI between midnight and 6:59 am and performed a next-day PCI between 7 am and midnight, and chronically sleep deprived if they had performed multiple middle-of-the-night PCI procedures during the previous 7 days.
Results Only 2.4% of all daytime PCI procedures were performed by operators who had performed at least 1 middle-of-the-night PCI procedure earlier that day. In adjusted analyses, when comparing procedures performed by acutely sleep-deprived with non–sleep-deprived operators, there were no significant differences in mortality (odds ratio [OR]: 1.02, 95% confidence interval [CI]: 0.94 to 1.12; p = 0.61) or bleeding (OR: 1.03, 95% CI: 0.98 to 1.08; p = 0.19). However, a greater degree of chronic sleep deprivation was associated with a higher adjusted risk of bleeding (OR: 1.19, 95% CI: 1.05 to 1.34; p = 0.007).
Conclusions Daytime PCI procedures are uncommonly performed by sleep-deprived operators. We found no signal of increased complications when acutely sleep-deprived operators performed PCI but an increased risk of bleeding associated with procedures performed by operators with greater degrees of chronic sleep deprivation.
Given the importance of primary percutaneous coronary intervention (PCI) in the treatment of ST-segment elevation myocardial infarction (1), interventional cardiologists often perform PCI while on call. In other settings, extended work hours among physicians have been associated with poor psychomotor performance (2), reduced alertness (3), increased likelihood of medical errors (4), occupational injuries (5), and motor vehicle accidents (6). It is not clear, however, how frequently or how safely interventional cardiologists perform PCI on the day following a middle-of-the-night PCI procedure. One small, single-center study did not find a statistically significant increase in risk-adjusted mortality when PCI was performed by interventional cardiologists who were post-call; however, the study was underpowered and the estimated effect size (odds ratio [OR]: 6.8, 95% confidence interval [CI]: 0.66 to 30.6) was large (7). It is plausible that sleep deprivation could adversely affect cognitive or psychomotor function, translating into an increased risk of life-threatening procedural complications and/or altering the operator’s threshold for employing bleeding avoidance strategies, ultimately culminating in increased risk of bleeding. If PCI performance following middle-of-the-night procedures were associated with worse outcomes, this would have major patient safety and policy implications. To address this existing gap in knowledge, we used data from the NCDR (National Cardiovascular Data Registry) CathPCI registry to examine next-day PCI outcomes among operators who had performed middle-of-the-night PCI procedures.
The CathPCI registry, an initiative of the American College of Cardiology Foundation and the Society for Cardiovascular Angiography and Interventions, collects detailed demographic, clinical, process, and in-hospital outcome data for patients undergoing PCI at participating academic, community, for profit, and not-for-profit hospitals (8). Approximately, 85% of all PCI procedures performed in the United States are captured by CathPCI (9). Complete data element definitions for CathPCI registry version 4 are available online at the NCDR website.
Between July 1, 2009, and June 30, 2012, 1,869,997 PCI procedures were entered into the registry. We excluded 1,836 operators who never performed a procedure between midnight and 7 am; 1,945 operators who performed at least 1 procedure between midnight and 7 am, but performed no procedures after 7 am the subsequent day; and all operators without a valid national provider identifier. This yielded a final study population of 1,509,096 PCI procedures performed by 5,014 operators between 7 am and midnight.
The overall study aim was to determine whether daytime procedures performed by sleep-deprived operators (so-called sleep-deprived procedures) were associated with increased mortality or bleeding complications when compared with daytime procedures performed by operators who were not sleep-deprived. In the primary analysis, operators were considered acutely sleep-deprived if they began a middle-of-the-night PCI between midnight and 6:59 am and performed a next-day PCI between 7 am and midnight on the same date. In addition to our more coarse definition of “sleep-deprived,” we also sought to examine the degree of acute sleep deprivation prior to the conduct of a procedure. The degree of acute sleep deprivation following middle-of-the-night PCI was approximated by the next-day PCI procedure start time. Because we assumed that no sleep opportunity existed after completing a middle-of-the-night PCI but before beginning a next-day PCI procedure, we considered later procedural start times after a middle-of-the-night case as indicating greater degrees of acute sleep deprivation. We also categorized the degree of chronic sleep deprivation by the number of days (≥2 vs. 1) on which a middle-of-the-night PCI procedure was performed during the week preceding a daytime sleep-deprived PCI procedure. More frequent middle-of-the-night PCI procedures during the week preceding a sleep-deprived procedure were assumed to indicate a greater degree of chronic sleep deprivation.
