Association of non-standard working time arrangements with safety incidents: a systematic review

Abstract

Objective

To systematically review the evidence on the association between non-standard working time arrangements (such as night work or shift work) and the occurrence of safety incidents.

Design

Systematic review conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines and using a structured narrative approach and the Synthesis Without Meta-analysis framework to evaluate and summarise findings.

Data sources

MEDLINE, Embase, PsycINFO, Web of Science and ProQuest Health and Safety Science Abstracts were searched through February 2024.

Eligibility criteria for selecting studies

We included peer-reviewed English-language studies of paid workers (18–70 years) that examined the association between non-standard working time arrangements and safety incidents (accidents, near-accidents, safety incidents or injuries), excluding cross-sectional designs and studies on unpaid workers, athletes or military personnel.

Data extraction and synthesis

Two reviewers independently extracted data and assessed risk of bias using standardised forms, extracting study characteristics (author, year, country, sector and population), working time arrangements and exposure assessment, outcomes and their assessment, and reported risk estimates. We conducted a narrative synthesis, classifying studies into three exposure contrasts (shift worker versus non-shift worker, time-of-day and shift intensity), and summarised risk estimates using forest plots without calculating pooled effects.

Results

A total of 13 569 records were screened, and 24 studies met the inclusion criteria. The results indicated that shift workers generally had an elevated safety incident risk compared with non-shift workers (risk estimates ranged from 1.11 to 5.33). Most of the included studies found an increased risk of safety incidents during or after night shifts. Accumulated exposure to evening or night shifts increased the risk of safety incidents during the following 7 days. However, bias and heterogeneity across studies in design, populations and outcome measures resulted in an overall low to very low certainty of the evidence.

Conclusions

Non-standard working time arrangements, including night and evening shifts, appear to increase the risk of occupational safety incidents. Despite the low certainty of evidence, the findings highlight a potential area for preventive measures in work scheduling. Future longitudinal studies using individual data on daily working hours are needed.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Grouping studies by three exposure contrasts (shift worker status, time-of-day and shift intensity) enabled coherent synthesis of diverse evidence on safety incidents.

• Excluding cross-sectional studies for a better assessment of temporal associations.

• High heterogeneity in study designs and definitions of exposure precluded meta-analysis.

• The low certainty of the evidence limits the generalisability of the findings.

• Limiting the inclusion to only English-language publications potentially increased the risk of publication bias.

Introduction

It is estimated that between 10% and 30% of the workforce in industrialised societies engage in non-standard work schedules (shift work),1,3 typically defined as any work outside the traditional daytime hours of 07:00 to 18:00.⁴ Shift work encompasses a variety of evening and night shift arrangements. Broadly, these may be characterised by four patterns: length of working hours, time-of-day, shift intensity and social aspects of the working hours.⁵ Shift work, along with other types of non-day work, notably affects alertness and sleep by causing changes in sleep–wake and light–dark cycles. This may result in circadian misalignment, where internal biological clocks fall out of sync with the external environment.⁶ Suggested mechanisms for this association are outlined in the three-phase model of sleep, wake and alertness.⁷

Shift work, especially during night and early morning hours, may lead to sleep loss⁸ and other sleep-related problems such as shift work disorder, a condition characterised by insomnia and excessive sleepiness.⁹ One major cause of shift work-induced sleepiness is disruption of circadian rhythms, which can lead to a decline in performance and alertness.⁶ Furthermore, simulation studies based on the three-phase model indicate that a complex interplay of factors, such as sleep duration, recovery speed during sleep, time spent awake, napping, circadian phase and overall performance is connected.⁸ These factors are all affected by shift features such as time-of-day, shift length and breaks. Several features related to shift and night work may therefore reduce alertness and impair performance,^{10 11} thus increasing the risk of occupational accidents and injuries.^{6 12} Adding to the complexity, poor metabolic health with a higher energy intake,¹³ increased risk for cardiovascular disease¹⁴ and diabetes¹⁵ are observed among shift workers and can negatively impact cognitive function.¹⁶

Occupational injury is defined by the International Labour Organisation as ‘any personal injury, disease or death resulting from an occupational accident’. Occupational accidents, whether fatal or non-fatal, are unexpected incidents caused by external factors in the workplace that result in employee injury, disease or death.^{17 18} In this review, we define the term ‘safety incident’ as any workplace incident that results in non-fatal or fatal injuries from unexpected events (accidents), as well as any reported near-miss incident that has the potential to cause injury. These outcomes were grouped together because the primary mechanisms by which shift work is believed to increase accident risk (sleepiness, diminished alertness and impaired cognitive functioning)^{8 10 11 19 20} are likely to affect the likelihood of an incident occurring, rather than the severity of the resulting injury.

