Article Text
Abstract
Background: Healthcare providers work increasingly under a variety of shift work systems to cover the continuous care required by patients. However, the effects of shift work on patient and provider outcomes in healthcare settings has not been systematically evaluated.
Objective: To identify and analyse the available evidence on the effect of shift length (8-h vs 12-h shifts) on quality of patient care and healthcare provider outcomes.
Methods: Systematic searching of eight online databases, key governmental/organisational websites and academic journals with ancestry search of relevant articles (limited to articles published in English and Spanish).
Results: Of 562 articles that were retrieved from 20 446 titles identified through database and manual searches, 27 satisfied the inclusion criteria, of which 15 were rejected because of low methodological quality. The 12 final studies included cross-sectional/survey (7), before–after (3) and prospective cohort (2) designs. The main primary outcomes evaluated were: (1) quality of patient care and (2) healthcare provider outcomes. The results were equivocal. With respect to the effect of shift length on quality of patient care, two studies found that errors and near errors were associated with working longer shifts, and another study reported decreased patient complications and length of stay with longer shifts. Specific healthcare provider outcomes such as health complaints, well-being, drug and alcohol consumption, stress and job satisfaction were mostly evaluated by single studies and therefore there was insufficient evidence from which to draw conclusions.
Conclusions: Methodological quality of the studies generally was low and results equivocal with insufficient evidence to determine the effects of shift length on quality of patient care and healthcare provider outcomes. Clearly, robust well-designed studies are needed to examine the effect of shift length on patient and healthcare provider outcomes.
Statistics from Altmetric.com
Diversification and change in working systems have characterised the past two decades. One of these changes has been the incorporation of flexibility in work schedules and the implementation of shift work systems. Shift work refers to patterns that extend beyond the conventional 8-h work day. Typical shift work systems divide a 24-h day in two (12-h) or three (8-h) shifts.12 According to a survey performed in Europe,34 33% of the active population in the European Union work irregular hours outside the normal 8-h day and 13% work within shift work systems. The negative effects of shift work on workers’ physical and mental health, sleep, job performance and psychosocial well-being have been well chronicled but not empirically verified.56
The majority of studies examining the effects of shift work have been carried out in industry settings with factory workers. Smith and colleagues2 for example, conducted a narrative review and reported little difference between 8-h and 12-h shifts with respect to shift worker outcomes and safety. Although healthcare workers work in a variety of shift work systems that are similar to industry workers, they cannot be readily compared to factory shift workers for several reasons. For example, healthcare workers are in continuous contact with psychological strains such as acute illnesses, human suffering and death. Additionally, the context (ie, work environment) for healthcare workers differs from that of factory workers.
Healthcare environments are complex and dynamic, and workers must be continually responsive to ever-changing needs of human beings. This is in contrast to a factory environment where human life is not at stake. Furthermore, healthcare environments are less structured and controlled then factory settings, necessitating increased flexibility and adaptation among care providers to meet the requirements of their role. These factors make it difficult to extrapolate results from industry to healthcare environments.
Since the mid 1970s, several individual studies have been carried out in the health sector comparing the traditional 8-h shift with extended shifts, mainly 12-h shifts. These studies were done to evaluate the effects of different shift lengths both on the quality of care patients in these settings receive and on healthcare provider outcomes. Two systematic reviews in the health field on the effect of shift length were located. Poissonnet and Véron5 examined the effects of shift length on healthcare workers’ health while Fletcher and colleagues7 examined the effects of resident (doctor) work hours on patient safety. Findings from the Poissonnet and Véron5 review were inconclusive while the Fletcher et al7 review concluded that shorter shift lengths for residents do not result in improved patient outcomes. Both of these reviews, however, searched a limited set of databases (MEDLINE, EMBASE and Current Contents databases). The Fletcher et al7 review was also limited to a single population, medical residents. In addition, the Poissonnet and Véron5 review did not account for the methodological quality of the studies assessed in their report. We have expanded on these reviews by including all healthcare providers, covering additional databases, and performing methodological quality assessment on relevant studies.
In summary, healthcare providers such as nurses, allied health professionals and doctors work increasingly under a variety of shift work systems in order to cover the continuous care requirements of patients.8 Information regarding the effects of shift length on patient and provider outcomes is important from both quality of care and patient safety perspectives. However, we located no reports of studies to date that rigorously and systematically assessed the evidence on the effects of shift length in healthcare settings. The objective of the systematic review reported in this paper was to evaluate the existing evidence on the effects of different shift lengths on quality of patient care and healthcare provider outcomes.
