Current practice and experience in drug and alcohol testing in the workplace
I. History of drug testing in the United States
II. Rationale for drug screening in the workplace
III. Employer drug-testing programmes: practice and attitudes in the United States
IV. The effectiveness of workplace drug screening in reducing drug usage in the workforce
V. Drugs, alcohol and occupational injury
VI. Drug testing and other adverse employment outcomes
VII. Cost-benefit analysis
VIII. Future trends and policy concerns
Author: C. ZWERLING
Pages: 155 to 196
Creation Date: 1993/01/01
The present paper presents a review of the current practice and experience in drug and alcohol testing in the workplace, focusing primarily on the United States of America. After reviewing the history of workplace drug screening, the author describes the growth and impact of the drug-testing industry. He outlines the four most common rationales for workplace drug testing: safety, productivity, decreasing drug use and legislative/regulatory requirements. He summarizes the best studies on the prevalence of workplace drug testing in the United States and describes employer attitudes in that country. The author reviews in some detail the association between drugs, alcohol,-occupational injuries and other adverse employment outcomes. He then reviews the literature on cost-benefit analysis of workplace drug testing.
The author concludes that relatively little is known about the role of alcohol and drugs in the workplace. The important association between substance abuse and occupational injury has not been established. If there is such an association it is much weaker than previously believed. The contradictory findings in different studies suggest that substance abuse may well play different roles in different occupational and cultural settings. Thus, caution should be exercised in transposing results from one setting to another. Finally, the review of cost-benefit analyses suggests that any economic analysis of workplace drug screening is likely to be greatly influenced by the prevalence of drugs in the population screened.
*This work was supported in part by a grant from the Centers for Disease Control and Prevention.
The present paper examines the current practice and experience in drug and alcohol testing in the workplace, focusing primarily on the United States of America, where there has been the most experience. Section I presents a review of the history of workplace drug screening in the United States, emphasizing the close ties between workplace drug screening and concerns about safety, as well as describing the growth of the drug-testing industry. Section II presents four rationales for workplace drug testing: safety, productivity, decreasing drug use and legislative/regulatory requirements. Section III presents a summary of the best studies on the prevalence of workplace drug testing in the United States, in addition to examining some employer attitudes. Section IV describes the impact of drug testing in decreasing drug use in the United States military, while pointing out the difficulties in generalizing that experience to the civilian workforce. Section V presents a review of the literature on the association between alcohol, drugs and occupational injuries. Section VI presents a review of the literature on alcohol, drugs and other adverse employment outcomes. In both sections V and VI, special attention is given to assessing the strengths and weaknesses of existing literature. Section VII presents a discussion of the factors involved in conducting cost-benefit analyses of workplace drug testing. Finally, in section VIII, there is a discussion of future trends in workplace drug testing and their implications for policy makers. Throughout the paper, the emphasis is placed on describing what is known, where the data are inadequate and what types of research could fill those gaps.
The emergence of workplace drug testing in the United States can be viewed in the context of changing patterns of drug use over the last three decades. In the period from 1955 to 1980, the United States experienced a 20-fold increase in the non-medical use of drugs  . During the 1960s, this "epidemic" was characterized by large-scale experimental use of marijuana, lysergic acid diethylamide (LSD) and other hallucinogens by middle-class and upper-class youth  . During the 1970s, drug use among young people in the United States increased in all social strata.
In the United States, national surveys can be used to track patterns of self-reported drug use. The National Household Survey on Drug Abuse began in the period 1971-1972 and has continued ever since. The national surveys show that marijuana use increased over the decade of the 1970s, reached a plateau in the 1980s and began to drop in the late 1980s and early 1990s. Cocaine use grew through the 1970s and early 1980s but then reached a plateau and also began to decrease. Among Americans over the age of 12, the number currently using illicit drugs (i.e. having used them in the past month) dropped from 23 million in 1985 to 13 million in 1991, or by more than 40 per cent  . During the same period, current cocaine use dropped by about two thirds. In 1991, among full-time employees, 6 per cent were current users of illicit drugs. Drug use varied significantly from industry to industry. Among full-time employees aged 18-34, current drug use ranged from .15.4 per cent among construction workers down to 6.5 per cent among professionals. In most industries, the percentage of current drug users in this age group has been decreasing since 1988  . The annual Monitoring the Future Survey conducted by the Institute for Social Research at the University of Michigan showed a similar pattern of drug use among secondary school pupils aged 17-18 in the period 1975-1990  . The number of such pupils who had used marijuana in the last year increased from 40 per cent in 1975 to 51 per cent in 1980 but subsequently dropped to 27 per cent in 1990. Similarly, cocaine use increased from 6 per cent in 1975 to 12 per cent in 1981 but dropped to 5 per cent in 1990.