Previous reviews have examined specific professions, such as the 2010 systematic review of 13 studies on healthcare workers,²¹ which did not find a relationship between shift work and work-related injuries. Subsequently, two reviews focusing on several occupations indicated that some shift work characteristics were associated with a higher risk of safety incidents.^{22 23} One review, including 14 studies, found that longer shift length and night shifts increased incident risk.²³ The latest study from 2017,²² comprising 16 studies, concluded that night shifts, but not evening shifts, increased the incident risk. A recent scoping review of nurses from 2023²⁴ reported that long working hours and rotating shifts increased the risk of injuries. The rationale for conducting a new systematic review, in addition to the ones mentioned, is threefold. First, since Härmä and coworkers published their paper on register-based measures for assessing working time patterns a decade ago,⁵ there has been an increasing number of studies using precise exposure data to evaluate their association with accidents and safety incidents. Second, previous systematic reviews also included cross-sectional studies. Finally, aside from the scoping review by Imes et al from 2023,²⁴ the earlier reviews are relatively old (the latest from 2017). Collectively, this led us to conduct a reassessment of the risks associated with various shift schedules and work patterns across different sectors and occupations.

We based the present review on studies that met the specific inclusion criteria and evaluated the risk of bias for each study. The aim of this systematic review was to assess the potential effects of non-standard working time arrangements on occupational accidents, near accidents, injuries and safety incidents by considering the overall certainty of evidence (CoE).

Methods

In this review, we followed all the recommended steps in conducting systematic reviews,²⁵ including registering the protocol in PROSPERO (CRD42020134823; access at https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=134823). During screening, the scope was amended by splitting the review into two separate publications to improve clarity. This review focuses on non-standard working exposures; the other, already published, addresses long or extended working hours.²⁶

Inclusion and exclusion criteria

We included peer-reviewed articles written in English that investigated paid workers aged 18–70 years. The studies had to contain explicit measures of exposure (working time outside day work) and outcomes (safety incidents), along with a statistical measure of the relationship between exposure and outcome. We excluded studies with a cross-sectional design as well as those that focused solely on non-paid or volunteer workers, professional athletes or military personnel.

Procedure

Exposure and outcome search terms were selected following a thorough review of key articles and reports to identify relevant keywords.^{22 27} Two preliminary searches were performed, followed by adjustments to the search terms. We identified studies addressing the association between non-standard work schedules and safety incidents published until 26 June 2019,²⁶ by searching five databases: Medline, Embase, PsycINFO, Web of Science and ProQuest Health and Safety Science Abstract (the search strategy can be found in online supplemental appendix 1). The search string for Medline, Embase and PsycINFO was also published at https://www.crd.york.ac.uk/PROSPEROFILES/134823_STRATEGY_20190625.pdf. An updated search using the same strategy was performed to include studies published up to 5 March 2024. From a total of 13 569 studies, we identified 12 332 studies in the initial search,²⁶ and 3805 studies in the updated search, after removing duplicates (figure 1).

Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart illustrating the selection process and the studies included. The updated database search is the same search strategy from Matre et al, covering a more recent time period.

Title and abstract screening were performed by all authors using Covidence systematic review software (Veritas Health Innovation, Australia, http://www.covidence.org/). Two reviewers independently screened each study, and any disagreements in the abstract screening were resolved by a third reviewer. At this stage, 13 224 studies were excluded, leaving 345 for full-text evaluation.

For full-text screening, two authors independently decided on the inclusion or exclusion of each study in Covidence. Any disagreements were resolved between the two reviewers or by a third reviewer before the final decision. After the full-text evaluation, 24 studies were included, while 321 were excluded with reasons. Subsequently, we manually searched the reference lists of the included papers, papers citing the included papers and relevant reviews for additional potential studies; however, no further eligible studies were found. From each included study, we extracted data on the author and publication year, the study design, work sector, age and sex distribution, and working time exposure characteristics of the workers, and outcomes (table 1).

Table 1. Study characteristics (n=24).