Primary outcomes of interest included in-hospital all-cause mortality and in-hospital bleeding. By the NCDR definitions, bleeding was defined as any of the following within 72 h of the PCI procedure: intracranial hemorrhage; cardiac tamponade; non–coronary artery bypass graft–related transfusion of whole or packed red blood cells excluding patients with pre-procedure hemoglobin level ≤8 g/dl, or absolute hemoglobin decrease from pre- to post-PCI ≥3 g/dl excluding patients with pre-procedure hemoglobin >16 g/dl (10). A number of pre-specified secondary in-lab and in-hospital outcomes were also examined using standard NCDR definitions. Secondary in-lab outcomes included lesion crossing with a guidewire, device deployment, vessel dissection, vessel perforation, and successful lesion dilation. Secondary in-hospital outcomes included the composite endpoint death, myocardial infarction, or stroke, as well as the individual endpoints of periprocedural myocardial infarction, stroke, coronary artery bypass graft, tamponade, vascular complication, blood transfusion, bleeding event within 72 h, access site hematoma, and retroperitoneal bleed.
Patient characteristics and procedural outcomes are compared according to whether PCI was performed by a sleep-deprived versus a non–sleep-deprived operator. Categorical variables are presented as frequencies (percentages), and in unadjusted analyses, between-group differences are assessed using the chi-square or Fisher exact tests where appropriate. Continuous variables are presented as medians with interquartile ranges (IQR) and in unadjusted analyses these are compared using the Wilcoxon rank-sum test.
Unadjusted and adjusted OR with 95% CI are presented for each outcome (bleeding and mortality) for procedures performed by sleep-deprived versus non–sleep-deprived operators using logistic regression with generalized estimating equations to account for within-operator clustering. In the bleeding models, adjustments were made for covariates in the revised NCDR CathPCI bleeding risk model (10). In the mortality models, adjustments were made for covariates from the NCDR CathPCI mortality risk model (11). Variables included in these models are listed in Online Table 1.
To assess the association between greater degree of acute sleep deprivation and outcome, we stratified cases after a middle-of-the-night procedure by whether or not the case started after noon, and we assumed that later cases indicated greater acute sleep deprivation. Similarly, we assessed chronic sleep deprivation by categorizing next-day procedural outcomes according to whether operators had performed ≥2 versus 1 middle-of-the-night PCI procedure during the past 7 days preceding a sleep-deprived procedure.
A p value of <0.05 was considered statistically significant in all analyses. All statistical analyses were performed using SAS (version 9.3, SAS Institute, Cary, North Carolina).
Given the lack of a finer measure to assess the degree of sleep deprivation, we conducted several sensitivity analyses to confirm our primary findings. In the first sensitivity analysis, middle-of-the-night PCI was defined as occurring between 2 and 5:59 am, and next-day PCI procedures included those started between 6 am and midnight on the same date. In the second sensitivity analysis, middle-of-the-night PCI was defined as occurring between midnight and 6:59 am, and next-day PCI included only those PCI procedures started between 7 am and 5 pm on the same date. In the third sensitivity analysis, we defined middle-of-the-night PCI procedures as those started between midnight and 6:59 am and next-day PCI as those procedures started between 7 am and midnight on the same date, but we further characterized sleep status according to sleep opportunity, where sleep opportunity was defined by the number of middle-of-the-night PCI procedures performed by an operator (0, 1, 2, or more) on the same date.
Finally, the primary analyses were repeated using a within-operator propensity-matched sample. Variables included a priori in the propensity model are listed in Online Table 1. We aimed to use the greedy match algorithm to pair 2 control cases (non–sleep-deprived procedures) to every case (sleep-deprived procedure) within operator, on the logit of the propensity score using the publically available gmatch macro (12) within a pre-specified caliper of 0.2× the standard deviation of the logit of the propensity score. This approach results in estimates of the treatment effect with lower mean squared error (13). If there were no control procedures with a propensity score within the caliper of a given sleep-deprived procedure, then that sleep-deprived procedure was not included in the matched sample. To estimate the sleep-deprived versus non–sleep-deprived effect on outcomes among the propensity-matched sample, we estimated the OR from a logistic regression model stratified by matched group. We were able to match at least 1 control for 99.7% of cases (35,949 of 36, 049) and 2 control cases for 98.4% (35,461 of 36,049) of cases.
