School of Medicine, College of Medicine, National Taiwan University,Taiwan (R.O.C.)
Department of Emergency Medicine, National Taiwan University Hospital, and National Taiwan University, College of Medicine,Taipei,Taiwan (R.O.C.)
School of Medicine, College of Medicine, Erasmus University Rotterdam,The NetherlandsNetherlands
School of Medicine, College of Medicine, Osaka University, JapanJapan
Also available as PDF (figures and tables appear in the PDF).
Aim Inspired by the famous Libby Zion case in 1984, which revealed the underlying flaw in the medical system at the time──the exploitation of junior medical staff and the inadequate surveillance by attending physicians, authors of this systemic review aim to investigate the potential consequence of overworking in medical scenarios. Methods Excerpta medica database (EMBASE) and PubMed had been systematically searched, bibliographies of relevant studies additionally reviewed. Four cohorts and three randomized controlled trials were selected and quality-assessed with Newcastle-Ottawa Scale and the Cochrane risk of bias tool, respectively. Literature were extracted and discussed. Results Two randomized controlled trials have concluded that residents or interns with longer consecutive working hours per shift in the division of internal medicine are prone to more medical errors. One cohort study has shown a significant association between longer weekly working hours and worsened quality of patient care (intensive care unit (ICU) transfer rates and in-hospital mortality) in the setting of internal medicine. However, none of the 3 studies which were conducted in surgical departments suggests a further restriction on weekly working hours or shift length. Conclusion The specialty-specific policy is recommended based on our between-study comparison. Specifically, with respect to the incidence of medical errors or adverse events, weekly hours or shift length are weakly recommended to be regulated in the department of internal medicine, but there is no recommendation to surgical departments. However, one must consider all respects of the impact brought by any alternation of working policy before the actual implementation.
Keywords: Physicians; Residents; Interns; Duty hours; Working hours; Accreditation council for graduate medical education (ACGME); Complication; Medical error; Adverse events; Patient Safety
Excerpta medica database (EMBASE) and PubMed had been systematically searched, bibliographies of relevant studies additionally reviewed. Four cohorts and three randomized controlled trials were selected and quality-assessed with Newcastle-Ottawa Scale and the Cochrane risk of bias tool, respectively. Literature were extracted and discussed.
Two randomized controlled trials have concluded that residents or interns with longer consecutive working hours per shift in the division of internal medicineare prone to more medical errors. One cohort study has shown a significant association between longer weekly working hours and worsened quality of patient care (intensive care unit (ICU) transfer rates and in-hospital mortality) in the setting of internal medicine. However, none of the 3 studies which were conducted in surgical departments suggests a further restriction on weekly working hours or shift length.
The specialty-specific policy is recommended based on our between-study comparison. Specifically, with respect to the incidence of medical errors or adverse events, weekly hours or shift length are weakly recommended to be regulated in the department of internal medicine, but there is no recommendation to surgical departments. However, one must consider all respects of the impact brought by any alternation of working policy before the actual implementation.
Physicians; Residents; Interns; Duty hours; Working hours; Accreditation council for graduate medical education (ACGME); Complication; Medical error; Adverse events; Patient Safety.
esearch by Grober and Bohnen1 defined the medical error as “an act of omission or commission in planning or execution R that contributes or could contribute to an unintended result.” Similarly, the definition of an adverse event is an “unintended injury to patients caused by medical management (rather than the underlying condition of the patient) that results in measureable disability, prolonged hospitalization, or both.” Not all medical errors necescc 2019 Chao W-Y. This is an open-access article distributed under Creative Commons Attribution 4.0 International License (CC BY 4.0), which allows to copy, sarily result in adverse events and vice versa. Any occurrence of a medical error or a preventable adverse event may reveal underlying problems in any section before the final execution of a medical order. Overworking, for instance, is regarded as the culprit for the incidence of medical errors or adverse events. One prospective 2-session within-subject study shows that residents with heavy night-call duty and placebo drug ingestion have cognitive performance comparable to residents following light-call duty with alcohol intake, blood-alcohol concentration falling in the range of
0.04~0.05 g%.2 An other comparative researchindicated interns with significantly more weekly working-hours and significantly less sleeping-hours had more attentional failure.3 Furthermore, having discovered more recent articles, we aim to make an update of the concerned topic by conducting a literature systematic review, investigating whether long working hours and related factors will lead to higher medical-error rates or worsened patient outcomes.