Author, year, country	Participants; shift workers/injured (n)	Sector	Study population (sex, composition), age, working hour arrangement	Study design	Exposure (exposure assessment)	Outcome (outcome assessment)
Barreto, 1997, Brazil43	177; 141	Manuf	Employees at a steel plant (age distribution not reported, only men)	Nested case–control study	Two or three- versus one-shift work schedule (personnel records)	Death by work-related injury (database)
Bun et al, 2023, France44	730; 237	Health	Healthcare workers (≤25 to >55 years, 83.6% women)	1-year follow-up prospective study	Nightly or rotating shift versus daily shift (survey)	Occupational exposure to blood (survey)
Dembe, 2006, USA45	3834; 765	Gen	General working population (aged 22–42 years, 41% women)	Longitudinal 1987–2000	Shift work schedules (survey)	Injuries and illness (survey)
Dembe, 2009, USA46	10 793; 2339	Health	Healthcare workers (aged 22–43 years)	Retrospective review	Shift work schedules (survey)	Injuries and illness (survey)
D’Ettorre, 2017, Italy56*	765; 545	Health	Nurses (age distribution not reported, 72% women)	Cross-sectional nested case–control analysis*	Number of night shifts last 7 days (registry) (survey)	Needlestick and sharp injuries (database)
Ftouni et al, 2013, Australia58	27; 27	Health	Nurses (mean age 41.6, 85% women)	Case–crossover study	Night shift only versus rotating three-shift (survey)	On-road events before/after shift (driving log)
Govindarajan et al, 2015, Canada59	1448; 1448	Health	Physicians (mean age 46.3, both men and women)	Retrospective matched cohort	Physician treating patients after midnight or not (health databases)	Complications of the patients treated by the physicians (health databases)
Hopcia et al, 2012, USA42*	502; 502 injured and controls	Health	Hospital employees (median age 40 years, 93% women), daily and weekly hours	Cross-sectional nested case–control study*	Number of night shifts last 7 days (registry)	Injury (database)
Horwitz and McCall, 2005, USA49	5619; 1931	Gen	General working population (categorical age distribution, 17% women)	Retrospective study over 5 years	Evening or night work versus day work (registry)	Chemical and heat burns (registry)
Härmä et al, 2020, Finland36	18 700; 18 700	Health	Hospital employees (mean age 43 years, 91% women). Total daily and weekly hours	Case–crossover study	Morning, evening or night shift. Cumulative count of night or evening shift (daily working hours)	Injury (registry)
Lam, 2004, Australia50	7923; 2123	Transp	Taxi drivers (categorical age distribution, 5% women)	Exploratory study, 1996–2000	Nighttime driving (22:00–5:59 versus day (registry)	Death/injury to the driver or non-casualty crashes (registry)
Larsen et al, 2017, Denmark37	150 438; 19 212 night workers	Gen	General working population (age 20–59 years, 47% women), long weekly hours	Retrospective, longitudinal study (1999–2013)	Night work versus non-night work (Danish survey)	Accidental injuries causing hospital contact or death (registry)
Liddell, 1982, Canada48	7634; 347 case/controls	Gen	General working population (categorical age distribution, 31% women)	Retrospective cohort study over 3 years	Days only, nights only or irregular or rotating shift (survey)	Motor vehicle accidents (records of accidents)
Neuberger et al, 1984, USA31	286; 310	Health	Nurses, housekeepers, laboratory technicians (age distribution not reported, 81% women)	Retrospective study over 29 months	Evening or night work versus day work (register)	Needle stick injury (compensation reports, logbook)
Nielsen et al, 2018, Denmark57	69 200; 40 646 (evening) 24 515 (night)	Health	Hospital and administrative workers (mean age 40.5 years, 76% women)	Longitudinal, dynamic cohort	Cumulative count of night or evening shift (daily working hours‡)	Injury and death (registry)
Nielsen et al, 2019, Denmark51	16 059; 3279	Health	Hospital employees (aged 18–65, 80% women)	Multiple interval case–crossover	Last shift before injury (daily working hours‡)	Injury registration at the emergency department (registry)
Ott et al, 2009, Germany47	31 346; 14 128	Manuf	Chemical Industry workers, aged 16–24 years at study entry, 100% men	Cohort study over 11 years	Rotating shift versus day work (electronic work history files)	Work-related injury (medical examinations)
Smith et al, 1994, UK52	4645; 4250	Manuf	Employees (mean age 30.7 years, 4.5% women)	Register-based over 1 year	Day, evening and night work (working records)	On-duty accidents (injury accident reports)
Stevenson et al, 2013, Australia53	1047; 334	Transp	Heavy-vehicle drivers, only men, time since start of drive	Case–control study	Time of crash (interview)	Non-fatal, non-severe crashes (police reports)
Stutts et al, 2003, USA38	1403; 150	Gen	Drivers (varying age and sex in cases and controls), weekly hours	Case–control study	Morning/evening or night work versus regular shift (survey)	Car crash (police register)
Trinkoff et al, 2007, USA39	2624; Not reported	Health	Nurses (age and sex distribution not reported), total daily/weekly hours	Prospective study with follow-up at 6 and 15 months later	Other than day versus daytime work (survey)	Needle stick injury (survey)
Violanti et al, 2012, USA54	419; 245	Police officers	Police officers, mean 43 years, 25.5% women,	Retrospective longitudinal survey over 16 years	Day, evening and night work (survey and registry)	Injury (registry)
Wong et al, 2014, Canada40†	19 131; 9183	Gen	General working population (16–55+ years, ∼50% women), weekly hours	Prospective study	Non-standard shift (evening, night) versus day shift (survey)	Injury (survey)
Åkerstedt et al, 2002, Sweden41†	47 860; 42 407	Gen	National sample, 16–50+ years, ∼50% women, weekly hours	Repeated cross-sectional surveys	Non-daytime work versus daytime work (survey)	Occupational fatal accident (registry)

^*These studies were included because the nested case–control component made them eligible, unlike strictly cross-sectional studies.