INCLUSION CRITERIA FOR STUDIES
Types of study
Randomised controlled trials, clinical trials and observational studies (ie, cohort, case–control, cross-sectional and survey designs) that evaluated the effect of shift length on quality of patient care and/or healthcare provider outcomes were eligible for inclusion. Case reports and literature reviews were excluded. Studies were limited to those published in English and Spanish languages as both languages were represented on the research team. There were no restrictions on the basis of country of origin or when the study was undertaken.
Types of participant, intervention and outcome: inclusion criteria
We considered studies that examined healthcare providers (ie, nurses, allied health professionals, doctors) who: (1) worked shift work, and (2) were part of unit-based hospital staff or a long-term care facility employing continuous (24-h) patient care. The outcomes of interest were: (1) quality of patient care (eg, errors, patient injury and nurse perception of quality of care), and (2) healthcare provider outcomes (eg, overall well-being, fatigue, drug/alcohol use, stress, physical/mental health complaints, job satisfaction). Measures of both quality of patient care and healthcare provider outcomes needed to be expressed quantitatively.
Search strategy for identification of studies
For this review, we searched the literature according to methods outlined by Dickersin and Lefebvre.9 We designed the search strategy in consultation with a health sciences librarian. We searched the following bibliographic databases: EMB Reviews-Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, MEDLINE, EMBASE, ISI Web of Science, Health Star, PsychINFO and CINAHL. Keywords and medical subject headings related to shift work, quality of patient care, and healthcare providers were identified prior to initiating the search (table 1). Searches of key government and regulatory and professional provider organisation websites such as the Canadian Medical Association (CMA), Alberta College of Physicians and Surgeons, Institute of Nursing, Agency of Health Care Research and Quality (AHRQ), National Institute of Nursing Research, Canadian Institute of Health Information (CIHI), Australian Adverse Drug Reactions Bulletin and World Health Organization (WHO) Pharmaceutical Newsletter were also undertaken. Key journals (ie, Ergonomics and the Journal of Occupational and Environmental Medicine) as well as the bibliographies of primary research articles were handsearched.
METHODS
Study identification
One investigator (SAO) screened the titles and abstracts of studies identified by the search strategy for potentially relevant studies. If the investigator felt that any published article potentially met the inclusion criteria, or if there was inadequate information to make a decision, a copy of the article was obtained.
Quality assessment
Two independent reviewers (JS and SAO) assessed the methodological quality of all articles meeting the review’s inclusion criteria (n = 27). To assess methodological quality of the different study designs (cross-sectional studies/surveys, and cohort/before–after studies) we adapted two previously used tools. These tools were developed based on the Cochrane Collaboration guidelines and medical literature1011 and have been used in other systematic reviews by our group.1213 Each tool contains a maximum of either 16 (cross-sectional studies/surveys) or 17 (cohort/before–after studies) total points. In order to derive the final score for each article, the amount of points scored was divided by the total of possible points. Thus, a maximum score of 1 could be obtained for each scale. Each study was then classified as weak, moderate or strong as follows:
<0.50: weak study;
0.50–0.74: moderate study;
0.75–1: strong study.
The rating system was based on a similar system developed by de Vet et al14 and Reish et al.15 All discrepancies in quality assessment were resolved through consensus. The κ coefficient was used to evaluate agreement between the two reviewers.
Data extraction
One independent reviewer (SAO) extracted data from all articles included in the review. A second reviewer (JS) reviewed extracted data for accuracy. Data were extracted on study design, sample/subject characteristics, setting, shifts analysed, measurement tool(s), reliability and validity, key findings (ie, quality of patient care and healthcare provider outcomes such as fatigue, job satisfaction and health complaints), statistical tests used and associated statistical and/or clinical significance. All discrepancies in data extraction were resolved through consensus.
Data analysis
We synthesised the data extracted from the studies. Two comparisons were established a priori for separate analyses: (1) effect of shift length (8-h vs 12-h) on quality of patient care and (2) effect of shift length (8-h vs 12-h) on healthcare provider outcomes. Findings from the review are summarised according to these two a priori comparison groups.
