Random drug and alcohol testing for preventing injury in workers

Abstract

Background

Drug‐ and alcohol‐related impairment in the workplace has been linked to an increased risk of injury for workers. Randomly testing populations of workers for these substances has become a practice in many jurisdictions, with the intention of reducing the risk of workplace incidents and accidents. Despite the proliferation of random drug and alcohol testing (RDAT), there is currently a lack of consensus about whether it is effective at preventing workplace injury, or improving other non‐injury accident outcomes in the work place.

Objectives

To assess the effectiveness of workplace RDAT to prevent injuries and improve non‐injury accident outcomes (unplanned events that result in damage or loss of property) in workers compared with no workplace RDAT.

Search methods

We conducted a systematic literature search to identify eligible published and unpublished studies. The date of the last search was 1 November 2020. We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, Embase, two other databases, Google Scholar, and three trials registers. We also screened the reference lists of relevant publications known to us.

Selection criteria

Study designs that were eligible for inclusion in our review included randomised controlled trials (RCTs), cluster‐randomised trials (CRTs), interrupted time‐series (ITS) studies, and controlled before‐after (CBA) studies. Studies needed to evaluate the effectiveness of RDAT in preventing workplace injury or improving other non‐injury workplace outcomes. We also considered unpublished data from clinical trial registries. We included employees working in all safety‐sensitive occupations, except for commercial drivers, who are the subject of another Cochrane Review.

Data collection and analysis

Independently, two review authors used a data collection form to extract relevant characteristics from the included study. They then analysed a line graph included in the study of the prevalence rate of alcohol violations per year. Independently, the review authors completed a GRADE assessment, as a means of rating the quality of the evidence.

Main results

Although our searching originally identified 4198 unique hits, only one study was eligible for inclusion in this review. This was an ITS study that measured the effect of random alcohol testing (RAT) on the test positivity rate of employees of major airlines in the USA from 1995 to 2002. The study included data from 511,745 random alcohol tests, and reported no information about testing for other substances. The rate of positive results was the only outcome of interest reported by the study.

The average rate of positive results found by RAT increased from 0.07% to 0.11% when the minimum percentage of workers who underwent RAT annually was reduced from 25% to 10%. Our analyses found this change to be a statistically significant increase (estimated change in level, where the level reflects the average percentage points of positive tests = 0.040, 95% confidence interval 0.005 to 0.075; P = 0.031). Our GRADE assessment, for the observed effect of lower minimum testing percentages associating with a higher rate of positive test results, found the quality of the evidence to be 'very low' across the five GRADE domains. The one included study did not address the following outcomes of interest: fatal injuries; non‐fatal injuries; non‐injury accidents; absenteeism; and adverse effects associated with RDAT.

Authors' conclusions

In the aviation industry in the USA, the only setting for which the eligible study reported data, there was a statistically significant increase in the rate of positive RAT results following a reduction in the percentage of workers tested, which we deem to be clinically relevant. This result suggests an inverse relationship between the proportion of positive test results and the rate of testing, which is consistent with a deterrent effect for testing. No data were reported on adverse effects related to RDAT.

We could not draw definitive conclusions regarding the effectiveness of RDAT for employees in safety‐sensitive occupations (not including commercial driving), or with safety‐sensitive job functions. We identified only one eligible study that reflected one industry in one country, was of non‐randomised design, and tested only for alcohol, not for drugs or other substances. Our GRADE assessment resulted in a 'very low' rating for the quality of the evidence on the only outcome reported. The paucity of eligible research was a major limitation in our review, and additional studies evaluating the effect of RDAT on safety outcomes are needed.

Plain language summary

Does random drug and alcohol testing prevent injuries in workers?

Background

Workplace accidents and injuries happen more often when people's physical abilities and judgement are impaired by drugs or alcohol. The workplace is not usually a place where research is conducted. There are many factors that make it difficult for an employer to measure the impact of a workplace drug testing program on the overarching goal of ensuring a safe workplace.

Some employers — particularly in sectors where safety is very important, such as the commercial driving and airline industries — choose to test workers randomly for drugs and alcohol (random drug and alcohol testing (RDAT)). Through such testing, employers hope to deter employees from inappropriate use of these substances. However, we do not know if RDAT produces the desired effect.

Review question

We wanted to know whether RDAT in the workplace prevents injuries and unplanned events that result in damage or loss of property (non‐injury accidents) compared with no RDAT.

Search date

The evidence in this Cochrane Review is current to 1 November 2020.

Study characteristics

We wanted to include all of the relevant research about RDAT in the workplace in our review. We looked for different kinds of published studies that measured how RDAT affected workplace safety. We excluded research on RDAT in commercial drivers, because another Cochrane Review covering these studies has already been published.

Two authors from our team examined all of the references identified by our search, but they found only one study that met our selection criteria and could be included in the review.

This study investigated random alcohol testing in airline employees in the USA whose jobs included safety‐related tasks. The study did not test employees for drugs. Airlines are required by law to test and report on a randomly selected sample of their employees. The study used testing data from between 1995 and 2002. A total of 511,745 random alcohol tests were performed on airline employees.

Key results

From 1995 to 1997, random tests for alcohol included 25% of the relevant airlines' workforce each year. During this period, the average percentage of employees who tested positive for alcohol was 0.07%. From 1998 to 2002, the proportion of the workforce tested each year dropped to 10%. During this period, the average percentage of employees who tested positive for alcohol increased to 0.11%.

This means that when the airlines randomly tested a larger percentage of employees per year, a smaller proportion of them tested positive for alcohol. This relationship between the frequency of alcohol testing and the proportion of positive tests is what we would expect to see if testing has a deterrent effect, though one study alone cannot prove that there is a deterrent effect.

This study did not provide any information for other areas of interest to us, specifically:

‐ fatal injuries;

‐ non‐fatal injuries (in which people are physically injured, but do not die);

‐ 'non‐injury accidents'; that is, accidents in which people are not injured, but property, processes, materials, and/or the environment are damaged;

‐ absenteeism; and

‐ unwanted, or adverse, events associated with RDAT, including impacts on privacy, confidentiality and employee perceptions.

Quality of the evidence

Two of our team rated our confidence in the evidence, based on factors such as number of studies, study size and methods. Overall, our confidence in the evidence was very low. This means we cannot rely on what this study reported to make generalisations about the effectiveness of random alcohol testing alone, or random alcohol testing combined with drug testing (RDAT), in the workplace. We need researchers to do more studies to find the answers.

Study funding sources

The one study we included in our review was funded in part by two grants: a grant from the National Institute on Alcohol Abuse and Alcoholism, and a grant from the Centers for Disease Control and Prevention.

Summary of findings

Summary of findings 1. The effects of random alcohol testing (RAT) on workplace safety outcomes and adverse events.

Population: Workers in positions with safety‐sensitive functions, other than commercial drivers Intervention: RAT Comparison: Comparing two frequencies of RAT, 25% or 10% of the workforce annually
Outcomes	Assumed risk	Corresponding risk	Relative effect	No. of random tests (studies)	Quality of the evidence (GRADE)	Comments
Rate of positive results found by RAT among airline employees in the USA Random tests were conducted by certified technicians using National Highway Traffic Safety Administration approved devices according to the US Department of Transportation Procedures for Transportation Workplace Drug Testing Programs (Department of Transportation (US) 2019). A positive result was defined as a test result where a blood alcohol concentration was ≥ 0.04%, or when a worker refused to be tested. Testing data were retroactively provided by the US Federal Aviation Administration (FAA) to the study authors some time after 2002, the final year of the study's data.	Rate of positive results found in FAA‐regulated aviation personnel performing safety‐sensitive functions from 1995 to 1997 (25% of workplace required to be tested) was 0.07%.	Rate of positive results found in FAA‐regulated aviation personnel performing safety‐sensitive functions from 1998 to 2002 (10% of workplace required to be tested) was 0.11%.	59.7%	511,745 random tests (1 study)	Very lowa	When the review authors re‐analysed the results, they found a statistically significant increase (estimated change in level: 0.040, 95% CI 0.005 to 0.07; P = 0.031)
Fatal injuries — not measured	‐	‐	‐	‐	‐	No data reported for this outcome in the included study
Non‐fatal injuries — not measured	‐	‐	‐	‐	‐	No data reported for this outcome in the included study
Non‐injury accidents — not measured	‐	‐	‐	‐	‐	No data reported for this outcome in the included study
Absenteeism — not measured	‐	‐	‐	‐	‐	No data reported for this outcome in the included study
Adverse events associated with RAT — not measured	‐	‐	‐	‐	‐	No data reported for this outcome in the included study

Summary of findings table created with guidance from Chapter 14, Section 14.1 of the Cochrane Handbook of Systematic Review of Interventions (Higgins 2020) and Ryan 2016b.