^†In Wong et al and Åkerstedt et al, the lower age range was 16 years. These studies were included nevertheless, because the majority of the participants were aged between 18 and 70 years.

^‡Information on daily working hours was obtained from a registry, which includes the exact time of start and end of each shift.

Constr, construction; Gen, general working population; Manuf, manufacturing; Transp, transportation.

Risk of bias assessment

We evaluated the risk of bias using a scoring system developed by the Norwegian National Institute of Occupational Health (STAMI).²⁶ Each article was scored by a pair of team members, and any disagreements were resolved through discussion. The scoring system evaluated three types of bias: selection bias, information bias and confounding factors.^{25 26} Each type of bias was qualitatively assessed based on a set of specific criteria (online supplemental table S1). According to the scores, the risk of bias for each study was categorised as low, moderate or high: a study was rated as low risk if all three sources of bias were considered low, as high risk if any of the three sources were rated as high and moderate risk for any other combination.

Narrative synthesis of results

This review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (http://www.prisma-statement.org/) and the Synthesis without Meta-analysis (SWiM) checklist https://www.equator-network.org/reporting-guidelines/synthesis-without-meta-analysis-swim-in-systematic-reviews-reporting-guideline/.

Using a data-driven approach, the included studies were classified into three exposure contrasts based on their reported shift timing and intensity. (1) Shift worker versus non-shift worker: studies were classified under this contrast when there was no information about the timing of the incident (ie, whether it occurred during or after an evening or night shift). The exposure in these studies was any non-standard working time arrangement. (2) Time-of-day: studies were included in this contrast if the safety incident occurred during or immediately after evening or night shifts, thereby allowing for an assessment of risks associated with specific times of the day. (3) Shift intensity: this contrast included studies in which the safety incident occurred during or after several consecutive evening or night shifts. These studies generally reported the number of shifts worked within a defined time frame (eg, the past 7 days).

Given the variations across studies regarding sample populations, definitions of working time exposure, risk estimates and methods, we used a narrative synthesis approach in accordance with the ‘SWiM‘ () guidelines.²⁸ SWiM is intended for systematic reviews where exposure, outcomes or study designs are too diverse to provide a meaningful overall estimate of the effects.²⁸

To summarise the results, we made a forest plot for each exposure contrast. These plots were created using the ‘metafor’ package²⁹ in R-studio (V.2024.04.2+764), with R V.4.4.1³⁰ (the script for making the plots is provided in the online supplemental appendix 2). The forest plots include effect sizes extracted from each study, OR, incidence rate ratio, HR or relative risk (RR). However, we did not calculate pooled effect estimates. The difference in average incidence rate between shift types was reported in one study³¹ for which we subsequently calculated the OR. Maximally adjusted risk estimates were extracted for all studies (online supplemental table S2).

Certainty of evidence

We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines^{32 33} to assess the overall CoE for the association between working hour arrangements and incident risk for each contrast. As no meta-analysis was performed, we applied GRADE in the context of a narrative synthesis without a single estimate.³⁴ Additionally, we adapted the GRADE framework^{32 33} to assess the CoE for observational epidemiological studies, given the lack of consensus on how to evaluate such studies.³⁵ The CoE was graded for each exposure contrast and classified as ‘high’, ‘moderate’, ‘low’ or ‘very low’. Since no randomised controlled trials were included in this review, we used ‘low’ evidence as the starting point when evaluating all the contrasts. We downgraded the CoE for studies with the following characteristics: (1) high overall risk of bias, (2) inconsistency, (3) indirectness, (4) imprecision and (5) publication bias. Conversely, we upgraded the CoE if the effect size was large or a dose–response relationship was reported. GRADE assessments were made independently by two reviewers, and disagreements were resolved through discussion.

