Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
6 Contributors to Error in the Training Environment Residents can make errors, but the proportion of errors they make relative to those of other healthcare workers is unknown. Inexperience, fatigue, inadequate supervision, workload intensity, and other work system fac- tors (poor handover practices, inadequate medication labeling) contribute to errors by residents as they may for all health care workers. Data are insufficient to determine the relative contribution of each of these factors. Because residents are in supervised training programs and work within teams, many mistakes can be intercepted before they can harm patients. Uncertainty surrounds the impact of the 2003 reduction of resident duty hours on patient safety (adverse patient outcomes) and whether further adjustments to duty hours might diminish unsafe conditions (e.g., sleep deprivation) and reduce errors. The few national studies that have at- tempted to capture the impact of duty hour reform show no evidence of harm as measured by mortality rates. A well-designed randomized trial in two intensive care units of a single institution found a reduction in rates of serious medical error committed by first-year residents when their extended duty periods (up to 30 hours) were reduced to 16 hours, total weekly work hours were also reduced, and they obtained more sleep. The study found no statistically significant difference in unit-wide preventable adverse events or patient mortality between the reduced duty hour and standard hours. Nor was it able to isolate the effect of the shorter shift from reduced total workweek hours, increased sleep, having an additional intern, or increased handovers. A larger-scale, multicenter trial with suf- ficient statistical power would be necessary to confirm the positive findings in other settings and for residents in other training years. 179
180 RESIDENT DUTY HOURS This chapter examines what is known about the relationship between resident duty hours and patient safety. By definition the performance of trainees is imperfect as they learn, and they, just as other healthcare profes- sionals, will make errors. The response of the system to those errors and its actions to prevent future errors determine the safety of patients. First, this chapter discusses what is known about the overall frequency of medi- cal errors in hospitals by all staff and the resulting patient harm. Then it examines what evidence is available on the relative contribution of residents to the overall patient safety burden in teaching hospitals, and examines whether the degree to which resident fatigue contributes to the occurrence of error can be ascertained. The chapter continues with a discussion of the results of two natural experiments (the 1989 New York State and the 2003 Accreditation Council for Graduate Medical Education [ACGME] national duty hour reforms). Then a detailed review follows of the effects of an interventional study in which both total duty hours and the 30-hour duty period were further constrained from the limits allowable under the 2003 ACGME duty hour rules. Finally, literature on how other factors contribute to hospital errors, including the influence of poorly designed work systems on individual performance is considered. The discussion that follows presents research that helps answer five broad questions: 1. Do residents make errors that contribute to patient harm? 2. Is resident fatigue from long duty hours among the most significant risks to patient safety? 3. Did the 2003 reduction in resident duty hours affect patient safety? 4. Would further reductions in resident duty hours improve patient safety? 5. What factors in the resident work and learning environment con- tribute to error? The committeeâs answers to these questions will be drawn together in this chapter in a final section of conclusions. The next chapter (Chapter 7) looks to the human performance and sleep literature on how adults perform under scheduling practices that contribute to sleep deprivation, and con- tains the committeeâs recommendations on adjustments to duty hours. MEASURING HOSPITAL-BASED ERROR RATES AND RESIDENT INVOLVEMENT This Institute of Medicine (IOM) study grew out of questions about how significant a part residents play within the universe of hospital errors
contributors to error 181 that affect inpatients and to what degree the long duty hours and associated fatigue contribute to making errors (Dingell et al., 2007). The purpose here is to determine what is known scientifically about resident-associated errors and the degree to which fatigue and sleep deprivation of residents affect patient safety. Lessons learned from resident errors may reveal approaches for improving overall patient safety. Evidence on the subject is limited to a few studies. Measuring Patient Safety Before beginning, it is important to understand basic terms and ap- proaches used in discussing and measuring patient safety. Defining Medical Errors A spectrum of medical errors may occur during the treatment and care of hospital patients. If it is a very serious error, death, injury, or other preventable harm (e.g., delays in treatment, extended days in hospital, complications) could result if an error is not intercepted and corrected. Other errors may have no or very little impact on a patientâs condition or may be intercepted before they reach the patient and cause harm. The 2000 IOM report To Err Is Human: Building a Safer Health System presents an extensive analysis of safety and errors, based in large part on the research of James Reason and Charles Perrow. The framework, terms, and definitions used here are from that report (see Box 6-1). Measuring Medical Errors The measurement of patient safety is neither easy nor cost-free, and the ideal method for system-level surveillance has not been established. There are several types of measures commonly found in the literature that are used to assess patient safety (freedom from accidental injury). These include measuring the following: â¢ The occurrence of errors, â¢ The occurrence of adverse events (AEs) and preventable adverse events (PAEs), and â¢ Patient outcomes such as injury or death or length of stay in the hospital. Errors with the potential to harm patients tend to be classified in stud- ies according to their seriousness and category (e.g., medication, diagnostic, procedural, or other errors). Different approaches to collecting data both
182 RESIDENT DUTY HOURS BOX 6-1 Taxonomy of Errors Error: â.ââ.ââ.ââfailure of a planned action to be completed as intended (i.e., error of execution) or the use of a wrong plan to achieve an aim (i.e., error of planning)â (p. 28). An error of execution could be an error of omission of an essential step, a critical piece of data, etc.; could be caused by a poorly designed system requiring staff to âwork aroundâ the design fault or miscommunications; an error of planning could result from a misdiagnosis or lack of knowledge about the patientâs medical problem. Some errors are caught and corrected before they harm the patient. Harm or adverse event: An unintended physical injury resulting from or contrib- uted to by medical care rather than the underlying condition of the patient, that requires additional monitoring, treatment, or hospitalization or results in death. Not all adverse events are caused by errors. Preventable adverse event (PAE): âAn adverse event attributable to errorââ.ââ.ââ.â (p. 28). Sentinel event: An unexpected occurrence (which may or may not result from an error) in a hospital patientâs case, including actual or risk of death or serious physical or psychological injury (Joint Commission, 2007). Negligent adverse event: A subset of preventable adverse events that satisfy a legal standard of negligence (i.e., the care provided did not meet the standard of care reasonably expected of an average physician qualified to care for the patient) (p. 28). Safety: â.ââ.ââ.ââfreedom from accidental injuryâ (p. 58). SOURCE: IOM, 2000. for internal hospital quality improvement efforts and for research purposes capture different pieces of data but not a whole picture of patient safety or the universe of error. Data sources include (1) voluntary reporting by patients and families; (2) mandatory or voluntary but facilitated reporting systems for healthcare workers; (3) direct, prospective observation of work being done in the hospital; (4) retrospective review of medical records us- ing formal criteria or a âtrigger toolâ approach (i.e., clues in data that help predict adverse events) (Classen et al., 2008; Griffin and Classen, 2008); (5) use of administrative data on average length of stay, complication rates, readmission rates, and mortality; and (6) hybrid approaches that combine two or more of these methods.
contributors to error 183 No one method of data collection is ideal. The method used to identify medical errors and assess the preventability of a patientâs death in the stud- ies that produced the early IOM estimates used trained physicians conduct- ing a structured implicit review of medical records. This method has been shown to have a low interrater reliability and other limitations (Hayward and Hofer, 2001), although other studies have found similar rates of pre- ventable deaths. In recognition of this fact, institutions and researchers are increasingly employing a combination of different methods for collecting data on errors and analyzing them (Bates et al., 1995; Rothschild et al., 2005). In fact, one study that observed staff in a medical care unit and a coronary intensive care unit (ICU) reported that 62 percent of identified incidents were found through direct observation, 49 percent through chart review, 15 percent through solicited staff reporting, 7 percent through pharmacy reports including adverse drug event monitoring, and 4 percent through formal incident reporting (Rothschild et al., 2005). Only 23 per- cent of these events were identified by more than one approach. The common feature of these methods is the reliance on frontline pro- vider knowledge and description of the patientâs treatment and condition to inform voluntary or mandatory reporting systems, or to record direct or indirect observations of care (e.g., medical records, non-participant observ- ers). The reproducibility and precision of measurements of AEs and PAEs are limited (Classen et al., 2008; Hayward and Hofer, 2001). In particular, the determination of preventability is subjective and can change based on the state of medical knowledge available at the time of assessment. Error-Reporting Systems While national data on errors and PAEs are nearly nonexistent, more information exists at the hospital level since most now have voluntary error-reporting systems. The Joint Commission requires hospitals seek- ing accreditation to implement a voluntary reporting system for sentinel events, to conduct a root-cause analysis of reported events, and to prepare a corrective action plan to avoid similar incidents in the future (Joint Com- mission, 2007). These error-reporting systems can provide useful data, but they do not define the universe of errors, only those events recognized as problematic and reported by an observer or participant. Underreporting appears to be a common problem; such systems may detect fewer than 10 percent of adverse events (Classen et al., 2008; Rothschild et al., 2005), but the data provided nonetheless can have important uses to the reporting facility when they are embedded in a vigorous error elimination program. Such voluntary systems focus on the circumstances surrounding the adverse event and the systems involved, rather than identifying the individuals in- volved. Hence, even well-supported reporting systems do not typically note
184 RESIDENT DUTY HOURS whether a resident was involved with the patientâs care. Also, because of the complexity associated with some adverse events, it may be difficult to attribute the event to a specific individual or even to know exactly when it was committed. Compliance with voluntary reporting systems by physicians and other clinicians depends in part on the importance given to safety issues by the organizationâs leadership, whether such data (when gathered) are actually used in respected improvement efforts, and importantly whether workers feel safe to discuss errors without fear of punishment, retribu- tion, or other negative consequences (Garbutt et al., 2008; Kaldjian et al., 2008). These issues are discussed in Chapter 8. Although voluntary reporting systems cannot be used to define the frequency of harmful and other medical errors, they can be an important source of information to hospital leaders for identifying vulnerabilities in their systems that should be considered for corrective action. Along with risk management reports, patient complaints, error reports, quality assurance audits, and quality improvement reports, such systems can indicate areas for more detailed retrospective review, which can identify many more adverse events (Griffin and Classen, 2008). Error-reporting systems can provide data to assist in priority setting for quality improvement projects. The committee believes strongly that they can also be of educational value to doctors in training and should become an integral part of residency programs, as discussed in Chapter 8. Determining the Universe of Errors and PAEs with Limited Data As background for the committeeâs study of the impact of residentsâ duty hours on patient safety, it would be useful to follow a chain of inquiry and quantify, in order, the universe of medical errors, medical errors made in hospitals, medical errors made by residents, and medi- cal errors made by residents in which fatigue is a contributing factor. The universe of medical errors affecting patient safety would encompass PAEs as defined earlier, including both fatal preventable errors and the larger number of nonfatal preventable errors. The data to determine the universe of errors and the subelements in the above-mentioned hierarchy are not available to present a full picture. This lack inhibits the ability of the medical community to track and guide progress on patient safety. It has constrained the ability of the committee to answer fully some of the important questions put forth by the sponsors of this inquiry. Nonetheless, this section of the chapter gathers available data to paint a partial picture of the relationship between residents, errors in hospitals, and patient safety.
