Submission to the Society of Energy Professionals and the Power Workers’ Union
Comment on the Canadian Nuclear Safety Commission discussion paper
Fitness for duty: Proposals for strengthening alcohol and drug policy, programs and testing
By
Scott Macdonald
August 10, 2012.
My background
Attached is my curriculum vitae, which provides a summary of various
accomplishments over my career. My educational background includes degrees in
Psychology (BSc, University of Victoria), Criminology (MA, University of Toronto) and
Epidemiology and Biostatistics (PhD, University of Western Ontario). In 1978, I started
work as a Research Assistant at the Addiction Research Foundation in Toronto. In
1985, I became a Scientist with responsibility for independent research. I remained at
the Addiction Research Foundation (which later became the Centre for Addiction and
Mental Health) until 2005, when I took on a new position as Assistant Director at the
Centre for Addictions Research of BC and Associate Professor, School of Health
Information Science, University of Victoria. In 2009, I was promoted to a Professor.
At all these positions, I conducted social epidemiological research related to
substance use and abuse. A primary focus of my research has been on the
relationships
between
substance
use
and
injuries.
These
studies
include
injuries/accidents in the workplace and crashes. I have over 20 peer-reviewed
publications on this subject and 8 peer-reviewed grants and contracts, specifically on
this topic. I have considerable expertise in drug testing programs, which includes 18
published works, a peer reviewed grant from the Social Sciences and Humanities
Research Council of Canada and contracts with the International Labour Office and
Transport Canada. As well, I have conducted program evaluation studies of workplace
programs.
Social epidemiology is the study of the causes of diseases that have a social
dimension. This discipline focuses on assessing the distribution and determinants of
health-related states or events, including injuries, in populations and recommending
approaches to alleviate these health problems (Last, 1995). A major purpose of
epidemiology is to collect and analyze data that provide an accurate assessment of
risks associated with diseases. Analytic epidemiology involves the assessment of risks
of diseases from various exposures – for example, the risk of job accidents from
substance use.
2
An emphasis of epidemiology is on methodological issues in the collection of data that
could potentially bias findings, such as:
•
Measurement error – Are the instruments/scales valid and reliable and
measure what we wish to measure?
•
Internal validity – Does the study design allow strong inferences of cause
and effect?
•
External validity – Do the results of the study apply to the “real world”?
•
Selection bias – Does the study have systematic differences in how data
was collected in the cases (e.g. injury subjects) and control (e.g. noninjury subjects) that may inflate or deflate the estimates of risks associated
with exposure (e.g. alcohol or drug use)?
•
Confounders – Are there other third variables that may better explain the
findings?
•
Threats to internal validity (i.e. the degree to which we can make strong
inferences between cause and effect). Threats to internal validity
represent various explanations of how the results of a study might not be
internally valid, based on the designs. These include potential flaws, such
as selection (differences found are due to fundamental differences in the
group),
history
(competing
interventions
caused
the
change),
experimenter bias, testing effects (both practice effects and Hawthorne
effect) and statistical regression.
Introduction
This report contains my opinions of the discussion paper document in response
to requests from Sonia Pylyshyn at the Society of Energy Professionals and Emily
Lawrence, Counsel for the Power Workers' Union. I was asked to comment on different
3
drug testing approaches - specifically their accuracy and what they show. I was also
asked to report on the scientific literature on drug use and safety risk, issues of
deterrence, and evaluation studies of drug testing (with emphasis on job accidents).
Finally, I was asked to comment on studies showing possible negative consequences of
drug testing and alternatives to drug testing.
In this review, I will first examine the accuracy, strengths and limitation of
different biochemical approaches for determining drug use (Section 1). In the following
section, research on the acute effects of different drugs will be briefly reviewed (Section
2). Next, will be an analysis of studies that address the issue of whether drug users
represent a safety risk in the workplace (Section 3). Then the question of whether drug
testing is a deterrent is addressed (Section 4). In the final sections, studies on the effect
of drug testing on safety (Section 5), potential negative consequences of testing
(Section 6) and alternative approaches (Section 7) are addressed. The report
culminates in recommendations with respect to the Nuclear Commissions discussion
document.
My main conclusions are as follows. The discussion document contains
recommendations that will not achieve the stated goal of identifying individuals who are
unfit for duty. Commonly used biochemical drug tests cannot identify whether
employees are under the influence of a drug at the time of the test and therefore cannot
be used to identify those unfit for duty. As well, drug testing has not been scientifically
shown to improve work safety. In most Canadian workplaces, drug use at work
represents an insignificant problem.
1. Strengths and limitations of different drug testing approaches.
An important issue is the accuracy of the drug tests in detecting the drugs in
question. Sensitivity and specificity are terms often used to describe the accuracy of
individual tests. Sensitivity refers to the proportion of people who test positive divided by
the number of true drug positives. Specificity refers to the proportion of people who test
negative divided by the number of true drug negatives. Potential errors that can occur
are false positives (indicating an individual has used a drug but he/she has not) or false
4
negatives (indicating an individual has NOT used a drug but he/she has). These types
of errors are typically determined using studies, where drug use of individuals is known,
and compared against results from drug testing.
Two major analytic approaches
Two main analytic approaches exist for assessing drug use: (1) immunoassay or
(2) mass spectrometry in combination with gas/liquid chromatography (MS/GC).
Immunoassay screens use antibodies to detect the presence of a drug or its metabolites
(Dyer & Wilkinson, 2008). Immunoassays lack accuracy and are unable to definitively
identify the presence or absence of a specific drug, hence their use as a screening
device only (Dyer & Wilkinson, 2008). Although false negatives are more common, false
positives can also occur where a molecule with a similar structure to the drug of interest
is present (Dyer & Wilkinson, 2008). Most immunoassay tests are conducted in
laboratory conditions and therefore results are not immediate.
An exception is point of collection tests, an immunoassay approach that can
provide findings of drug use (typically from saliva) within 10 minutes. There is
approximately 12 point of collection tests on the market but none provide a high degree
of accuracy (Dyer & Wilkinson, 2008). In a recent study, Blencowe et al. (2011)
compared the accuracy of the point of collection tests with laboratory methods using
MS/GC. While accuracy varied across devices, none of the tests had perfect sensitivity
for any drug. The average sensitivity for amphetamines was 60% and specificity was
97%. Accuracy was particularly poor for cannabis at 73%, with average sensitivity of
38% and specificity of 95%. For cocaine, these tests had excellent specificity of close to
100%, but average sensitivity across all tests was only 36%.