Much of the early experience in drug screening large numbers of people came from the United States military. The military had initially introduced urine drug screening to identify heroin users returning from military duty in Viet Nam in the late 1960s and early 1970s. That programme was extended to screening soldiers reporting for active duty in the early 1970s. In 1980, the United States Department of Defense published a survey of substance abuse among active duty, military ,personnel  . Overall drug use in the military services was reported at 26 per cent. Among young enlisted men aged 18-25, usage was as high as 47 per cent in the United States Navy and Marine Corps. In May of 1981, a Marine Corps aircraft crashed aboard the aircraft carrier Nimitz. Of the 14 people who died, 9 revealed evidence of cannabinoids in the autopsy. In addition, the pilot was taking a prescribed antihistamine without the knowledge of his commanding officer or flight surgeon. The publicity surrounding the crash accelerated the decision by the Navy to implement across-the-board drug screening.
In addition, the increasing use of urine drug testing in the 1980s was made possible by technological improvements in testing. In the 1960s and 1970s, the screening method of choice was thin-layer chromatography (TLC). TLC is inexpensive and relatively quick and permits the simultaneous detection of many substances in a single test run. However, it requires a skilled analyst to read and is not easily adapted to mass screening programmes. In the late 1970s, radioimmunoassay (RIA) and the enzyme-multiplied immunoassay technique (EMIT) began to appear  . The immunoassays were more easily automated, allowing specimens to be screened at a lower cost than was possible with TLC. The first equipment enabled a technician to double the number of tests per- formed in a day [6). Today's more sophisticated equipment can process 4,000-7,000 urine samples per hour, reducing the costs for immunoassay substantially. The availability of this technology allowed the United States Department of Defense to plan for the rapid expansion of its testing programme  .
In 1986, the United States Government began full-scale efforts to advocate urine drug testing in the workplace. In the spring of 1986, the President's Commission on Organized Crime released its report on habits of Americans: drug abuse, drug trafficking and organized crime. After outlining the relation between organized crime and illegal drug use, the Commission turned towards solutions. Since attempts to limit the supply of drugs had failed, the Commission advocated a series of measures to decrease demand. In particular, it called upon the Government to "provide an example of the unacceptability of drug use. The President should direct heads of all federal agencies to formulate immediately clear policy statements with implementing guidelines, including suitable drug testing, expressing the utter unacceptability of drug use by federal employees"  . On 15 September 1986, President Ronald Reagan issued Executive Order 12564 on the Drug Free Federal Workplace. President Reagan stated that drugs were causing billions of dollars of lost productivity each year. In particular, he stated that federal employees using illegal drugs were less productive, less reliable, and prone to absenteeism. Moreover, he asserted that "the profits from illegal drugs provide the single greatest source of income for organized crime, fuels for violent street crime, and otherwise contribute to the breakdown of our society". He then called on all federal employees to refrain from using illegal drugs and mandated each executive agency to establish a programme to "test for the use of illegal drugs by employees in sensitive positions"  . Federal agencies moved quickly to set up drug-screening programmes.
In January 1987, a passenger train crashed at Chase, Maryland, killing 16 passengers, injuring 174 and causing millions of dollars in property damage. The urine specimens from both the engine-driver and brakeman of the train were positive for -marijuana. On 21 January 1987, the United States Department of Transportation proposed rigorous drug- testing programmes, requiring pre-employment, post-accident and random testing of airline pilots, railroad workers, air traffic controllers and other employees in safety-related positions  .
The move towards drug testing in the federal workplace was challenged by several organizations. Both the American Civil Liberties Union and a variety of workers' organizations raised concerns about drug tests being invasions of employee privacy.. From 1987 to 1991, the United States Department of. Justice spent 725,000 United States dollars (US$) defending litigation related to drug testing  . Two of those cases reached the United States Supreme Court, where they were decided on 21 March 1989. In one case, the National Treasurer's Employees Union challenged the United States Customs Services requirement involving the drug screening of certain customs employees. In the other, railway workers' organizations sued to enjoin regulations: promulgated by the Federal Railroad Administration which governed the drug and alcohol testing of railway employees. In both cases, the Supreme Court upheld the legality of the drug-screening programme of the Government. In both cases, the Supreme Court balanced the invasion of employees' privacy against "other compelling government interest".