Abbreviations

FAA: United States Federal Aviation Administration

RAT: random alcohol testing

Grade assessment: the evidence quality began at a 'low' rating, as the included study was non‐randomised. The evidence was further downgraded because of a rating of 'serious' for indirectness. This judgement of 'serious' was given because: only two random testing regimes (10% and 25%) were investigated; only one industry was investigated; only random alcohol testing was evaluated; and the comparison of one random testing regime against another is not as high quality as comparing a random testing regime against no random testing regime.

Background

Humans have been consuming psychoactive substances since prehistoric times (Guerra‐Doce 2015). Before the advent of the industrial revolution, the consumption of alcohol in some workplaces was viewed as normal, and alcohol was often used as a substitute for water—at times even offered to workers as payment (Trice 1981). A normative shift occurred in the early 20th century, with decreasing acceptance of alcohol consumption in the workplace (Taylor 1915). At the beginning of the 21st century, attention to the use of drugs other than alcohol led to concern about the potential adverse impact of these substances on safety in the workplace (Frone 2013). Today, among researchers, policymakers and employers, there is a widely held view that the consumption of substances by employees is a risk factor for injuries and workplace accidents (Frone 2013). Currently, substance use represents a leading cause of preventable death and disease, as well as a significant global public health concern and economic burden.

Description of the condition

Numerous factors, both occupational and non‐occupational, contribute to workplace injuries (Dong 2015). The consumption of psychoactive substances, among other factors, may result in occupational impairment and, hence, occupational injury risk. Abuse of alcohol, cocaine, cannabis and other substances is associated with workplace injuries (Chau 2009; Dong 2015; Pollack 1998; Shipp 2005). Substance abuse has become a growing concern for employers given their obligation to maintain a safe and healthy workplace for their employees. The 2015 National Survey on Drug Use and Health in the USA found that approximately 70% of adults who reported using illicit substances in the month before the survey were employed (CBHSQ 2016). Approximately 5% to 18% of full‐time adult workers in the USA met the criteria for a substance use disorder in 2015 (Bush 2015).

The economic impact of workers' alcohol and drug use is also an important concern. The Alberta Alcohol and Drug Abuse Commission in Canada estimated that in 2002, in the Alberta workforce, lost productivity related to alcohol use was valued at more than CAD 32 million, and drug use at more than CAD 13 million (AADAC 2003).

It is well established that alcohol or substance use may impair skills related to operating a motor vehicle or machinery. The cognitive, motor and other skills required for safety‐sensitive and decision‐critical duties overlap to varying degrees with the skills required to operate safely a motor vehicle or machinery (Hegmann 2014). Driving can therefore be viewed as a proxy for the prediction of impairment for other safety‐sensitive and, by extension, some decision‐critical tasks (Hegmann 2014). Alcohol and other substance use both contribute to the risk of an accident, and the risk of being fatally injured is increased when drivers ‐ whether they have consumed alcohol or not ‐ test positive for another drug (Romano 2014).

Sources of data that address work‐related injuries and accidents come primarily from medical examiner records and workplace drug testing programmes (Frone 2013). Estimates suggest that alcohol‐related impairment occurs in approximately 5% and 10% of non‐fatal and fatal work injuries, respectively (Zwerling 1993). In 2018, based on analysis of nearly nine million urine tests, the annual rate of positive test results found by workforce drug testing in the USA was 4.4%, the highest in the past 15 years (Quest Diagnostics 2019).

Despite being illegal in most jurisdictions, cannabis is the most commonly used psychoactive substance globally (Els 2019). Cannabis was also the most commonly found substance in a workplace drug‐testing context in Canada (Els 2016), and that was before its legalisation for recreational use in Canada in 2018. In North America, with the ongoing progressive legalisation of cannabis and the opioid use epidemic, we consider the corresponding potential for a serious adverse impact on occupational health and safety to be likely, substantial and foreseeable.

With the increased focus on the impact of alcohol and substance use both in and outside the workplace, greater attention is being paid to interventions to mitigate the risk of harm, especially in safety‐sensitive work settings. Approximately 20% of workers in the USA are employed in safety‐sensitive positions (CBHSQ 2016). A salient example where alcohol may have contributed occurred on 24 March 1989, where the Exxon Valdez oil tanker was responsible for the then largest single oil spill in USA coastal waters (NSCEP 1989), which may have been associated with the ship captain’s reported alcohol abuse (Brown 2013). The magnitude of the environmental disaster resulting from this incident, along with other critical accidents involving drugs or alcohol in the 1980s and early 1990s, contributed to the introduction of workplace drug and alcohol testing in safety‐sensitive settings (Brown 2013). This introduction of testing occurred despite a dearth of empirical evidence of its effectiveness in preventing occupational accidents and injuries (Murray 1996; Murray 2013; Whiteford 2013).

Description of the intervention

Several different approaches and interventions have been utilised to mitigate the occupational risk from alcohol or drug‐related impairment (Dyck 2013). Interventions include: voluntary peer‐based assistance programs (Golan 2010); employee assistance and aftercare programs offered by the employer (Waehrer 2016); training supervisors to identify impairment (Cenovus 2011); worker education programs on substance abuse (Cook 2003); drug‐free workplace policies (ACCA 2010) (with or without drug testing (Huestis 2007), discipline, counselling (Knudsen 2004), or rehabilitation (AHRC 2012)); and random drug and alcohol testing (RDAT), which can be combined with other measures or used as a stand‐alone intervention (VicRoads 2015).

Drug and alcohol testing can be conducted in a variety of contexts, including: pre‐employment; for reasonable cause; post‐incident; as part of follow‐up monitoring after treatment for a substance use disorder; prior to a return‐to‐work; or as random testing.

Despite variability in policies, practices and implementation, RDAT programs usually have a number of features in common. Thus, Coates defines random testing as "the unscheduled, unannounced drug testing of randomly selected employees by a process designed to ensure that selections are made in a nondiscriminatory manner" (Coates 2014). Employers typically decide what percentage of employees are tested annually depending on the needs of the company. At least 50% of the workforce, tested annually, is suggested as a reasonable baseline target (Frone 2013). The Department of Transportation in the USA recommends testing on at least a quarterly basis (Department of Transportation (US) 2015). It provides guidance and best practices that are widely used in the USA and Canada (COAA 2018).

How the intervention might work

Workplace drug testing aims to detect and deter drug use in workers. As the workplace is usually not a place of research, there are many factors that may make it difficult for the employer to measure the impact of a workplace drug testing program on the overarching goal of ensuring a safe workplace. That said, research indicates that workplace drug testing is most likely to be a deterrent in more addicted or very frequent drug users (Frone 2013). Theoretically, employees who consume substances in violation of punitive workplace drug‐free policies (which specify that workers will be disciplined, sanctioned or discharged following a positive alcohol or drug test) should be motivated to discontinue consumption. A study of drug testing in the US Navy, where there is a zero‐tolerance policy to drug use, estimated that RDAT would deter almost 60% of potential drug use (Borack 1998). A study of mandatory alcohol testing for large commercial truck and bus drivers found the risk of alcohol involvement in fatal crashes dropped by 23% (Brady 2009). In theory, if RDAT is found to deter employees from using alcohol or drugs, this reduction in use may, in turn, reduce the associated risk of occupational accidents and injuries.