contributors to error 185 Estimates of PAEs U.S. short stay, non-federal hospitals treated and discharged 35 million inpatients in 2006 (DeFrances et al., 2008) and can produce miraculous cures, but an estimated 44,000-98,000 patients die from preventable errors (IOM, 2000). The broad range of that estimate reflects, in part, the meth- odological challenges mentioned above. The estimate of deaths was based on studies in which researchers examined hospital medical records from large samples of admissions in New York, Colorado, and Utah to determine whether the patients had experienced AEs as a consequence of medical er- rors (Brennan et al., 1991; Leape et al., 1991). A later study determined that 2.9 percent of admissions in Utah and Colorado and 3.7 percent of admissions in New York State experienced an AE; that 53 percent of Utah and Colorado events and 58 percent of the events in New York were at- tributable to errors and therefore were PAEs (Thomas et al., 1999). Another study by Thomas and colleagues determined that the AE rates in Utah and Colorado varied by teaching status: 4.0 percent in major teaching hospitals, 3.9 percent in minor teaching hospitals, and 2.5 percent in non-teaching and private hospitals. The study did not focus on case mix differences among individual hospitals or categories of hospitals. The researchers did not present sufficient data to explain the variation based on their available data (Thomas et al., 2000a). The estimated number of deaths resulting from PAEs was extrapolated from 1992 data by applying the death rates due to errors in the three states noted to the total of national hospital admissions in 1997. The committee uses the Thomas study (1999) as the basis for cost estimates of PAEs discussed in Chapter 9. Experts believe that the rate of preventable deaths has not improved substantially since the report To Err Is Human brought these issues to the publicâs attention in 2000 (Leape and Berwick, 2005). A significant and unsatisfactory level of errors is also indicated by several smaller studies of medical errors in a single hospital or hospital service since that time (AHRQ, 2002; Forster et al., 2003; Hayward and Hofer, 2001; IOM, 2006; Leape and Berwick, 2005; Rothschild et al., 2005). No recent estimate of the universe of errors nationwide exists, and because studies use different definitions of errors and PAEs and a variety of inconsistent methodologies for identifying PAEs and calculating error rates, their results cannot be aggregated. Assessing Patient Safety and Quality In the absence of a national error-reporting system, several commercial organizations as well as the Centers for Medicare and Medicaid Services (CMS), the Agency for Healthcare Research and Quality (AHRQ), and the
186 RESIDENT DUTY HOURS Commonwealth Fund have developed alternative methods for assessing quality and safety using existing data sources. CMS posts provider-level quality measures, including indicators for hospitals, nursing homes, home health providers, and dialysis facilities to help consumers make more in- formed choices (HHS, 2008). AHRQ created national estimates of hospital quality from existing data sources for its annual National Healthcare Qual- ity Report, which includes some indicators of safety, but not errors. For example, a composite indicator of selected generally avoidable postopera- tive complications shows that such adverse events occurred in 6.55 percent of cases in 2005, and that nearly one-quarter of surgical patients did not receive appropriately timed antibiotics (AHRQ, 2007). The improvements in quality according to a variety of ambulatory and hospital indicators used in AHRQâs National Healthcare Quality Reports amounted to only 1.5 percent per year between 2000 and 2005 (Brady et al., 2008). The Com- monwealth Fund uses a safety indicator for U.S. hospitalsâa construction of unexpected mortality, calculated by Jarmanâthat it tracks over time (Commonwealth Fund, 2008). The U.S. rate shows an improvement of 19 percent in the 2004-2006 period compared to 2000-2002. Nonetheless, both of these quality reports indicate the persistence of significant hospital mortality and injury related to conditions that generally should be avoid- able or should be caught and treated before the patient dies, indicating the continuing need for improvement in patient care. Errors and PAEs Involving Residents The above âclassicâ studies involving statewide hospital AEs do not re- port errors or PAEs that were related specifically to residentsâ care although there would appear to be higher AE rates in teaching hospitals based on these data alone (Brennan et al., 1991; Leape et al., 1991; Thomas et al., 1999, 2000b). A more recent set of papers by Rothschild, Landrigan, Lock- ley, and colleagues examined resident error through a randomized trial in two critical care units at a single institution (Landrigan et al., 2004; Lockley et al., 2004; Rothschild et al., 2005). This section discusses the studies with a focus on the baseline incidence of errors while a later section of this chap- ter examines the effect of a scheduling intervention on error and PAE rates. Malpractice negligence claims provide another source of data (Gandhi et al., 2006; Regenbogen et al., 2007; Singh et al., 2007). Incidence of Error and PAEs in ICUs Rothschild (2005) and colleagues conducted a prospective observa- tional study of two critical care units at a major urban teaching hospital. This study focused on errors made by all caregivers when first-year residents
contributors to error 187 were following a traditional duty hour schedule. The authors found that 20.2 percent of patients suffered at least one AE and 45 percent of those AEs were found to be preventable (Rothschild et al., 2005). The authors note that their definition of an AE is more inclusive than the earlier study by Brennan et al. (1991) cited above and that the ICU setting of their trial would be expected to have higher medical error rates than other areas (Beckmann et al., 2003). The unit-wide error rates per 1,000 patient-days were 80.5 for all AEs, 36.2 for PAEs, and 149.7 for serious errors. Serious errors did not always result in harm to patients âeither because the patient had sufficient reserve to buffer an error (nonintercepted serious error) or because the error was caught before reaching the patient or before harm developedâ (Rothschild et al., 2005, p. 1697). The Rothschild data along with the national reports from AHRQ, CMS, and the Commonwealth Fund support the committeeâs conclusion that 8 years after publication of the IOM report To Err Is Human (2000), patient safety remains a serious issue in the United States (AHRQ, 2007; Commonwealth Fund, 2008; HHS, 2008). The complementary article by Landrigan et al. (2004) reporting on data collected in the same setting but for a slightly shorter period describes differences in error rates unit-wide and for first-year residents. It found the rates per 1,000 patient-days involving all staff unit-wide were 38.6 for PAEs and 193.2 for serious errors. Incidents involving first-year residents working a schedule with overnight call every third night appear to make up a substantial portion of the reported errors, including 20.9 per 1,000 patient-days for PAEs and 136.0 per 1,000 patient-days for serious errors (Landrigan et al., 2004). Rothschild notes that compared to the unit-wide data, the âdata on interns were somewhat more comprehensive because of the presence of the observersâ who kept the interns under direct continuous observation, but that the unit-wide results were within the range identified by other studies (Rothschild et al., 2005, p. 1695). Thus, the error rates for other workers may have been underestimated relative to the error rates of first-year residents. Errors and PAEs in Malpractice Claims Another study that identified errors associated specifically with doctors in training (both residents and fellows) is based on 1,452 closed malpractice claims from five liability insurers in different parts of the country (Singh et al., 2007). Malpractice claims represent only a small proportion of er- rors and AEsâthe more serious AEs for which negligence is assessed. It is unclear in what other ways these data might differ from the universe of PAEs. Singh identified 889 cases that reviewers determined to have included both an error and an adverse outcome; 240 (27 percent) involved trainees.
188 RESIDENT DUTY HOURS Residents were involved with 87 percent of the 240 cases involving trainees, and fellows were involved with 13 percent of those cases. Multiple train- ees could have been involved in a single case, with interns involved in 13 percent of the 240 cases. The studyâs physician reviewers considered these doctors in training to have had at least a moderately important contributory role in those cases with a PAE. A study of 307 diagnosis-related ambulatory care malpractice claims closed between 1984 and 2004 found that 181 such claims involved diag- nostic errors that led to adverse outcomes (Gandhi et al., 2006). Of the 181 cases, trainees (intern, resident, or fellow) were identified as involved in 20 percent of them by trained reviewers. The study also identified several causes of breakdowns in the diagnostic process and concluded that multiple factors were involved. Researchers in a different study examined surgi- cal malpractice claims, selecting a random sample of 444 cases for closer study. Among the 52 percent (n = 133) that included technical errors, the researchers determined that 9 percent involved poorly supervised residents (Regenbogen et al., 2007). Conclusion About Whether Residents Make Errors These studies provide enough evidence to answer the question: Do residents make errors that contribute to patient harm? Common sense and these studies lead to the conclusion that the answer is, Yes, they do. Additional information from resident surveys confirms this as well (Jagsi et al., 2005, 2008; Wu et al., 2003). Without more quantitative data, it is impossible to determine what proportion of all errors or what proportion of PAEs involve residents. Consequently, the magnitude of the impact of residents on patient safety is unknown. FATIGUE AS A CONTRIBUTOR TO ERROR A principal aim of this study is to determine the degree to which resi- dent fatigue from long duty hours poses a significant risk to patient safety and whether there are interventions that might reduce that risk. As ÂHoward and colleagues have observed, âcontinuous operational demands [of pro- viding access to health care in hospitals 24 hours a day] present unique physiologic challenges to the humans who are called on to provide safe operations within these systemsâ (Howard et al., 2002b, p. 1281). While long work hours and fatigue appear to play a role, other systemic factors also contribute. Resident reports give some insight into how great a factor they believe fatigue to be. In a survey of two large teaching institutions just before the required 2003 ACGME duty hour limits were in force, medi- cal and surgical specialty and subspecialty residents were asked what the
contributors to error 189 contributing factors were for mistakes related to AEs. They reported that long work hours were a contributing factor in 19 percent of the mistakes observed, but they also noted that lack of supervision (20 percent), faulty handovers (15 percent), large patient caseloads (12 percent), and cross- covering too many patients (5 percent) were important factors (Jagsi et al., 2005). Working more than 80 hours in the past week was a significant predictor of caring for a patient with an AE in the last week (odds ratio 1.8) (Jagsi et al., 2005). Chapter 7 details the evidence base that establishes the link between fatiguing aspects of resident work-rest schedules and what is known about how fatigue affects human performance and the propensity for error. Assessing Incidence of AEs Involving Fatigue This section examines data from the U.S. Department of Veterans Affairs (VA) and from malpractice claims to evaluate the contribution of fatigue as a factor. The VA offers residency training through approximately 8,800 residency positions in its facilities (9 percent of U.S. total), and be- cause residents from other facilities rotate through the VA, this training reaches about one-third of residents in training in any single year (Chang, 2007). The VA has a heavy emphasis on patient safety and has trained its staff in the value of reporting both AEs and close calls. The system has ac- cumulated more than 10,000 root-cause analyses (RCAs) of individual seri- ous incidents or groups of events since its inception in 1999. The analyses tend to look beyond the individuals involved with an AE to the underly- ing systemic causes. The database is not designed to identify the specific involvement of residents. It does, however, include fatigue as a âcauseâ choice on its structured data collection tool. Fewer than 4.5 percent of the VA RCA reports included fatigue as an associated factor and 0.7 percent included a more extensive discussion of fatigue-related causation. A review of a random sample of 4,742 reports drawn from approximately 180,000 reports from the same time period concerning less serious safety incidents showed that 1.0 to 3.3 percent included fatigue-associated causes., It is unknown what percent of those cases associated with fatigue included fatigued residents because the VA does not routinely track residency status of the involved parties. Fatigue related to medical errors is recorded in some cases in the Singh study of malpractice claims discussed above: 5 percent (n = 12) of the trainee â Personal communication, J. P. Bagian, Director, VA National Center for Patient Safety, Department of Veterans Affairs, February 11, 2008. â Personal communication, J. P. Bagian, Director, VA National Center for Patient Safety, Department of Veterans Affairs, February 14, 2008.