In the National Treasury Employee's Union (No. 86-1879), the court upheld the constitutionality of the programme by a 5-to-4 majority, arguing that drug-using customs officers would be vulnerable to corruption and blackmail and might have unsound judgement. The dissenters replied that there was no incident cited in which any of those adverse outcomes had occurred: no example of bribe-taking, improper handling of a firearm, or compromising of classified information associated with drug use. In Skinner v. Railroad Labor Executives Association (No. 87-1555), the Supreme Court upheld by 7 votes to 2 the Federal Railroad Administration regulations. The majority depended heavily on the "surpassing safety interests served by toxicological tests in this context". The majority in the safety-related railroad case (7 votes to 2) was larger than in the law enforcement driven customs case (5 votes to 4), suggesting that the Supreme Court found the safety arguments more compelling.
Until recently, drug testing in the workplace in the Unites States had been primarily limited to illegal drugs. Recent developments, however, suggest that there will be increasing testing for alcohol as well. Again, the new initiative in testing arose from a much publicized accident, the wreck of the Exxon Valdez, and the widespread concern for the role that alcohol use played in that incident. The Omnibus Transportation Employee Testing Act of 1991 mandates pre-employment, random, reasonable suspicion and post-accident testing for alcohol, as well as for controlled substances. It requires the Secretary of Transportation to design and implement a pilot programme for the random testing of operators of commercial motor vehicles for alcohol, as well as for controlled substances.
The increasing use of drug tests in the workplace in the United States has led to the emergence of a large drug-testing industry. It has recently been estimated that the manufacture of equipment and chemicals used in drug testing is an industry worth US$ 300 million  . But the drug- testing industry extends beyond the manufacturers of the equipment and chemicals. As the industry has grown, so has the number of people and organizations with a stake in workplace drug testing  . A small number of large pharmaceutical companies market most of the chemicals and equipment to the drug-testing laboratories, in addition to selling directly to firms doing on-site testing. The-laboratories that carry out the testing take in hundreds of millions of dollars a year  . In addition, there are services necessary to keep the laboratories going, such as continuing medical education courses offered by the American Association of Analytical Chemists. An entire group of medical review officers has arisen to review the results of the tests. They have recently formed their own association, which sponsors continuing medical education courses and certification examinations. Currently, over 500 have been certified  . In addition, the American College of Occupational and Environmental Medicine has been offering similar courses and examinations. In implementing the drug-testing programme, private industry has often called upon consulting firms to advise them on how to comply with the law. Some of the best-known consultants to industry were once officials active in anti-drug activities. Because many employers refer first-time positive employees to employee assistance programmes, increased drug testing has also increased referrals to privately owned substance abuse treatment clinics. Such clinics increased in number and in volume of admissions during the 1980s  .
In addition to all these organizations with a stake in workplace drug screening, the National Institute on Drug Abuse (NIDA) has played a large role in expanding this programme. NIDA has sponsored a series of national conferences on the topic, has provided a toll-free helpline to assist private employers designing anti-drug programmes and has published a wide variety of monographs and booklets supporting workplace drug screening  ,  . It is important to understand the size of the drug-testing industry, the variety of organizations that have a stake in drug testing, and the large number of consultants and experts who benefit financially from drug testing. All these factors must be kept in mind when reviewing the drug-testing literature and formulating policy.
Safety concerns have played a crucial role in the development of workplace drug testing in the United States. As noted above, several of the main advances in workplace drug testing took place shortly after, and in response to, well-publicized accidents in which drugs or alcohol were implicated. The aircraft crash on the Nimitz,the train crash in Maryland and the wreck of the Exxon Valdez are just three examples. Those accidents have had a great impact on public policy because they loom large in the minds of citizens.
Workplace drug testing always involves a trade-off: workers relinquish part of their privacy and submit to invasive tests in return for some benefits. The benefits that weigh most heavily on the minds of the public are safety benefits, a fact recognized early on by the supporters of workplace drug testing. In 1986, Bensinger  argued that 'the principal concern to industry must be the health and safety of its employees". Similarly, safety arguments were weighed most heavily by the Supreme Court in its first workplace drug-testing case, Skinner v. Railroad Labor Executives Association.