The gold standard of drug tests is MS/GC - considered near perfect for drug
detection and identification (Jaffee et al, 2008) and can be conducted with any type of
biological sample. This approach requires sophisticated laboratory equipment.
Immunoassay screens are often used as a screening device with confirmation of
positive tests with MS/GC. However, the immunoassay approach with confirmation with
MS/GC still likely results in some false negatives, as confirmation tests will only be
5
conducted on samples that initially test positive with immunoassay screens (known for
low sensitivity).
Since the accurate testing procedures require laboratory conditions, chain of
custody procedures are required to ensure that biological samples are not adulterated
or misplaced. Proper quality control involves documentation of specimen collection to
the final report (Normand et al., 1995; p.184-190).
Positive drug tests using MS/GC still can result from licit sources. For example, it
is possible that a positive test for morphine could be due to certain poppy seed cakes
(Cone and Huestis, 2007). A positive test for cannabis could be due to pharmaceutical
THC (Sativex) (Cone and Huestis, 2007) or over the counter hemp oil products (U.S
Department of Transportation, 2004). Both drug testing methods have limitations as in
most instances (described under biological specimens) cannot be used to determine
whether individuals are under the influence of the drug or have a drug problem.
Different biological samples
Different types of biological samples can be used to test for alcohol and drugs.
For the detection of drugs, hair, urine, saliva or blood can be used. Urine testing is the
most common approach used in Canadian workplaces, and more recently saliva testing
is being used. Hair testing is used by some employers the United States to detect prior
drug over a long time frame (months) but since it is not common in Canadian
workplaces, it will not be reviewed here. For alcohol detection, breathalyzer tests are
most common. In this section, first the general strengths and limitations of urine and
saliva testing will be reviewed. Since the Commission’s recommendations describe
fitness for duty as impairment due to the presence of alcohol and/or drugs, in the next
section, comparisons are drawn between normal periods where people are under the
influence of different drugs and the detection windows for these drugs with urine or
saliva tests. Finally, the strengths and limitations of breath and blood tests are reviewed.
Urine tests detect inert metabolites associated with ingestion of each drug being
tested and not the psychoactive ingredients of each drug. These metabolites are
6
eliminated at different rates, depending on the drug and usage (Moeller, Lee, & Kissack,
2008). For each drug, the detection periods for urine far exceed the periods for which
individuals could be considered under the influence and in every case, exceed the
detection times for saliva tests (with the exception of PCP where no studies were
found). Impairment, dependence, or frequency of use cannot be determined by urine
tests (Kapur, 1994). The most common procedure in the workplace is to first conduct a
immunoassay screen, and positive tests are verified with MS/GC procedures. Urine
tests are invasive because visual monitoring of urination is needed to prevent
adulteration of the samples.
Saliva tests for drugs detect the psychoactive components (i.e. the parent drug),
with the exception of cocaine where both the parent drug and inert metabolite can be
detected (Drummer, 2006). Although saliva tests detect the psychoactive components
of the drug, conclusions of impairment cannot be made based on test results
(Kadehjian, 2005; Dyer and Wilkinson, 2008). A positive saliva test can only be used to
determine prior use of drugs within certain detection periods that vary, depending on the
drug. The detection time (i.e. the time from last use to the test) for all drugs will vary
among individuals regardless of the testing technology, depending on the dose taken,
frequency of use, route of administration, duration of use, tolerance, metabolic rate and
cut-off levels of the tests (Verstraete, 2004; Dyer and Wilkinson, 2008). Saliva tests can
also be used to detect alcohol levels within two hours after consumption, which closely
correlate to those in blood (Center for Substance Abuse Treatment, 2006). However,
since the oral cavity is easily contaminated; the time window for detecting ethanol in
saliva is short (Center for Substance Abuse Treatment, 2006) and saliva tests are rarely
used for alcohol.
Comparisons of time periods for being under the influence of various
drugs and the detection windows for urine and saliva tests.
Table 1, column 1, shows the time periods in which a person consuming the drug
under typical conditions would normally be considered under the influence of each drug.
For smoking cannabis, the maximum period of being under the influence is estimated to
be about 4 hours (Grotenhermen, 2003; Ramaekers et al., 2004). Typically, a peak high
7
is experienced within 30 minutes which lasts approximately 2 hours Grotenhermen,
2003). After eating the drug, a peak high typically is noted between 1 to 3 hours which
declines to low levels within 6 hours (Grotenhermen, 2003). For snorting cocaine, a high
will typically last 15 to 30 minutes and for oral ingestion about 1 to 2 hours (U.S.
Department of Transportation, 2004). For amphetamines, typical effects last 4 to 8
hours, with residual effects possible up to 12 hours (U.S. Department of Transportation,
2004). Peak effects for morphine occur under 60 minutes, and residual effects up to 6
hours; whereas for heroin, peak effects occur under 2 hours and overall effects
dissipate after 5 hours (U.S. Department of Transportation, 2004).
Column 2, Table 1, shows the detection periods after last use for saliva tests.
Detection periods with saliva tests vary, depending on the reference source and other
variables, such as the cut-off values and the analyte being targeted. Regardless of
reported differences in maximum detection times among authors, the detection period
with saliva tests is longer than the period that individuals can be considered under the
influence for every drug. For cannabis, the detection period from last use is about 34
hours (Verstraete, 2004). Dyer and Wilkinson (2008) differentiate between a single use
(4 to 14 hours) and chronic use (4 to 30 hours); however, test results cannot be used to
differentiate occasional from heavy users.
For amphetamines, the detection time
indicated noted by Verstraete is 20 to 50 hours, whereas for Dyer and Wilkinson it is
less than 24 hours.
Also, differences exist for cocaine, depending on whether the
parent drug (5 to 12 hours) or metabolite (12 to 24 hours for benzoylecgonine) is being
detected (Verstraete, 2004). Research on detection times for PCP is not available
(Cone and Huestis, 2007). The tests are a poor instrument for assessing workplace
performance. Column 3 of Table 1 shows the detection periods for urine tests. For every
drug, the detection periods for urine exceed those of saliva tests.