Patient and procedural characteristics
In this study, 2.4% (n = 36,049) of all 1,509,096 daytime PCI procedures were performed by operators who had also performed at least 1 middle-of-the-night PCI earlier that day. Patient and procedural characteristics for the primary analytic cohort appear in Table 1. Median (IQR) patient age in the overall cohort was 65 (56 to 74) years, more than two-thirds were men, and nearly 9 of 10 were Caucasian. Most patient characteristics were similar for sleep-deprived and non–sleep-deprived procedure groups, although patients who underwent daytime PCI by a sleep-deprived operator were more likely to present with ST-segment elevation myocardial infarction (18.4% vs. 14%) or to have Canadian Cardiovascular Society angina class IV angina (33.2% vs. 28.6%). The predicted probabilities for both in-hospital bleeding and mortality were higher among patients undergoing PCI performed by sleep-deprived operators.
Patients who underwent PCI performed by sleep-deprived operators were less likely to undergo an elective procedure (39.4% vs. 44.4%), less likely to have TIMI (Thrombolysis In Myocardial Infarction) flow grade 3 pre-procedure (52.1% vs. 56.4%), more often treated with unfractionated heparin (54.3% vs. 51.2%), and less often treated with bivalirudin (53.9% vs. 57.3%). Although bifurcation PCI was performed equally often in both groups, chronic total occlusion PCI was performed less often in procedures performed by a sleep-deprived operator. Bleeding avoidance therapies, including transradial access, bivalirudin, and/or vascular closure devices were employed slightly less often among sleep-deprived than non–sleep-deprived operators (77% vs. 78.6%).
Unadjusted in-lab and in-hospital outcomes appear in Tables 1 and 2⇓, respectively. In-lab outcomes, including coronary artery dissection and perforation, as well as successful lesion treatment occurred with similar frequencies in both groups. Unadjusted in-hospital mortality (1.6% vs. 1.3%; p < 0.001) and bleeding (6.1% vs. 5.4%; p < 0.001) occurred significantly more frequently among patients treated by acutely sleep-deprived versus non–sleep-deprived operators. However, after adjustment for bleeding and mortality risks using the NCDR risk-adjustment models, there were no significant differences in either mortality (OR: 1.02, 95% CI: 0.94 to 1.12; p = 0.61) or bleeding (OR: 1.03, 95% CI: 0.98 to 1.08; p = 0.19). A composite of all registry-measured PCI complications, including dissection, perforation, myocardial infarction, cardiogenic shock, heart failure, stroke, tamponade, new dialysis requirement, blood transfusion, bleeding event within 72 h, other vascular complication, emergent coronary artery bypass graft, or death, occurred more often among PCI procedures performed by sleep-deprived than by non–sleep-deprived operators in unadjusted analyses (9.3% vs. 8.6%, p < 0.0001), but after multivariable adjustment, there were no significant differences in the incidence of this endpoint (OR: 1.01, 95% CI: 0.97 to 1.05; p = 0.78). In sensitivity analyses, there were no significant differences in adjusted mortality or bleeding when an acutely sleep-deprived operator was defined as one who began middle-of-the-night PCI during a more narrow time window, when next-day procedures included only those during “normal business hours,” when procedures were stratified according to the number of middle-of-the-night PCI procedures begun earlier on the same date, or when sleep-deprived procedures were matched to non–sleep-deprived procedures within operator (Figures 1 and 2⇓). Among procedures performed by sleep-deprived operators, there was no significant relationship between acute sleep deprivation and adjusted risk of bleeding (OR: 0.90, 95% CI: 0.80 to 1.01; p = 0.06) or mortality (OR: 0.88, 95% CI: 0.70 to 1.12; p = 0.3). In contrast, while procedures performed by operators with greater degrees of chronic sleep deprivation (1.3% of all daytime procedures in this cohort) were not associated with increased adjusted risk of mortality (OR: 0.81, 95% CI: 0.62 to 1.05; p = 0.1), they were associated with a significantly greater adjusted risk of bleeding (OR: 1.19, 95% CI: 1.05 to 1.34; p = 0.007) (Table 3).
We examined the safety of PCI procedures performed the day after a middle-of-the-night PCI procedure. In the largest study of its kind to date, we observed that only ∼1 in 40 daytime PCI procedures were performed by operators who had also performed middle-of-the-night PCI earlier that day. Importantly, we found no significant adverse association with adjusted in-hospital mortality or bleeding when comparing procedures performed by acutely sleep-deprived versus non–sleep-deprived operators. However, among chronically sleep-deprived operators performing middle-of-the-night PCI on 2 or more nights during the 7 days prior to a sleep-deprived procedure (an uncommon occurrence), there was a statistically significant but small excess risk of bleeding, but not of mortality.