Early advocates of workplace drug testing stated that drug-abusing employees had 3-4 times more accidents at work than other employees  ,  . In addition, they asserted that drug abusers had five times more compensable injuries  . Similarly, they stated that 40 per cent of industrial fatalities and 47 per cent of industrial injuries could be traced to alcohol abuse  .
Although none of these early claims were backed by substantive empirical evidence, they appeared plausible. It is well known that the use of alcohol and other drugs can impair motor coordination and perceptual abilities. Thus, the argument that the elimination of drugs in the workplace will reduce injury rates is, on the surface, plausible. More- over, people value health and safety so much that many are willing to tolerate a certain invasion of privacy in order to protect themselves and others from injury. Because of the importance of these safety arguments, both in the mind of the general public, in the minds of policy makers and in the deliberations of the United States Supreme Court, the evidence on the association between occupational injuries and alcohol and other drug use in the workplace is reviewed at length in section V below.
A second justification of workplace drug screening is to increase productivity in the workplace by minimizing absenteeism, turnover, and other gauges of poor work performance. While these arguments are less convincing to the general public and play a minor role in the legal defence of drug testing, they appear to loom large in the business community.
In 1983, when he was a United States Senator, Dan Quayle stated  that the rate of growth of productivity in the United States had decreased dramatically since 1977. After listing nine potential legal and regulatory policies that might be related to productivity, he turned to alcoholism and drug abuse and suggested that reduction of their negative impact could most easily increase productivity in the workplace. Framing the argument in terms of cost to productivity, Quayle stated that lost productivity due to alcohol and drug abuse amounted to US$ 30.1 billion a year. He went on to argue that employees with a drinking or drug problem were absent 16 times more than the average employee and that such impaired workers functioned in slightly more than half their normal capacity  .
These arguments have been used extensively as a rationale for preemployment drug screening. Some of the evidence relating absenteeism and turnover to drug and alcohol use is examined in section VI below.
A third goal of workplace drug screening is to reduce the widespread use of illicit drugs in society. This was most clearly stated in the report of the President's Commission on Organized Crime  , in which the Commission drew attention to the need to reduce demand for illicit drugs as part of the "war on drugs". Workplace drug screening was considered a tool in demand reduction. That concern continues to be a major aspect of United States federal drug policy. In a manual entitled Building a Drug Free Workforce, the Office of National Drug Control of the United States asserts that "because 70 per cent of all drug users are employed, the workplace may be the most strategic point in society from which to combat the scourge of drugs"  .
As noted in section I above, illegal drug use in the United States has declined in recent years. It is difficult to assess what role workplace drug screening has played in that process. Although workplace drug screening may have accelerated the downward trend, it appears that drug use was declining before the widespread introduction of such testing.
Finally, many employers adopt workplace drug-screening programmes because they are mandated to do so. In the United States, such regulations affect the transportation industry, the nuclear industry and the federal Government. In those industries, employers must screen applicants for certain positions and some categories of employees irrespective of the utility of such screening in improving safety or productivity. Through detailed government regulations, screening in the regulated industries plays a large role in setting the technical standards for the drug-testing industry.
Since most of the experience with workplace drug testing has been in the United States, it is worth while to summarize what is known about .the practices and attitudes there. Relatively little national information is available concerning private-sector efforts to deal with drug abuse in the United States. Although privately financed surveys have been conducted, they suffer from methodological problems, usually focusing on relatively small segments of the private sector and using samples that are not representative of employers as a whole  . However, one comprehensive, scientifically conducted survey was carried out by the Bureau of Labor Statistics in the-United States in the summer of 1988.
The 1988 survey by the Bureau of Labor Statistics showed that per cent of non-agricultural establishments in the United States. had drug-testing programmes; those establishments employed 20 per cent of the non-agricultural workforce. A follow-up survey in 1990  showed no statistically significant change in the percentage of establishments with drug-testing programmes (4.0 per cent), but did note an increase in the percentage of large employers testing. Of all establishments with 250 or more employees, 32 per cent tested in 1988, compared with 46 per cent in 1990.