Breath tests (i.e. breathalyzers) can only be used to detect alcohol. They
measure the concentration of ethanol in the breath, which is highly correlated to the
level of ethanol in the blood (Drummond & Ghodse, 1999). The advantages of using
breathalyzers are that they provide rapid results, are non-invasive, and provide accurate
information on impairment by alcohol (Drummond & Ghodse, 1999). Sources of error in
8
breathalyzer use include failure to follow manufacturer’s instructions for regular
calibration and ensuring that alcohol has not been consumed within the last 15 minutes
(including alcohol-containing mouthwashes), smoking cigarettes or exposure to certain
types of paints (Drummond & Ghodse, 1999). Aside from these potential limitations,
false positives are unlikely (Hannuksela et. al., 2007).
Blood tests are another approach that can be used for the detection of drugs;
however, they are invasive. There currently is no accepted relationship between drug
concentrations in the blood (alcohol excepted) and degree of impairment. Preliminary
consensus cut-off levels in the blood for cannabis impairment have been prepared by a
consensus group (Grotenhermen et al, 2007), but there are no internationally accepted
standards. For other drugs, no accepted standards for impairment levels have been
proposed.
2. The acute effects of drugs
Experimental laboratory studies have been conducted to understand how the
level of drugs in the human body is related to performance. The acute effects of drugs
(i.e. observed when the psychoactive ingredients are present) are best assessed in
experimental studies by measuring the performance of subjects randomly assigned to
drug and no drug conditions. A clear conclusion drawn from a multitude of laboratory
studies is that the pharmacological properties and effects on psychomotor performance
for the drugs vary considerably. Research shows the acute effects of some drugs (i.e.
cannabis, PCP and opiates) can negatively affect performance, such as the ability to
operate equipment (McKim, 1986). Some improvement in cognitive skills, such as visual
spatial and attention has been noted with methamphetamines (Hart et al., 2012). An
important limitation of experimental studies is that the results may not be applicable to
real world conditions; that is, the tasks used in a laboratory setting may not reflect
activities conducted in the workplace. For example, most laboratory studies measure
the effects of a drug on peak performance as opposed to typical performance. In some
studies participants are asked to drive around an obstacle course as quickly as possible
and deficits have been observed. In real driving situations these deficits might not be
severe enough to increase collision risk because people do not need to drive as fast as
9
possible. Alternatively, those who use drugs may be unlikely to use during activities
such as driving or while at work.
3. Do drug users represent a safety risk in the workplace?
Given that drug tests are only useful for detecting drug users (as opposed to
detecting impairment), one question posed by many is whether drug users are more
likely to have accidents and injuries than non-users. Two main types of epidemiological
studies have been conducted to answer this question: (1) those comparing accidents or
collisions of people who test positive for drugs versus those who test negative; and (2)
those comparing accident rates of people who self-report drug use versus those who do
not. These studies have been conducted on two population groups, employees and
drivers. Conclusions from these studies are summarized below.
Studies of employees
Some epidemiological studies were found that compared on the job accident
rates of employees who tested positive versus negative for drug use. Only one study
was found where those testing positive had significantly higher accident or injury rates
than negatives (Zwerling et al., 1990). In this study, U.S. postal workers who tested
positive for marijuana were significantly more likely to have reportable accidents and
work injuries. Those who tested positive for cocaine had significantly more injuries, but
no statistical difference was found for job accidents. After two years of follow-up, risks of
adverse outcomes declined among drug positives (Ryan et al., 1992). The authors
acknowledge a limitation of their study is that alcohol use associated with drug use was
not investigated. In addition, the results may not be replicable in other countries or for
other types of workers.
Other studies failed to show a relationship between positive drug tests and
accidents or injuries. In a longitudinal study of 5,465 job applicants, Normand et al
(1990) found no significant differences between drug positives and negatives in terms of
job injuries. Similarly, Crouch et al (1989) found that drug positives did not have higher
rates of job accidents than drug negatives. Overall, there is not enough research
10
evidence to conclude that those who test positive for drugs represent an increased risk
for work injuries or accidents, which is a fundamental requirement of causation. Some
studies have compared job injuries of drug users and non-users as determined through
self-reports. A U.S. study of 1,740 employed adults indicated that drug users were 1.7
times more likely to be involved in injuries on the job (Hingson et al., 1985). More
recently, Shipp et al. (2005) found the odds of workplace injury increased with the
frequency of drug use for marijuana and cocaine. By contrast Hoffman and Larison
(1999) did not find a relationship between occupational injuries and use of either
marijuana or cocaine. In a survey of Ontario adults using similar methods, those who
used either illicit or licit drugs in the prior year were significantly more likely to have job
injuries than non-users (Macdonald, 1995). However, additional analyses controlling for
potential confounding variables, pointed away from illicit drug use as a causal agent in
job injuries.
The existence of a statistical relationship between drug use and job injuries
should not be interpreted that drug use causes job accidents/injuries. For example, illicit
drug use is related to age and sleep problems, which in turn are significantly related to
job injuries (Macdonald et al., 2009). Other researchers have suggested that risk taking
behaviour is a plausible explanation for the relationship between drug use and
accidents; we do not know whether drug users might have elevated rates of job
accidents because they use drugs or because they tend to be risk takers, a
characteristic also related to job accidents (Newcomb, 1994; Spicer, Miller & Smith,
2003).
In summary, some studies found drug users have a higher risk of job accidents,
while other studies failed to find significant relationships. All studies failed to show that
drug use was a cause of workplace accidents, as statistical relationships between drug
users and accidents/injuries might be explained by other variables. One study
suggested the relationship between drug use and job injuries could be explained by
other factors correlated with drug use, such as sleep problems, smoking, shift work, and
age (Macdonald, 1994). In most studies, several drugs with different pharmacological
properties were grouped into one category (presumably to increase statistical power);
11
consequently, very little is known about specific drugs (see Hingson et al., 1985,
Normand et al., 1990; Crouch et al., 1989). The methods and operational definitions of
drug use used varied considerably, making it difficult to meaningfully compare the
results of the studies.
To assess causality, the strength and significance of a statistical relationship
between positive tests and job accidents/injuries should first be determined. This is
typically accomplished by calculating odds ratios (OR) where the proportion of those
who test positive in the accident/injury group is compared to the proportion in a noninjury control group. Factors such as sex, age or risk taking propensity (called
confounders), should be ruled out as possible explanations for the results found.
Confounders are typically addressed through matching of cases and controls in the
design of a study, or through stratification or multivariate statistics in the analysesprocedures rarely used in the drug testing and job accident research.
Studies of drivers
Such studies can be found in the traffic safety literature, and are reviewed below.