This systematic review is in accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines.4 Ethical Committee approval is not required. The primary author (C W-Y) searched excerpta medica database (EMBASE) and PubMed in January 2018. The searching strategies of both EM- BASE and PubMed are attached in the Appendix. The author also manually reviewed bibliography of relevant studies and other systematic reviews to ensure the completeness the literature search.
The primary author (C W-Y) firstly screened titles and abstracts from primary records. Duplicates, unrelated studies and studies without full-acccess were excluded. The inclusion criteria are as follows: the population should be physicians, residents, house-staff, or interns, and the comparison of working schedules between the intervention and the control groups should be demonstrated in the study. Outcomes encompass the incidence of medical errors, adverse events, quality indicators of healthcare, or post-operation complications. Outcomes such as the records of errors in surgical simulations or training, cognitive performance, and records of attentional failure are excluded from consideration.
Within articles extracted from the database, many articles investigating the effectiveness of either the Code 405 in New York State, or 2003, 2011 accreditation council for graduate medical education (ACGME) act of regulation on the duty hour in the U.S. were retrieved. Among the primarily retrieved record from the database, there are many studies investigating the effectiveness of either the Code 405 in New York State, or 2003, 2011 accreditation council for graduate medical education (ACGME) act of regulation on the duty hour in the US., demonstrating crude data of mortality or complication rates. Even so, these studies without presentation the difference of working-hours or the compliance of each resident training program to their investigated policies is not accepted. Systematic reviews, letters, editorials, and surveys are excluded. Only cohort studies and randomized controlled trials are eventually included. Studies sourced through the bibliography-review were documented in the Additional records of our PRISMA 2009 Flow diagram.
The primary author (C W-Y) adopted the Newcastle-Ottawa quality assessment and The Cochrane risk of bias tool (modified by Dr. Lee C-C) to conduct the quality assessment for cohort studies and randomized controlled trials, respectively.
Information obtained from all studies includes (1.) Location where studies were conducted (2.) Investigated specialties (3.) Patient numbers (4.) Schedules of the intervention groups and the control groups (5.) Study types and designs (6.) Periods between the beginning and the end of each research (7.)Characteristics of patients and participants at baseline (9.) p value presentation of schedule difference (10.) Postgraduate trainees’ involvement in patient contact (11.) Supervision involvement (12.) Kappa statistics to assess discrepancy of review’s opinions on incident reports (13.) Methods used to document data of working hour, medical errors, adverse events, and patient outcomes to determine the validity of both the exposure and the outcome.
Terms to assess the risk of bias for cohort studies include (1.) Representatives to the exposed cohort (2.) Selection of the non-exposed cohort (3.) Ascertainment of exposure (4.) Demonstration that outcome of interest was not present at the start of the study (5.) Comparability (6.) Assessment of outcome (7.) Was follow-up long enough for outcomes to occur, and (8.) Adequacy of follow up of cohorts. Answers to each question are ranked by 0, 1, or 2 stars and are categorized into “Selection, Comparability, Exposure, and Outcome.”
Terms to assess the risk of bias for randomized controlled trials include (1.) Random sequence generation (2.) Allocation concealment (3.) Blinding of a participants and personnel (4.) Blinding of outcome assessment (5.) Incomplete outcome and (6.) Selective reporting. Each question is assessed as “High Risk”, “Low Risk”, or “Unclear” based on the methodology of collected studies.
Discrepancies were resolved through discussions between the author and the third investigator, Dr. Lee C-C. The recommendation of this article was based on the evaluation of the strength and the weakness of each included studies with grading of recommendations assessment, development and evaluation (GRADE) guideline. Additionally, the critical appraisal skills programme (CASP) checklist for systematic reviews5 are referenced to comment on one another systematic review, presented in the discussion. Literature extraction is presented in Tables 1-7.
Numbers of excluded articles are reported with reasons in the PRISMA flow diagram (Figure 1). Primary records include 502 articles and 25 from bibliography review. Following the deletion of duplicates and the abstract screening, 122 eligible articles undergo the process of full-text review. With the inclusion criteria matched, 4 retrospective cohort studies and 3 randomized controlled trials were selected. Among the 7 studies, 3 studies were conducted in surgery departments with the rest 4 belonging to the category of internal medicine. Six studies were conducted in the U.S. and one in the U.K., and there are 6 single-center studies and 1 multi-center studies (Tables 1 and 2).