The 1988 study was designed to estimate the number of private, non - agricultural establishments with drug-testing programmes, by employment size class, major industry division, and multi-state geographical region  . The survey was a one-time probability sample survey of 7,502 private non-agricultural establishments in the United States with one or more employees in the first quarter of 1987. It was based on a sampling frame constructed from the 1987 Unemployment Insurance Address File, containing approximately 4.5 million establishments and accounting for about 85 million employees. The establishments were divided into 400 sample strata within five geographical regions, 10 standard industrial classification (SIC) groupings and 8 employment size classes. Response rates for the drug-testing questions were high, ranging from 84.5 to 92.4 per cent. In sum, the survey provided the most scientifically collected and analysed data on the scope and nature of workplace drug testing in the United States.
The survey clearly demonstrated that the most important determinant of the incidence of drug testing was the establishment size as measured by the number of employees in the establishment. The larger establishments were more likely to have drug- testing programmes than smaller ones (see figure I). Thus, among the largest establishments in the United States, with more than 1,000 employees each, 43 per cent had drug-testing programmes. In contrast, among the smallest establishments, with fewer than 50 employees, only 2 per cent had drug-testing programmes. Since small workplaces make up the overwhelming majority of a nation's businesses, only 3 per cent of establishments overall had drug-testing programmes.
There were also differences in rates of drug testing by industry, but they were not as marked as differences by size or establishment (see figure II). The rates of drug testing were lowest in retail trade (0.7 per cent), service industries (1.4 per cent) and construction (2.3 per cent); they were highest in mining (21.6 per cent), communications and public utilities (17.6 per cent) and transportation (14.9 per cent). In part, some of the differences may have been accounted for by difference in size of establishment.
Differences between geographical regions were minimal. The pro- portion of companies with testing programmes ranged from 2 per cent in the north-east to around 4 per cent in the south and Middle West. Of the establishments with testing programmes, about 85 per cent tested job applicants; 64 per cent tested current employees.
In private companies, relatively few current employees were actually tested for drugs. In 1987, in firms with testing programmes, just under I million workers were tested, or about I per cent of all the eligible workers. Of the employees tested, about 9 per cent were positive for some illicit drug. Of the 3.9 million applicants who were tested, 12 per cent were positive for drug use. The proportion of employees testing positive varied from 3 per cent in service industries to 20 per cent in wholesale trade. Among job applicants, the positive rate varied from 6 per cent for communications and public utilities to 24 per cent in the retail trades.
Figure 1. Share of United States non-agricultural firms with drug-testing programmes, by number of employees In firm, 1988.
Figure II. Share of United States non-agricultural firms with drug-testing programmes, by number of employees in firm, 1988.
In 1990, the Bureau of Labor Statistics conducted a follow-up survey using as a sample 749 of the establishments that had participated in the 1988 survey  . Overall, the survey found no statistically significant change in the incidence of drug-testing programmes: 3 per cent in 1988 compared with 4 per cent in 1990. -While many larger firms adopted drug testing between 1988 and 1990, fully one third of the establishments with drug-testing programmes in 1988 had discontinued them by 1990. Most of the establishments that had discontinued drug-screening programmes were small businesses.
Although no equivalent to the Bureau of Labor Statistics survey is available for other countries,- in 1990 the International Labour Office carried out a study  of 53 large enterprises with greater than 1,000 employees in seven countries. Canada, Germany, Netherlands, Norway, Sweden, United Kingdom of Great Britain and Northern Ireland and United States. Since the International Labour Office study was not based on a random sample, results cannot be generalized validly to all companies in those seven countries. Discussions among the collaborators from the seven countries suggested, however, that drug testing was an important component of workplace drug and alcohol programmes only in the United States. It was suggested that multinational corporations might be expected to introduce drug testing in their facilities outside the United States.
A survey sponsored by the Conference Board, an industry sponsored group  , provides some insight into employer attitudes towards urine drug testing. In interpreting the results of that survey, however, it must be kept in mind that only 26 per cent of the 2,675 questionnaires mailed out to senior human resource officers in large United States companies were returned. This contrasts markedly with the return rate of 85 per cent in the Bureau of Labor Statistics survey described above. As might be expected, the attitudes of executives of firms with drug- testing programmes differed from those of executives of firms without such programmes. In firms that did not test, fully 63 per cent of their executives felt that their primary substance abuse problem was alcohol. By comparison, in firms that did have a drug-testing programme, only 25 per cent of the executives felt that alcohol was the primary substance abuse problem. Among both groups, there was agreement that illegal drugs were more of a problem than five years before. Both groups also agreed that alcohol abuse, although still a problem, was no worse than it had been five years before.