Alternatively, a positive alcohol or drug test may trigger a comprehensive medical assessment, followed by rehabilitative measures, such as entry into addiction treatment or detoxification. Non‐negative or positive drug or alcohol tests in the workplace may serve as a mechanism for the early identification of workers at risk of, or affected with, an addiction or substance use disorder, who can then be referred to appropriate interventions.

We have developed a logic model to illustrate the mechanism by which the complex intervention of RDAT programs might work (Table 2). This approach has been used in other contexts (Anderson 2011; Baxter 2010; Pigott 2013). Contextual factors comprise three main domains: 1) company characteristics (size, location, industry, organisational climate); 2) job characteristics (types of positions and work content, i.e. job demands, decision latitude, effort, and schedule); and 3) employee characteristics (especially socioeconomic status, age, sex, and tobacco smoking).

1. Logic model.

Context Company characteristics: size location industry organisational climate Job characteristics: types of positions work content Employee characteristics: socioeconomic status age sex or gender tobacco smoking previous history of addiction or substance use disorder(s)
Inputs	Intervention	Intermediate outcomes	Longer‐term outcomes
Identification of need for RDAT: safety‐sensitive work demonstrated drug and/or alcohol problem in the workplace desire to reduce workplace injuries and accidents Resources (supplies, personnel, monetary)	Testing: number of tests completed number of positive test results number and percentage of employees tested schedule of testing Service provision: employee assistance program drug and alcohol education drug and alcohol treatment Data collection: fatal injury rate non‐fatal injury rate non‐injury accident rate absenteeism adverse events associated with RDAT	Deterrence of use or nonuse related to: punitive action (discipline, sanction, or discharge penalties); rehabilitative action (receiving treatment for an addiction or substance use disorder); receiving accommodations for substance use as a disability	Injuries: changes in fatal injury rate changes in non‐fatal injury rate changes in non‐injury accident rate Changes in rate of absenteeism Adverse events associated with RDAT

Abbreviation

RDAT: random drug and alcohol testing

Why it is important to do this review

RDAT is a controversial subject with implications in a number of domains. For example, in the unionised context, establishing whether or not RDAT is effective at preventing workplace injuries and accidents may inform the balancing of rights between employers' right and legal obligation to impose rules relating to workplace safety and employees' right to privacy. In the human rights context, establishing whether or not RDAT is effective may inform the question as to whether RDAT is reasonably necessary to meet the employer's goal of increasing workplace safety and reducing occupational accidents.

Despite the ongoing and often vigorous debate over its benefits and risks, as well as the limitations, the effectiveness of RDAT as a safety strategy has received limited research attention. Although RDAT as an intervention is claimed to have a deterrent effect on drug and alcohol use, no one has conducted a methodologically rigorous systematic analysis of the evidence in recent years.

Objectives

To assess the effectiveness of workplace RDAT to prevent injuries and improve non‐injury accident outcomes (unplanned events that result in damage or loss of property) in workers compared with no workplace RDAT.

Methods

Criteria for considering studies for this review

Types of studies

Given the infeasibility of conducting randomised studies (especially with randomisation of the individual participant) in this area, we considered eligible for inclusion in our review all randomised as well as non‐randomised studies of the following types.

• Randomised controlled trials (RCTs), defined as studies in which participants are randomly allocated into groups to receive an intervention. Identical treatment is provided to all groups with the exception of the intervention the research is designed to study (Hammond 2015).

• Cluster‐randomised trials (CRTs); that is, RCTs that involve groups of participants, as opposed to individuals, as the unit of randomisation. Comparisons are then made between these clusters rather than between individuals (Kaura 2015).

• Interrupted time‐series (ITS) studies, where observations of a group are taken repeatedly over time and used to establish an underlying trend, which is then interrupted by an intervention. The analysis of ITS data provides statistical evidence about whether changes in the trend represent real increases or decreases (Bernal 2017).

• Controlled before‐after (CBA) studies, in which outcomes of interest are measured in both intervention and control groups before and after an intervention has been performed (EPOC 2017a).

In order to be inclusive and to facilitate capturing all relevant data, we considered studies published in the peer‐reviewed literature as well as unpublished data from clinical trial registries to be eligible for inclusion.

While the highest‐quality (lowest risk of bias) results could be expected from RCTs, we reasoned that restricting the review to these trials only might neglect evidence in the form of studies without individual participant randomisation. Indeed, for RDAT, as previously mentioned, randomisation of the individual participant may not be practical. Therefore, we made the decision to include other study designs (CRTs, ITS studies and CBA studies), which are easier to conduct in an occupational health setting, even though they are more prone to bias. Such studies could provide important, practically relevant information on the effectiveness of RDAT. We have taken the higher risk of bias inherent in these study designs into account in our analysis and conclusions.

Types of participants

We considered studies conducted on adult workers in any occupation as eligible for inclusion, with the exception of commercial drivers, as this is the subject of another Cochrane Review (Cashman 2009).

Types of interventions

Studies that evaluated the effectiveness of workplace RDAT according to the following criteria were eligible for inclusion.

• Randomness: each worker studied had to have an equal likelihood of being chosen for testing, and with the choice made through a probabilistic method (e.g. using a random‐number table, a computer‐generated list of random numbers, or drawing numbers out of a hat to select from a list of consecutively numbered names).

• Substances tested: random testing in the study was performed to detect:

  • alcohol;

  • illicit or prescription substances other than alcohol; or

  • both alcohol and illicit or prescription substances.

• Frequency of testing: random testing in the study was performed on at least a quarterly basis or more frequently, with scheduled retesting at regular intervals as a planned part of the program.

• Proportion of workers tested: random testing in the study could be performed on any proportion of the worker population, as long as test results from at least 10 subjects were described in each contributing study.

• Setting: the study could be conducted in any work‐related setting, excluding those where the tested employees were commercial drivers (as this is covered in another Cochrane Review, Cashman 2009).

We considered studies that investigated co‐interventions to be eligible for inclusion, provided that they were implemented in both the RDAT and comparator arms.

Types of outcome measures

Studies that evaluated the effectiveness of workplace RDAT by using any suitable measures of any of the outcomes below were eligible for inclusion.

Primary outcomes

• Fatal injuries: according to Ehnes 2012, an occupational injury is "... any personal injury, disease or death resulting from an occupational accident. An occupational accident is an unexpected and unplanned occurrence, including acts of violence, arising out of or in connection with work which results in one or more workers incurring a personal injury, disease or death. An occupational injury is therefore distinct from an occupational disease, which is a disease contracted as a result of an exposure over a period of time to risk factors arising from work activity".

• Non‐fatal injuries (defined as above, but excluding incidents that result in death).

• Non‐injury accidents: according to Binch 2007, non‐injury accidents are "...any unplanned event that results in damage or loss to property, plant, materials, the environment, and/or a loss of business opportunity but does not result in injury".

Secondary outcomes

• Rate of positive results found by RDAT

• Absenteeism (reported as days absent per time period)

• Adverse events associated with RDAT: we considered any reported adverse events such as impacts on privacy and confidentiality, including employee perceptions of intrusiveness, separately.

Search methods for identification of studies

Electronic searches

We conducted a systematic literature search to identify all published and unpublished studies that could be considered eligible for inclusion in this review. We adapted the search strategy we developed for MEDLINE (Appendix 1) for use in the other electronic databases. We did not restrict the search by date or language of publication.

We searched the following databases and registers to identify potential studies.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 11) in the Cochrane Library (searched 4 November 2020).

• MEDLINE Ovid (1946 to 30 October 2020) (searched 1 November 2020 (Appendix 1).

• PsycINFO Ovid (1806 to October week 4 2020) (searched 2 November 2020) (Appendix 2).