Table 1. Background Settings (Locations and the Specialty); Schedule or Policy Settings;
Anderson et al The U.S. Surgery N=198 N=185 <80 <80
Bhavsar et al The U.S. Internal Medicine N=431 N=572 <80 <80
12 The U.S. Internal Medicine N=1,411 N=2,039 ≥80 Not specified <80 Not specified
15 The U.S. Surgery
9 The U.K. Internal Medicine N=238 N=244 43.2 9-11 52.4 9.9-12.5
Bilimoria et al The U.S. Surgery N=65,849 N=72,842 <80 No restriction <80
14 The U.S. Internal Medicine N=227 N=354 60-63 7-16 77-81 7-29
Figure 1. Flowchart of the Literature Search and Numbers of Patients
On call: 26-28 Interns:16 Not on call: 10-14 Residents:24 Interns&:35 Interns:30 Residents:35 Residents:14
N=878 (-A) N=708 (-A) Not specified <16 Not specified >16 N=2,522 (-C) N=386 (-C)
Interns≤16 Residents≤24
Table 2. Study Types; Designs; Periods of the Research; p value of Schedule Difference; Specific Patient Groups; Characteristics Demonstration of the Participants and the Patients at Baseline; PGs’(Postgraduate Trainees’) Involvement in Patient Contact; whether Medical Action was Under Reviewers’ Opinions; Data Collection of Working Hours, Medical Errors and Patient Outcome
Retrospective cohort 10 Crossover 2 years study 13 Retrospective cohort Bhavsar et al Crossover 2 years study 12 Retrospective cohort Ouyang et al Parallel 1 year study 15 Retrospective cohort Yahoubian et al Parallel 7 years study 9 Retrospective cohort Cappuccio et al Parallel 12 weeks study 11 Randomized controlled Bilimoria et al Parallel 1 year trial 14 Randomized controlled Landrigan et al Parallel 1 year trial
10 81.4% Not specified Not specified 13 100% Yes Not specified 12 100% Yes Not specified 15 100% Yes Not specified 9 100% Not specified 0.8 11 Not specified Not specified Not specified 14 100% Yes 0.8-0.9
All patients from the exposed group in 4 cohort studies truly represent the community of interests. Non-exposed participants were drawn from the identical community as the exposed. Exposure ascertainment is evaluated by the method to obtain figures of the working-hour. Only in the Ouyang et al6 the electronic record was used to document the time during work: the time difference between the first and the last medical action made. By contrast, the compliance rate was unclear in other cohorts. Violation of the duty hour policy results in exposure validity unless the record was secured by electronic data. Similarly, how the data of the outcome was retrieved in a study affect the outcome validity. Participant residents were unable to be independently blinded from the
Table 3.1. Quality Assessment for Cohort Based on the Newcastle-Ottawa Scale (NOS)
team-Supervision; Kappa Statistic for Assessment of Discrepancy among Incident Report
No No Not specified / Not specified
No Yes Not specified / Yes
No No Not specified / Yes
No Yes Yes / Yes
Yes Yes Yes / Yes
No Yes Yes / Yes
No No Yes / Yes
Not specified Self-report Self-report Not specified Not applicable Electronic records Electronic records Not applicable Electronic records Not specified Not applicable Electronic records Self-report Electronic records Electronic records Not specified Not applicable Electronic records Self-report Electronic records Electronic records
knowledge to the exposure state. Besides, Anderson et al7 collected data via self-reporting when there is record linage in Ouyang et al,6 8 9 Yahoubian et al8 and Bhavsar et al9 The follow-up was long enough for the outcome to be visible in the 4 cohorts. On the other hand, no statement of the adequacy of the follow-up was noted. Patient outcomes may be associated with the pre-admission condition such as inherent systemic or chronic diseases. However, the outcome of interests at the entry endpoint was not discussed in these studies. Primary exposures are either shift-length or weekly working hours. Other variables in the intervention design may also include the length for clinical education, hand-over between shifts, sleeping hours as expected to be associated with a change of schedule (Tables 3.1 and 3.2).