Patient and public involvement

As this was a systematic review, patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Results

Overview of the included studies

The 24 included studies were published between 1982 and 2023, with sample sizes ranging from 27 to 1 50 438 (table 1). The study populations comprised workers from various sectors, including healthcare (n=11), the general working population (n=7), manufacturing (n=3), professional drivers (n=2) and police officers (n=1). The study designs included prospective and retrospective longitudinal studies, case–control studies and case–crossover studies. No randomised trials were identified. Seven studies explicitly examined the length of the workday or workweek, in addition to the time of the day.36,42 We assessed the risk of bias as ‘high’ for one study,⁴³ and ‘moderate’ for the remaining 23 studies (online supplemental table S2). Of the included 24 studies, 11 were classified under the ‘shift worker versus non-shift worker’ without incident timing. Eight studies looked at incidents occurring during or after evening or night shifts (‘time-of-day’). Four studies focused on incidents after a specified number of evening/night shifts in a defined time frame (‘shift intensity’). See also ‘narrative synthesis of results’ under the Methods. Two studies could not be categorised into any of the three contrasts.

Contrast shift worker versus non-shift worker

11 studies examined the risk associated with being a shift worker. Nine of these studies reported incidents occurring in the workplace3739,41 43 (figure 2, online supplemental table S3), while two examined incidents outside the workplace^{38 48} (online supplemental table S3). The study populations included individuals from the general workforce, the manufacturing industry and the healthcare sector. The risk of bias was assessed as moderate in 10 studies and high in one study (online supplemental table S2).

Figure 2. Overview of the HR, OR or RR of an occupational safety incident for being a shift worker compared with non-shift workers. RR, relative risk ratio.

Risk of incidents at the workplace

Out of 14 different exposure/outcome (E/O) combinations with moderate bias, nine indicated an increased risk of safety incidents, ranging from 1.11 to 5.33 (figure 2, online supplemental table S3).

The significant work schedule exposures were identified as shift work,⁴⁷ non-standard shifts,⁴⁰ night work,^{37 45 46} evening work^{45 46} or non-daytime work,⁴¹ rotating shifts44,46 or working shifts other than day shifts.³⁹

Risk of incidents outside the workplace

One study found that working night shifts, compared with regular shifts, was significantly associated with an increased risk of sleep-related car crashes at any time point (OR 5.55 (95% CI 3.1 to 9.88)).³⁸ Another study reported that the ‘irregular shift’ work status at the beginning of the study was associated with an increased risk of car accidents at any time point (RR 2.04) when compared with not working shifts.⁴⁸

Summary of findings

Being a shift worker was associated with an increased risk of safety incidents in 9 of 14 E/O combinations when compared with day workers, excluding one study with a high risk of bias. The strength of the evidence for this contrast was rated as very low due to risk of bias (based on our classifications), substantial heterogeneity, indirectness related to variations in outcomes and populations, and imprecision bias, and variations in outcomes and populations (online supplemental tables S2 and S4).

Contrast time-of-day

Eight studies investigated the risk of safety incidents occurring during or after night or evening shifts.3136 49,54 Nine E/O combinations examined night shifts, and eight focused on evening shifts. In addition, one study investigated late-day shifts⁵³ in comparison to day shifts (figure 3, online supplemental table S5). The study populations included the general working population and specific groups, such as manufacturing workers, healthcare professionals, police officers and professional drivers. The risk of bias was assessed as moderate in all eight studies (online supplemental table S2).

Figure 3. Overview of OR, IRR or RR for accidents happening during an evening or a night shift compared with a day shift. IRR, incidence rate ratio; RR, relative risk ratio.

Incident occurring during night or evening work

Significantly higher incident risk during night shifts compared with day shifts was identified in five out of nine E/O combinations.^{36 51} Additionally, one study found that driving at night shift posed a higher risk of a safety incident compared with shifts without night shift,⁵⁰ with risk estimates ranging from 1.23 to 3.42 (figure 3). Among the studies, chemical burns (but not heat burns) were found to be significant in one study.⁴⁹ Regarding evening shifts, four out of eight E/O combinations indicated a higher incidence risk compared with day shifts,^{36 51 55} with risk estimates ranging from 1.09 to 1.69. Conversely, one study reported a lower risk during evening shifts compared with the day shift³¹ (figure 3).

Incident occurring after night work

In addition to the results presented in figure 3, one study identified a significantly increased risk of incidents following an evening or night shift³⁶ showing increased risk following night shift (OR 1.33 (95% CI 1.17 to 1.52)), but not evening shift (OR 1.01 (95% CI 0.95 to 1.07)) (online supplemental table S5). Another study⁵¹ examined transportation and leisure incidents occurring after working evening or night shifts and found a heightened risk of leisure injuries after evening shifts (OR 1.54 (95% CI 1.42 to 1.65)) (online supplemental table S5).

Summary of findings

Five of the nine E/O combinations showed an increased risk of incidents during night shifts, while four of the eight E/O combinations reported an increased risk of incidents during or after evening shifts. Five studies indicated a higher risk during night shifts compared with evening shifts3149 50 52,54; while one study observed the opposite⁵¹ (based on visual observation of figure 3). The CoE for these comparisons was rated as very low, due to inconsistency, with wide variation in effect estimates (online supplemental table S4).