These studies have failed to find a relationship between positive tests results and
collisions. The best methodological studies using oral fluid or urinalysis techniques to
test drivers are analytic epidemiological studies that utilize either the case-control
method (Ferrara et al., 1994; Marquet et al., 1998) or methods used to ascribe crash
responsibility (Longo et al., 2000ab). Statistical significance was found in only one study
with one drug - cannabis (Marquet et al., 1998).
A study by Longo et al (2000ab) used drug tests of 2,500 injured drivers. Blood
samples were analyzed to determine the concentration of each drug present and
culpability for the crashes was assessed. The authors did not find that drivers who
tested positive for cannabis were any more likely to be culpable for their crash than
drug-free drivers. Many of the studies cited above also investigated whether the
presence of cocaine metabolites was significantly related to the likelihood of crashes;
but none found a statistical relationship. Thus epidemiological studies have failed to
demonstrate a statistical relationship between drug positives and driver related
12
collisions, a fundamental requirement of causation. One reason for the failure to find
significance could be the inaccuracy of the tests, which cannot detect impairment by
drugs.
In stark contrast to these studies, analytic epidemiological studies on alcohol
impairment and collisions are persuasive that alcohol is a cause of collisions (see
Macdonald, 1997 for a summary of this evidence).
4. Is drug testing a deterrent for drug use and drug related job accidents?
Quest diagnostics is the largest provider of drug testing in the United States and
conducted about 6.4 million tests in 2011 (Quest Diagnostics Incorporated, 2012). The
proportion of employees who test positive has dropped considerably from 13.6% in
1988 to 3.5% in 2011. This decline might be interpreted as an indication of a deterrent
effect of drug use among employees in companies where testing is conducted.
However, other explanations are possible. As noted by Walsh (2008), there has been
an exponential growth of alteration in substitution products developed to produce
negative results, which might help to explain the decline. The most experienced users
may be the most likely to use such methods. Although overall research indicates that
companies that drug test generally produce a lower proportion of positive tests over
time, this does not mean that fewer people are impaired by drugs at work. If some
employees change their drug using behaviours, the intervention may have a greater
effect on recreational users, who are unlikely to use at work. Testing does not appear to
have had any impact on overall drug use in the US population. US surveys indicate a
30% increase in drug use between 1988 and 2004 (Walsh, 2008). Various explanations
are possible for this discrepancy. First, employees who use drugs may choose to work
in companies that do not test for drug usage. The lower rate of positive tests for preemployment testing (3.5%) is suggestive that companies that drug test may deter
prospective employees from applying to vacant openings.
Research has not shown that the decline in positive rates corresponds to safer
workplaces. One approach that could be used to assess the likely impact on workplace
safety is to compare the percentage of employees who test positive from random testing
13
to those who test positive from post-accident testing. If drug use is a major cause of
workplace accidents then one would expect that the percentage of employees testing
positive after accidents would be much higher than those from random tests. From 2009
to 2011, 5.3% of the US employees tested positive with post-accident testing (Quest
Diagnostics Incorporated, 2012). This compares with random testing at 5.4% in 2009,
5.3% in 2010, and 5.2% in 2011. Given that the percentages are nearly identical
suggests that the causal impact of drugs on workplace accidents is negligible. The
aforementioned evidence does not indicate that reductions in positive test results over
the past decade have translated into fewer job accidents.
Although research has shown that per se laws for alcohol (i.e. laws prohibiting
driving at certain BAC levels) and strong enforcement have been beneficial in reducing
collisions caused by alcohol (Mann et al., 2001), it is not a good analogy to suggest that
drug testing will achieve the same result among employees as there are many
differences between the two. First, there is a substantial difference between the
prevalence of use for alcohol and illicit drugs in our society. For example the most
recent Canadian survey shows that 78% of Canadians used alcohol in the past 12
months, which compares to 9.1% for cannabis, .9% for cocaine and the numbers were
too small to report for methamphetamine (Health Canada, 2011). Second, these
aforementioned percentages represent overall usage and not use at work. Alcohol and
drug use in the workplace is much less likely than during leisure hours (Frone, 2006).
Finally, the epidemiological research shows a strong link between BAC levels and
collision risk. Virtually every large case-control study conducted has shown a strong
relation between blood alcohol content (BAC) levels and the likelihood of collision. For
example, compared to drivers with BAC = 0, the adjusted relative risk of collisions was
1.4 for drivers with a BAC= 0.05%, 2.7 for drivers with BAC = 0.08%, 22.1 for BAC =
0.15%, and 81.8 for BAC = 0.20% (Bloomberg et al. 2009). By contrast, comparable
road research using urinalysis has failed to find any relationship at all (see review of
driver studies above).
14
5. Does drug testing improve workplace safety?
Many studies of the impact of drug testing on job accidents have methodological
limitations (Macdonald, 1997; Kraus, 2001). Kraus (2001) noted that “despite the
extensive use of and management support for worksite based drug testing, the
published evidence for the effects, such as reduced injury or accident rates lacks
scientific detail. Better studies and careful reassessment of the issue appear
warranted.”
Numerous studies have a methodological limitation in that other safety initiatives
were implemented at the same time as a drug testing program. Therefore, the specific
impacts of drug testing on injury/job accident rates cannot be separated from these
other initiatives. Several studies are illustrative of this flaw. Taggert (1989) reported
substantial reductions in job accidents subsequent to drug testing being instituted;
however, this study has been severely criticized because major safety improvements
occurred at the same time that testing was implemented, making any conclusions
unreliable (Jones, 1990). In another study, Ozminkowski et al. (2003) concluded that
there was a significant relationship between testing and decreased injury rates, but they
did not control for other safety initiatives. Similarly, another study failed to examine the
impact of other safety programs and additionally had an extremely low response rate of
17% (Gerber and Yacoubian, 2001). Finally, a study of Peer Care training and random
drug testing found reduced injuries (Miller et al., 2007) but the specific effects of drug
testing could not be separated from other simultaneous interventions. Additionally,
industries that had not introduced drug testing had similar reductions. Jacobson (2003)
found a significant reduction in truck fatalities in 13 U.S. states that mandated drug
testing compared to other states but there are other plausible explanations for this
finding other than drug testing. For example, the states with workplace drug testing
were twice as likely to have Employee Assistance Programs compared to non-testing
mandated states.