10 *** ** * 6; High quality 13 ** ** *** 7; High quality 12 *** * * ** 7; High quality 15 ** * ** 5; Medium quality
Table 3.2. Quality Assessment for Cohort Based on the Newcastle-Ottawa Scale
True representative of the average Drawn from resident(including interns) in the community community as the exposed cohort 13 True representative of the average Drawn from resident(including interns) in the community community as the exposed cohort 12 True representative of the average Drawn from resident(including interns) in the community community as the exposed cohort 15 True representative of the average Drawn from resident(including interns) in the community community as the exposed cohort
Study controls for call shift length, additional 10 Self-report Anderson et al controlled factor is time off between shifts or after 24-hour shifts Study controls for call shift length, additional 13 Record linage Bhavsar et al controlled factor is time off between shifts or after 24-hour shifts 12 Record linage Ouyang et al Study controls for call shift length 15 Record linage Yahoubian et al Study controls for call shift length
With respect to random sequence generation, high risk of bias in Cappuccio et al9 is attributed to the lack of interest in participation of interns in the intervention arm due to the physical and mental exhaustion induced by the rapid change of circadian rhythm. As a result, the selection bias existed during the research. Secondly, the allocation concealment was attained in all studies. Blinding of the participants, however, was not feasible to be attained. On the other hand, following the participant interns’ schedules, the observers (the person who observes the participant interns and records down each occurrence of medical errors or adverse events) in Landrigan et al11 were not blinded. The medical record was further reviewed by two blinded independent reviewers. Results of
Table 4. Quality Assessment for Randomized Controlled Trial
High (due to working schedule design, some 11 Low 14 Unclear
9 Low Low (intension-to-treat analysis) Unclear 11 Low Low (intension-to-treat analysis) Unclear 14 Low (Un-blinded observers; blinded medical
the same Unclear Not applicable / Unclear
the same Unclear Not applicable / Unclear the same Secure records Not applicable / Unclear the same Unclear Not applicable / Unclear
Yes No statement Blinding: Not applicable
Yes No statement Blinding :Not applicable
Yes No statement Blinding: Not applicable Yes No statement Blinding: Not applicable
most studies were assessed manually by reviewers. However, only 9 11 Cappuccio et al9 andLandrigan et al11 provided kappa statistics to evaluate the discrepancy in the preliminary reviewing process, so-called inter-rater reliability. Disagreements were resolved by either re-discussion or seeking the third opinion in studies selected within the present review.
10 12 Both Cappuccio et al10 and Bilimoria et al12 used intention-to-treat analysis to address missing data, while Landrigan et al11 didn’t provide any information on how missing data were addressed; however, Landrigan et al11 did note that only one intern was observed at a time because there were no enough observers, missing some dataevenly in both the intervention and the control arms. (Table 4).
Low High interns were reluctant to join.)
Low High Low High
Unclear Unclear record reviewers)
12 7 Bilimoria et al12 and Anderson et al7 adopted the identical schedule policy in accordance with 2011 ACGME reform and the flexibility in duty hour requirements for surgical trainees (FIRST) trial, which waived maximum shift-length and transitional periods between shifts──the reform limited weekly working hours to less than 80, shift-length to less than 24 to 28 hours for residents (4 hours for transition) and to less than 16 hours for interns, the transitional periods between shifts exceeding at least 8 to 10 hours and at least 14 hours after 24-hour on-call shifts. A randomized controlled trial owned profound evidence: Bilimoria et al12 affirms that there was no significant difference in the in-hospital complication rate, or the rate of the 30-day post-operative mortality or serious complication.
Anderson et al7 shows the insignificant difference of the incidence of medical errors in surgical departments of the university of California Davis Medical Center between the group following the FIRST policy and the other of the standard policy, thus no association is indicated. By contrast, significantly more “total adverse events,” including unpreventable adverse events (the unavoidable injury resulting from appropriate medical care), were
Table 5. Results of Medical Error Rates Across Studies
observed in the FIRST trial than in the group of the standard policy.
In Yahoubian et al,8 surgeries were stratified according to the starting time into day-time surgery (from 6 A.M. to 10 P.M.) and night-time surgery (10 P.M. to 6 A.M.). The night-time surgeries were conducted by residents working beyond 16 hours. postgraduate year (PGY) levels were fairly distributed between groups and the operation time were similar as well. A policy was also implemented: Residents who were on the call was not allowed to operate after 6 A.M. in the next morning, thus distinguishing the night-time group from the day-time group. Between-group differences of complication rates were both insignificant in Yahoubian et al.8
In Cappuccio et al10 the intervention group followed the European Working Time Directive (the EWTD) schedule, working 43.2 hours on average, while the control followed the traditional working hour in the U.K., working 52.4 hours in a week on average. Results demonstrated the significant between-group difference of weekly hours of work and a higher rate of intercepted adverse events (p=0.002) in the traditional working group compared to the group of EWTD schedule (Tables 5, 6.1 and 6.2).