• Embase Ovid (1974 to 29 October 2020) (searched 4 November 2020) (Appendix 3).

• Google Scholar (https://scholar.google.ca/; searched 3 November 2020).

• Occupational Safety and Health (OSH) Update (19th century to 24 May 2019). We previously had access to this database through the Cochrane Work Editorial Group, but did not have such access for our search update in November 2020.

• OSH References Collection (19th century to October 2020).

• ClinicalTrials.gov (www.ClinicalTrials.gov; searched 3 November 2020).

We searched the World Health Organization International Clinical Trials Registry Platform (www.who.int/ictrp/en) in May 2019. Due to high traffic related to COVID‐19, this site was not searchable despite multiple attempts when we updated our search in November 2020, and so we searched the ISRCTN Registry and EU Clinical Trials Register (searched 4 November 2020) instead.

Searching other resources

We searched the reference lists of original studies and review articles for additional references and continued this in an iterative fashion until no new studies were identified. We also searched the publication history of frequently cited authors known to us.

Data collection and analysis

Selection of studies

We used Covidence 2020 to assist with study selection. We conducted the selection of eligible studies in two stages. First, two review authors (out of CE, TJ, MM, GW, DS, DD) independently screened the titles and abstracts of the systematic search results to identify studies for inclusion. Each reviewer assessed studies as potentially eligible or ineligible. We excluded, as ineligible, studies that clearly did not fulfil our inclusion criteria or definitely fulfilled one or more exclusion criteria. At the second stage, we retrieved the full‐text publications of the potentially eligible studies and two review authors (out of CE, TJ, DK, GW, DS) independently assessed these and screened studies for final inclusion. We recorded reasons for exclusion of studies assessed as full‐texts in Table 3. We resolved any disagreement through discussion or, if required, we consulted another review author (SS).

2. Reasons for exclusion.

Reason for exclusion	Excluded studies
Ineligible study design	Buchanan 1988; Carpenter 2007; Crouch 1989; DuPont 1995; Elliott 2006; Fitzsimons 2008; Fox 2014; French 2004; Li 2010; Li 2011; Maretha 2016; Marques 2014; McFadden 1997; Miller 2007; Ozminkowski 2003; Price 2012; Price 2014a; Price 2014b; Price 2015; Price 2016; Schofield 2011; Schofield 2013; Vignali 2013
Primary research not reported	Carlton 2004; Grabowski 1989

Data extraction and management

We used a data extraction form for study characteristics and outcomes. Two review authors (TJ and DK) independently extracted study characteristics from the included study. Disagreements were resolved by consensus, or if required, by consulting another review author (SS). Where available, we extracted the following study characteristics.

• Study information: study design; study duration; and study location.

• Participant information: number of participants; number of withdrawals; reason for withdrawals; sex or gender; mean age or age range; race or ethnicity; occupation(s); employer or company; and study inclusion/exclusion criteria.

• Interventions: screening methods; whether screening was for alcohol, non‐alcoholic substances, or both; frequency of screening; number of employees tested; and total number in testing pool.

• Outcomes: fatal injuries; non‐fatal injuries; non‐injury accidents; rate of positive results found by RDAT; absenteeism; and adverse events.

We noted in the 'Characteristics of included studies' table if outcome data were not reported in a usable way.

We took occupations specifically identified for the study population and coded these occupations into the 10 broad occupational structure categories and the unit group codes of Canada's National Occupational Classification (Statistics Canada and ESDC 2018). We also reported the industry employing the study population as a code in the North American Industry Classification System (NAICS) (Statistics Canada 2012). Two review authors (TJ, GW) independently performed and reconciled this coding, resolving disagreements by consensus or by involving another review author (SS).

We did not enter any data into the Data and analysis section of Review Manager 5.4 (RevMan 2020). Had data been available, one review author (TJ or GW) would have entered data into Review Manager 5.4. A second review author (CE) would have confirmed that the data were entered correctly by spot‐checking study characteristics for accuracy against the study reports, and comparing the data presented in the systematic review with those in the study reports. Had we decided to include studies published in languages in which our author team is not proficient, we would have arranged for a native speaker or a sufficiently proficient translator to complete a data extraction form for us.

Assessment of risk of bias in included studies

Two review authors (CE, GW) independently assessed the risk of bias in the included study. We resolved any disagreements by discussion or by involving another author (SS).

For the ITS study we included, we used the risk of bias criteria developed by the Cochrane Effective Practice and Organisation of Care (EPOC) Group (EPOC 2017b). Thus, we graded each potential risk of bias as 'high', 'low', or 'unclear' in each of the following domains.

• Was the intervention independent of other changes?

• Was the shape of the intervention effect prespecified?

• Was the intervention unlikely to affect data collection?

• Was knowledge of the allocated interventions adequately prevented during the study?

• Were incomplete outcome data adequately addressed?

• Was the study free from selective outcome reporting?

• Was the study free from other risks of bias?

Had it been necessary, we would have assessed the risk of bias in RCTs and CRTs according to the following standard domains, grading each potential risk of bias as 'high', 'low', or 'unclear'.

• Random sequence generation

• Allocation concealment

• Blinding of participants and personnel

• Blinding of outcome assessment

• Incomplete outcome data

• Selective outcome reporting

• Other biases

Had it been necessary, we would have assessed CBA studies using the risk of bias criteria as given in Sterne 2016, grading each potential risk as 'low', 'moderate', 'serious', 'critical', or 'no information', in each of the following domains.

• Bias due to confounding

• Bias in selection of participants into the study

• Bias in classification of interventions

• Bias due to deviations from intended interventions

• Bias due to missing data

• Bias in measurement of outcomes

• Bias in selection of the reported result

Potential confounding domains that we anticipated would be relevant to included studies were socioeconomic status, age, sex, and tobacco smoking.

Assessment of bias in conducting the systematic review

We conducted the review largely according to the published protocol. We reported any deviations from the protocol in the 'Differences between protocol and review' section of this review.

Measures of treatment effect

Had we found suitable data, we would have entered the outcome data for each study into the data tables into Review Manager 5.4 (RevMan 2020), in order to calculate the treatment effects.

For the one included ITS study, we extracted and re‐analysed their data to assess differences in average percent of positive tests (level) and in the time trend (slope). To do this, we followed the procedure described in Ramsay 2003. We performed multiple checks for auto‐correlation and used the most appropriate regression model to assess pre‐ and post‐intervention level and time‐trends. Each data point was defined as an annual period, starting with period 1. Periods 1, 2, and 3 were pre‐intervention, and periods 4, 5, 6, 7, and 8 were post‐intervention.

The general model tested was:

Y = β₀ + β₁t + β₂I(t ≥ 4) + β₃tI(t ≥ 4) + ε,ε~N(0,σ)

Period 4 is the beginning of the intervention, and the indicator variable I(・) is one when the condition in the parenthesis is true, and zero otherwise. The parameters have the following interpretation:

β₀ is the pre‐intervention intercept

β₁ is the slope of the regression pre‐intervention

β₂ is the level change pre‐ and post‐intervention

β₃ is the change in slope post‐intervention

We performed the statistical analysis in Stata 15 for Windows (Stata 2017).

We allowed for the possibility that the errors were auto‐correlated, and in that case, we would have used the Prais‐Winston first‐order auto‐correlation version of generalised least squares (GLS) regression estimation, otherwise reverting to ordinary least squares. We defined a change in level as an immediate shift in the regression line post‐intervention, and a change in slope post‐intervention representing a gradual response to the intervention.

Unit of analysis issues

Had we included studies that employed a cluster‐randomised design and reported sufficient data to be included in the meta‐analysis, but did not make an allowance for the design effect, we would have calculated the design effect based on a fairly large assumed intracluster correlation of 0.10. We based this assumption of 0.10 being a realistic estimate on an analogy with studies on implementation research (Campbell 2001). We would have followed the methods described in the Cochrane Handbook for Systematic Reviews of Interventions for the calculations (Higgins 2020).