190 RESIDENT DUTY HOURS cases (less than 2 percent of PAE claims studied, n = 889) and 1 percent (n = 6) of the nontrainee PAE cases included fatigue as a factor. Since the fatigue of the provider is not routinely noted in medical records and legal case notes, it is not possible to know how frequently it was a factor but not noted as such; thus, the actual percentage of negligence cases in which the trainee was affected by fatigue is unknown. It is also possible that fatigue would be noted more frequently if healthcare workers were more aware of the role of fatigue and how to assess its role in creating unsafe conditions, and if workers were informed about the importance of adequate sleep. Such courses have been developed (e.g., those by NTSB ) in response to fatigue-related incidents in other industries (Rosekind et al., 1994). Better Conditions for Patient Safety Through Reducing Fatigue One survey-based study and one prospective observational study of ICUs suggest that shorter work hours may lead to less fatigue and, as a result, to better patient safety (Jagsi et al., 2008; Landrigan et al., 2004). The survey incorporated questions on the relationship of duty hours and fatigue to the quality of care delivered, patient safety, and AEs. Responses of residents in 76 specialty and subspecialty programs at two institutions were obtained before and after the 2003 ACGME duty hour reforms. Residents in programs that reduced their workweek by at least 5 hours were found to be less likely to violate the 80-hour limit than prior to 2003 (16.6 percent vs. 44.0 percent) and less likely to have worked more than 30 continuous hours in the past week (11.4 percent vs. 40.8 percent). Days of significant fatigue in the past 4 weeks remained but were less (6.5 vs. 8.7). Fewer residents reported that âfatigue frequently or always affected the quality of care they providedâ (14.6 percent vs. 9.2 percent) and that âfatigue frequently or always impacted the safety of patients that they cared forâ (7.0 percent vs. 2.9 percent) and these differences are significant when compared to programs that did not reduce work hours (Jagsi et al., 2008, p. 496). The ICU environment examined in the Rothschild study was the subject of a change in the resident duty schedule that resulted in fewer work hours per week (a mean 19.5 hours less per week) and a shorter consecutive duty period (no shifts over 16 hours). This allowed more hours of sleep (mean 5.8 hours per week) and presumably more rested interns (Landrigan et al., 2004; Lockley et al., 2004). Interns working on the intervention work schedule made 36 percent fewer serious medical errors (p = .001), but the difference in rates of PAEs was not statistically significant in comparing the two groups (Landrigan et al., 2004). Interns on the intervention schedule also had fewer attentional failures as measured by slow rolling eye move- ments (Lockley et al., 2004). Rothschild et al. (2005) found that 53 percent
contributors to error 191 of the performance errors were slips (unintended acts) or lapses (omitted acts) rather than knowledge-based or rule-based errors (e.g., not follow- ing a protocol). Because sleep and fatigue were not the only factors that changed during the intervention, it is not possible to attribute all of the reduction in errors to reduced fatigue, but this study provides a substantial contrast to the VA and Singh data, which had found relatively low rates of fatigue noted in relation to PAEs. Conclusion About Whether Fatigue Is a Significant Factor in Error Clearly fatigue is a factor in some of the errors by medical workers in general and by residents in particular given their work-rest schedules. It is unresolved exactly what percentage of all errors that fatigue-based errors compose, with these sources reviewed suggesting a wide range from 5 to 36 percent. The potential impact of fatigue in the ICU studies (Landrigan et al., 2004; Lockley et al., 2004; Rothschild et al., 2005) was substantial, whereas the VA and the malpractice studies noted relatively little mention of fatigue as a factor in error reports. Overall, the committee concludes that the existing data are insufficient to determine if the current duty hours of residents and the fatigue resulting from them are the most significant causal factors for errors committed by residents or if resident errors occur more frequently than errors committed by other health workers. Assessing Fatigue and Performance After Extended Duty Periods A number of studies have noted poorer performance by residents post- call, but others find no difference. Friedman and colleaguesâ classic study showed that interns made almost twice as many errors when reading elec- trocardiograms after being up for 24 hours than when they had a night of sleep (Friedman et al., 1971). Additional studies also point out increased technical errors in simulated laparoscopic surgical skills after being up all night (Eastridge et al., 2003; Grantcharov et al., 2001), decrease in cogni- tive skills (Jacques et al., 1990; Robbins and Gottlieb, 1990) as well as in memory attention and coordination in surgical residents post-extended duty period (Kahol et al., 2008), and reduced psychological well-being and problems with alertness and coordination after an extended shift (Leonard et al., 1998). When Jacques et al. (1990) examined the effects of sleep loss on resident performance on the American Board of Family Practice in- training examinations, they found that the difference in test scores after a night without sleep was equivalent to the difference between third-year and first-year residentsâ performance. Further, researchers have found that even with a call frequency no more often than every fourth night, which is typical of call schedules under the
192 RESIDENT DUTY HOURS ACGME 80-hour limit, residents do not fully recover between nights of overnight call (Saxena and George, 2005). This may imply that residents are not sufficiently recovering their lost sleep time between extended duty periods (Saxena and George, 2005). Howard et al. (2002a) have confirmed that residents prior to the 2003 duty hour limit were as sleepy before and after extended duty shifts and that their level of sleepiness matched that of persons with clinical sleep disorders. Obtaining sufficient sleep returned the residents to normal sleep levels (i.e., 2 hours more sleep per day over 4 days). Saxena et al. (2005) suggest that it is unlikely in an emergency that resident judgment would be impaired even in a sleep-deprived state, but that more routine tasks (e.g., medication reorders) might be missed, potentially leading to more serious consequences later. Indeed, a number of studies have found that medication errors are among the most com- mon errors that residents make (Jagsi et al., 2005; Landrigan et al., 2004; R Â othschild et al., 2005). Other studies report no deficit in resident performance after being up all night (Ellman et al., 2005; Howard et al., 2002b; Jakubowicz et al., 2005). These results are not consistent with the extensive literature on human performance and acute sleep deprivation presented in Chapter 7. Several review articles note that research examining the effects of fatigue on the performance of healthcare personnel, not just residents, do not always come to the same conclusion, and these articles ascribe this to definitional and methodological differences (Howard et al., 2002a; Veasey et al., 2002; Weinger and Ancoli-Israel, 2002). Differences in reported results may have to do with the way sleep deprivation is defined, the degree of chronic sleep deprivation present both pre- and post-call (i.e., extended duty periods), and the presence of compensating factors that may have helped mitigate the performance of sleep-deprived residents (e.g., presence of rested and expe- rienced team members). For example, in a retrospective review of 10 years of cases, Ellman et al. (2005) concluded that thoracic residents in an acute sleep-deprived state had patient outcomes (morbidity or mortality) and operative efficiency comparable to those who had not been on call the pre- vious evening. This matches their previous findings for attending physicians (Ellman et al., 2004). In these retrospective studies, the definition of acute sleep deprivation was based on whether a resident or attending had started or ended an operation the previous night; there was no determination of the chronic sleep deficit for either population although it was assumed that in a university-based teaching program the attendings would not have a chronic sleep deficit. Just 3 percent of resident cases were performed by residents who met the acute sleep deprivation criteria. The report does not indicate whether the resident was assisted by an attending, which may have buffered the effects of sleep deprivation for the resident or to what degree resident errors were intercepted by an attending (Ellman et al., 2005). One
contributors to error 193 could assume that a rested attending surgeon supervising these procedures would compensate for suboptimal resident performance and help prevent errors or surgical inefficiency. Veasey and colleagues (2002) have commented on how residents were chronically sleep deprived in their baseline state prior to the 2003 duty hour reform. Kiernan et al. (2006) have suggested that some problems, such as declines in mood, previously associated with acute sleep deprivation may have been ameliorated by the 2003 duty hour limits. Their rationale is that the limits may have helped to reduce chronic sleep deprivation, giving resi- dents sufficient reserve to better tolerate a night without sleep. Yet others, such as Saxena and colleagues, suggest that even on an every fourth night overnight call schedule, residents do not sufficiently recover from their sleep loss (Kiernan et al., 2006; Saxena and George, 2005). As Chapter 7 discusses, the buildup of sleep loss contributes more heavily to impaired states of alertness, cognition, and performance. IMPACT OF REDUCED DUTY HOURS ON ERROR RATES AND PATIENT SAFETY This section addresses two central questions: (1) Did the 2003 reduc- tion in resident duty hours improve patient safety? (2) Would further reductions in resident duty hours improve patient safety? As noted above, attempting to isolate the effect of reducing resident duty hours on patient outcomes is difficult. For studies on a national level, drawing a direct link between hours worked by residents and patient outcomes is problematic, given available data, and for studies at individual institutions, it is difficult to obtain a sample size sufficiently powered to find statistically significant differences in mortality or other patient safety measures. The relationship between duty hours and patient safety has been a central element in the debate over reduced hours. Expectations were raised in 2003 that reducing duty hours would improve patient outcomes and safety, but others pre- dicted that reducing hours could negatively affect patients in the short and long term because it required more frequent handover of patients from one resident to another. Such handovers, or âhandoffs,â are considered poten- tially risky for loss of information and continuity of care, and could also lessen the overall learning experience of residents (Fischer, 2004; Petersen et al., 1994, 1998). Natural Experiments A number of studies look at the effect of implementation of duty hour reforms throughout a single state or nationwide without controlling for the specifics of a scheduling intervention. As shown in Chapter 3, there
194 RESIDENT DUTY HOURS are many different approaches to schedule resident hours, and there can be variations even within a single resident team that suit the specialty, patient characteristics, size of the resident team, system supports, and other fac- tors. Thus, these studies examine outcomes of practices that have naturally evolved within and across teaching institutions. On the positive side, the datasets in these studies are sufficiently large to detect changes in mortal- ity. Patient mortality became the center of attention in To Err Is Human and subsequent patient safety campaigns (e.g., the Institute for Healthcare Improvementâs [IHIâs] 100,000 Lives and 5 Million Lives campaigns) with projections of the number of lives potentially saved if there were focused attention on healthcare quality improvement (IHI, 2008; IOM, 2000; M Â cCannon et al., 2007). These studies do not provide information on how many hours per week residents actually work or sleep. New York State Studies When New York State instituted an 80-hour workweek and limited extended duty periods to 24 + 3 hours in 1989, it provided a laboratory for implementation and evaluation of such changes and led the way for eventual adoption of an 80-hour week nationwide (Howard et al., 2004; Laine et al., 1993). In their systematic review of the literature available before implementation of the 2003 ACGME rules, Fletcher and colleagues found few duty hour and patient safety studies that adequately addressed their two criteria: (1) examined a system change to address work hours, fatigue, or sleep deprivation and (2) included an outcome directly related to patient safety (e.g., mortality, morbidity, error) (Fletcher et al., 2004). Three well-designed, but not randomized, studies found that duty hour re- duction led to (1) more complications and test delays but neither increased nor decreased mortality (Laine et al., 1993), (2) a decrease in mortality in teaching hospitals that was equally apparent in non-teaching hospitals (Howard et al., 2004), and (3) more PAEs attributed specifically to hando- vers and cross-coverage of unfamiliar patients (Petersen et al., 1994). The first two studies of New York State are discussed below; the Petersen study is described within the context of continuity of care in Chapter 8. Summaries of the two New York studies looking at patient outcomes before and after the 1989 New York State duty hour limitations follow. Laine et al. (1993) examined patient outcomes in a teaching hospital with residents on a general medical service by comparing all admissions during October of 1988 (pre) and October of 1989 (post) (Laine et al., 1993). They found an increased number of patients with at least one complication (35 percent vs. 22 percent; p = .002) and delays in residents ordering diagnostic tests (17 percent vs. 2 percent; p < .001) after the reduction in duty hours. However, these decreases in quality-of-care metrics did not result in more
contributors to error 195 serious outcomes for patients. No difference was found for transfers of pa- tients to intensive care, length of stay, or disposition at discharge; the study did not have sufficient power to detect a statistically significant change in mortality. Howard and colleagues (2004) found that mortality declined for congestive heart failure, acute myocardial infarction, and pneumonia after New York Stateâs duty hour change, but this could not be attributed with confidence to duty hour reduction alone because mortality rates declined at both teaching and non-teaching hospitals in the state between 1988 and 1991; moreover, the study assigned teaching status to the hospital as a whole rather than to specific patients cared for by residents (Howard et al., 2004). Another large-scale study examining surgical mortality in New York also found no change over time (Poulose et al., 2005). Poulose and col- leagues examined changes in surgical outcomes based on five patient safety indicators and found no substantive change for New York State teaching hospitals compared to two control groups: non-teaching hospitals in the same state and teaching hospitals in another state that had not yet imple- mented an 80-hour limit (Poulose et al., 2005). The authors also suggest that this can best be interpreted as examining the systemâs global response to duty hour limits (e.g., schedule and supervision changes, substitution by others for residentsâ time) not resident work hours alone. The data examined were from the Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS) from 1995 to 2001; the choice of sur- gical patient safety indicators was deliberate because of a higher level of agreement between administrative data and actual occurrences in surgery than medicine (73-81 percent vs. 32-70 percent). Chapter 2 has noted the poor state of adherence to duty hour rules in New York State in the early years of implementation (DeBuono and Osten, 1998; Kennedy, 1998). Low levels of adherence suggest that the results of these otherwise well-designed studies comparing pre- and post-duty hours reform should be examined with caution since there may indeed have been little actual change in hours worked by residents. Additionally, New York State hospitals had very long lengths of stay and low managed care pen- etration, giving other reasons why the findings might not be generalizable outside of New York. Nationwide Studies Now this section turns to studies of the impact of the ACGME 2003 duty hour reforms. Three recent studies by Shetty and Bhattacharya (2007) and Volpp et al. (2007a,b) examining national mortality trends showed some improvement in mortality for medical patients but not surgical pa- tients after duty hour regulations (Shetty and Bhattacharya, 2007; Volpp