Total errors (p=0.596) n=488 n=472 Preventable adverse events (p=0.618) 3.9 % of all operations during the period 3.7% of all operations during the period Cognitive errors (p=0.360) n=114 n=115 Judgement errors (p=0.196) n=95 n=101 Technical errors (p=0.196) n=95 n=101 Team failure errors (p=0.719) n=64 n=63 Overall Medical errors (p=0.006) 27.6 medical errors happened in 1,000 patient-days 41 medical errors happened in 1,000 patient-days Serious medical errors (p< 0.001) 100.1 medical errors per 1,000 patient-days 132 medical errors per 1,000 patient-days Preventable adverse events (p=0.21) 16.5 medical errors per patient-days 20.9 medical errors per 1,000 patient-days Intercepted serious errors (p=0.02) 55 medical errors per 1,000 patient-days 70 medical errors per 1,000 patient-days Nonintercepted serious errors (p<0.001) 28.6 medical errors per 1,000 patient-days 44.8 medical errors per 1,000 patient-days Serious medication errors (p=0.03) 82.5 medical errors per 1,000 patient-days 99.7 medical errors per 1,000 patient-days Serious procedural errors (p=0.34) 6.6 medical errors per 1,000 patient-days 8.5 medical errors per 1,000 patient-days Serious diagnostic errors (p<0.001) 3.3 medical errors per 1,000 patient-days 18.6 medical errors per 1,000 patient-days
Table 6.1. Results of Patient Outcomes (Adverse Events and Complication Rates) Across Studies
13 Cardiac arrest
11 Serious complications
Table 6.2. Results of Patient Outcomes (Mortality) Across Studies
12 In-hospital mortality 9 In-hospital mortality 11 2 Death rate within 30 days after operations 14 Not applicable
Total adverse events (p=0.026) 6.6% of all operations during the period 7.8% of all operations during the period Preventable adverse events (p=0.618) 3.9 % of all operations during the period 3.7% of all operations during the period In-hospital adverse events (unadjusted) Congestive heart failure/ Pulmonary edema OR=1.14, 95% CI: [0.59 to 2.21],p=0.694 Cardiogenic shock OR=1.06, 95% CI: [0.57 to 1.96],p=0.85 OR=0.77, 95% CI: [0.42 to 1.43],p=0.409 In-hospital mortality (unadjusted) OR=0.65, 95% CI: [0.32 to 1.32],p=0.237 MACE at 6 months following discharge MACE at 6 months following discharge (unadjusted) OR=0.76, 95% CI: [0.50 to 1.18],p=0.226 MACE at 6 months following discharge (adjusted) OR=1.29, 95% CI: [0.76 to 2.20],p=0.35 Composite endpoint OR=1.21, 95% CI: [1.02 to 1.43],p=0.027 Transfer to ICU OR=1.62, 95% CI: [1.13 to 2.32],p=0.008 30-Day readmission OR=1.07, 95% CI: [0.88 to 1.29],p=0.513 In-hospital mortality OR=1.44, 95% CI: [0.99 to 2.10],p=0.057 Overall complication rate (Appendectomy) (p=0.1) 1% 2% Overall complication rate (Cholecystectomy) (p=0.9) 2.3% 2.2% Preventable adverse events (p=0.68) 1.6 medical errors happened in 1,000 patient-days 2.2 medical errors happened in 1,000 patient-days Intercepted potential adverse events (p=0.002) 3 medical errors happened in 2467 patient-days 16 medical errors happened in 2315 patient-days Non-intercepted potential adverse events (p=0.067) 41 medical errors happened in 2467 patient-days 56 medical errors happened in 2315 patient-days OR=0.94 (unadjusted), 95% CI: [0.84 to 1.06] Any complications OR=0.96 (adjusted), 95% CI: [0.89 to 1.04] OR=0.96 (unadjusted), 95% CI: [0.86 to 1.06] OR=0.96 (adjusted), 95% CI: [0.90 to 1.04] 30-day rate of post-operative death or serious OR=0.96 (unadjusted), 95% CI: [0.87 to 1.06] complications OR=0.96 (adjusted), 95% CI: [0.95 to 1.04] Preventable adverse events (p= 0.21) On average 16.5 medical errors per 1,000 patient-day On average 20.9 medical errors per 1,000 patient-day
In hospital death (p=0.479) 12.6% in-hospital death rate 15.7% in hospital death rate In-hospital mortality(unadjusted) OR=0.65, 95% CI:[ 0.32 to 1.32],p=0.237 In-hospital mortality(adjusted) OR=0.47, 95% CI: [0.18 to 1.20],p=0.11 Mortality at 6 months following discharge (unadjusted) OR=0.40, 95% CI: [0.22 to 0.71],p=0.002 Mortality at 6 months following discharge (adjusted) OR=0.53, 95% CI: [0.28 to 0.99],p=0.05 OR=1.44, 95% CI: [0.99 to 2.10],p=0.057 Not eligible (casenumber=1, out of 1,587 patients in appendectomy ) Not eligible (case number=1, out of 1,666 patients in cholecystectomy) OR=1.13 (age-adjusted), 95% CI : [0.88 to 1.44] ,p=0.340 OR=0.95 (Adjusted), 95% CI : [0.83 to 1.10]