In the re‐analysis that we did on data reported in the included study, given that these data were reported on an annual basis, the unit is the proportion of positive outcome randomised drug testing.

Dealing with missing data

We would have employed a conservative approach for dealing with missing data, preferring baseline observation carried forward over last observation carried forward, if both were reported. If it had been possible and appropriate to calculate values for missing data from other statistics reported in studies, we would have done so. If such computation were not possible, we would have contacted authors to request additional data and reported which study analyses made use of unpublished data.

Assessment of heterogeneity

We would have used the I² statistic to assess statistical heterogeneity among the studies in each meta‐analysis. We would have discussed any substantial statistical or clinical heterogeneity.

We would have considered the following three substance groupings as being clinically heterogeneous.

• Alcohol only

• Illicit or prescription substances but no alcohol

• Alcohol and any illicit or prescription substances

We would have provided further detail about the planned assessment of clinical heterogeneity in the section 'Subgroup analysis and investigation of heterogeneity'.

Assessment of reporting biases

Had we been able to pool data from more than 10 studies in any single meta‐analysis, we would have created and examined a funnel plot to explore possible reporting biases.

Data synthesis

As we included only one study, we could not pool data nor perform a meta‐analysis.

'Summary of findings' table

We created 'Table 1' using the following outcomes.

• Fatal injuries

• Non‐fatal injuries

• Non‐injury accidents

• Rate of positive results detected by RDAT

• Absenteeism

• Adverse events associated with RDAT

We completed Table 1 with guidance from Chapter 14, Section 14.1, of the Cochrane Handbook of Systematic Review of Interventions (Higgins 2020), and Ryan 2016b.

We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of the evidence from the included study. We used GRADEpro GDT 2015 to conduct the assessment. When conducting the assessment, we used the methods and recommendations described in Ryan 2016a. When determining whether or not to use the ROBINS‐I tool in GRADEpro GDT, we referred to Sterne 2016 and Chapter 14, Section 14.2, of the CochraneHandbook (Higgins 2020).

Subgroup analysis and investigation of heterogeneity

We planned to conduct subgroup analyses if there were sufficient data. We would have examined the effects of RDAT according to the presence or absence of four elements:

• safety‐sensitive work;

• manual labour;

• testing for cannabinoids; and

• testing for opioids.

We would have treated studies of different designs separately; that is, we would not have combined data from the four eligible study designs. We would have discussed the implications of 'high' or 'low' risks of bias, inherent in these study designs and specific to the studies.

Further, we would have pooled outcome data for three comparable time points, defining short‐term follow‐up to be up to and including one month, medium‐term to be greater than one month and less than one year, and long‐term as one year or greater.

Sensitivity analysis

We would have performed a sensitivity analysis to assess the robustness of our conclusions by omitting studies with a 'high' risk of bias, if studies with both 'low' and 'high' risks of bias had been present.

Summary of findings and assessment of the certainty of the evidence

We constructed a 'Summary of findings' table that presents our GRADE judgements concerning the certainty of the evidence.

Reaching conclusions

We based our conclusions only on the results of our analysis of the data presented in the included study. Our implications for research suggest priorities for future research and outline the remaining uncertainties in the area.

Results

Description of studies

We obtained all studies from the electronic database searches.

Results of the search

We identified a total of 4574 articles, which was reduced to 4198 articles after eliminating duplicates. Two researchers (out of CE, TJ, MM, GW, DS, DD) screened these 4198 articles by titles and abstracts, excluding 4098 studies based on relevance. Two researchers (out of CE, TJ, MM, GW, DS) then assessed the remaining 100 articles via full‐text screening, with one article ultimately meeting all of the inclusion criteria. The reasons for exclusion for the other 99 assessed articles are detailed in Table 3. The entire process is illustrated in Figure 1.

1.

Study flow diagram.

Included studies

We included one study (Li 2007) from the USA. This study analysed data from random alcohol testing (RAT) as reported to the US Federal Aviation Administration (FAA) by major airlines, in an ITS design for the period of 1995 to 2002. Tests were conducted on employees with safety‐sensitive functions working for major airlines.

For context, the FAA stated in 1994 that it might set the minimum percentage of employees required to be tested to 10 per cent, if the rate of violations for the previous two years was less than 0.5 per cent (FAA 1994). Accordingly, in 1998, when the violation rates for two consecutive years had been less than 0.5 per cent, the minimum percentage of employees required to be tested was reduced (FAA 1997).

According to the North American Industry Classification System (NAICS) (Statistics Canada 2012), the aviation industry that this study reported on is best defined under the code 4811 for scheduled air transportation. According to Canada's National Occupational Classification (Statistics Canada and ESDC 2018), the occupations on which this study reported are best identified and coded as shown in Table 4.

3. National occupational classifications of employees in Li 2007.

Occupation	Broad occupational category (BOC)	BOC code	Unit group title	Unit group code
Flight crew	Natural and applied sciences and related occupations	2	Air pilots, flight engineers and flying instructors	2271
Flight attendants	Sales and service occupations	6	Pursers and flight attendants	6522
Flight instructors	Natural and applied sciences and related occupations	2	Air pilots, flight engineers and flying instructors	2271
Aircraft dispatchers	Natural and applied sciences and related occupations	2	Air traffic controllers and related occupations	2272
Maintenance personnel	Natural and applied sciences and related occupations	2	Aircraft instrument, electrical and avionics mechanics, technicians and inspectors	2244
Aviation screeners	Trades, transport and equipment operators and related occupations	7	Aircraft mechanics and aircraft inspectors	7315
Ground security co‐ordinators	Sales and service occupations	6	Security guards and related security service occupations	6541
Air traffic controllers	Natural and applied sciences and related occupations	2	Air traffic controllers and related occupations	2272

The occupations listed have been classified according to Canada’s National Occupational Classification (Statistics Canada and ESDC 2018).

Excluded studies

The reasons we excluded the 99 studies assessed at the full‐text level are listed in Characteristics of excluded studies and Table 3. The reasons for exclusion were:

• ineligible study design (63 studies);

• primary research not reported (23 studies);

• insufficient detail reported (9 studies);

• ineligible participant population (2 studies); and

• testing frequency not reported (2 studies).

Risk of bias in included studies

When considering intervention effects, we used the risk of bias criteria developed by the Cochrane Effective Practice and Organisation of Care (EPOC) Group (EPOC 2017b). For all of the risk of bias judgments, we summarised evidence reported in the study, together with a justification for our judgments, which is presented in the 'Risk of bias' table (included with the 'Characteristics of included studies' table). We judged the study to have a 'low' risk of bias overall, as the majority of domains had a 'low' risk of bias.

Was the intervention independent of other changes?

The study authors indicated it is unlikely that the intervention occurred independently of other changes, saying, "... it is probable that a small portion of the observed increase in alcohol violations during 1998 and 2002 was due to confounding effects from extraneous variables, such as demographic changes in aviation employees" (Li 2007). We therefore assessed the study to be at 'high' risk of bias for this domain.

Was the shape of the intervention effect prespecified?

The study's point of analysis was the point of intervention, and the study authors provided a rational explanation of the shape of the intervention effect. We therefore assessed the study to be at 'low' risk of bias for this domain.

Was the intervention unlikely to affect data collection?

Federally mandated testing procedures were consistent for all testing that was conducted throughout. We therefore assessed the study to be at 'low' risk of bias for this domain.

Was knowledge of the allocated interventions adequately prevented during the study?

We do not know whether or to what extent employees were aware of the reduction in percentage of workers randomly tested. We therefore assessed the study to be at an 'unclear' risk of bias for this domain.

Were incomplete outcome data adequately addressed?

The relevant outcome for this review ‐ annual violation rates for random alcohol testing ‐ was specified by Li and colleagues in their Methods section and fully reported in their Results section. We therefore assessed the study to be at 'low' risk of bias for this domain.

Was the study free from selective outcome reporting?