196 RESIDENT DUTY HOURS et al., 2007a,b). A fourth study by Prasad (2008) finds no change in ICU mortality after the duty hour reforms. Summaries of these studies follow. None of these studies were able to document actual hours worked by resi- dents in the facilities studied. Shetty and Bhattacharya (2007) compared changes, if any, in patient outcomes at teaching hospitals that should have been affected by the duty hour rule changes versus non-teaching hospitals. They examined a repre- sentative national dataset, the HCUP NIS, which is of sufficient size to have enough statistical power to detect changes in mortality. The researchers used non-teaching hospital services as a control (963,916 non-teaching patients and 548,029 teaching patients). To ensure that the teaching cases were indeed on teaching services, a cross-match was made to the presence of specific residency programs for each type of patient examined (e.g., in- ternal medicine, orthopedics). After the 2003 duty hour limitations, there was a small but statistically significant improvement for medical but not surgical cases on teaching services, specifically a â0.25% decrease in the absolute risk for death (p = .043), which corresponded to a 3.75% decrease in relative risk in medical patients per hospitalizationâ (p. 76). The oldest patients, those more than 80 years old, and those with infectious diseases were most likely to benefit in the period after duty hour changes. Mortality decreased as the number of residents in a facility increased. The authors offer several possible explanations of why there were no observed changes for surgical patients in spite of the fact that the 2003 duty hour reforms reduced surgical resident hours most substantially. Their reasons included a smaller set of surgical cases, which may have limited the statistical power to detect change for this type of patient and the possibility that work con- ditions at least in the operating room may not have changed. One critique of the study is the nature of the dataset since the HCUP NIS looks at dif- ferent hospitals each year and the dataset does not allow one to distinguish between single and multiple admissions for the same condition (Volpp et al., 2007b). Since mortality may occur not long after discharge from a hospital, Volpp and colleagues scrutinized both in-hospital and post-discharge rates. They looked at mortality for Medicare beneficiaries and VA patients in acute care hospitals in the first 2 years after implementation of the 2003 limits (Volpp et al., 2007a,b). The main outcome measure is all-location mortality within 30 days of a first hospital admission for acute myocardial infarction, stroke, gastrointestinal bleeding, congestive heart failure, general surgery, orthopedic surgery, or vascular surgery. These are the AHRQ Qual- ity Indicators that use mortality as an outcome measure. Each hospital is compared with itself over time. For Medicare beneficiaries, no significant change was found in the odds of risk-adjusted mortality in either the first or the second year of duty hour reforms based on data from the Medicare Provider Analysis and Review File (MEDPAR). There was a small increase
contributors to error 197 in the relative mortality for stroke in more intensive teaching hospitals, but it appeared to be part of an ongoing trend that is divergent from non-teach- ing hospitals and a trend that started before reform. In more teaching-intensive VA hospitals, there was a significant im- provement for AMI on its own and for the other medical conditions com- bined (with or without AMI) in the second year of duty hour reform but no improvement for surgery. The VA health system is the largest single pro- vider of residency training in the country. The authors suggest that a reason for the difference between the VA and Medicare patient outcomes may be a dose-response effect due to the markedly higher resident-to-bed ratios in VA hospitals than non-VA teaching hospitals, as well as differences in staffing models. They suggest that this may create a different balance between the consequences of resident fatigue and discontinuity in patient care in VA fa- cilities versus others so that the VA experience is not generalizable to other facilities. Also, there may be other nonmeasured factors that contribute to the difference; for example, VA hospitals have electronic medical records, which may help diminish communication problems in transitions of care. Although not stated explicitly, this study like the one by Shetty and Bhat- tacharya documents the advantage of being a patient in a teaching hospital; risk-adjusted mortality rates were generally lower for hospitals with higher resident-to-bed ratios. Prasad (2008) examined the pattern of adult in-hospital mortality in ICUs from July 1, 2001, through June 30, 2005, using the APACHE IV da- tabase, a voluntary multicenter ICU clinical registry, to determine whether there was an effect of the 2003 duty hour reforms on ICU mortality. They found a significant decline in risk-adjusted mortality in both teaching and non-teaching hospitals from before the implementation of duty hour rules to after. They eliminated patients from the sample whose care straddled the start date of the rules. The difference between the two settings over time was not significant: the adjusted odds ratio for mortality after the regulations was 0.89 (95 percent CI 0.87, 0.92; p < .001) overall, 0.88 (95 percent CI 0.85, 0.92; p < .001) in teaching hospitals, and 0.91 (95 percent CI 0.87, 0.95; p < .001) in non-teaching hospitals. The authors conclude that duty hour reforms did not have a positive or negative effect on major patient outcomes and it is possible that the positive and negative effects of reform may have offset each other, that mortality in the ICU environment may not have been sensitive to resident staffing patterns, or that ICUs may have made other compensating changes to maintain and improve patient outcomes (Prasad, 2008). Institution-Specific Studies Few studies to date have an adequate control group to isolate the specific effects of duty hours (e.g., Horwitz et al., 2007; Howard et al.,
198 RESIDENT DUTY HOURS 2004; Landrigan et al., 2004; Poulose et al., 2005). Studies comparing data both before and after the 2003 ACGME limits without control groups may falsely read improvement in error rates or patient outcomes as being related to duty hour reform when in fact they may have nothing to do with resident work hours but reflect national trends toward improved quality of care (Horwitz et al., 2007). Additionally, smaller institution-specific studies often have insufficient statistical power to detect changes in mortality (e.g., Landrigan et al., 2004). To control for temporal trends in practice and patient outcomes, Horowitz et al. (2007) carried out a retrospective single-center cohort study comparing outcomes for medical patients on a resident-hospitalist teaching service (n = 708) to a non-teaching service run by hospitalists (n = 2,954) (Horwitz et al., 2007). No adverse consequences for patients occurred under their new scheduling plan in which residents have no overnight call. The teaching service had a significant improvement relative to the hospital- ist service from 2002-2003 to 2003-2004 on three measures of mean net adjusted change: ICU utilization decreased by 2 percent, discharge to home or rehabilitation versus elsewhere increased 5 percent, and pharmacist in- terventions to prevent error decreased by 1.9 interventions per 100 patient- days. Readmission rates, length of stay, and medication interactions were not found to be significantly different. There was also insufficient statistical power to detect changes in mortality. The remaining variables (length of hospital stay, 30-day readmission rate, and drug-drug interactions) were consistent across both services. Bhavsar and colleagues performed a retrospective analysis of patient outcomes at a single facility for those with acute coronary syndrome, before and after duty hour regulation (Bhavsar et al., 2007). They assert that their program maintainedâif not enhancedâthe level of care because they had more well-rested residents to handle discharge planning as a result of their scheduling response to duty hour limits (an incremental increase in residents available at discharge through use of day float). No significant difference was detected for in-hospital patient mortality (4.2 percent before vs. 2.8 percent after, p = .23), but 6-month mortality (8.0 percent vs. 3.8 percent, p = .007) and risk-adjusted 6-month mortality (OR 0.53; 95 percent CI 0.28, 0.99, p = .05) improved. At the same time, there was increased adherence to quality prescribing practices for cardiac care at discharge and mean length of stay was reduced. There was no control group in this study and the cardiac care quality improvement program instituted at this facility may be at least part of the reason for the improvements rather than duty hours. Surgical programs traditionally have had much longer duty hours than medical programs, so the adjustment to 80 hours was expected to be more difficult for these programs. The choice of 80 hours was seen by some stakeholders as arbitrary and not responsive to the special demands
contributors to error 199 of surgery (Fischer, 2004). Despite these concerns, reports from several surgical programs found no change in mortality and other patient indica- tors or increased errors with the reduced work hours (de Virgilio et al., 2006; Kaafarani et al., 2005; Vaughn et al., 2008). DeVirgilio et al. (2006) examined mortality and morbidity for trauma patients at one institution before (July 1998-June 2003) and after (July 2003-June 2005) duty hour changes; adjusting to the reduced hours required an increase in their resi- dent complement and hiring others to do some of the tasks previously done by residents (de Virgilio et al., 2006). They conducted a pre-post study without a comparison group and found no significant difference in patient mortality during the periods before and after the implementation of duty hour rules, despite a larger volume of patients, a higher injury severity score (7.9 to 9.6, p < .0001), and a greater portion of penetrating trauma (14.85 to 17.6 percent, p < .0001) among patients.Â Morbidity and raw mortality data come from their Trauma and Emergency Medicine Information System (TEMIS).Â They also observed no decline in operative experience for the residents or in their success rate in passing the General Surgery Board Ex- amination. Thus, reassurance is given that there were no overt downturns in patient outcomes in this surgical program despite the reduction to an 80-hour week. Kaarafani and colleagues (2005) similarly found no worsening in mor- tality and morbidity in either vascular or general surgery at a single institu- tion based on surgical outcome data from the VA National Surgical Quality Improvement Program. Pre-intervention hours were longer (87-92 hours without a consistent 24 hours off every week) than the post-intervention period (80-87 hours from October 1, 2002, until January 1, 2003, and af- ter that 80-hour weeks until September 30, 2003) (Kaafarani et al., 2005). The number of cases with an attending present increased. In the same way, another redesigned surgery program using a combination of apprenticeship, small-team, and night-float models was able to increase operative volume, improve ABSITE scores for PGY-1s and PGY-2s, and maintain previous patient mortality levels. In addition to remodeling its schedule, the program added 0.2 FTE (full-time equivalent) of physician assistant and nurse posi- tions per resident (Schneider et al., 2007). Conclusion About Patient Outcomes After Implementation of 80-Hour Duty Week Smaller institution-specific studies allow easier identification of the actual duty hours worked by residents, how fatigued they may be, and the multiple programmatic changes made that help balance the reduction in resident hours (e.g., hire additional staff, remodel their education program, increase attending presence). These studies illustrate the complexity of teas-
200 RESIDENT DUTY HOURS ing out not only the impact of duty hours alone, but also the impact on pa- tient outcomes of different staffing configurations and scheduling practices whether in medical or surgical settings. Few studies to date have any type of concurrent control group (Horwitz et al., 2007; Howard et al., 2004). The national studies of mortality show that there is no evidence of wide- spread harm occurring after implementation of the limits (i.e., duty hour restrictions did not lead to an increase in mortality rates for the common conditions studied) and there may be modest improvements for medical if not surgical patients. Interventional StudyâReducing Intern Duty Hours in the ICU Setting The most rigorous scientific data on the direct impact of duty hours on patient safety available to the committee comes from three publications that describe overlapping aspects of the same prospective 1-year random- ized trial in 2002 and 2003 (Landrigan et al., 2004; Lockley et al., 2004; Rothschild et al., 2005). This trial compared âthe rates of serious medical errors made by interns while they were working according to a traditional schedule with extended (24 hours or more) work shifts every other shift (an âevery third nightâ call schedule) and while they were working according to an intervention schedule that eliminated extended work shifts and reduced the number of hours worked per weekâ (Landrigan et al., 2004, p. 1838). The âinterventionâ schedule had shifts with a maximum of 16 consecutive hours. The study followed a sample of 20 interns who were randomly as- signed to work 3-week rotations on both schedules in two ICUsâessentially a crossover experimental design. In contrast to many of the retrospective studies cited earlier, the authors carefully observed medical errors in real time, monitored and recorded actual hours worked, recorded hours slept, and measured intern fatigue. This was a well-designed and well-executed randomized controlled trialâalthough the randomization was only partial and the evaluations of medical errors could not be fully blinded. Medical error detection was by multiple means: primarily trained physician observ- ers, with voluntary staff reporting, chart review, and computerized event detection monitors. In this chapter, the focus is on the error reduction and patient safety aspects of this trial (Landrigan et al., 2004). Chapter 7 contains discussion of the associated sleep and fatigue data (Lockley et al., 2004). The study found that the intervention schedule with its shorter shifts resulted in more intern sleep time, decreased intern fatigue, and significantly fewer serious â addition to its close reading of the three published reports, the committee benefited from In testimony by some of the studyâs principal investigators and from follow-up written and oral communications with Dr. Landrigan, lead author on the Landrigan et al. (2004) paper.