RESULTS
Description of studies
The database and manual searches yielded 20 446 titles and abstracts. Of these 20 446 published articles, 562 were identified as being potentially relevant after a title and abstract review. Foreign languages (eg, Japanese, Chinese, German, Portuguese, Italian and French), reviews, letters, comment articles and duplicates were excluded, leaving 303 articles to be formally analysed using an inclusion and exclusion criteria tool developed by the research team for the review. From these 303 articles, 716–22 could not be obtained by our library systems and an additional 269 articles did not meet the inclusion criteria. This resulted in the selection and analysis of 27 articles. Another 1523–37 articles were excluded due to low methodological quality (ie, a quality score <0.50), resulting in a final sample of 12 studies38–49 (fig 1). The weak studies had several serious methodological flaws (eg, non-probabilistic sampling method, small sample size, lack of instrument reliability and validity, and inappropriate statistical analyses). Due to the presence of such methodological flaws, as well as a lack of statistically significant differences in outcome findings between the weak studies and moderate/strong studies, we elected to focus our report on the moderate/strong studies only.
Methodological quality of included studies
Methodological quality of the 12 final studies included in this review is reported in supplementary tables A and B (see online supplementary data). Of the 12 final studies, all used an observational design; 7 used a cross-sectional/survey design,394044–4749 3 used a before–after design,414248 and 2 used a prospective cohort design.3843 Methodological quality of the studies was assessed using two quality assessment tools. Only studies rated as moderate or strong were included in the final sample of 12. Of the 12 included studies, 1 study38 was rated as “strong”, and the remaining 11 studies were rated as “moderate”.39–49 Inter-rater reliability during this stage of the review, assessed by the κ coefficient, was κ = 0.97 for cross-sectional studies and surveys, and κ = 0.94 for cohort and before–after studies. Discrepancies related primarily to the appropriateness of statistical analyses, and were settled through consensus.
Study outcomes
Methodological weaknesses, varied outcome measures across different healthcare contexts and small samples prevented meta-analysis. Instead we present narrative results. The findings of the included studies are summarised in table 2. For characteristics of the included studies, see appendix A online.
Quality of patient care
Six articles384142444547 investigated the relationship between shift length and quality of patient care. A wide range of quality of patient care activities were assessed: patient recovery,38 patient mortality,38 length of hospital stay,38 documentation and observation of patient care provided,4142 incident reports,47 errors and near errors4445 and nurses’ perceptions of patient care quality.47 Bollschweiler and colleagues38 concluded that quality of patient care (recovery, mortality and length of hospital stay) was significantly better when healthcare providers worked 12-h shifts compared with 8-h shifts. On the other hand, Reid et al42 through an analysis of nurses’ activities, found that nurses who worked 12-h shifts provided lower quality patient care compared with nurses who worked 8-h shifts. Similar findings supporting 8-h shifts were also reported by Rogers et al44 and Scott et al,45 whereas Mills and colleagues41 found no significant difference in the quality of patient care delivered before and after the implementation of a 12-h shift system.
The relationship between shift length and the number of errors was examined in two studies.4445 Both studies found a significant relationship between shift length and the numbers of nursing errors, with more errors occurring on longer (12-hs +) shifts. Rogers et al44 found the likelihood of making an error was three times higher when nurses worked shifts 12.5 h or greater in comparison with nurses who worked less than 12.5 hours (odds ratio = 3.29). Similarly, Scott et al45 found the risk of making an error almost doubled when nurses worked 12.5 or more consecutive hours (odds ratio = 1.94).
Healthcare provider outcomes
Well-being
One study43 examined the relationship between shift length and healthcare provider well-being. Roberson43 found nurses who worked a compressed schedule (ie, work 12 h/day ×7 days, then off ×7 days) reported significantly higher levels of overall well-being than nurses who worked other shift schedules (ie, standard 8-h shifts, standard 12-h shift schedule with 2–3 days on, or a revised 8-h shift schedule working 4 days in a week).
Health complaints (physical and psychological)
Four studies39404347 investigated the relationship between shift length and healthcare provider health complaints (physical and/or psychological).
Physical complaints
Lipscomb and colleagues40 examined musculoskeletal complaints of the neck, shoulder and back such as pain, numbness, tingling, aching, stiffness and burning. Findings showed that working >12 h/day when combined with working >40 h/week was significantly associated with a higher number of musculoskeletal complaints (odds ratios ranged from 2.30 to 2.67). However, neither working >12 h/day nor >40 h/week by itself increased the risk of reported musculoskeletal disorders.