In a study by Wickizer et al., (2004) occupational injury rates of companies
enrolled in a drug free work program were compared to 20,500 companies without
15
these programs. The drug free programs included formal written substance use policies,
drug testing and Employee Assistance programs. Although the authors indicate the drug
free program was associated with selective reductions in work injuries, we cannot
conclude that the reduction were due to drug testing because the programs included
other interventions. Furthermore, companies that enrolled in the program had
substantially higher injury rates (21.18 per 100 person years after intervention) than the
non-intervention group (13.82 per 100 person years).
In another study by Morantz and Mas (2008), occupational injury claims of
Divisions of a large retail chain that implemented post-accident drug testing is compared
to Divisions without testing before and after program implementation. Although declines
were noted for some types of compensation claims in Divisions with testing, the rate of
claims was still higher in the testing. Divisions (than non-testing) for 3 of 4
compensation indicators. For example, a rate of 9.63/100,000 hours first aid reports was
found in the after period among drug testing Divisions compared with 5.54/100,000 hrs
in the non-testing Divisions (see Table 1, p.264). Such a finding raises the issue of
regression to the mean, where sites with higher claims rates were chosen for
implementation of post-accident testing. Reductions to the level of non-testing Divisions
may have occurred without implementation of the testing programs.
Furthermore, other studies using a cross-sectional design have been used;
however, conclusions drawn from such designs should be considered tentative, due to
the existence of possible alternative explanations for the findings. For example, in a
study by Carpenter (2007), data from two U.S. national household surveys of 57,397
people was analyzed to assess whether drug testing was related to lower rates of
marijuana use. Companies with drug testing programs, employee assistance programs,
and written policies on substance use had lower prevalence of drug use among
employees than those without these policies. Separation of the unique effects of drug
testing from other programs was not possible.
Feinauer & Havlovic (1993) found no differences in workplace accidents and
injuries at companies with drug testing compared to 36 non-testing businesses. They
16
report significant reductions in accident and illness rates for post-accident testing but no
change for reasonable cause testing. A limitation of this study was that data was
obtained through a self-administered questionnaire raising issues of both accuracy of
the data and possible selection bias (companies with lower incident rates may be more
likely to respond).
A common practice of employers is to dismiss employees who test positive for
drugs. Interestingly, no evaluation study has been located in the literature on the
performance of such employees prior to their dismissal compared to matched control
subjects.
Little research has been conducted that compares the different situations that
employees could be tested. As pointed out in the report by the Canadian Nuclear Safety
Commission, the courts have found that random alcohol testing is acceptable but
random drug testing is not acceptable. No persuasive evidence is provided in the
document to indicate potential benefits in terms of workplace safety. In light of the
material reviewed in this document, the preponderance of the evidence does not
indicate that measurable improvements in safety will be achieved with random testing.
One conclusion drawn from existing research is that there is no credible evidence
that drug testing programs reduce job accidents. Although some studies have noted
reductions of work accidents/injuries following the implementation of drug testing
programs, others have not. Methodological flaws are common – in particular, history
(competing safety initiative at the same time as the drug testing program may have
caused the change), and statistical regression (companies tend to implement drug
testing when work accidents/injuries are particularly high) are common.
6. Potential negative impacts of drug testing
Few studies exist that investigate the potential negative consequences on
employee behaviour or attitudes related to drug testing. Employee drug testing has
been theorized to be an instrument of social control, a technique for defining and
responding to the use of illegal drugs (Brunet, 2002). Conflict theorists see drug testing
is an issue of power rather than one that relates directly to drug use in the workplace
17
(Gerber et al., 1990). Most of the social control literature on drug testing is theoretical,
and empirical studies that substantiate these theories could not be found. In
experimental studies, subjects had a more positive attitude and intentions toward
companies that did not have drug testing programs than ones that did (Crant and
Bateman, 1990). A possible impact of testing, given this study is a decline of the total
number of job applicants (Macdonald and Wells, 1994). Others have argued that testing
can produce declines in employee morale or undermine labour-management relations
(see Macdonald and Wells, 1994). Overall, too little research has been conducted to
provide conclusions regarding negative impacts of drug testing on employees.
7. Alternative approaches for addressing workplace substance use
An approach implemented in Canada for impaired driving by drugs is the Drug
Recognition Expert (DRE) Program, also known as the Drug Evaluation Classification
(DEC) Program. This program has evolved since initially being developed in the late
1970's by the Los Angeles Police Department, (Beirness et al., 2007; Porath-Waller, et
al., 2009). A standardized procedure has been developed to identify both individuals
under the influence of drugs, and the type of drug causing the observable impairment
(Beirness et al., 2007; Porath-Waller, et al., 2009). The process involves a series of
physical and psychomotor tests, and concludes with toxicological testing of a sample of
a blood, urine or oral fluid (Beirness et al., 2007; Porath-Waller, et al., 2009). The
procedures have been applied to other areas where identification of the drug-impaired
individual is desirable (Porath-Waller, 2009).
Beirness et al (2007) conducted a review of existing studies evaluating the DRE
program. Laboratory studies of subjects administered impairment doses of different
drugs produced
ranges (depending on the study) in terms of sensitivity
of DRE
assessments: (cannabis – 30.4 to 53.1%, cocaine – 13.2% (one study), amphetamines
– 4.2 to 10%, opiates – 65.4 to 75.9%, PCP – 75.3%) and specificity (cannabis – 59.1 to
86.4%, cocaine – 61.1%, amphetamines – 79.2 to 91%, opiates – 90.3 to 97.9%, PCP –
89.5%)
(see Beirness et al., 2007). After initial behvaioural assessments of drug
impairments, drug tests can be conducted. The DRE approach is advantageous over
18
drug testing for assessing fitness of duty in that behavioural symptoms consistent with
impairment are used rather than prior drug use.
In the early 1990’s the Ontario Law Reform Commission (1992) recommended
behavioural performance testing, such as computerized tests, to assess psychomotor
skills in safety sensitive positions, rather than drug testing. In relation to work, a possible
advantage of these tests is they can pick up a wider variety of performance deficits
beyond those that are drug related, such as those caused by fatigue, and thus could
potentially be more directly linked to overall fitness for work. At that time, a review of
computerized behavioural testing approaches was conducted by Butler and Tranter
(1994). Behavioral performance tests have been developed to assess various
dimensions of human functioning, such as psycho motor skills, memory, divided
attention, decision-making and critical tracking, which are often used to assess drug
related deficits in laboratory studies. The importance of each dimension for the
workplace will depend largely on task requirements and the specific job. The degree to
which the test results accurately relate to varied complex requirements of different jobs
is not adequately known, with the possible exception of some specialized positions (e.g.
airline pilot). Additionally, practice effects with repeated administrations tend to create
improvements of scores so the cut-off levels change over time. Another issue is that
some individuals do better than others on these tests, necessitating the development of
different baseline measures for each person. These variations in standards across and
within individuals are inconsistent with the principle of specific requirements for each
job.