contributors to error 201 errors by interns. While there was a reduction in intern-related PAEsâa measure of harm that reached the patientâthis outcome was not statisti- cally significant. Among the interns participating, serious medical errors decreased by 36 percent (136 vs. 100 per 1,000 patient-days, p < .001), while intern PAEs declined by 27 percent (20.9 to 16.5 per 1,000 patient- days, p = .21). The authors did not report the proportion of patients in the two arms of the trial who suffered PAEs. Improvements occurred across the various categories of medical errors observed for interns working on the intervention schedule of 16-hour shifts; they made statistically signifi- cantly fewer serious diagnostic and medication errors but not fewer serious procedural errors. The committee also noted that intern-related diagnostic errors showed the greatest reduction, from 18.6 to 3.3 per 1,000 patient- days (Landrigan et al., 2004). In addition to collecting data on errors made by interns, the study reported overall unit-wide error rates (serious errors 193.2 per 1,000 p Â atient-days on the traditional schedule versus 158.4 on the intervention schedule, p < .001), but the intense real-time error monitoring processes were applied only to interns so the overall error data may be less com- plete. Patient populations were similar in volume, severity, and complexity across the two schedules, but the study did not detect effects on patient mortality and unit-wide PAEs (not just those by interns), which remained the same (38.6 vs. 38.5 per 1,000 patient-days). The authors suggest that larger-scale, multicenter trials would be needed to gain sufficient power to confirm their findings. The committee concludes that, because of careful experimental design, the reported 36 percent reduction in the internâs rate of serious medical er- rors and other performance improvements appear to be largely a result of the intervention, rather than a result of confounding influences. The com- mittee also notes that the schedule intervention actually incorporated five changes, each of which may have contributed to error reduction: 1. Total duty hours per week were reduced from about 80 to about 60 hours. 2. The duration of long duty periods was reduced from about 30 to about 16 hours. 3. Sleep was significantly increased by an average of 5.8 hours per week. 4. Workload per intern was reduced under the intervention because the bed census, severity and complexity of patients, and number of ad- â There are slight differences in the hours of work reported in the Landrigan and Lockley papers, from 79 to 63 hours or 84.9 to 65.4 hours, respectively.
202 RESIDENT DUTY HOURS missions were similar across the two schedules, but the intervention schedule used four rather than three interns to handle the work. 5. The number of handovers during the intervention increased, but during the intervention schedule there also was a designed increase in the overlap time between tours in order to perform handovers. In the complementary paper by Rothschild et al. (2005), about 53 per- cent of errors were judged to be slips (unintended acts) and lapses (omitted acts) rather than rule-based errors (e.g., not following a protocol). Since the frequency of such errors tends to be increased by sleep loss, sleep depriva- tion may be a more important factor than hours worked. As might be ex- pected, the duty hour reduction in this study did not provide an equivalent increase in sleep time: 19 additional minutes of sleep occurred per hour of duty hour reduction for a total increase of 5.8 hours per week, while the mean decrease in work was 19.5 hours (Lockley et al., 2004). Further, Lockley et al. (2004) state that the 16-hour shift schedule was still âlong enough to indeed produce significant decrements in neurobehavioral per- formance owing to sleep deprivationâ and required interns to ârise between 4 a.m. and 6 a.m., the time of maximal sleep propensity and efficiency in this age group, to review their patientsâ progress before morning roundsâ (p. 1836). Still the shorter intern schedule was associated with less fatigue, more sleep overall, increased numbers of shifts where the intern had more sleep in the preceding 24 hours, and fewer electro-oculography (EOG)- defined attentional failures. During the intervention, the interns may have committed fewer errors for several reasons. In addition to providing more sleep overall (5.8 hours per week), the intervention had an increased number of shifts in which the intern had had more than 4 hours of sleep in the previous 24 hours and had less fatigue as measured by fewer EOG-defined attentional failures (âde- fined as intrusion of slow-rolling eye movements into polysomnographically confirmed episodes of wakefulness during work hoursâ). The scientific rigor of the study results and the significance of its find- ings do not imply that simply changing residentsâ work schedules along the lines of the authorsâ intervention schedule would guarantee a similar 36 percent reduction in resident-caused serious medical errors across the spec- trum of U.S. medical residency programs. It is not known to what extent the results of Landrigan and colleagues (2004) can be used to represent other medical or surgical subspecialties, and how well the results from this single center represent effects in other teaching hospitals. The committee has a number of concerns about how replicable and generalizable the results of this important and seminal study were:
contributors to error 203 â¢ Would the 36 percent reduction in serious medical errors hold up under an actual long-term implementation in this same setting? It is often the case that a long-term implementation of a managerial or technical intervention, such as described in the study, loses its efficacy over time. The long-term effectiveness would likely depend on the dedication and intensiveness of the ongoing supervision and management of the intervention. The committee notes that notwithstanding its beneficial impact, the 16-hour intervention schedule was not continued at original study sites when the ex- periment ended. Thus, its long-term efficacy cannot be determined. This cautionary observation is not uniquely applicable to this in- tervention. The effect of scaling up from a laboratory trial to a full implementation is almost always fraught with difficulties, and it is not uncommon that upon full implementation the results are less dramatic than estimated from the initial trial. â¢ Would or could this model be replicated with similar results in other clinics of the same type in other hospitals? The committee notes that this study focused on interns, not residents in general, and was conducted in intensive care environmentsâwhere many hospitals do not assign interns to work at all. Moreover, cultures and systems differ from hospital to hospital, indeed from service to serviceâand culture and systems can be either a major enabler or a major barrier to the effectiveness of an intervention. The reported experiences of Dr. Peter Pronovost in exporting his anti-line infec- tion checklist methodology from the original ICU setting at Johns Hopkins to other hospitals and services gives some basis for both optimism and caution about the exportation of safety-oriented managerial innovations in medicine. Dr. Pronovostâs work has demonstrated that while dissemination is possible, it has been pain- fully slow and difficult (Gawande, 2007; Pronovost et al., 2006). â¢ Does the 36 percent reduction in serious medical errors apply to other residency services and to residents in other years of train- ing? The studies focus on the residents in the first year of graduate medical training (interns), those with the least experience and the greatest propensity for error, and therefore, the results are unlikely to be indicative of the error rates of more senior residents. ICUs were an appropriate site for this research precisely because ICUs may be more vulnerable to fatigue-driven errors and are environs in which patients experience more frequent AEs. On the other hand, the typical ICU has a redundancy that could facilitate the intercep- tion of errors before they affect the patientâas seen in this study. Residents who are less supervised on the service floors of hospitals
204 RESIDENT DUTY HOURS may make the same or fewer errors, but there may be fewer protec- tions to keep errors from reaching the patient. â¢ Do reduced serious errors translate into improved patient safety? Making a medical error is the first step in a chain of events that can lead to harm to a patient. Errors are precursors of AEs, so reduc- tions in errors would appear to hold the promise of improvements in patient safety. The reduction in serious medical errors found in this study is scientifically valid and is consistent with evidence from Chapter 7. However, no differences were found in PAEs and ICU mortality. For this reason, and for the reasons offered above, it is problematic to project the benefit to patients from this intervention. Larger-scale trials will be needed to evaluate these outcomes. The detail of the error data reported in the Landrigan paper did not enable the committee to judge the causal, but potentially off-setting, roles of the two key factors of fatigue and handovers in the generation of errors. A motivation for this study was clearly the hypothesis that fatigued interns will make more errors. While the Lockley paper isolated sleep data on an intern-by-intern basis, the authors of the Landrigan paper did not report error data by intern, and unfortunately, there is no analysis of the timing of the serious medical errors that occurred. Moreover, the committee was interested in the question of whether certain individuals in this population of interns made the preponderance of errors either due to lack of knowledge and supervision or whether error incidence was related to their sleep pat- terns. The committee advocates a systems approach to error reduction, so the intent and spirit of this query is not on seeking out and blaming interns as individuals if the circumstances are beyond their individual control, but on the potential to detect whether error frequency was related to sleep pat- terns and how to address that through scheduling modifications. A frequently offered counterargument to the benefits of shorter work schedules is that the concomitant increase in patient handovers could actu- ally increase risk. The committee noted that there is also no discussion in the papers of whether any of the serious medical errors were attributable to handovers and communication failures. The authorsâ attempt to institute new and improved sign-out practices was met with resistance and eventu- ally abandoned by the ICU (Landrigan et al., 2004). Conclusions Relative to Further Restrictions of Duty Hours The study by Landrigan and colleagues is a carefully conducted ex- periment that demonstrates remarkable improvements in the two services studied (Landrigan et al., 2004; Lockley et al., 2004; Rothschild et al., 2005). Notwithstanding the caveats raised above, it demonstrates that a
contributors to error 205 substantial reduction in error rates appears possible through such duty hour interventions and increased opportunity for sleep. Together with earlier reports on errors made by fatigued physicians (e.g., Friedman et al., 1971; Grantcharov et al., 2001) and the literature on the impact of fatigue and sleep deprivation on human performance (see Chapter 7), this study lends critical support to the hypothesis that long work hours, including long con- secutive duty periods that are accompanied by acute sleep loss, can put pa- tients at risk for errors that could lead to harm. The fatigue associated with long work hours and subsequent propensity for errors is what Bernstein and Etchells (2005) call a âlatent hazard.â Furthermore, as reported in the Landrigan et al. (2004) study, interns worked beyond their scheduled hours; their colleagues recommend that any maximum hours prescribed in rules account for this inevitability (Lockley et al., 2004). Chapter 7 details additional evidence and suggests ways to reduce acute and chronic sleep deprivation among residents, and Chapter 8 addresses handovers of care because they constitute a period when errors may occur and shortening the length of duty periods increases the number of transitions. Finally, the committee notes that error rates by residents were high even during the intervention schedule (Landrigan et al., 2004; Rothschild et al., 2005). For example, during the intervention period, interns committed 16.5 errors (PAEs) per 1,000 patient-days. Unit-wide PAE rates, committed by all staff in the unit, were even higher (and almost equal at 38.5 per 1000 patient-days) during both phases of the study. It is noteworthy that these high error rates occurred despite the fact that the subjects were participat- ing in a research program whose ultimate aim was error reduction. The fact that error rates remained high under the intervention schedule suggests to the committee that factors besides work hours, workload, and sched- ules contribute substantially to the error rates of both interns and others working in the units. Similarly, another study has found that hospital-wide adverse drug events remained the same after duty hour reform (Mycyk et al., 2005). The committeeâs conclusion is that a vigorous, systematic effort must be made to identify the root causes of medical errors by residents and others in addition to any adjustments in duty hours. OTHER CONTRIBUTORS TO ERROR Resident reports teach us about their experiences with error, how they learn from them, and where systems change might most effectively address the potential for intercepting resident errors. Residents do not want to make mistakes and often feel great anguish upon making an error, and the poorer the patient outcome is, the more intense is their reaction (Engel et al., 2006). Residents see both positive and negative results from the 2003 duty hour reforms with respect to patient safety (Fletcher et al., 2008; Lin et