Psychological complaints: stress and emotional exhaustion
Two studies3943 analysed the association between shift length and stress. According to Hoffman and Scott,39 registered nurses who worked 12-h shifts experienced significantly higher levels of stress compared to those who worked 8-h shifts. However, when differences in experience were controlled for, similar levels of stress were found. Contrary to Hoffman and Scott,39 Roberson43 found that stress symptoms were significantly improved 2 months after the implementation of a 12-h shift system and that the stress symptoms were significantly less pronounced in the 12-h shift systems compared to traditional 8-h shift systems. Similarly, Stone et al47 also found that nurses who worked 12-h shifts had significantly less emotional exhaustion than nurses who worked 8-h shifts.
Fatigue
One study43 analysed the relationship between nurses’ fatigue and differences in shift length. Roberson43 found that nurses working a compressed 12-h shift system (ie, work 12 h/day ×7 days, then off ×7 days) had significantly lower levels of fatigue 2 months after implementation of the compressed 12-h shift system. However, these changes were not maintained in the long term; 13 months after implementation of the compressed 12-h shift system, fatigue levels were comparable with baseline levels obtained prior to implementation of the new shift system.
Drug and alcohol use
One study49 examined the relationship between shift length and use of drugs and alcohol by healthcare providers. Trinkoff and Storr49 found that nurses working rotating or night shifts in combination with shifts longer than 8 h had the greatest risk for alcohol use. Similarly, nurses working night shifts longer than 8 h also had the greatest risk for smoking cigarettes.
Job satisfaction
Five articles394346–48 analysed the relationship between shift length and healthcare provider job satisfaction. Three of these studies394346 found no significant differences in levels of job satisfaction among nurses working 8-h compared with 12-h shifts,4648 or before and after the implementation of a 12-h shift system.43 One study48 found that nurses who worked 8-h shifts had significantly higher levels of job satisfaction than nurses who worked 12-h shifts, whereas another study47 found that nurses who worked 12-h shifts were more satisfied with their jobs than those who worked 8-h shifts.
DISCUSSION
The effect of shift length on quality of patient care and healthcare provider outcomes is an issue of ongoing and intense debate. The primary concern raised to date with healthcare workers working extended hours has been the potential (negative) impact on the quality of patient care and on patient safety. Errors during medication administration, as well as during procedures and charting have been suggested as reasons to abandon the 12-h shift system. An early narrative review performed by Smith and colleagues2 reported that job performance (quality and quantity) tended to decrease under the 12-h shift system in comparison to the 8-h shift system. They argued that the negative impact of the 12-h shift system related to fatigue and safety. However, findings from this review were equivocal with respect to patient safety and length of shift. Only one study38 in our review offered support for 12-h shifts over 8-h shifts with respect to better patient care. The remaining studies either found no significant association between shift length and the quality of care patients received41 or favoured the 8-h shift.424445
Adverse effects in the physical health of shift workers such as musculoskeletal complaints, poor sleep quality, disturbances in the circadian rhythm, and infertility have been described in the literature.150 Mental health problems such as stress, anxiety and burnout syndrome have also been documented in both healthcare and industry shiftworkers.851 However, this review found little evidence of significant effects of shift length on psychosocial well-being or physical health. These findings are comparable to an earlier review conducted in the industrial setting.4
Findings for the effect of shift length on specific health provider outcomes, such as drug and alcohol consumption, stress, fatigue, and job satisfaction, were inconclusive. Similar findings were reported by Poissonnet and Véron5 in an earlier review on the effects of irregular schedules on healthcare professionals. Although Poissonnet and Véron5 found no statistical evidence in favour of 8-h or 12-h shifts, they did recommend avoiding longer shifts as much as possible.
The findings from the review reported in this paper suggest that there is insufficient evidence available to conclude that shift length (8-h or 12-h) has an impact on patient or provider outcomes. With respect to quality of patient care, two studies4445 found that errors and near errors were associated with working longer shifts while another study38 found decreased patient complications and length of stay with longer shifts. It is important to note that these two different results came from two different populations, settings and study designs. For example, the study by Rogers and colleagues44 was conducted with a sample of nurses using survey (indirect measurement), whereas the study by Bollschweiler and colleagues38 was conducted with a sample of surgeons using objective measures of quality of patient care routinely collected in intensive care units. Thus, the setting, professional group analysed and data collection measures used may have influenced the differences reported in these studies with respect to quality of patient care and shift length. Studies were inconsistent and insufficient in number to draw any firm conclusions about specific healthcare provider outcomes such as health complaints, well-being, drug and alcohol consumption, stress, fatigue and job satisfaction.