Employee assistance programs (EAPs) are another approach used by a large
proportion of Canadian employers to improve employee health and well-being, and to
reduce performance problems (Macdonald et al., 2006). A few reviews have been
conducted on the effectiveness of EAPs in achieving these objectives (McLeod, 2010;
Macdonald et al., 1998, Normand et al., 1994) and all the reviews arrived at similar
conclusions. Although definitive conclusions cannot be drawn based on the
methodology of the studies in the heterogeneity of EAPs themselves, the
19
preponderance of the evidence suggests that these programs are beneficial in
alleviating psychological problems and can have a positive effect on work performance.
8. Overall assessment of the policy proposal
In the discussion paper by the Canadian Nuclear Safety Commission, fitness for
duty is described as the capacity of employees to effectively perform job duties without
risk to their own or others' health and safety. Several components of fitness are
described as medical, psychological, occupational fitness and behavioural-performance.
With respect to alcohol or drugs, nuclear power plant licensees would be required to
take steps to prevent workers from bringing or consuming alcohol or illicit drugs at work,
or working while under the influence of alcohol or any drug that impairs their ability to
perform his or her duties safely. These goals are reasonable as the acute (i.e.
immediate effects) of alcohol and many psycho-active drugs cause decreases in job
performance. They further suggest that biochemical substance testing should be used
to determine impairment due to the presence of alcohol and/or drugs (Section 3.0).
Unfortunately, drug tests with saliva or urine cannot accurately determine whether
employees are under the influence during work hours and therefore do not closely
measure fitness for work. Furthermore, the drug tests cannot be used to assess
whether employees have a substance use problem or dependence. The discussion
paper provides no justification of random drug testing and given the research evidence
outlined in this report, there is not a strong evidence base to support this
recommendation. The exception is the breathalyzer for alcohol, which is an accurate
device to assess whether employees are under the influence of alcohol at the time of
the test and could be used randomly, if there is a legal justification for doing so. Saliva
and urine tests, and the use of random drug testing in particular, will not achieve the
goals of assessing fitness for duty in relation to substance use as defined by the
Commission. These components do not have a sufficient evidence base to justify their
use.
20
Table 1: Typical periods subjects considered under the influence and maximum
detection period (or ranges) for saliva and urine tests
Drug
Under the
Influence* (depends
on dose and route
of admin.)
Amphetamine < 12 hours
Maximum detection times (ranges)
Saliva
Amphetamines <24 to < 50 hours
Urine
Amphetamines < 34 hrs to < 9 days
Methamphetamines – Methamphetamines –
<24 to <48hr
< 6 days
MDMA - < 24hr
MDMA - < 4 days
Cannabis
< 4 hours after
smoking
THC - <30 to <34hr
<3 days to <5 weeks
(depends on frequency
of use)
Cocaine
< 2 hours
< 3 to < 22 days
Opiates
< 6 hours
Cocaine - <8 to <24hr;
Benzoylecgonine(metabolite) <24hr
Morphine - < 24hr;
Heroin - <8 to <24;
Codeine - <36hr
PCP
< 6 hours (smoked, Unknown (Cone &
Phencyclidine snorted or orally
Huestis, 2007)
ingested)
18 hours (injected)
21
Morphine - < 48hr to 11
days;
Heroin - < 54hr
<10 days
References
Berghaus, G., & Guo, B.L. (1995). Medicines and driver fitness - findings from a metaanalysis of experimental studies as basic information to patients, physicians and
experts. In: Kloeden, C.N., McLean, A.J. (Eds.), Alcohol, Drugs, and Traffic
Safety - T’95: proceedings of the 13th International Conference on Alcohol, Drugs
and Traffic Safety, Adelaide, pp. 295-300.
Bierness, D.J., LeCavalier, J. & Singhal, D. (2007). Evaluation of the drug evaluation
and classification program: a critical review of the evidence. Traffic Injury
Prevention, 8, 368-376.
Blencowe, T., Pehrsson, A., Lillsunde, P., Vimpari, K., Houwing, S., Smink, B., Mathijssen,
R., Van der Linden, T., Legrand, S., Pil, K., & Verstraete, A. (2011). An analytical
evaluation of eight on-site oral fluid drug screening devices using laboratory
confirmation results from oral fluid. Forensic Science International, 208, 173-179.
Blomberg RD, Peck RC, Moskowitz H, Burns M, Fiorentino D. The Long Beach/Fort
Lauderdale relative risk study. Journal of Safety Research. 2009;40(4):285-92.
Butler,B. & Tranter, D. (1994) Behavioral tests to assess performance. In: Macdonald,
S. and Roman, P. (Eds). Research Advances in Alcohol and Drug Problems.
New York: Plenum Press.
Brunet, JR (2002) Employee drug testing as social control. Review of Public Personnel
Administration. 22(3): 193-215. A
Carpenter C. S. Workplace drug testing and worker drug use. Health Services Research
2007; 42: 795–810.
Clarke, J., & Wilson, J.F. (2005). Proficiency testing (external quality assessment) of
drug detection in oral fluid. Forensic Science International, 150, 161-164.
Coambs, R.B., & McAndrews, M.P. (1994). The effects of psychoactive substances on
workplace performance. In: Macdonald, S & Roman, P. (Eds.), Drug testing in
the workplace. New York: Plenum Press.
Compton, R.P., Blomberg, R.D., Moskowitz, H., Burns, M., Peck, R.C. & Fiorentino, D.
(2002). Crash risk of alcohol impaired driving. Proceedings of the 16th
International Conference on Alcohol, Drugs and Traffic Safety (CD-ROM).
Montreal, Canada.
Cone, E., Johnson, R.E., Darwin, W.D., Yousefnejad, D., Mell, L.D., & Paul, B.D.
(1987). Passive inhalation of marijuana smoke: urinalysis and room air level of
delta-9 tetrahydrocannabinol. Journal of Analytical Toxicology, 11: 89-96.
Cone, E.J. (2001). Legal, workplace, and treatment drug testing with alternate biological
matrices on a global scale. Forensic Science International, 121, 7-15.