Contrast shift intensity

Four studies examined the risk of safety incidents associated with cumulative exposure to several night or evening shifts over the preceding 7 or 28 days.^{36 42 56 57} In two of the studies, the exposure was categorised (eg, 4–8 night shifts)^{42 56} (figure 4A), while the other two studies treated the number of exposure days the preceding 7 days as a continuous variable, ranging from 1 to ≥5 evening or night shifts^{36 57} (figure 3, online supplemental table S6). The risk of bias was moderate in all four studies (online supplemental table S2). Although these studies specified the number of shifts within a defined period (eg, the past 7 or 28 days), they generally did not report details on shift types or rotation direction (eg, forward versus backward rotation), with the exception of one study.⁵⁶

Figure 4. (A, B) Overview of OR or IRR for shift intensity. The exposure is defined as the number of shifts worked in the last seven or 28 days. IRR, incidence rate ratio.

Preceding night shift exposure

A significantly increased incident risk was observed among individuals working 6–12 night shifts⁴² or working more than eight night shifts in the past 28 days,⁴⁸ and 3–6 night shifts in the last 7 days^{42 56} (figure 4A). If individuals worked five or more night shifts in the 7 days prior to the incident, both studies^{36 57} reported a significantly higher risk even after just one night. However, as the number of night shifts increased, the risk appeared to gradually decline, showing elevated risk in only one of the studies.⁵⁷ One of the studies found a reduced risk associated with working three night shifts in the preceding 7 days.³⁶ This study revealed a U-shaped pattern, showing a decline in risk from the first to the third night shift, followed by an increase in risk from the third to the fourth night shift and beyond (ie, after more than five night shifts) (figure 4B).

Preceding evening shift exposure

One study indicated an increased risk after working three or more evening shifts during the preceding 7 days.³⁶ According to the other study, risk increased in a near-linear pattern with the number of evening shifts (1 to >5) during the preceding 7 days⁵⁷ (figure 4B).

Summary of findings

Overall, the results support the notion that exposure to three or more evening shifts in the previous 7 days increases the risk of safety incidents in the following week. Regarding night shifts, the previous 7 days, exposure to one night shift or to three or more night shifts seemed to increase the risk of safety incidents. We note, however, that the strength of the evidence from these studies was low (online supplemental table S4).

Other studies

Two studies were not categorised into any of the three exposure contrast categories (online supplemental table S7). The risk of bias was assessed as moderate for both studies (online supplemental table S2). One of these studies⁵⁸ examined safety incidents among shift workers during their commutes and found that the risk of incidents significantly increased when driving home after night shifts, compared with driving to work (OR 8.05 (95% CI 3.17 to 20.5)). The other study⁵⁹ compared physicians performing surgeries the day after a period of rest with those operating after midnight. This study found no increased risk of patient safety incidents among physicians who performed surgeries after midnight (OR 0.99 (95% CI 0.95, 1.03)).

Summary of findings by each exposure contrast

A summary of the number of studies, key findings, risk estimate ranges, populations studied and overall CoE for the three exposure contrasts is presented in Table 2. Detailed findings and risk estimates can be found in online supplemental tables S3–S7 and figures24.

Table 2. Summary of findings on the association between each exposure contrast and safety incidents.

Exposure contrast	Number of studies (E/O combinations)	Main findings (number/E/O combinations)	Risk estimates (range)	Populations studied	CoE
Shift worker versus non-shift worker	11 studies (14*)	Increased risk 9/14* One study showed a higher risk for day work	OR/HR/RR 1.11–5.33	General workforce, healthcare workers and manufacturing	Very low
Time-of-day (evening/night versus day shifts)	Night: 8 studies (9)	Increased risk 6/9	OR/RR/IRR 1.23–3.42	Healthcare, police, drivers and manufacturing	Very low
	Evening: 7 studies (8)	Increased risk 4/8	OR/RR/IRR 1.09–1.69
Shift intensity (≥1 night/evening shifts previous week)	Night: 4 studies	Increased risk (2/4) 1 night shift Increased risk (4/4) ≥3 night shifts	OR/IRR	Healthcare workers	Low
	Evening: 2 studies	Increased risk (2/2) ≥3 evening shifts

^*Only studies with a moderate risk of bias were included.

E/O, exposure/outcome; IRR, incidence rate ratio; RR, relative risk.

Discussion

Summary of findings

The present review included 24 original studies grouped into three sets of exposure contrasts. Overall, we found an association between shift work and an increased risk of safety incidents. However, the level of evidence was assessed as either low or very low, suggesting that future studies may alter this conclusion.