206 RESIDENT DUTY HOURS al., 2006). On the positive side, well-rested residents find their clinical deci- sion making is improved especially on post-call days, working conditions are better, and they have a generally improved sense of personal well-being. They report downsides including that hour limits are inflexible, patient care can be rushed under the compressed duty hours, treatment decisions are sometimes delayed, and information can be lost in handoffs, thus creating fragmented and less patient-centered care. From the residentâs perspective, duty hours alone are not the only issue when it comes to making errors (Jagsi et al., 2005, 2008; Lin et al., 2006). A systems view of AEs in hospitals and other nonmedical environments recognizes the organizational contribution to a chain of events that can lead to error rather than blaming the individual (Barach and Small, 2000; Leape, 1994; Shojania et al., 2002; Volpp and Grande, 2003). Residents often blame their inexperience and faulty judgment for making errors (e.g., did not ask for advice, missed patient warning signs, had never seen a patient with an atypical presentation of a certain condition, hesitated to act for too long) (Wu et al., 2003). Yet just as frequently they note job overloadâtoo much work to do within the time allotted (Jagsi et al., 2008; Wu et al., 2003). Adverse events are âmore likely when suboptimal working conditions occurâ (Tibby et al., 2004, p. 1160). Vidyarthi and colleagues (2007), in their analysis of a cross-sectional survey of internal medicine residents (n = 125), found that a multifactorial work stress factor (fatigue, excessive workload, inadequate time, distractions, and stress) (mean = 2.92, SD = 0.67 on a 5-point Likert scale) contributes more often than an in-Â tellectual stress factor (inadequate knowledge, inadequate supervision) (mean = 2.39, SD = 0.54, p < .0001) to errors. Resident use of suboptimal care practices (e.g., working while fatigued, forgetting to transmit informa- tion during sign-out) was the only significant feature predictive of error (p < .0001). These internal medicine residents also report that they make cognitive errors more often than administrative errors or procedural ones. Other specialties make procedural errors more often (Jagsi et al., 2005). Jagsi and colleagues (2008) later surveyed residents in 76 different residency programs at two major teaching hospitals before and after imple- mentation (n = 684/801 residents) of the 2003 duty hour limits to look for contributors to error. In the post-duty hour reform period, similar propor- tions of residents respond as to what the contributing factors for errors are whether they are in programs that reduced their total weekly work hours (e.g., reduced by 5 or more hours) or made no change in work hours. The values, respectively, for the reduced hours group and the other programs follow: poor handoffs (63.5-61.6 percent), working too many hours (44.0- 45.4 percent), carrying or admitting too many patients (47-51.8 percent),
contributors to error 207 cross-covering too many patients (46.9-45.9 percent), or inadequate super- vision (24.7-34.1 percent). Studies of resident errors should identify how the work system itself contributes to resident errors. Rothschild et al. (2005) point out that most of the errors in which residents were involved occurred during treatments involving medications and in procedures (78 percent of incidents) and communication (13.7 percent), and these can be system-level problems not just individual performance issues. It is unreasonable to expect residents not to make mistakes in unreliable work settings. For example, medication vials that look almost identical increase the risk of a mistake. Improving systems (e.g., changing paging practices to decrease interruptions, improved handover procedures, computerized orders to avoid illegible handwriting, better supervision) can improve the performance of residents and improve patient safety (Volpp and Grande, 2003). Wu says that residents need help: âalthough patients are the first and obvious victims of medical mistakes, doctors are wounded by the same er- rors; they are the second victimsâ (Wu, 2000, p. 358). West and colleagues confirm this observation, finding that errors appear to beget increased burnout and depression and that these, in turn, may set up a continuing cycle as burnt-out residents make errors more frequently (West et al., 2006). Fahrenkopf and colleagues also report that depressed pediatric residents make 6.2 times more medication errors than those who are not depressed (Fahrenkopf et al., 2008). Burnout in residents is discussed more fully in Chapter 5. Learning from Errors Wu and colleagues (2003, republished from 1991, p. 221) argue that mistakes can be âpowerful formative experiencesâ and ideally should be used as teaching tools. They queried internal medicine residents (n = 114) at three large tertiary care facilities about the most significant medical mistake they ever made and how they responded to it. Mistakes were defined as âan act or omission for which the resident felt responsible that had serious or potentially serious consequences for the patient and would have been judged wrong by knowledgeable peers at the time it occurred.â The most significant mistakes reported by residents fell into several categories (33 percent diagnosis, 29 percent prescribing, 21 percent evaluation, 11 percent procedural, 5 percent communication) and the majority occurred in the first year of residency. Residents perceived that 90 percent of the patients involved had adverse outcomes as a result of their mistake (e.g., physical discomfort, additional procedure, prolonged hospital stay, death). In June 2003, Jagsi and colleagues surveyed medical and surgical resi-
208 RESIDENT DUTY HOURS dents doing clinical training in 15 specialties at two major teaching hospi- tals about their exposure to errors made during the delivery of patient care by themselves or others (Jagsi et al., 2005). More than half of the surveyed medical and surgical residents (55 percent) reported that they had cared for a patient who had experienced an AE sometime during their training, with the residentsâ most recent AE âexposureâ (median time since last event = 21 days) being related to procedures (31 percent), adverse drug events (21 percent), and infections (11 percent). The categories of error are consis- tent with medical records review studies (Gawande et al., 1999; Leape et al., 1991; Neale et al., 2001; Thomas and Brennan, 2000; Thomas et al., 2000a). Eighteen percent of these residents reported exposure to an AE in the past week in a patient that they cared for, and about one-third of these residents felt that they had, at least in part, been responsible (Jagsi et al., 2005). The percentage of those who report AEs caused by mistakes that they felt at least partially responsible for varied by specialty (surgical 10.9 percent, medical 4.7 percent, hospital based such as radiology or anesthe- siology 3.4 percent), procedural specialty (yes 8.0 percent, no 3.7 percent), and year of training (PGY-1 8.2 percent, PGY-2 or more 5.4 percent). This high level of self-reported exposure in this study illustrates the key role residents could play in the reduction of errors if error reporting and system quality improvement were integrated into residency programs. In Chapter 8, the committee recommends changes in error-reporting systems to enhance the opportunity for teaching and learning when errors occur. Conclusion About Other Factors The committee concludes that a number of factors can contribute to resident errors (whether errors of commission or omission) and that it is not just a matter of hours worked or length of shift. Because first-year residents tend to work longer hours than residents in other years, more frequently violate duty hours, and appear to be more vulnerable to making mistakesâand yet can be reluctant to reach out for helpâthe committee has recommended in Chapter 4 the particular need to increase supervision for these trainees. Additionally, the committee has concluded in Chapters 3 and 4 that excessive workload creates pressure to violate work hours and can limit learning. The resident self-report studies discussed in this section examine the experiences of residents at a small number of major teaching institutions. As noted earlier in this chapter, clearly, residents make mistakes during patient â Note that these are not considered rates of âresident-committed errorsâ because the study questioned exposure to events and thus could be double counting errors due to cross-coverage of patients by different residents.
contributors to error 209 care and these can result in harm to patients, but research studies to date do not allow us to determine with precision the frequency and the severity of those mistakes across all specialties or how often they lead to adverse patient effects that would be preventable. However, first-year residents appear particularly vulnerable to these mistakes or near misses although they occur with residents of all training years, and the types of mistakes (diagnosis, delays in treatment, and performing procedures) are ones that better supervision would help address (Jagsi et al., 2005; Wu et al., 2003). Many of the perceived causes of the mistakes that residents make appear avoidable not only by better supervision but also by workload reduction, more rest, better handovers, and other changes in the work environment. Summary This chapter has examined five questions that are central to the debate on the scope of resident errors while in training, the extent to which duty hour reforms have already made a difference, and the potential contribution of further duty hour reductions. 1. Do residents make errors that contribute to patient harm? Resi- dents do make errors that contribute to patient harm (Jagsi et al., 2005, 2008; Landrigan et al., 2004; Rothschild et al., 2005; Wu et al., 2003). However, data are too limited to determine what por- tion of errors in training facilities are due to residents and what portion of errors result in preventable adverse events that contrib- ute to patient harm. 2. Is resident fatigue from long duty hours among the most significant risks to patient safety? There is evidence that residents can expe- rience fatigue under the current ACGME duty hours (2003) and that fatigue may derive from a number of factors, one of which is lengthy duty hours. There is also evidence that schedules that induce fatigue can result in increased medical errors by residents, which are a potential risk to patientsâ safety. The one randomized controlled trial of duty hour reduction reported to date found that serious medical errors (including medication and diagnostic errors) and non-intercepted serious errors were significantly higher with longer duty hours and less sleep (Landrigan et al., 2004). However, they did not find a statistically significant difference in patient safety as directly measured by PAEs (Landrigan et al., 2004). Con- sequently, while resident fatigue might pose a risk to patient safety, it is not possible to determine the extent of this risk. 3. Did the 2003 reduction in resident duty hours affect patient safety? The national studies of mortality, at the very least, show that
210 RESIDENT DUTY HOURS there is no evidence of widespread harm occurring after imple- mentation of the limits (i.e., 2003 duty hour restrictions did not lead to an increase in mortality rates for the common conditions studied) and there may be modest improvements for medical if not surgical patients (Landrigan et al., 2004; Prasad, 2008; Shetty and ÂBhattacharya, 2007; Volpp et al., 2007a,b). The results from n Â ational studies as well as smaller institution-specific studies indi- cate how difficult it is to scientifically substantiate the conventional wisdom that reduced hours would clearly result in improved pa- tient care. Based on the available data, the committee concludes that movement toward the 80-hour week has not had an adverse effect on patient outcomes. It also recognizes that all training pro- grams in the country have not actually achieved compliance with the 80-hour week consistently. 4. Would further reductions in resident duty hours improve patient safety? At this point, no study indicates that 80 hours or some other lower duty hour total is optimal for patient safety. A num- ber of studies of individual programs have found that they have been able to accommodate to the 80-hour week, even in surgical programs, without sacrificing educational or patient outcomes or increasing error (e.g., de Virgilio et al., 2006; Vaughn et al., 2008). The study by Landrigan and colleagues tested in an ICU setting an intervention with a shorter workweek, shorter shift lengths, and more sleep for interns. This study suggests that further reductions in resident work hours could potentially improve conditions for patient safety by reducing errors although the reduction in PAEs was not statistically significant. As noted by Landrigan et al. (2004, p. 1844), âTherefore, it remains to be determined whether the de- crease in the rate of serious medical errors by interns will translate into a reduction in the rate of adverse events.â Although Landrigan and colleagues conducted a well-designed study, there are a number of questions about its generalizability to other settings, specialties, and years of training. Chapter 7 examines evidence from the hu- man performance literature on the contribution of shift length, night work, and amount of sleep in order to help identify the factors that contribute to diminished performance and to identify opportunities for preventing and mitigating fatigue. 5. What factors in the resident work and learning environment con- tribute to error? Numerous factors can contribute to resident er- rors. The causes of resident errors as well as those of other clinical staff are not one-dimensional but include multiple factors in ad- dition to fatigue: a work and learning environment with insuffi- cient staffing and heavy workload, inadequate supervision, mental
contributors to error 211 health (e.g., burnout, depression), level of skills and knowledge, complexity of patientâs clinical condition, communication problems between team members, language barriers with patients, and inher- ent system failures (Carayon and Gurses, 2008; Dean et al., 2002; Fahrenkopf et al., 2008; West et al., 2006; Wu et al., 2003). The committee encourages additional research on the questions in this chapter. Identifying ways to prevent resident fatigue and the risks it poses to patient safety requires a more systematic understanding of the extent to which fatigued residents are causing patient harm and, if so, under what conditions. For example, the following information would help identify how to best protect patients from errors by residents: When during shifts are errors made? Are many errors made by a few residents or are all residents equally likely to commit errors? What types of errors are made, and how serious and preventable are they? To what extent are errors cor- rected by other clinicians and systems, and to what extent could more be prevented by the committeeâs recommendations for changes in supervision, handovers, and protected sleep? Larger samples of residents from a greater variety of programs and institutions would provide a better population es- timate for identifying best practices to prevent risks to patient and resident safety. Notwithstanding some of the excellent research that has been done in recent years, multi-institutional studies would also have the power to detect changes in preventable adverse errors and mortality as a function of changes in duty hours and any resultant increases in handovers, and would provide data on what kinds of situations need to be targeted to reduce risks to patients and residents. While the research studies discussed in this chapter concerning resi- dents, duty hours, and patient safety generally have limitations and are less conclusive about the effects of duty hours on patient safety, the research discussed in Chapter 7 presents strong evidence that sleep deprivation, which can result from some aspects of current duty hours, can cause fatigue, which contributes to reduced well-being, increased errors, and accidents. The evidence presented in the next chapter provides the basis for the committeeâs recommendations concerning changes in duty hours to prevent fatigue. REFERENCES AHRQ (Agency for Healthcare Research and Quality). 2002. Medical schools and residency programs should provide more training on preventing adverse drug reactions. Research Activities 263:8. âââ. 2007. National healthcare quality reportâchapter 3: Patient safety. Rockville, MD: U.S. Department of Health and Human Services.