All outcomes specified in the Methods section of the study were reported by Li and colleagues in their Results section. We therefore assessed the study to be at 'low' risk of bias for this domain.

Was the study free from other risks of bias?

There is no evidence of additional risks of bias. We therefore assessed the study to be at 'low' risk of bias for this domain.

Effects of interventions

See: Table 1

Li and colleagues presented data relevant to the 'rate of positive test results found by RDAT' outcome (Li 2007). They calculated the prevalence of alcohol violations, defined as a blood alcohol concentration (BAC) level greater than or equal to 0.04%, or a refusal to submit to testing. Their study reported data on 511,745 random alcohol tests; 111 were not completed due to refusal to test, and 329 had a BAC of 0.04% or more. According to the analysis presented by the authors, when the rate of alcohol testing was decreased, the prevalence of alcohol violations increased. The years 1995 to 1997 ‐ during which a minimum of 25% of the eligible workforce was randomly selected for testing ‐ saw an average prevalence of alcohol violations of 0.07%, while the years 1998 to 2002 ‐ during which a minimum of 10% of the eligible workforce was randomly selected for testing ‐ saw an average prevalence of alcohol violations of 0.11%. This difference was described as statistically significant (P < 0.001) by the study authors, based on a Chi² test.

As our single included study was of an ITS design, we extracted and re‐analysed the data to assess differences in level and slope (Table 5). Our analysis was conducted according to the procedure described in Ramsay 2003. We extracted the data (the alcohol violation rate for each year) from the graph published in the study using the Plot Digitizer program (Plot Digitizer 2015).

4. Results from re‐analysis of Li 2007.

Pre‐int level (SE)	Change level (SE)	Estimated change in level	95% Confidence interval	P value	Pre‐intervention slope (SE)	Change slope (SE)	Estimated change in slope	95% Confidence interval	P value
0.067 (0.011)	0.040 (0.013)	0.040	0.005 to 0.075	0.031	0.051 (0.019)	‐0.0057 (0.0063)	‐0.006	‐0.022 to 0.010	0.41

The general model tested was:

Y = β₀ + β₁t + β₂I(t ≥ 4) + β₃tI(t ≥ 4) + ε,ε~N(0,σ)

We performed the statistical analysis in Stata 15 for Windows (Stata 2017). We tested for equality of variances before and after the intervention, and found no significant change (F = 1.83, P = 0.383). Since there was no evidence for either auto‐correlation or heteroscedasticity, we used standard Ordinary Least Squares (OLS). The results, however, barely changed between standard OLS and OLS with Newey‐West standard errors. The change in slope was not statistically significant (estimated change in slope: ‐0.006, 95% CI ‐0.022 to 0.010; P = 0.41), and, therefore, we excluded the interaction term. The change in intercept was significant for this reduced model, using both the OLS (estimated change in level: 0.04, 95% CI 0.005 to 0.075; P = 0.031) and Newey‐West standard errors (P = 0.021).

There is therefore no evidence for a change in slope, but there is a statistically significant increase in level.

The included study did not include data relevant to fatal injuries, non‐fatal injuries, non‐injury accidents, absenteeism, or adverse events associated with RDAT.

Discussion

Summary of main results

We found one study eligible for inclusion, which used an ITS design, and no other eligible studies.

Our independent analysis of the data published in Li 2007 showed that reducing the annual minimum percentage of airline workers who were required to undergo random testing from 25% to 10% was associated with an increase in the prevalence of positive test results in the worker population.

Apart from the prevalence of positive test results, our single included study does not address any of our other primary or secondary outcomes.

Overall completeness and applicability of evidence

From the available data, the increased prevalence of alcohol violations following the reduction in the percentage of workers who underwent RAT is consistent with a deterrent effect. This is a statistically significant finding, which we deem to be of clinical relevance for the context in which the one included study was conducted. However, this finding should not be generalised outside of the aviation industry in the USA, nor should it be viewed as definitive, due to a number of issues. Firstly, as previously specified, we are unable to contextualise the data from the Li 2007 study, as no other study met our criteria for inclusion. Secondly, only one of our six review outcomes was discussed, leaving us unable to perform a more robust analysis of RDAT. Thirdly, the study dealt only with two different random alcohol testing regimes, so we cannot comment on random drug testing at all. Finally, our evidence for the only reported outcome was rated as being of 'very low' quality, due to issues related to indirectness as well as the study being of a non‐randomised design. In totality, these issues indicate the evidence we have found is not sufficiently complete to allow definitive conclusions to be drawn about the effectiveness of RDAT.

Quality of the evidence

Three main factors informed our quality evaluation of the review evidence. Firstly, we considered that the source of our evidence was a study of ITS design, which typically provides lower quality evidence than an RCT would, due to a lack of randomisation. Secondly, we evaluated the risk of bias for our included study, ultimately determining the study was at 'low' risk of bias (Characteristics of included studies). Finally, a GRADE assessment was conducted for the study outcome 'rate of positive test results found by RDAT'. Our judgements for four of the five GRADE domains (imprecision, risk of bias, inconsistency, and publication bias) were 'not serious'. The evidence was assessed as having a serious risk related to indirectness, which was explained as follows: only two random testing regimes (10% and 25%) were investigated, only one industry was investigated, only RAT was tested, and the comparison of one random testing regime against another is not as high quality as comparing a random testing regime against no random testing regime. Due to these concerns, we downgraded the study from the starting point of 'low' quality evidence (the standard baseline rating for evidence from non‐randomised studies) to 'very low'.

While the fact that the study was at low risk of bias is a positive indicator of the quality of the evidence, the 'very low' GRADE assessment and study design considerations lead us to the overall conclusion that the quality of the evidence is poor. This result further reinforces that conclusions cannot be drawn about the general effectiveness of RDAT based on our review's evidence.

Potential biases in the review process

Biases in the review process were minimised by adhering to standard Cochrane methodology. Review authors made subjective decisions when conducting the risk of bias assessment and the GRADE assessment. The decision to re‐analyse the data from Li 2007 was made after discussion with Cochrane editors.

Agreements and disagreements with other studies or reviews

Only a limited number of trials have been conducted in this field, and the scope of reviews is narrow. An earlier Cochrane Review suggested there was insufficient evidence to offer advice either for or against the use of RDAT as a sole measure for preventing injuries in commercial drivers (Cashman 2009). Our study, investigating a similar question in a work sector outside that of commercial drivers, supports the notion that there is insufficient evidence to either support or reject RDAT alone as a measure to reduce or prevent workplace accidents or injuries. Another review suggested that effectiveness of drug testing in addressing drug issues is weak (Pidd 2014), and this was echoed by a later study (Pidd 2016), which suggested that alcohol or drug testing policies are not significantly associated with the substance use behaviours of workers. However, Pidd and colleagues concluded that there was empirical support for the value and efficacy of comprehensive policies (as opposed to testing only) in reducing drug and alcohol problems (Pidd 2016). They concluded that drug or alcohol testing policies did not significantly change workers' drug consumption behaviours. Our study, in contrast, suggests a decreased alcohol use in the airline industry with increased random testing, but we did not investigate the effectiveness of comprehensive policies.

Authors' conclusions

Implications for practice.

We were able to include one study in this review, which was conducted in the aviation industry in the USA. While this study appears to support the premise that decreasing the frequency of testing increases the use of alcohol, and thus testing may be an important factor in reducing accidents, the overall paucity of data means we cannot conclude that random drug and alcohol testing (RDAT) is generally effective, nor can we conclude the opposite. We could not report on any adverse effects related to RDAT, as these were not reported. The one included study presented results that were consistent with a deterrent effect of RDAT.

Implications for research.