22
Cone, E., Huestis, M. (2007) Interpretation of oral fluid tests for drugs of abuse. Annals
of the New York Academy of Science, 1098, 51-103.
Crant, MJ & Bateman, TS (1990) An experimental test of the impact of drug testing
programs on potential job applicants’ attitudes and intentions. Journal of Applied
Psychology 75(2) : 127-131.
Crouch, D., Vebb, D., Peterson, L., Buller, P., & Rollins, D. (1989). A critical evaluation
of the Utah Power and Light Company's substance abuse management program:
absenteeism accidents, and costs. In: Gust SW, Walsh JM, eds. Drugs in the
workplace: research and evaluation data, NIDA Research Monograph, No. 91.
Washington, DC: US Government Printing Office.
Drummer, O.H. (2006). Drug testing in oral fluid. The Clinical Biochemist Reviews, 27,
147-159.
Dyer, K.R., & Wilkinson, C. (2008). The detection of illicit drugs in oral fluid: another
potential strategy to reduce illicit drug-related harm. Drug and Alcohol Review,
27, 99-107.
Feinauer D. M., Haviovic S. J. (1993) Drug testing as a strategy to reduce occupational
accidents: a longitudinal analysis.J Safety Res 24: 1–7.
Ferrara, S.D., Giorgetti, R., & Zancaner, S. (1994). Psychoactive substances and
driving: State of the art and methodology. Alcohol, Drugs and Driving, 10, 1-55.
Frone, MR (2006) Prevalence and distribution of illicit drug use in the workforce and in the
workplace: Findings and implications from a US national survey. Journal of
Applied Psychology 91(4) , 856-869.
Gerber J., Yacoubian G. (2001) Evaluation of drug testing in the workplace: study of the
construction industry. Journal of Construction Engineering Management 127:
438–44.
Gerber, J., Jensen, EL, Schreck, M Babcock, GM (1990) drug testing and social control:
Implications for state theory. Contemporary Crisis, 14: 243-258.
Grotenhermen, F. (2003) Pharmacokinetics and pharmacodynamics of cannabinoids.
Clinical Pharmacokinetics, 42(4), 328-360.
Grotenherman F., Leson G., Berghaus G., Drummer O., Krüger H., Longo M., et al.
Developing limits for driving under cannabis. Addiction 2007; 102: 1910-17.
Hart, CL, Marvin, CB, Silver, R & Smith, EE (2012) is cognitive functioning impaired in
math and amphetamine users? A critical review. Neuropsychopharmacology,
37(3), 586-608.
Hayden, J.W. (1991). Passive inhalation of marijuana smoke: a critical review. Journal
of Substance Abuse, 3, 85-90.
23
Health Canada (2011) Canadian Alcohol and Drug Use Monitoring Survey Summary of
Results for 2011, http://www.hc-sc.gc.ca/hc-ps/drugs-drogues/stat/_2011/summarysommaire-eng.php
Hingson, R., Lederman, R., & Walsh, D. (1985). Employee Drinking Patterns and
Accidental Injury: A Study of Four New England States, 46(4), 298-303.
Hoffman, J.,& Larison, C. (1999) Drug use, workplace accidents and employee
turnover. Journal of Drug Use, 29(2), 341-364.
Huestis, M.A., & Cone, E. (1995). Drug test findings resulting from unconventional drug
exposure. In: R.H. Liu, & B. Goldberger eds. Handbook of Workplace Drug
Testing. Washington, DC, AACC Press: 289-320. in Clinical False Positives....
Jaffee, W.B., Trucco, E., Teter, C., Levy, S., & Weiss, R.D. (2008). Ensuring Validity in
Urine Drug Testing. Psychiatric Services, 59, 140-142.
Jacobson M. (2003). Drug testing in the trucking industry: the effect on highway safety.
Journal of Law and Economics, 46, 131-56.
Jones, J. P. (1990). Drug Testing Did Not Reduce Southern Pacific's Accident Rate.
Forensic Urine Drug Testing, 2-4.
Kadehjian, L. (2005). Legal issues in oral fluid testing. Forensic Science International, 150,
151-160.
Kapur, B. (1994) Drug testing methods and interpretations of drug testing results. In:
Macdonald, S. and Roman, P. (Eds). Research Advances in Alcohol and Drug
Problems. New York: Plenum Press.
Kraus, J.F. (2001). The effects of certain drug-testing programs on injury reduction in the
workplace: an evidence based review. International Journal of Occupational and
Environmental Health, 7(2), 103-108.
Last, J. M. (1995). A dictionary of epidemiology (3rd ed). New York: Oxford University
Press.
Longo, M.C., Hunter, C.E., Lokan, R.J., White, J.M., & White, M.A. (2000a). The
prevalence of alcohol, cannabinoids, benzodiazepines and stimulants amongst
injured drivers and their role in driver culpability. Part I: The prevalence of drug
use in drivers and characteristics of drug positive group. Accident Analysis and
Prevention, 32, 613-622.
Longo, M.C., Hunter, C.E., Lokan, R.J., White, J.M., & White, M.A. (2000b). The
prevalence of alcohol, cannabinoids, benzodiazepines and stimulants amongst
injured drivers and their role in driver culpability. Part II: The relationship between
drug prevalence and drug concentration, and driver capability. Accident Analysis
and Prevention, 32, 623-632.
24
Macdonald, S., Lothian, S. and Wells, S. (1998) Evaluation of an employee assistance
program at a transportation company. Evaluation and Program Planning,
20(4):495-505.
Macdonald, S. (1995). The role of drugs in workplace injuries: Is drug testing
appropriate? Journal of Drug Issues, 25(4), 703-722.
Macdonald, S. (1997). Work-place alcohol and other drug testing: A review of the
scientific evidence. Drug and Alcohol Review, 16, 251-259.
Macdonald, S., Csiernik, R., Durand, P., Rylett, M. and Wild, T.C. (2006) The
prevalence and factors related to Canadian workplace health programs.
Canadian Journal of Public Health, 97(2):121-125.
Macdonald, S., Hall, W., Roman, P., Stockwell, T., Coghlan, M. and Nesvaag, S.
(2009). Testing for cannabis in the workplace: A review of the evidence.
Addiction, 105(3), 408-416.
Mann RE, Macdonald S, Stoduto G, Bondy S, Jonah B, Shaikh A. The effects of
introducing or lowering legal per se blood alcohol limits for driving: an
international review. Accident Analysis & Prevention. 2001;33(5):569-83.