Whereas previous studies have found that operator sleep deprivation resulting from extended work hours results in impaired operator cognitive and psychomotor function and the occurrence of serious medical errors (2–4), the relationship between middle-of-the-night on-call procedures and subsequent day procedural outcomes has not been as clear. Surgical simulator studies of physicians and physician trainees have yielded mixed results. One study found no impairment in psychomotor skills, but a reduction in cognitive function among post-call physicians (14). Another found no decrement in psychomotor or cognitive skills following relative sleep deprivation (15). In contrast, a third study found similar technical but improved cognitive function following acute partial sleep deprivation (16). Studies of thoracic surgeons (17) and thoracic surgical residents (18) have not demonstrated higher cardiac surgical morbidity or mortality among surgeons who operated during the previous night. Surgeons and obstetrician/gynecologists who operate on the day following middle-of-the-night surgery have also not experienced higher surgical complication rates, unless there was <6 h of opportunity for sleep before beginning the next operation (19). Finally, no difference in readmission rates was observed among patients operated on by general surgeons after providing a trauma call (20). When considered alongside these data, our findings are reassuring for patients and practitioners alike that middle-of-the-night PCI does not appear to be associated with adverse patient outcome during subsequent day procedures, except in rare circumstances. It is also reassuring that only a small proportion of PCI are performed by operators who also performed middle-of-the-night PCI procedures earlier on the same day.
Middle-of–the-night PCI could potentially affect each of the 4 major physiologic determinants of fatigue—circadian time of day, sleep inertia, acute sleep deprivation, and chronic sleep deprivation (21)—culminating in impaired psychomotor and cognitive performance, and theoretically, translating into adverse patient outcomes. As the primary aim of our study was to relate the performance of middle-of-the-night PCI to next-day PCI outcomes, middle-of-the-night PCI procedural outcomes were not examined, thereby precluding any investigation into the impact of circadian time of day or sleep inertia (22). Acute sleep deprivation leads to a sigmoidal decline in cognitive function with increasing time awake (23). If relevant, the effects of acute sleep deprivation following middle-of-the-night PCI might be most apparent in association with next-day PCI procedures having later start times (assuming there were no sleep opportunities before beginning the next day’s procedures). We did not observe such a relationship, but given that sleep opportunity was not directly measured, we cannot rule out that such a relationship exists. Chronic partial sleep deprivation leads to more frequent lapses in psychomotor performance with less time-in-bed sleep opportunity (24). This scenario might occur when operators perform middle-of-the-night PCI procedures on days that are in close proximity to one another over relatively short blocks of time; we observed a greater likelihood of bleeding when operators had performed middle-of-the-night PCI procedures on at least 2 nights during the 7 days preceding a sleep-deprived procedure. This observation should be confirmed independently before efforts to limit middle-of-the-night PCI call frequency are entertained, given the potential impact on access to timely life-saving patient care. Finally, it remains possible that any or all of the physiologic determinants of fatigue negatively affected operator psychomotor performance but that these changes did not translate into adverse patient outcomes. Some data suggest that more experienced operators may be less susceptible to the effects of sleep deprivation (14).
Although we cannot determine whether case selection was altered following middle-of-the-night PCI, it is noteworthy that patients who underwent PCI performed by sleep-deprived operators were less often elective. It is possible that sleep-deprived operators are less likely to schedule or more likely to defer elective procedures on post-call days. Regardless, a shift from elective to urgent/emergent procedures is likely responsible for the greater prevalence of acute coronary syndromes observed among patients who underwent PCI at the hands of sleep-deprived operators.
The CathPCI registry does not collect data pertaining to an operator’s sleep status or whether an operator was awake during the preceding night for other reasons (e.g., performing diagnostic cardiac catheterization that did not culminate in PCI, temporary pacemaker placement, pericardiocentesis, or engaged in other clinical responsibilities), nor can we account for cases performed at non-CathPCI registry participating centers. It was not possible to determine whether operator age or other demographic characteristics modified the relationship between sleep status and outcome as data on these were not available. This analysis compares “case mix” between sleep-deprived and non–sleep-deprived operators, but it cannot account for “case selection” per se, as we cannot account for PCI procedures that may have been canceled or rescheduled after diagnostic catheterization in this dataset. Sleep deprivation may be subtle in its manifestations and could result in longer procedure times and/or greater equipment use. Current CathPCI registry data elements do not allow for analyses of these endpoints. Finally, although we referred to operators performing middle-of-the-night PCI as “sleep-deprived,” we did not measure the extent of sleep deprivation and it remains possible that operators obtained enough sleep before or after the procedure to be fully rested at the time of next-day PCI procedures.