Most of the companies with drug-testing programmes had written substance abuse policies that had been carefully crafted. On the average, executives from at least four functional areas within the corporation were involved in the preparation of the substance abuse policies. Ninety -four per cent of the time, policies were written with the involvement of human relations staff. Three quarters of the time, the legal staff was involved. About one half of the time, the chief executive officer, medical staff, labour relations staff or employee assistance programme (EAP) staff were involved. In only one seventh of the cases were unions involved in the formulation of the policy.
Executives in companies that implemented urine drug-testing programmes felt that the most compelling evidence for doing so included evidence of drugs in the workplace, a sense that drug testing was an early detection procedure, concern for legal liability for the action of drug- impaired employees, evidence of drugs in the community and evidence that drugs were costing the company money. The executives at those companies that did not have a drug-testing programme cited the threat of legal action challenging testing as the most important deterrent. Other reasons for not testing included concern with the accuracy of the test, the inability of drug testing to measure impairment, and potentially negative effects on employee relations. Union opposition was reported as "not at all important" by two thirds of the responding firms, probably a reflection of the absence of organized labour at many non-testing firms. In the financial services industry, the incompatibility of drug testing with the corporate culture and philosophy was perceived as a stumbling-block.
Of the firms that were drug testing, 12 per cent reported that their programmes had been challenged in the courts. Twenty- three per cent of .the programmes had been challenged by unions and brought to arbitration. In 20 per cent of the companies testing, employee resentment was a concern, as were problems with cheating in the testing process.
The Conference Board report included several warnings. Some corporate executives had tempered their observations with cautionary notes. Several firms, for example, had advised companies to first take stock of their own unique work conditions, business needs, and mode of operation. 'A company should not start a programme just because .everyone else is doing it'", a human resource manager in a corporation at St. Louis had said. Companies should not overemphasize the benefits of drug testing. An Atlanta manufacturer had cautioned: "Rather than being swept along by the media, companies should carefully evaluate their objectives and culture before taking steps. Education and employee assistance are far more powerful than drug testing." Going even further, an aircraft manufacturer had stated: "Drug testing is a very limited deterrent. A value system that does not tolerate substance abuse and a capable EAP are the only means to deal with (substance abuse) issues." And there had been a down-to-earth reminder from a corporate medical director in a consumer goods firm in the Middle West: "Drug testing is not the solution to drug problems. It is only a piece of technology that supplies information for making intelligent decisions  ."
Workplace drug screening is aimed at reducing injuries, and increasing work productivity by reducing drug usage among the workforce. However, there are few data showing the effects of workplace drug screening on drug prevalence in the workforce. Furthermore, obtaining scientifically valid data is quite difficult. One way to obtain such data would be a randomized controlled study by a large corporation. In such a study, similar plants would be randomized to either a drug-screening programme or no such programme and levels of drug use among the workforce in the two plants would need to be monitored subsequently. A review of the literature on the subject revealed no such studies. In the absence of a randomized, controlled study, historical controls must be relied upon to evaluate the efficacy of drug screening in reducing drug usage. This is problematic, especially over the last decade when, as noted above, there has been a significant decrease in drug usage. Studies showing a decrease in drug usage after the institution of a drug programme may only be showing the effect of decreasing drug usage in society at large rather than the effect of the programme. In addition, such studies are limited by the difficulty in obtaining accurate data on prevalence of drug use in working populations. Survey data are subject to reporting bias if the employee is not confident that confidentiality will be respected. More and more sophisticated techniques are being used by workers to avoid submitting positive samples in urine drug screens  .