Significant uncertainty exists regarding the effectiveness of RDAT. Thus, there is a need for further studies to determine whether RDAT is effective for preventing workplace accidents or improving other non‐injury outcomes. Ideally, future studies should be randomised or cluster‐randomised, and should include assessments of random testing for drugs as well as alcohol. In addition, such studies should be conducted in different workplace settings, industries, and regulatory environments to produce results that may be generally applicable. The outcomes considered in this review (fatal injuries, non‐fatal injuries, non‐injury accidents, rate of positive results found by RDAT, absenteeism, adverse events associated with RDAT) would serve as useful metrics for evaluating RDAT in future investigations.

History

Protocol first published: Issue 1, 2018
Review first published: Issue 12, 2020

Notes

Parts of the methods section and Appendix 1 of this review are based on a standard template established by the Cochrane Work Editorial Group.

Appendices

Appendix 1. MEDLINE search strategy

1	exp Alcohol Drinking/
2	exp Alcoholism/
3	exp Alcoholic Intoxication/
4	exp Ethanol/
5	or/1‐4
6	exp Inhalant Abuse/
7	exp Substance Abuse, Intravenous/
8	exp Substance‐Related Disorders/
9	exp Opioid‐Related Disorders/
10	drug abuse.mp. or intoxication.mp.
11	or/6‐10
12	exp Substance Abuse Detection/
13	exp Breath Tests/
14	exp Hematologic Tests/
15	exp Urinalysis/
16	toxicology screen*.mp.
17	random test*.mp.
18	drug test*.mp.
19	exp Mandatory Testing/
20	employment test*.mp.
21	or/12‐20
22	exp Employment/
23	exp Occupations/
24	exp Workplace/
25	exp Work/
26	exp Work Performance/
27	employ*.mp.
28	work*.mp.
29	profession*.mp.
30	or/22‐29
31	exp Narcotics/
32	exp Analgesics/ or exp Analgesics, Short‐Acting/ or exp Analgesics, Opioid/
33	exp Codeine/
34	(demerol or pethidine).mp. or exp Meperidine/
35	dilaudid.mp. or exp Hydromorphone/ or exp Hydrocodone/ or Vicodin.mp. or tramadol.mp. or exp Tramadol/
36	exp Fentanyl/
37	exp Heroin Dependence/ or exp Heroin/
38	exp Methadone/
39	exp Mitragyna/ or kratom.mp.
40	exp Morphine/ or exp Morphine Derivatives/
41	opiate.mp. or exp Opiate Alkaloids/
42	exp Buprenorphine/ or exp Butorphanol/ or exp Tramadol/
43	exp Oxycodone/
44	exp Central Nervous System Depressants/
45	exp Barbiturates/
46	exp Benzodiazepines/
47	exp "Hypnotics and Sedatives"/
48	rohypnol.mp. or exp Flunitrazepam/
49	exp Sodium Oxybate/
50	tranquilizer.mp. or exp Tranquilizing Agents/
51	valium.mp. or exp Diazepam/
52	exp Central Nervous System Stimulants/
53	exp Amphetamines/
54	exp Cocaine/ or exp Crack Cocaine/
55	crystal meth.mp.
56	khat.mp. or exp Catha/
57	exp Methamphetamine/
58	exp Methylphenidate/
59	exp Hallucinogens/
60	exp Dextromethorphan/
61	exp Ketamine/
62	LSD.mp. or exp Lysergic Acid Diethylamide/
63	exp Mescaline/
64	exp N,N‐Dimethyltryptamine/ or DMT.mp.
65	exp Phencyclidine Abuse/ or exp Phencyclidine/ or PCP.mp.
66	exp Psilocybin/
67	exp Salvia/
68	exp Street Drugs/
69	(marijuana or THC).mp. or exp Cannabis/ or exp Marijuana Abuse/ or exp Medical Marijuana/ or exp Marijuana Smoking/ or exp Dronabinol/
70	exp Psychotropic Drugs/
71	exp Designer Drugs/
72	exp 3,4‐Methylenedioxyamphetamine/ or exp N‐Methyl‐3,4‐methylenedioxyamphetamine/ or ecstasy.mp. or MDMA.mp.
73	(synthetic cathinones or bath salts).mp.
74	or/31‐73
75	5 or 11 or 74
76	21 and 30 and 75
77	remove duplicates from 76

Appendix 2. PsycINFO search strategy

1	exp ALCOHOLISM/
2	exp Alcohol Intoxication/
3	exp Alcohol Rehabilitation/
4	exp ETHANOL/
5	or/1‐4
6	exp Drug Rehabilitation/
7	exp Drug Addiction/
8	substance abuse.mp. or exp Drug Abuse/
9	exp Drug Dependency/ or exp Drug Addiction/
10	intoxication.mp.
11	or/6‐10
12	exp Drug Usage Screening/
13	exp URINALYSIS/
14	exp Employment Tests/
15	or/12‐14
16	exp DRUGS/
17	exp HALLUCINOGENIC DRUGS/ or exp DESIGNER DRUGS/ or exp ANALGESIC DRUGS/ or exp NARCOTIC DRUGS/ or exp NONPRESCRIPTION DRUGS/
18	exp MORPHINE/
19	exp CODEINE/
20	exp MEPERIDINE/
21	exp TRAMADOL/
22	(hydromorphone or demerol or dilaudid or hydrocodone or vicodin).mp.
23	exp FENTANYL/
24	exp HEROIN ADDICTION/ or exp HEROIN/
25	exp METHADONE/
26	(kratom or mitragyna).mp.
27	exp OPIATES/
28	exp BUPRENORPHINE/
29	butorphanol.mp.
30	exp TRAMADOL/
31	oxycodone.mp.
32	exp CNS Depressant Drugs/
33	exp BARBITURATES/
34	exp BENZODIAZEPINES/
35	exp HYPNOTIC DRUGS/
36	exp SEDATIVES/
37	rohypnol.mp. or exp Flunitrazepam/
38	exp Gamma Hydroxybutyrate/
39	exp Tranquilizing Drugs/ or exp MINOR TRANQUILIZERS/
40	valium.mp. or exp Diazepam/
41	exp CNS Stimulating Drugs/
42	exp AMPHETAMINE/
43	exp CRACK COCAINE/ or exp COCAINE/
44	exp Methamphetamine/ or crystal meth.mp.
45	(catha or khat).mp.
46	exp METHYLPHENIDATE/
47	exp N‐Methyl‐D‐Aspartate/ or dextromethorphan.mp.
48	exp KETAMINE/
49	exp Lysergic Acid Diethylamide/ or LSD.mp.
50	exp MESCALINE/
51	DMT.mp.
52	(N,N‐Dimethyltryptamine or DMT).mp.
53	exp PHENCYCLIDINE/
54	PCP.mp.
55	exp PSILOCYBIN/
56	salvia.mp.
57	street drugs.mp.
58	exp Cannabinoids/ or exp Cannabis/ or exp Tetrahydrocannabinol/ or dronabinol.mp. or exp Marijuana/
59	exp Designer Drugs/
60	mdma.mp. or exp Methylenedioxymethamphetamine/
61	or/16‐60
62	5 or 11 or 61
63	62 and 15