Being a shift worker

The initial contrast revealed that a working arrangement involving non-day work shifts was linked to a higher risk of safety incidents, particularly when excluding one study with a high risk of bias. Since the timing of these incidents was not defined, the risk was attributed solely to being a shift worker. Although this finding is novel, it has its limitations; it fails to identify which specific characteristics of shift work influence factors such as sleepiness, alertness and the risk of incidents. Additionally, the exposure periods varied greatly across studies, ranging from the previous 4 weeks³⁷ to 6 weeks³⁹ or even up to 12 months,⁴⁰ and some not defined at all. None of the nine studies examining this contrast reported daily working hours, making it unfeasible to calculate essential working time characteristics such as shift duration, time-of-day, number of consecutive shifts, rest periods or breaks—factors that likely impact incident risk.²² While we cannot definitively establish a causal link between any specific characteristics of shift work and an increased risk of incidents, the results do suggest that simply being a shift worker is associated with an elevated risk. To our knowledge, no previous systematic review has identified a relationship between just being a shift worker and an increased likelihood of safety incidents. The consistency of findings across various study populations, including the general workforce, healthcare settings and industrial environments, increases the external validity of our results.

Time-of-day

In evaluating the second contrast, regarding the time-of-day, our findings indicated an increased risk of safety incidents during or after night shifts.^{36 49 51 55} However, the association between incident risk following and during evening shift^{36 49 51 55} was more mixed. These results are largely consistent with those of two previous systematic reviews,^{22 23} although one of them did not assess evening shifts.²³ A strength of the studies in this contrast was the use of reliable exposure data from registries, with two studies providing detailed data on daily working hours.^{36 51} Both studies used a case–crossover design, which is advantageous for examining transient exposures and acute events by using subjects as their own controls, thereby reducing confounding from time-invariant factors. Each study examined the associations between various working time exposures. In addition to the findings mentioned above, they found that shorter rest periods between consecutive work shifts⁵¹ and the overall length of shifts were associated with higher risk of incidents.³⁶

The similar findings observed across various study populations, including the general workforce, healthcare professionals, drivers and police officers, enhance the external validity of the results, particularly the increased risk associated with night shifts. This notion is further supported by experimental studies indicating that alertness gradually declines during night shifts, compared with day shifts.^{19 60}

Shift intensity

In relation to the third contrast concerning shift intensity, exposure to multiple evening or night shifts within 7 days appears to be associated with an increased risk of incidents in the following week. Among the four studies that examined shift intensity, the highest confidence should be placed in the two studies that used data on daily working hours.^{36 51} However, the Hopcia et al study⁴² provided relatively robust exposure data, while the d’Ettore-study⁵⁶ may have been affected by recall bias. The findings regarding night shifts are consistent with one of the previous reviews,²² which indicates a dose-dependent relationship between the number of night shifts and incident risk.²² The declining trend in risk with increasing number of night shifts has also been seen in experimental studies of consecutive shifts and indicates that some level of adaptation may occur over consecutive night shifts.⁶¹

Regarding consecutive evening shifts, the same review did not find evidence to support a dose–response relationship. A potential explanation for the divergent findings related to several evening shifts could be that the two included original studies^{36 57} were based on daily working hours, thus minimising recall bias. Therefore, confidence can also be placed in the findings from these two studies, that working a single night shift in the previous 7 days increased the incident risk the subsequent week. A limitation of these studies is that they do not specify the day of the week on which the night shift occurred, despite the assumption that this information should be available in the dataset.

Risk of bias in the included studies

Bias was systematically assessed by evaluating each study’s potential for selection bias, information bias and confounding (online supplemental tables S1 and S2). Several studies did not report the inclusion or exclusion criteria, or did not report attrition, which increased the risk of selection bias. In contrast, other studies had high response rates or employed large payroll databases, thereby reducing the risk of selection bias.

Daily working hours or other forms of registry data on working hours were available and used in 12 of the 24 included studies, which helped to mitigate the risk of information bias. However, there remains significant potential to enhance the precision of exposure assessments in future research, which will clarify the shiftwork characteristics that are sustainable. Another important aspect of information bias is that studies should consider safety incidents occurring during or after shifts, or both. Additionally, we recommend that future studies express incidents as rates (eg, events per hour worked) to better account for variability in work hours and provide a more accurate assessment of risk.