In Landrigan et al11 a trial of working schedule was performed: shift-length ranges from 7 to 16 hours, weekly working hours ranging from 60 to 63 in the intervention group. On the other hand, with the restriction lifted, the shift-length of the control group exceeded 29 hours and weekly working hours reached up to 77 to 81 hours during the trail. Results show insignificance between groups regarding the medication-error, procedural-error, and intercepted medical errors. However, serious medical error, the diagnostic medical error, and non-intercepted medical error in the intervention was significantly less than those of the control group. The incidence of preventable adverse events was not significantly different in Anderson et al,7 Cappuccio et al,10 and Landrigan et al.11
In Ouyang et al6 significantly higher intensive care unit (ICU) transfer rate, the incidence of 30-day readmission, and more in-hospital death was observed in 40% of the patients taken care of by medical teams with at least one resident working over 80 hours per week, compared to the rest 60% who were taken care of by residents working less than 80 hours per week.
Bhavsar et al9 a retrospective cohort study, investigates if the implementation of the 2003 ACGME policy changed the patient outcomes in the inpatient cardiology service. In the era prior to the policy implementation, all residents and interns had worked 35 hours overnight per shift. Within the 2-years, after the policy was implemented, on the other hand, interns were on-call from 7 a.m., however, to no more than 1 p.m. the next day while residents abided by the day-float system (14 hours at work in the daytime). Results of the study show in-hospital mortality rates were insignificantly associated with working schedule difference while the th mortality rate at the 6th month of the follow-up is significantly higher among patients taken care of by clinicians working longer hours per week. Nevertheless, the mortality at the 6th month seems unlikely to be highly related to the quality of inpatient care because of causes of the death were unclear from the contexts.
An unexpected result is indicated: the weekly time on the arranged working schedule in the Cappuccio et al9 a U.K. based study, was far too short compared to other studies conducted in the United States, though a significant between-group difference of outcomes of medical error rates was revealed in Cappuccio et
Table 7. Summary with the Quality of Evidence Abbreviation.: AEs, adverse events
10 Surgery √ 13 Internal Medicine √ 12 Internal Medicine Not specified √ 15 Surgery √ Not specified Cappuccio et al Internal Medicine √ √ 11 Surgery √
Landrigan et al Internal Medicine √ √ al9 while most studies in the U.S. was not. Nevertheless, since the shift length is similar between researches in the U.K. and the U.S., it is reasonable to expect that other factors such as the weekly hours at work or sleeping hours per week might play a role in the introduction of such a diverse result.
According to the present review, of no significance is one generalized suggestion for the working schedule to all specialties due to their different needs in the arrangement of staff-on-board. There- 7 10 fore, to conclude (Table 7), weekly working hours and shift-length less than 80 and 16 hours, respectively, are likely to the reduce medical error rate and optimize the patient outcomes in the division of internal medicine. On the other hand, researches concerning the surgical division reveal duty hour regulation on residents might not be associated with the incidence of medical errors, or affect patient outcomes, a similar result in Amiran et al with the Karolinska scale and other indicators concerning laparoscopic skills utilized to observe interns, residents, and attending surgeons before and after a 17-hour night shift.