Appendix 3. Embase search strategy

1	exp drinking behavior/ or exp alcohol consumption/
2	exp Alcoholism/ or exp alcohol intoxication/
3	exp intoxication/ or exp drug intoxication/
4	exp addiction/
5	exp drug abuse/ or exp inhalant abuse/ or exp drug abuse pattern/ or exp substance abuse/
6	or/1‐5
7	exp *breath analysis/
8	exp *blood alcohol level/
9	exp *blood examination/
10	exp *urinalysis/
11	toxicology screen*.mp.
12	random test*.mp.
13	exp *drug screening/ or exp *drug testing/ or exp *testing, drug/
14	exp *mandatory testing/
15	exp *preemployment medical examination/
16	or/7‐15
17	exp permanent employment/ or exp employment/ or exp parttime employment/ or exp temporary employment/ or exp full time employment/
18	exp occupation/
19	exp workplace/
20	exp work/
21	work performance.mp. or exp job performance/
22	exp employee/
23	profession*.mp.
24	or/17‐23
25	exp narcotic agent/
26	exp analgesic agent/
27	exp codeine/
28	exp pethidine/
29	exp hydromorphone/
30	exp tramadol/
31	exp fentanyl/
32	heroin.mp. or exp diamorphine/
33	exp methadone/
34	exp Mitragyna/ or exp mitragynine/ or kratom.mp.
35	exp Morphine/
36	exp opiate/
37	exp buprenorphine/
38	exp oxycodone/
39	exp central depressant agent/
40	exp barbituric acid derivative/
41	exp benzodiazepine derivative/
42	exp hypnotic agent/
43	exp sedative agent/
44	exp flunitrazepam/
45	exp oxybate sodium/
46	exp tranquilizer/
47	exp diazepam/ or exp diazepam derivative/
48	exp central stimulant agent/
49	exp amphetamine derivative/
50	exp cocaine derivative/ or exp cocaine/
51	exp methamphetamine/ or exp amphetamine/
52	exp Catha edulis extract/
53	exp Methylphenidate/
54	exp psychedelic agent/
55	exp dextromethorphan/
56	exp ketamine/
57	exp lysergide/
58	exp mescaline/
59	exp n,n dimethyltryptamine/
60	exp phencyclidine abuse/ or exp phencyclidine/ or exp phencyclidine derivative/ or exp phencyclidine dependence/
61	exp psilocybine/
62	exp Salvia divinorum/ or exp Salvia/
63	exp street drug/
64	exp cannabis derivative/ or exp cannabis addiction/ or exp "cannabis use"/ or exp cannabis/ or exp cannabis smoking/ or exp medical cannabis/
65	exp anandamide/ or exp tetrahydrocannabinol/ or exp dronabinol/ or exp cannabinoid/ or exp tetrahydrocannabinolic acid/ or exp cannabidiol/
66	exp psychotropic agent/
67	exp designer drug/
68	exp 3,4 methylenedioxyamphetamine/ or exp 3,4 methylenedioxymethamphetamine/ or exp 3, 4 methylenedioxymethamphetamine/
69	exp midomafetamine/
70	exp 4' methylmethcathinone/ or bath salts.mp.
71	exp alcohol/
72	or/25‐71
73	6 or 72
74	16 and 24
75	73 and 74
76	remove duplicates from 75

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Li 2007.

Study characteristics
Methods	The included study measured the effect of different random alcohol testing (RAT) regimes using an interrupted time‐series design. Random tests were conducted by certified technicians using National Highway Traffic Safety Administration‐approved devices according to the USA Department of Transportation Procedures for Transportation Workplace Drug Testing Programs (Department of Transportation (US) 2019). Testing data were retroactively provided by the Federal Aviation Administration (FAA) to the study authors some time after 2002, the final year of the study's data.
Participants	The unit of measurement for data for this study was 511,745 random test results, rather than individual participants (who were not identified, and may have been randomly selected for testing more than once). These tests were conducted on employees of major airlines who performed safety‐sensitive functions (flight crew, flight attendants, flight instructors, aircraft dispatchers, maintenance personnel, aviation screeners, ground security controllers, and air‐traffic controllers).
Interventions	From 1995 to 1997, major airlines randomly tested a minimum of 25% of their employees, who performed safety‐sensitive functions, for alcohol each year. From 1998 to 2002, the minimum proportion of aviation employees required to undergo RAT was reduced to 10% per year. The study measured how this reduction in the percentage of employees required to undergo RAT affected the rate at which employees tested positive. The study did not present data on any of our other outcomes of interest.
Outcomes	Prevalence of alcohol violations was assessed annually by the FAA. From 1995 to 1997, the average prevalence of alcohol violations was 0.07%. From 1998 to 2002, the average prevalence of alcohol violations was 0.11%.
Funding source	The study was funded in part by a grant from the National Institute on Alcohol Abuse and Alcoholism and a grant from the Centers for Disease Control and Prevention.
Declarations of interest	The authors did not declare conflicts of interest within the article. We (GW) contacted the lead study author, Dr Guohua Li, who informed us that the study authors did not have any conflicts of interest (Li 2020 [pers comm]).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Was the intervention independent of other changes?	High risk	The study authors indicate it is unlikely that the intervention occurred independently of other changes: "... it is probable that a small portion of the observed increase in alcohol violations during 1998 and 2002 was due to confounding effects from extraneous variables, such as demographic changes in aviation employees" (Li 2007, p 512).
Was the shape of the intervention effect pre‐specified?	Low risk	The study's point of analysis was the point of intervention, and the study authors provided a rational explanation of the shape of the intervention effect.
Was the intervention unlikely to affect data collection?	Low risk	Federally mandated testing procedures were consistent for all testing that was conducted.
Was knowledge of the allocated interventions adequately prevented during the study?	Unclear risk	We do not know whether or to what extent employees were aware of the reduction in percentage of workers randomly tested.
Were incomplete outcome data adequately addressed?	Low risk	The relevant outcome for this review, annual violation rates for random alcohol testing, was specified in the Methods section and fully reported in the Results section.
Was the study free from selective outcome reporting?	Low risk	All outcomes specified in the Methods section of the study were reported in the Results section.
Was the study free from other risks of bias?	Low risk	There was no evidence of additional risks of bias.

The risk of bias table reflects the EPOC 2017b risk of bias criteria for interrupted time‐series studies.

Abbreviation

RAT: random alcohol testing

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Buchanan 1988	Ineligible study design
Carlton 2004	Primary research not reported
Carpenter 2007	Ineligible study design
Crouch 1989	Ineligible study design
DuPont 1995	Ineligible study design
Elliott 2006	Ineligible study design
Fitzsimons 2008	Ineligible study design
Fox 2014	Ineligible study design
French 2004	Ineligible study design
Grabowski 1989	Primary research not reported
Li 2010	Ineligible study design
Li 2011	Ineligible study design
Maretha 2016	Ineligible study design
Marques 2014	Ineligible study design
McFadden 1997	Ineligible study design
Miller 2007	Ineligible study design
Normand 1990	Ineligible study design
Ozminkowski 2003	Ineligible study design
Price 2012	Ineligible study design
Price 2014a	Ineligible study design
Price 2014b	Ineligible study design
Price 2015	Ineligible study design
Price 2016	Ineligible study design
Schofield 2011	Ineligible study design
Schofield 2013	Ineligible study design
Vignali 2013	Ineligible study design

Differences between protocol and review

GW, DS, DD, and RH joined the project after the protocol was complete. Based on recommendations from the Cochrane Work Editorial Group, we consulted Chapter 14, Section 14.1 of the Cochrane Handbook of Systematic Review of Interventions (Higgins 2020) and Ryan 2016b when creating Table 1. Also, when conducting the GRADE assessment, we consulted Ryan 2016a, Sterne 2016, and Chapter 14, Section 14.2 of the Cochrane Handbook of Systematic Review of Interventions (Higgins 2020).

Contributions of authors

All authors have contributed to drafting this review.

Conceiving and designing the review: CE, SS.

Co‐ordinating the review: MM, GW, TJ.

Designing search strategies: GW, TJ, SS.

Abstract screening and study selection: CE, DK, DS, MM, GW, DD, TJ.

Analysis of data: RH.

Writing the review: CE, SS, GW, TJ, RH.

Sources of support

Internal sources

• No sources of support supplied

External sources

• Government of Alberta, Canada

  This study was supported by the Government of Alberta OHS Futures – Research Funding Program (www.work.alberta.ca/ohsfutures).

Declarations of interest

Charl Els: none known. Charl Els is a psychiatrist, addictions specialist, and occupational physician.

Mathew T Milen: none known.

Tanya D Jackson: none known.

Diane Kunyk: none known.

Daniel Sowah: none known.

Graeme Wyatt: none known.

Reidar Hagtvedt: none known.

Danika Deibert: none known.

Sebastian Straube: none known. Sebastian Straube is a specialist in occupational medicine.

New