In this review, we defined safety incidents as any workplace event resulting in non-fatal or fatal injuries from unexpected events (accidents), as well as reported near-miss incidents with the potential to cause injury. However, the definition of safety incidents varies widely across studies, which may influence risk estimates.^{36 47 49} For instance, while working shifts was associated with an increased risk of all injuries, it was not linked to a higher risk of the more serious injuries.⁴⁷

Confounding can occur if factors such as age, sex and socioeconomic status are not properly adjusted for; approximately half of the studies in this review accounted for these variables in their analyses (online supplemental tables S3, S5–S7). Additionally, potential confounders may act as effect modifiers, such as work content or other types of work exposure, which could influence outcomes. Unfortunately, this type of modifier was often not reported in the studies. Furthermore, differences in work demands across shift types and the nature of occupations are likely to impact the risk of incidents, and shift work tolerance has been shown to vary across occupations.⁶² Inadequate adjustment for these factors can lead to either an overestimation or an underestimation of incident risk. Other confounders that were not considered in the studies include circadian adjustments to night work⁶³ as well as individual factors such as chronotype^{16 64} and genetic susceptibility.⁶⁵

Strengths and limitations

We adhered to established standards for conducting systematic reviews by implementing rigorous inclusion and evaluation criteria to ensure a transparent assessment process. We followed SWiM guidelines^{28 66} for structured synthesis and applied the GRADE framework to evaluate the CoE in the context of occupational health. A strength of the study was that we grouped studies into three meaningful exposure contrasts, thereby synthesising heterogeneous evidence into a coherent framework. By grouping studies into three categories, shift worker versus non-shift worker, time-of-day and shift intensity, we captured distinct aspects of working time arrangements relevant to safety outcomes, despite limited and inconsistent reporting of specific shift characteristics in some of the included studies. It is a strength that cross-sectional studies were excluded, as this design cannot underpin causality.

There are some limitations that should be considered. First, only English-language publications were included, and grey literature was not searched, nor did we contact study authors for additional or unpublished data, possibly omitting relevant findings. Additionally, substantial heterogeneity in study designs, exposure definitions and reported outcomes prevented meta-analysis, precluding statistical pooling of effect sizes. Finally, while most included studies were longitudinal observational studies, it remains challenging to rule out confounding factors when interpreting cause-and-effect relationships.⁶⁷

Implications for research and societal impact

This systematic review has highlighted several important implications for future research. First, there is a need for increased utilisation of payroll-based data to better define the characteristics of working hours.⁵ Second, future studies should investigate combinations of exposure.²² Assessments of combinations of working-time exposure and psychosocial factors, such as job strain and emotional load, should be explored. Third, it is essential to examine how work schedules allow for sufficient time for restitution and sleep, as these factors may influence the risk of incidents. Finally, future research should include stratified analyses by sex and consider individual characteristics such as chronotype.

The prevalence of night and rotating shifts in society is likely to continue or even increase in the future, as these work schedules are essential for maintaining critical services such as healthcare and public safety,⁶⁸ as well as for meeting the demands of industries like logistics, transportation and retail. In this review, nearly half of the included studies (11 of 24) focused on the healthcare sector, which may limit the generalisability of the findings to other sectors. Differences in occupational exposures (eg, pathogens in healthcare versus physical hazards in manufacturing), national regulations and safety standards, and workforce demographics (eg, age and sex) could contribute to differences in incident risks or severity. Nevertheless, the proposed mechanisms through which shift work increases the likelihood of safety incidents, such as sleepiness,^{10 19} reduced alertness^{20 69} and impaired cognitive functioning,¹⁶ are expected to operate similarly across sectors and countries. Therefore, the external validity of our findings is likely robust. Overall, the evidence supports an association between non-standard shift arrangements and an increased risk of safety incidents.

The low certainty of the evidence in this systematic review limits the generalisability of its conclusions, and future studies could alter some of the conclusions. Additionally, the inability to conduct a meta-analysis due to significant heterogeneity among the studies limits the ability to draw robust and generalised conclusions about the effects of interventions or exposures in the occupational health field. Nonetheless, most studies suggest that non-standard work schedules have safety implications. Society should therefore seek to minimise its use and implement risk-reduction measures whenever such schedules are necessary. As VanderWeele noted a few years ago,⁷⁰ evidence may emerge from the aggregation of findings from several high-quality studies, even when these studies use diverse designs that are influenced by various types of biases.

Conclusion

In summary, the present systematic review suggests an association between non-standard working time arrangements and an elevated risk of safety incidents. First, the results suggest that being a shift or night worker is associated with an increased risk of safety incidents, compared with being a day worker. Second, most E/O combinations suggest an increased risk of incidents during or after night shifts. Finally, accumulated exposure to evening shifts over a 7-day period seems to increase the risk of incidents the following week.

To the best of our knowledge, no prior systematic review has indicated that merely being a shift worker is associated with an increased risk of safety incidents. Looking ahead, future research should include work demands and other relevant workplace exposures, and it should stratify analysis based on these and other characteristics, such as sex and chronotype. Furthermore, we recommend that future studies present incidents as rates (eg, events per hour worked) to more accurately account for variability in work hours and provide a clearer assessment of risk. Conducting high-quality, longitudinal, controlled studies with adequate sample sizes is essential to enhance the certainty of the evidence.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.