In Fletcher et al13 a literature systematic review, all studies were searched systematically, appraised critically with searching the strategy provided in the context, strength and weakness of each study discussed in detail. Similarly, no definite conclusion was attained due to the diverse settings across studies.
There are multiple explanations of the phenomena. Gonzalo et al18 points out that medical error rates or physician-patient satisfaction survey concerning the life quality after the change of working schedule might be more sensitive to the working hour schedule change instead of in-hospital mortality. Additionally, th the indirectness, unclear ratios of compliance to the designated schedule for interns21 or the unknown proportion of residents’ involvement in each surgery or patient care,10 may render the result invalid. Other similar concerns include the lack of significant difference between schedules of the intervention and that of the control group and the uncontrolled confounders such as residentto-bed ratio, the length of surgery time, the sleeping hours, or the discontinued patient care.
Insignificant association Insignificant association Medium Not applicable Insignificant association Medium Not applicable pro low weekly working hours Medium Not applicable Insignificant Low pro low shift length and Insignificant causality High weekly working hours Not applicable Insignificant causality High pro low shift length and Insignificant causality High weekly working hours
Several studies have shown that long working hours or sleep deprivation deteriorated cognitive function of human 2,4,14,15,19,20,23 19 brains. In a survey from Mustahsan et al, 20 % of the house officers and postgraduate trainees reflected that sleep deprivation was the risk factor of medical errors. On the other hand, dual positive results, improved sleep hygiene and less medical errors, were revealed after the implementation of the EWTD Rota in Cappuccio et al.9
Discontinued patient care is considered to be emerging risk factors of communicational medical errors following the di- 23 17 minishment of duty hours.23 However, O’Leary et al17 indicates otherwise: adverse events are not associated with the handover period because patients were taken care of by medical teams and all medical actions were under the surveillance of attending physicians. On the other hand, prolonged shifts or night shifts may decrease the cognitive function of residents or affects their circadian 24,25 rhythm. However, the question of the inconsistent result of the surgical department remains unsolved. Naylor et al22 for example, demonstrated that most procedure-related complication after surgery occurred during the first 18 hours of their duty, attributing the insignificant difference between control and the intervention group to several factors such as the cavalier attitude, more team surveillance, and the different degrees of difficulty in a surgery. Since it is a sophisticated question, more studies are expected for a conclusion to be made.
Regarding this study, all sources were retrieved from the EMBASE and PubMed with access authorized by the National Taiwan University. All screening and selection processes were accomplished by the primary author (C W-Y), some articles prone to subjective filtration. The result of our study may not be generalized to other clinical specialties because of post-graduate trainees and residents are primary subjects of interests among the included studies, similarly inapplicable in countries other than the U.S. or the U.K.
Given the lack of eligible studies and the diverse background settings across studies, meta-analysis is not possible to be proceeded. Nevertheless, literature extraction in this study has provided a lens by systematic approach into either factor involved and into how duty hour restriction policy impacts patient safety. More researches are awaited for the evidence-based suggestion for the schedule change to reduce the fatigue or the workload of the medical staff as well as to benefit patients.
The study stemmed from a research program of social medicine, which had been started up by Osaka University and National Taiwan University in October, 2017. Chao Wei-Yo (C W-Y): as the primary author, is responsible for the integrity of the article. Lee Chien-Cheng (L C-C), Chen Shyr-Chyr (C S-C) and Liprovided Cheng-Han (L C-H): substantial knowledge supports in biostatistics and epidemiology. Geerdes Emma (G E): revised the section of discussion of the manuscript. Yamazaki Mai (Y M), Tanoue Haruna (T H) and Hsiung Jo-Ching (H J-C): team members in the research program of social medicine, are appreciated for their contribution to the research and their support of the result from their research of “the association between mental depression and long working hours among physicians and nurses.” Searching strategy for the PubMed and EMBASE was advised by Mrs. Hsin-Ping Chiu from the National Taiwan University Medical Library. All articles were obtained with full access authorized by the National Taiwan University Medical Library.
This work is sponsored by MOT research grant of Taiwan.
The authors declare that they have no conflicts of interest.