212 RESIDENT DUTY HOURS Barach, P., and S. D. Small. 2000. Reporting and preventing medical mishaps: Lessons from non-medical near miss reporting systems. BMJ 320(7237):759-763. Bates, D. W., D. J. Cullen, N. Laird, L. A. Petersen, S. D. Small, D. Servi, G. Laffel, B. J. Sweitzer, B. F. Shea, and R. Hallisey. 1995. Incidence of adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Study Group. JAMA 274(1):29-34. Beckmann, U., C. Bohringer, R. Carless, D. M. Gillies, W. B. Runciman, A. W. Wu, and P. Pronovost. 2003. Evaluation of two methods for quality improvement in intensive care: Facilitated incident monitoring and retrospective medical chart review. Critical Care Medicine 31(4):1006-1011. Berstein, M., and E. E. Etchells. ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ï¿½ 2005. Does reducing internsâ work hours reduce the rate of medical errors? CMAJ: Canadian Medical Association Journal 172(4). Bhavsar, J., D. Montgomery, J. Li, E. Kline-Rogers, F. Saab, A. Motivala, J. B. Froehlich, V. Parekh, J. Del Valle, and K. A. Eagle. 2007. Impact of duty hours restrictions on quality of care and clinical outcomes. American Journal of Medicine 120(11):968-974. Brady, J., K. Ho, and C. Clancy. 2008. Progress slows in improving patient safety for all populations. Patient Safety & Quality Healthcare (July/August):6-7. Brennan, T. A., L. E. Hebert, N. M. Laird, A. Lawthers, K. E. Thorpe, L. L. Leape, A. R. Localio, S. R. Lipsitz, J. P. Newhouse, and P. C. Weiler. 1991. Hospital characteristics associated with adverse events and substandard care. JAMA 265(24):3265-3269. Carayon, P., and A. P. Gurses. 2008. Nursing workload and patient safetyâA human factors engineering perspective. In Patient safety and quality: An evidence-based handbook for nurses. Rockville, MD: Agency for Healthcare Research and Quality. Original edition, AHRQ Publication No. 08-0043. Chang, B. K. 2007. Presentation to the Committee on Optimizing Graduate Medical Trainee (Resident) Hours and Work Schedules to Improve Patient Safety, December 3, 2007, Washington, DC. Classen, D. C., R. C. Lloyd, L. Provost, F. A. Griffin, and R. Resar. 2008. Development and evaluation of the Institute for Healthcare Improvement Global Trigger Tool. Journal of Patient Safety 4(3):169-177. Commonwealth Fund. 2008. Why not the best? New York: The Commonwealth Fund. de Virgilio, C., A. Yaghoubian, R. J. Lewis, B. E. Stabile, and B. A. Putnam. 2006. The 80- hour resident workweek does not adversely affect patient outcomes or resident education. Journal of Surgical Education 63(6):435-439. Dean, B., M. Schachter, C. Vincent, and N. Barber. 2002. Causes of prescribing errors in hospital inpatients: A prospective study. Lancet 359(9315):1373-1378. DeBuono, B. A., and W. M. Osten. 1998. The medical resident workload: The case of New York State. JAMA 280(21):1882-1883. DeFrances, C. J., C. A. Lucas, V. C. Buie, and A. Golosinskiy. 2008. 2006 National Hospital Discharge Survey. Hyattsville, MD: National Center for Health Statistics. Dingell, J. D., J. Barton, B. Stupak, and E. Whitfield. 2007. Letter to William Munier, Acting Director, Agency for Healthcare Research and Quality. Washington, DC: U.S. House of Representatives Committee on Energy and Commerce. Eastridge, B. J., E. C. Hamilton, G. E. OâKeefe, R. V. Rege, R. J. Valentine, D. J. Jones, S. Tesfay, and E. R. Thal. 2003. Effect of sleep deprivation on the performance of simulated laparoscopic surgical skill. American Journal of Surgery 186(2):169-174. Ellman, P. I., M. G. Law, C. Tache-Leon, T. B. Reece, T. S. Maxey, B. B. Peeler, J. A. Kern, C. G. Tribble, and I. L. Kron. 2004. Sleep deprivation does not affect operative results in cardiac surgery. Annals of Thoracic Surgery 78(3):906-911. Ellman, P. I., I. L. Kron, J. S. Alvis, C. Tache-Leon, T. S. Maxey, T. B. Reece, B. B. Peeler, J. A. Kern, and C. G. Tribble. 2005. Acute sleep deprivation in the thoracic surgical resident does not affect operative outcomes. Annals of Thoracic Surgery 80(1):60-65.
contributors to error 213 Engel, K. G., M. Rosenthal, and K. M. Sutcliffe. 2006. Residentsâ responses to medical error: Coping, learning, and change. Academic Medicine 81(1):86-93. Fahrenkopf, A. M., T. C. Sectish, L. K. Barger, P. J. Sharek, D. Lewin, V. W. Chiang, S. Edwards, B. L. Wiedermann, and C. P. Landrigan. 2008. Rates of medication errors among de- pressed and burnt out residents: Prospective cohort study. BMJ 336(7642):488-491. Fischer, J. E. 2004. Continuity of care: A casualty of the 80-hour work week. Academic Medicine 79(5):381-383. Fletcher, K. E., S. Q. Davis, W. Underwood, R. S. Mangrulkar, L. F. McMahon, Jr., and S. Saint. 2004. Systematic review: Effects of residents work hours on patient safety. Annals of Internal Medicine 141(11):851-857. Fletcher, K. E., V. Parekh, L. Halasyamani, S. R. Kaufman, M. Schapira, K. Ertl, and S. Saint. 2008. Work hour rules and contributors to patient care mistakes: A focus group study with internal medicine residents. Journal of Hospital Medicine 3(3):228-237. Forster, A. J., H. J. Murff, J. F. Peterson, T. K. Gandhi, and D. W. Bates. 2003. The incidence and severity of adverse events affecting patients after discharge from the hospital. Annals of Internal Medicine 138(3):161-167. Friedman, R. C., J. T. Bigger, and D. S. Kornfield. 1971. The intern and sleep loss. New EnÂ gland Journal of Medicine 285:201-203. Gandhi, T. K., A. Kachalia, E. J. Thomas, A. L. Puopolo, C. Yoon, T. A. Brennan, and D. M. Studdert. 2006. Missed and delayed diagnoses in the ambulatory setting: A study of closed malpractice claims. Annals of Internal Medicine 145(7):488-496. Garbutt, J., A. D. Waterman, J. M. Kapp, W. C. Dunagan, W. Levinson, V. Fraser, and T. H. Gallagher. 2008. Lost opportunities: How physicians communicate about medical errors. Health Affairs 27(1):246-255. Gawande, A. 2007. The checklist: If something so simple can transform intensive care, what else can it do? New Yorker, December 17, pp. 86-101. Gawande, A. A., E. J. Thomas, M. J. Zinner, and T. A. Brennan. 1999. The incidence and nature of surgical adverse events in Colorado and Utah in 1992. Surgery 126(1):66-75. Grantcharov, T. P., L. Bardram, P. Funch-Jensen, and J. Rosenberg. 2001. Laparoscopic performance after one night on call in a surgical department: Prospective study. BMJ 323(7323):1222-1223. Griffin, F. A., and D. C. Classen. 2008 (August). Detection of adverse events in surgical Âpatients using the trigger tool approach. Quality and Safety in Health Care 17:253-258. Hayward, R. A., and T. P. Hofer. 2001. Estimating hospital deaths due to medical errors: Preventability is in the eye of the reviewer. JAMA 286(4):415-420. HHS (U.S. Department of Health and Human Services). 2008. Hospital compare. http://www. hospitalcompare.hhs.gov/Hospital/Static/About-Overview.asp?dest=NAV|Home|About| Overview#TabTop (accessed July 28, 2008). Horwitz, L. I., M. Kosiborod, Z. Lin, and H. M. Krumholz. 2007. Changes in outcomes for internal medicine inpatients after work-hour regulations. Annals of Internal Medicine 147(2):97-103. Howard, D. L., J. H. Silber, and D. R. Jobes. 2004. Do regulations limiting residentsâ work hours affect patient mortality? Journal of General Internal Medicine 19(1):1-7. Howard, S. K., D. M. Gaba, M. R. Rosekind, and V. P. Zarcone. 2002a. The risks and implications of excessive daytime sleepiness in resident physicians. Academic Medicine 77(10):1019-1025. Howard, S. K., M. R. Rosekind, J. D. Katz, and A. J. Berry. 2002b. Fatigue in anesthesia: Impli- cations and strategies for patient and provider safety. Anesthesiology 97(5):1281-1294. IHI (Institute for Healthcare Improvement). 2008. 100,000 lives campaign general strategy/overview calls. http://www.ihi.org/IHI/Programs/Campaign/100kCampaignStrategyOverviewCalls. htm (accessed August 12, 2008).