Although the experience of the United States military demonstrates that, in certain restricted circumstances, workplace drug screening can have an impact on the prevalence of drug use in the workforce, this experience cannot he generalized to the civilian workforce. In the course of the 1980s, the United States military carried out one of the most rigorous employer anti-drug policies  . That policy included pre- enlistment urine drug screens, as well as random drug screens on all servicemen. The random drug screens averaged more than one per serviceman per year, the goal in the Navy being three per serviceman per year. For example, in 1986, the military undertook about 2.9 million urine drug screens at a total cost of nearly US$ 52.4 million  . The effects of that rigorous programme can be evaluated by following the data from worldwide surveys of substance abuse and health behaviour among military personnel that were conducted at the request of the United States Department of Defense in 1980, 1982, 1985 and 1988  ,  ,  ,  ,  . The sampling frame for these well-designed studies consisted of all United States active-duty military personnel stationed across the world except recruits, service academy students and persons absent without leave. The probability sample was selected for the survey using a deeply stratified, two-stage design. The first-stage sampling units were major military institutions stratified by service and world region; the second-stage sampling units were individuals located at installations stratified by military pay grade. The response rate for eligible participants was 81 per cent. Drug and alcohol use were measured using a confidential questionnaire. Drug use was measured during the past 12 months and use of any drugs during the past 30 days. Using this last measure, drug use within the military decreased over five -fold from 1980 to 1988: the share of all servicemen who had used drugs within the last 30 days was 27.6 per cent in 1980,.19.0 per cent in 1982, 8.9 per cent in 1985 and 4.8,per cent in 1990. These results remained unchanged when rates were standardized to account for changes in age, -education and marital status in the military in the 1980s 1301. In part, these changes reflect similar changes occurring in the civilian labour force. Careful comparisons of the rate of change among military personnel and civilians, however, suggest that the rate of change of drug use was, greater among military personnel than among civilians  .
In reviewing these data, one must keep, in mind the rigour of military drug-testing programmes. Throughout most of the 1980s, urine drug screens were monitored by direct observation, which is, not acceptable in the civilian labour force. The programme involved frequent -random screenings. Most civilian drug screening programmes do not entail random screening except in selected industries. Even in those industries, there is much pressure from the industry to cut down on the frequency of the screenings as a cost-containment measure. Thus, generalizations cannot be made about drug-screening programmes involving the civilian labour force based on the military, experience. Specially, the military experience provides little evidence of -the capability of pre-employment drug screening alone to lower the prevalence of drug use among the civilian workforce.
Thus, it is still not known whether drug testing in the civilian workforce is effective in reducing drug usage, and if so, by how much. A randomized, controlled study would help answer that question.
To evaluate the possible benefits of workplace drug and alcohol screening, it is necessary to have data on the adverse employment out- comes associated with drug and alcohol use. In the opinion of both the public and the United States judicial system, the most important of the adverse outcomes are occupational injuries or injuries to the public. In reviewing the literature on the relation between drug and alcohol use and occupational injury, two related questions must be addressed. First, what proportion of occupational injuries is associated with drug or alcohol use? Secondly how strong is the association between substance use and injury? The first question is best addressed through medical examiners studies of fatal occupational injuries and emergency-room-based population studies. The second question is better approached by evaluating industry-wide studies and population-based survey data.
In the studies of fatal occupational injuries summarized in table 1, alcohol was detected in the blood of about 10 per cent of the cases. If only those cases with blood alcohol concentration (BAC) levels of greater than or equal to 0.08 per cent are considered, then the percentage of alcohol-related work injuries decreases to less than 7.3 per cent in all but one study. That study  , conducted in Fulton County, Georgia, showed that 22 per cent of the fatally injured workers had BAC levels greater than 0.1 per cent; however, it was based upon only 23 cases. Thus, alcohol plays a relatively restricted role in -fatal occupational injuries, especially when compared with other types of injuries such as motor vehicle crashes, in which about 40 per cent of all fatalities involve drivers with BAC levels of 0.1 per cent or higher  , or drownings, where 40-50 per cent of the fatalities are associated with high BAC levels  .
Only one United States study  provides data on the proportion of occupational injuries associated with alcohol-related impairment. In that study, alcohol levels, measured by a Breathalyser, were examined among a sample of 5,622 injury patients seen in the Massachusetts General Hospital emergency room over a six-month period from 1966 to 1967. Of the 969 patients who were seen for non-transport-related occupational injuries, 4.9 per cent had BAC levels of 0.05 per cent or greater. Another 10.6 per cent had BAC levels between 0.01 per cent and 0.04 per cent. Thus, the elimination of all alcohol impairment on the job would only reduce the number of occupational injuries by about 5 per cent. By comparison, 17.1 per cent of the transportation-related injuries were associated with BAC levels of 0.05 per cent or greater, as were 11.3 per cent of the home injuries, 38.8 per cent of the injuries related to fights or assaults and 13.2 per cent of all other injuries.