Our study suggests that only a small fraction of PCI procedures are performed by sleep-deprived operators. The overall outcome of these patients appears similar to that in patients undergoing PCI at the hands of non–sleep-deprived operators.
For a supplemental table, please see the online version of this paper.
This research was supported by the American College of Cardiology Foundation NCDR (National Cardiovascular Data Registry). Dr. Blankenship is the uncompensated principle investigator at Geisinger Medical Center for industry-funded multicenter trials; funding companies include Boston Scientific, Tryton Medical, Stentys, AstraZeneca, Abbott Vascular, Regardo Biosciences, and Abiomed. Dr. Czeisler has received consulting fees from or served as a paid member of scientific advisory boards for: Boston Celtics; Boston Red Sox; Citgo Inc.; Cleveland Browns; Merck; Novartis; Purdue Pharma LP; Quest Diagnostics, Inc.; Teva Pharmaceuticals Industries Ltd.; Valero Inc.; and Vanda Pharmaceuticals, Inc.; owns an equity interest in Lifetrac, Inc.; Somnus Therapeutics, Inc.; and Vanda Pharmaceuticals, Inc.; has received royalties from McGraw Hill, Penguin Press/ Houghton Mifflin Harcourt, and Philips Respironics, Inc. Dr. Czeisler has also received grants and research support from Cephalon Inc., National Football League Charities, Philips Respironics, ResMed Foundation, San Francisco Bar Pilots and Sysco; has received lecture fees from AASM (American Academy of Sleep Medicine); AADSM (American Academy of Dental Sleep Medicine); Harvard School of Public Health; Integritas Communications Group; Montefiore Medical Center; Stanford Center for Sleep Sciences and Medicine and the University of Buffalo. The Harvard Medical School Division of Sleep Medicine (HMS/DSM), which Dr. Czeisler directs, has received gifts from many outside organizations and individuals including: Concord Music Company, Delos Living, Flux Software, Jordan's Furniture, King Koil, Leggett & Platt, Merck Neurosciences, Metro Naps, Novartis Consumer Health, Optum, Patient Point, Philips Home Healthcare Solutions, ResMed, Simmons Bedding, Sleep Apnea Treatment Centers of America, Sleep Med, Turner Broadcasting, United Healthcare Clinical Services, and Vanda Pharmaceuticals. The HMS/DSM Sleep and Health Education Program has received Educational Grant funding from Cephalon Inc., Takeda Pharmaceuticals, Sanofi-Aventis, Inc. and Sepracor, Inc. Dr. Czeisler is the incumbent of an endowed professorship provided to Harvard University by Cephalon, Inc. and holds a number of process patents in the field of sleep/circadian rhythms (e.g., photic resetting of the human circadian pacemaker). Since 1985, he has also served as an expert witness on various legal cases related to sleep and/or circadian rhythms including matters involving Bombardier, Inc.; FedEx; Greyhound; Michael Jackson's mother and children; Purdue Pharma, L.P.; and United Parcel Service (UPS). Dr. Wang has uncompensated research relationships with Eli Lilly, Daiichi Sankyo, AstraZeneca, Gilead Sciences, and GlaxoSmithKline. Dr. Spertus has a contract with the American College of Cardiology Foundation to analyze NCDR data. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. The views expressed in this manuscript represent those of the authors, and do not necessarily represent the official views of the NCDR or its associated professional societies identified on the NCDR website.
- Abbreviations and Acronyms
- confidence interval
- interquartile range
- odds ratio
- percutaneous coronary intervention
- Thrombolysis In Myocardial Infarction
- Received August 7, 2014.
- Accepted August 28, 2014.
- American College of Cardiology Foundation
- O'Gara P.T.,
- Kushner F.G.,
- Ascheim D.D.,
- et al.
- Moussa I.,
- Hermann A.,
- Messenger J.C.,
- et al.
- Dehmer G.J.,
- Weaver D.,
- Roe M.T.,
- et al.
- Rao S.V.,
- McCoy L.A.,
- Spertus J.A.,
- et al.
- Brennan J.M.,
- Curtis J.P.,
- Dai D.,
- et al.,
- for the National Cardiovascular Data Registry
- ↵Bergstralh E, Kosanke J. Gmatch macro. October 2013. Available at: http://mayoresearch.mayo.edu/mayo/research/biostat/sasmacros.cfm. Accessed October 13, 2013.
- Jewett M.E.,
- Kronauer R.E.