Studies conducted to date in this field have several important limitations. First, few studies on the effects of shift length in healthcare settings are of moderate or high methodological quality, illustrating a clear need for well-designed, robust studies that examine the association between different shift lengths and quality of patient care/healthcare provider outcomes. Second, there is inconsistency in the current shift length literature in the outcome measures being used. By this we mean that we observed a lack of standard measures (eg, for quality of patient care) across studies. This absence of standard measures across studies makes it difficult, if not impossible, to build a consistent body of knowledge on the effects of shift length on outcomes.
Other major limitations in the shift length research to date include the lack of: (1) studies incorporating comparison units (eg, critical care units vs general medicine and surgical units; adult units vs paediatric units vs geriatric units) and comparison settings (acute care vs long-term care vs community/home health settings; urban vs rural settings), and (2) a failure to control for variables such as burden of work, unit complexity, and shift complexity. Combined, these limitations necessitate the need for a reformed research agenda in this field.
There is a clear need for research examining the association between different shift lengths and both quality of patient care outcomes and healthcare provider outcomes. Such studies must be both adequately powered and clinically meaningful, and use valid and reliable outcome measures that enable robust comparisons across settings and studies. Furthermore, since the relationships between shift lengths, quality of patient care and healthcare provider outcomes are multifactorial, future studies need to include multivariate analyses and incorporate the inter-relationships between patient/healthcare provider outcomes and related factors (ie, type of setting, professional type, type of unit, type of rotation) that could affect the endpoint results. Future studies should also investigate and control for additional factors that may influence the effect of shift length on patient and provider outcomes including:
personal characteristics of healthcare providers such as profession, years of experience, sleep patterns, level of concentration, cognitive processing;
contextual factors—that is, workplace setting or environmental factors such as unit complexity and type of facility.
Berwick52 discusses the complex nature of hospitals and how interventions to improve patient care within such settings are often dependent on the hospital’s local context. The effectiveness of interventions to improve patient care, he argues, is sensitive to an array of contextual factors including leadership, changing environments (ie, culture), incentives and organisational history. We also believe that such contextual factors affect provider outcomes. Therefore, future studies examining the effects of shift length should include an assessment of these contextual factors. We also recommend that patient and provider outcomes not be treated in isolation of one another in future studies. For example, current healthcare provider outcome variables such as stress and fatigue should be treated as covariates when examining quality of patient care. Few studies to date have made any attempt to disentangle the possible interactions that may exist between healthcare provider outcomes and patient outcomes. This is a potentially productive area for future research. Results from this type of investigation could have an important effect on administrative decision making processes.
The gold standard of evidence in healthcare intervention research is commonly held to be the prospective randomised controlled trial (RCT). RCTs can be either explanatory or pragmatic in nature. Explanatory trials test whether an intervention is efficacious (ie, whether it is beneficial in an “ideal” situation) while pragmatic controlled trials measure effectiveness (ie, the degree of beneficial effect in real practice). Explanatory controlled trials are most often conducted in large tertiary care health centres on homogeneous groups of individuals with demonstrated compliance, who are likely to remain in the study and have no medical condition other then the one under investigation.53 These requirements however often restrict the use of explanatory controlled trials and necessitate the use of pragmatic controlled trials. Pragmatic trials are conducted on individuals who represent the full spectrum of the population of interest. Unlike the explanatory trial, they include individuals with demonstrated variable compliance, and a number of pre-existing comorbid conditions.53 Hence, the pragmatic trial is more often a reflection of the “real world”.