Marquet, P., Delpla, P-A., Kerguelen, S., Bremond, J., Facey, F., Garnier, M., Guery, B.,
Lhermitte, M., Mathé, D., Pelissier, A-L., Renaudeau, C., Vest, P., & Seguela, JP. (1998). Prevalence of drugs of abuse in urine of drivers involved in road
accidents in France: a collaborative study. Journal of Forensic Science, 43(4),
806-811.
McKim, W. (1986). Drugs and behavior. Prentice-Hall, New Jersey.
McLeod, J (2010) The effectiveness of workplace counseling: A systematic review.
Counseling and Psychotherapy Research, 10(4): 238-248. R
Miller T. R., Zaloshnja E., Spicer R. S. (2007) Effectiveness and benefit-cost of peerbased workplace substance abuse prevention coupled with random testing.
Accid Anal Prev; 39: 565-73.
Moeller, K.E., Lee, K.C., & Kissack, J.C. (2008). Urine drug screening: practical guide
for clinicians. Mayo Clinic Proceedings, 83(1), 66-76.
Morantz, A.D. & Mas, A. (2008). Does post-accident testing reduce injuries? evidence
from a large retail chain. American Law and Economics Review, 10(2), 246-302.
Mule, S.J., & Casella, G.A. (1988). Active and realistic passive marijuana exposure
tested by three immunoassays and GC/MS in urine. Journal of Analytical
Toxicology, 12, 113-116.
Newcomb, M.D. (1994). Prevalence of alcohol and other drug use on the job: Cause for
concern or irrational hysteria? Journal of Drug Issues, 24(3), 403-416.
25
Niedbala, S., Karados, K., Salamone, S., Fritch, D., Bronsgeest, M., & Cone, E.J.
(2004). Passive cannabis exposure and oral fluid testing. Journal of Analytical
Toxicology, 28, 546-552.
Niedbala, S. Karados, K.W., Fritch, D.F., Kunsman, K.P., Blum, K.A., Newland, G.A.,
Waga, J., Kurtz, L., Bronsgeest, M. & Cone, E.J. (2005). Passive cannabis
smoke exposure and oral fluid testing II: two studies of extreme cannabis smoke
exposure in a motor vehicle. Journal of Analytical Toxicology, 29, 607-615.
Niedbala, R. S., Kardos, K.W., Fritch, D.F., Kardos, S., Fries, T., Waga, J., Robb, J., &
Cone, E.J. (2001a) Detection of marijuana use by oral fluid and urine analysis
following single-dose administration of smoked and oral marijuana. Journal of
Analytical Toxicology, 25, 289-303.
Niedbala, R.S., Kardos, K., Fries, T., Cannon, A., & Davis, A. (2001b). Immunoassay for
detection of cocaine/metabolites in oral fluids. Journal of Analytical Toxicology,
25, 62-68.
Normand, J., Salyards, S.D., & Mahoney, J.J. (1990). An evaluation of pre-employment
drug testing. Journal of Applied Psychology, 75, 629-39.
Normand, J., Lempert, R., O, & O'Brien, C. P. (Eds.). (1994). Under the Influence?
Drugs and the American Work Force. Washington, D.C.: National Academy
Press.
Ontario Law Reform Commission (1992) Report on drug and alcohol testing in the
workplace. Toronto: Law Reform Commission.
Ozminkowski, R. J., Mark, T.L., Goetzel, R.Z., Blank, D., Walsh, J.M. & Cangianelli, L.
(2003). Relationships Between Urinalysis Testing for Substance Use, Medical
Expenditures, and the Occurrence of Injuries at a Large Manufacturing Firm.
American Journal of Drug and Alcohol Abuse, 29, 151–167.
Porath-Waller, A.J., Bierness, D.J., & Beasley, E.E. (2009). Toward a more
parsimonious approach to drug recognition expert evaluations. Traffic Injury
Prevention, 10, 513-518.
Quest Diagnostics Incorporated, (2012) Drug Testing Index
Ramaekers, J.G., Berghaus, G. van Laar, M., & Drummer, O.H. (2004). Dose related
risk of motor vehicle crashes after cannabis use. Drug and Alcohol Dependence,
73(2), 109-19.
Ryan, J., Zwerling, C., & Jones, M. (1992). The Effectiveness of Preemployment Drug
Screening in the Prediction of Employment Outcome. Journal of Occupational
Medicine, 34(11), 1057-1063.
26
Shipp, E., Tortolero, S, Cooper, S., Baumler, E., Weller, N. (2005) Substance use and
occupational injuries among high school students in South Texas. American
Journal of Drug and Alcohol Abuse, 31(2),253-265.
Smiley, A. (1999). Marijuana: On-road and driving-simulator studies. In: Kalant, H.,
Corrigall, W., Hall, W., Smart R. (Eds.), The health effects of cannabis, Addiction
Research Foundation: Toronto, Ontario, pp. 173-191.
Smith, J.A., Hayes, C.E., Yolton, R.L., Rutledge, D.A., & Citek, K. (2002). Drug
recognition expert evaluations made using limited data. Forensic Science
International, 130, 167-173.
Spicer, R.A., Miller, T.R., & Smith, G.S. (2003). Worker substance use, workplace
problems and the risk of occupational injury: a matched case-control study.
Journal of Studies on Alcohol, 64(4), 570-578.
Taggert, R. Results of the drug testing program at Southern Pacific Railroad.Gust, S. W.,
& Walsh, J. M. (1989). Drugs in the Workplace: Research and Evaluation Data.
Rockville, MD: National Institute on Drug Abuse, NIDA Monograph 91.
U.S. Department of Transportation, (2004) Drugs and Human Performance Fact
Sheets, National Highway Traffic Safety Administration, DOT HS 809 725.
Verstraete, A.G. (2004). Detection times of drugs of abuse in blood, urine, and oral fluid.
Therapeutic Drug Monitoring, 26, 200-206.
Walsh, JM (2008) New technology and new initiatives in US workplace testing. Forensic
Science International, 174, 120-124.
Wickizer, T., Kopjar, G., Jutta, J (2004) Do drug-free workplace programs prevent
occupational injuries? Evidence fro Washington State. Health Services Research
39(1), 91-110.
Zwerling, C., Ryan, J., & Orav, E.J. (1990). The efficacy of pre-employment drug
screening for marijuana and cocaine in predicting employment outcome. JAMA,
264(20), 2639-2643.
27