214 RESIDENT DUTY HOURS IOM (Institute of Medicine). 2000. To err is human: Building a safer health system. Washing- ton, DC: National Academy Press. âââ. 2006. Preventing medication errors: Quality chasm series. Washington, DC: The National Academies Press. Jacques, C. H., J. C. Lynch, and J. S. Samkoff. 1990. The effects of sleep loss on cognitive performance of resident physicians. Journal of Family Practice 30(2):223-229. Jagsi, R., B. T. Kitch, D. F. Weinstein, E. G. Campbell, M. Hutter, and J. S. Weissman. 2005. Residents report on adverse events and their causes. Archives of Internal Medicine 165(22):2607-2613. Jagsi, R., D. F. Weinstein, J. Shapiro, B. T. Kitch, D. Dorer, and J. S. Weissman. 2008. The Accreditation Council for Graduate Medical Educationâs limits on residentsâ work hours and patient safety: A study of resident experiences and perceptions before and after hours reductions. Archives of Internal Medicine 168(5):493-500. Jakubowicz, D. M., E. M. Price, H. J. Glassman, A. J. G. Gallagher, N. Mandava, W. P. Ralph, and M. P. Fried. 2005. Effects of a twenty-four hour call period on resident performance during simulated endoscopic sinus surgery in an Accreditation Council for Graduate Medical Education-compliant training program. Laryngoscope 115:143-146. Joint Commission. 2007. Sentinel event policy and procedures. http://www.jointcommission. org/SentinelEvents/PolicyandProcedures/ (accessed July 25, 2008). Kaafarani, H. M. A., K. M. F. Itani, L. A. Petersen, J. Thornby, and D. H. Berger. 2005. Does resident hours reduction have an impact on surgical outcomes? Journal of Surgical Research 126(2):167-171. Kahol, K., M. J. Leyba, M. Deka, V. Deka, S. Mayes, M. Smith, J. J. Ferrara, and S. ÂPanchanathan. 2008. Effect of fatigue on psychomotor and cognitive skills. American Journal of Surgery 195(2):195-204. Kaldjian, L. C., E. W. Jones, B. J. Wu, V. L. Forman-Hoffman, B. H. Levi, and G. E. Rosenthal. 2008. Reporting medical errors to improve patient safety: A survey of physicians in teaching hospitals. Archives of Internal Medicine 168(1):40-46. Kennedy, R. 1998. Residentsâ work hours termed excessive in hospital study. New York Times, May 1, p. 1. Kiernan, M., J. Civetta, C. Bartus, and S. Walsh. 2006. 24 hours on-call and acute fatigue no longer worsen resident mood under the 80-hour work week regulations. Current Surgery 63(3):237-241. Laine, C., L. Goldman, J. R. Soukup, and J. G. Hayes. 1993. The impact of a regulation restricting medical house staff working hours on the quality of patient care. JAMA 269(3):374-378. Landrigan, C. P., J. M. Rothschild, J. W. Cronin, R. Kaushal, E. Burdick, J. T. Katz, C. M. Lilly, P. H. Stone, S. W. Lockley, D. W. Bates, and C. A. Czeisler. 2004. Effect of reduc- ing internsâ work hours on serious medical errors in intensive care units. New England Journal of Medicine 351(18):1838-1848. Leape, L. L. 1994. Error in medicine. JAMA 272(23):1851-1857. Leape, L. L., and D. M. Berwick. 2005. Five years after To Err Is Human: What have we learned? JAMA 293(19):2384-2390. Leape, L. L., T. A. Brennan, N. Laird, A. G. Lawthers, A. R. Lacolio, B. A. Barnes, L. Hebert, J. P. Newhouse, P. C. Weiler, and H. Hiatt. 1991. The nature of adverse events in hospi- talized patients. Results of the Harvard Medical Practice Study II. New England Journal of Medicine 324(6):377-384. Leonard, C., N. Fanning, J. Attwood, and M. Buckley. 1998. The effect of fatigue, sleep deprivation and onerous working hours on the physical and mental wellbeing of pre- registration house officers. Irish Journal of Medical Science 167(1):22-25.
contributors to error 215 Lin, G. A., D. C. Beck, and J. M. Garbutt. 2006. Residentsâ perceptions of the effects of work hour limitations at a large teaching hospital. Academic Medicine 81(1):63-67. Lockley, S. W., J. W. Cronin, E. E. Evans, B. E. Cade, C. J. Lee, C. P. Landrigan, J. M. Rothschild, J. T. Katz, C. M. Lilly, P. H. Stone, D. Aeschbach, C. A. Czeisler, and Harvard Work Hours Health and Safety Group. 2004. Effect of reducing internsâ weekly work hours on sleep and attentional failures. New England Journal of Medicine 351(18):1829-1837. McCannon, C. J., A. D. Hackbarth, and F. A. Griffin. 2007. Miles to go: An introduction to the 5 million lives campaign. The Joint Commission Journal on Quality and Patient Safety 33(8):477-484. Mycyk, M. B., M. R. McDaniel, M. A. Fotis, and J. Regalado. 2005. Hospitalwide adverse drug events before and after limiting weekly work hours of medical residents to 80. American Journal of Health-System Pharmacists. Neale, G., M. Woloshynowych, and C. Vincent. 2001. Exploring the causes of adverse events in NHS hospital practice. Journal of the Royal Society of Medicine 94(7):322-330. NTSB (National Transportation Safety Board). 2008. NTSB training center. http://www.ntsb. gov/TC/CourseInfo/IM303_2009.htm (accessed October 24, 2008). Petersen, L. A., T. A. Brennan, A. C. OâNeil, E. F. Cook, and T. H. Lee. 1994. Does housestaff discontinuity of care increase the risk for preventable adverse events? Annals of Internal Medicine 121(11):866. Petersen, L. A., E. J. Orav, J. M. Teich, A. C. OâNeil, and T. A. Brennan. 1998. Using a com- puterized sign-out program to improve continuity of inpatient care and prevent adverse events. Joint Commission Journal on Quality Improvement 24(2):77-87. Poulose, B. K., W. A. Ray, P. G. Arbogast, J. Needleman, P. I. Buerhaus, M. R. Griffin, N. N. Abumrad, R. D. Beauchamp, and M. D. Holzman. 2005. Resident work hour limits and patient safety. Annals of Surgery 241(6):847-856. Prasad, M. 2008. The effect of resident work-hour regulations on ICU mortality. Paper read at 2008 American Thoracic Society International Conference, May 21, 2008, Toronto, ON, Canada. Pronovost, P., D. Needham, S. Berenholtz, D. Sinopoli, H. Chu, S. Cosgrove, B. Sexton, R. Hyzy, R. Welsh, G. Roth, J. Bander, J. Kepros, and C. Goeschel. 2006. An intervention to decrease catheter-related bloodstream infections in the ICU. New England Journal of Medicine 355(26):2725-2732. Regenbogen, S. E., C. C. Greenberg, D. M. Studdert, S. R. Lipsitz, M. J. Zinner, and A. A. Gawande. 2007. Patterns of technical error among surgical malpractice claims: An analysis of strategies to prevent injury to surgical patients. Annals of Surgery 246(5):705-711. Robbins, J., and F. Gottlieb. 1990. Sleep deprivation and cognitive testing in internal medicine house staff. Western Journal of Medicine 152(1):82-86. Rosekind, M. R., K. B. Gregory, D. L. Miller, E. L. Co, and J. V. Lebacqz. 1994. Aircraft ac- cident report: Uncontrolled collision with terrain, American International Airways Flight 808, Douglas DC-8, N814CK, U.S. Naval Air Station, Guantanamo Bay, Cuba, August 18, 1993. Washington, DC: National Transportation Safety Board. http://human-factors. arc.nasa.gov/zteam/PDF_pubs/G_Bay/GauntanamoBay.pdf (accessed March 16, 2009). Rothschild, J. M., C. P. Landrigan, J. W. Cronin, R. Kaushal, S. W. Lockley, E. Burdick, P. H. Stone, C. M. Lilly, J. T. Katz, C. A. Czeisler, and D. W. Bates. 2005. The critical care safety study: The incidence and nature of adverse events and serious medical errors in intensive care. Critical Care Medicine 33(8):1694-1700. Saxena, A. D., and C. F. P. George. 2005. Sleep and motor performance in on-call internal medicine residents. Sleep 28(11):1386-1391.
216 RESIDENT DUTY HOURS Schneider, J. R., J. J. Coyle, E. R. Ryan, R. H. Bell, Jr., and D. A. DaRosa. 2007. Implementa- tion and evaluation of a new surgical residency model. Journal of the American College of Surgeons 205(3):393-404. Shetty, K. D., and J. Bhattacharya. 2007. Changes in hospital mortality associated with resi- dency work-hour regulations. Annals of Internal Medicine 147(2):73-80. Shojania, K. G., H. Wald, and R. Gross. 2002. Understanding medical error and improving pa- tient safety in the inpatient setting. Medical Clinics of North America 86(4):847-867. Singh, H., E. J. Thomas, L. A. Petersen, and D. M. Studdert. 2007. Medical errors involv- ing trainees: A study of closed malpractice claims from 5 insurers. Archives of Internal Medicine 167(19):2030-2036. Thomas, E. J., and T. A. Brennan. 2000. Incidence and types of preventable adverse events in elderly patients: Population based review of medical records. BMJ 320(7237):741-744. Thomas, E. J., D. M. Studdert, J. P. Newhouse, B. I. W. Zbar, K. M. Howard, E. J. Williams, and T. A. Brennan. 1999. Costs of medical injuries in Utah and Colorado. Inquiry 36(3):255-264. Thomas, E. J., D. M. Studdert, H. R. Burstin, E. J. Orav, T. Zeena, E. J. Williams, K. M. Howard, P. C. Weiler, and T. A. Brennan. 2000a. Incidence and types of adverse events and negligent care in Utah and Colorado. Medical Care 38(3):261-271. Thomas, E. J., D. M. Studdert, W. B. Runciman, R. K. Webb, E. J. Sexton, R. M. Wilson, R. W. Gibberd, B. T. Harrison, and T. A. Brennan. 2000b. A comparison of iatrogenic injury studies in Australia and the USA I: Context, methods, casemix, population, patient and hospital characteristics. International Journal for Quality in Health Care 12(5):371-378. Tibby, S., J. Correa-West, A. Durward, L. Ferguson, and I. Murdoch. 2004. Adverse events in a paediatric intensive care unit: Relationship to workload, skill mix and staff supervision. Intensive Care Medicine 30(6):1160-1166. Vaughn, D. M., C. L. Stout, B. L. McCampbell, J. R. Groves, A. I. Richardson, W. K. Thompson, M. L. Dalton, and D. K. Nakayama. 2008. Three-year results of mandated work hour restrictions: Attending and resident perspectives and effects in a community hospital. American Surgeon 74(6):542-546. Veasey, S., R. Rosen, B. Barzansky, I. Rosen, and J. Owens. 2002. Sleep loss and fatigue in residency training: A reappraisal. JAMA 288(9):1116-1124. Vidyarthi, A. R., A. D. Auerbach, R. M. Wachter, and P. P. Katz. 2007. The impact of duty hours on resident self reports of errors. Journal of General Internal Medicine 22(2):205-209. Volpp, K. G. M., and D. Grande. 2003. Residentsâ suggestions for reducing errors in teaching hospitals. New England Journal of Medicine 348(9):851-855. Volpp, K. G., A. K. Rosen, P. R. Rosenbaum, P. S. Romano, O. Even-Shoshan, A. ÂCanamucio, L. Bellini, T. Behringer, and J. H. Silber. 2007a. Mortality among patients in VA hospitals in the first 2 years following ACGME resident duty hour reform. JAMA 298(9):984-992. Volpp, K. G., A. K. Rosen, P. R. Rosenbaum, P. S. Romano, O. Even-Shoshan, Y. Wang, L. Bellini, T. Behringer, and J. H. Silber. 2007b. Mortality among hospitalized Medicare beneficiaries in the first 2 years following ACGME resident duty hour reform. JAMA 298(9):975-983. Weinger, M. B., and S. Ancoli-Israel. 2002. Sleep deprivation and clinical performance. JAMA 287(8):955-957. West, C. P., M. M. Huschka, P. J. Novotny, J. A. Sloan, J. C. Kolars, T. M. Habermann, and T. D. Shanafelt. 2006. Association of perceived medical errors with resident distress and empathy: A prospective longitudinal study. JAMA 296(9):1071-1078. Wu, A. W. 2000. Medical error: The second victim. BMJ 320(7237):726-727. Wu, A. W., S. Folkman, S. J. McPhee, and B. Lo. 2003. Do house officers learn from their mistakes? Quality and Safety in Health Care 12(3):221-226.