In the case of the research question discussed in this paper, an explanatory controlled trial would be the more “ideal” study design. In such a study all staff in a facility would be randomly assigned to work all 8-h or all 12-h shifts. However, there are serious practical limitations to such a design. For example, such a design would necessitate that all staff in a facility be agreeable to being randomly assigned to work an unknown shift length for a specified period of time and that all patients be similar with respect to comorbidities. A fairly extensive washout period prior to initiation of the study would also be required to ensure the effects on patient care and healthcare provider outcomes were due to the prescribed shift length. A more realistic study design would entail a pragmatic controlled trial using cluster randomisation. In this case, facilities (or units within facilities) could act as clusters (ie, 8-h or 12-h shift clusters). Ideally a large number of groups (eg, units within facilities across several provinces or states) would exist and one could randomise each unit to work all 8-h or 12-h shifts. If finer stratification were needed, a pair-matched design could be used. In this type of study, similar groups (or units) would be matched on factors of interest (eg, profession, age, health status) and then one member of each matched pair would be randomly assigned to an 8-h or 12-h shift group. In addition to controlling for extraneous and confounding factors, using cluster randomisation or a pair-matched design has the added benefit of decreasing sample size requirements for the study. According to Knight and colleagues54 the ideal population range for a successful community intervention trial is between 5000 and 15 000 individual participants. Thus, a major concern in conducting community randomisation is the availability and accessibility of participants as well as the cost involved in conducting the trial. However, cluster and pair-matched randomisation methods have been shown to provide sufficient statistical power with fewer population units.55
While the pragmatic controlled trial described above offers advantages over the explanatory trial, a critical issue that needs to be considered in pragmatic trials is the balance between external and internal validity. Pragmatic trials tend to maximise external validity to ensure that the results of the study can be generalised.53 The danger in this is that internal validity is comprised in the process. Godwin and colleagues53 suggest investigators employing pragmatic controlled trials take additional measures such as limiting observer and assessment bias, using automated outcome assessment data where possible, and blinding the data analysis to increase internal validity. In the case of the effect of shift length on quality of patient care, several automated measures are available and should be used in future studies: records of patient complications (eg, adverse events, mortality) and length of stay. With respect to healthcare provider outcomes automated data (in most cases) will be available on absenteeism and turnover. Other healthcare provider outcomes should be collected with standard stable instruments (for example, health status could be assessed using the SF-8TM Health Survey.56
In addition to cluster (or matched-design) randomised trials, observational studies (eg, cohort, case–control, cross-sectional) with organised systems for data collection could also be used to study the effects of shift length on quality of patient care and healthcare provider outcomes. Heller and Page57 argue that routine data collection using well-designed standardised forms would improve the implementation and success of observation studies in the community. In addition to the use of organised systems for data collection, such studies should incorporate advanced statistical methods such as multilevel modelling to account for differences between groups. Both pragmatic trials and observational studies with organised systems for data collection are potentially feasible (albeit still challenging) study designs to more effectively study the effects of shift length. Both designs require, however, high levels of healthcare provider cooperation and equally high levels of commitment and partnership from facility administrators.
REVIEWERS’ CONCLUSION
This review suggests that nurses are reporting a variety of health and well-being concerns and issues related to patient care quality and safety regardless of shift length. Synthesis of studies examining 8-h and 12-h shifts did not point to a clear relationship between shift length and health and quality care concerns. Further programmatic research in this area is needed. Such programs would have several concurrent streams examining, for example, different settings (acute care, long-term care, community/home healthcare), different groups of patients (paediatrics, adults, older people, homecare), and different groups of healthcare providers (non-professional providers such as healthcare aides in addition to various groups of professional healthcare providers, such as licensed practical nurses, registered nurses, doctors and allied healthcare professionals) in addition to various patient and provider outcomes.
The processes of healthcare, among them the interactions between caregivers and patients, influence patient and provider outcomes. Two general types of intervention that affect these processes are changing structural conditions that influence provider behaviour and affecting provider practice directly.58 Shift length is an example of an intervention that changes the structural conditions within an organisation and is potentially important from both quality of care and patient safety perspectives. While the findings from this systematic review did not point to a clear relationship between shift length and quality of care, they contribute to the quality and safety literature by pointing to the need for more rigorous research and for the careful consideration of the role of context.52 Healthcare organisations vary by their context of care and future research should examine both whether a relationship between shift length and quality of care and/or patient safety exists and how this relationship operates in different healthcare contexts. Berwick52 argued recently for the scientific community to embrace a wider range of methods and to question our reliance on the RCT as the gold standard. The question of the effect of shift length on outcomes is one of effectiveness we believe best answered using a pragmatic trial approach. Such an approach requires that shift length and its impact be given an unusually high priority status in the health system in order to command the levels of commitment necessary to mount even one large trial. This commitment is required on the part of decision makers, providers and researchers, and is unlikely in the face of other system priorities and in the absence of compelling evidence that it should receive such priority. As a result, answers may in the end depend on incrementally accumulated evidence from other design approaches and on system events that have little to do with science. Such answers will be satisfying and in the best case, the result of care improvements resulting from accumulated learning.52
REFERENCES
Footnotes
Tables A and B and appendix A are published online only at http://qshc.bmj.com/content/vol18/issue3
Funding: CAE holds a CIHR Canada Research Chair in Knowledge Translation. JES holds Killam, CIHR and AHFMR doctoral fellowships. GGC holds a CIHR New Investigator award and an AHFMR Population Health Investigator award.
Competing interests: None.