Table 1. Studies of alcohol and drug use among fatally Injured workers
(THC = tetrahydrocannabinol)
|Author (and year of study)||Place||Number of workers||Years covered by study||Blood alcohol concentration(percentage)||Share of subjects(percentage)||Type of drug or work||Share of subjects(percentage)|
Other illicit drugs
Maryland, United States
Fulton County, Georgia,
United States of
Miami, Florida, United
States of America
Harris County, Texas,
United States of
Other illicit drugs
States of America
States of America
North Carolina, United
States of America
Sources:E. C Alleyne, P. Stewart and R. Copes, "Alcohol and other drug use in occupational fatalities", J ournal of Occupational Medicine (Baltimore), vol. 33, No. 4 (1991), pp. 496-500, S. P. Baker and others, "Fatal occupational injuries", Journal of the American Medical Association (Chicago), vol. 248, No. 6 (1982), pp. 692-697, R. L, Berkelman and others, "Fatal injuries and alcohol", American Journal of Preventive Medicine (New York), vol. 1, No. 6 (1985), pp. 21-28; A. R. Copeland, "Fatal occupational accidents: the fiveyear Metro Dade Country experience, 1979-1983", Journal of Forensic Sciences (Philadelphia), vol. 30, No. 2 (1985), pp. 494-503; R. J. Lewis and S. P. Cooper, "Alcohol, other drugs and fatal work related injuries", Journal of Occupational Medicine (Baltimore), vol. 31, No. 1 (1989), pp. 23-28; D. K Parkinson, W. F. Gauss and E. S. Perper, "Traumatic workplace deaths in Allegheny County, Pennsylvania, 1983 and 1994", Journal of Occupational Medicine (Baltimore), vol. 28, No. 2 (1986), pp. 100-102; C. C. Robinson, L. H. Kuller and J. Perper, "An epidemiologic study of sudden death at work in an industrial county, 1979-1982', American Journal of Epidemiology (Baltimore), vol. 128, No, 4 (1988), pp. 806-820; H. S. Shannon and others, "Fatal occupational accidents in Ontario, 1986-1989", American Journal of Industrial Medicine, vol. 23, No. 4 (1993), pp. 253-264; and J. P- Sniezek and T. M. Horiagon, "Medical examiner reported fatal occupational injuries: North Carolina, 1978-1984', American Journal of Industrial Medicine (New York), vol. 15, 1989, pp. 669-678.
a/ For the period 19&3-1986 only.
b/ N= 9.
c/ N= 65.
d /N= 104.
A more recent French study  involved a sample of 4,796 injury patients who had been treated in the emergency units of 21 French hospitals in 1982 and 1983. Of the 882 men treated for non- transport work-related injuries, 8.3 per cent had BAC levels above 0.08 per cent; of the 111 women treated, 0.9 per cent had BAC levels above 0.08 per cent. Thus, the elimination of all alcohol impairment on the job would only reduce the number of occupational injuries by about 7.5 per cent. By comparison, among men 30.9 per cent of the motor vehicle injuries were associated with BAC levels of 0.08 per cent or greater; as were 25.0 per cent of the home injuries and 56.4 per cent of the injuries related to fights.
Similarly, a survey of a -nationally representative sample of hospital emergency rooms  showed that non-work-related injuries were seven times as likely to be associated with alcohol as work-related injuries. However, the study was based primarily upon reports of alcohol intoxication in medical charts rather than on systematic alcohol levels in all patients. Thus, the study underestimates the incidence of alcohol impairment in injury victims by about two orders of magnitude and the possibility of differential reporting among different -types of injuries cannot be ruled out.
The evidence suggests that acute alcohol impairment is present in 5-10 per cent of occupational injuries. Thus, it is less of a problem in relation to occupational injuries than it is in relation to motor vehicle crashes and intentional injuries. However, acute alcohol impairment is not the only way in which alcohol may be related to occupational injury. It is possible that alcoholics may have neurologic impairments that put them at risk for occupational injury even when sober. Furthermore, it is possible that employees suffering from hangovers after drinking the day before could be at risk for occupational injuries. These issues have been dealt with in those studies, with varied results (see table 2). Many of the studies suffer from serious methodological problems because of the way the cohort was chosen, because of the definition of the high-risk group and because of the way outcomes were measured. Several studies  ,  ,  reported on "problem drinkers" as defined by company officials or company medical personnel. Because they used only that small group of "problem drinkers", the studies could be criticized for their serious selection bias. For example, it is quite likely that company officials or medical personnel became aware of drinking problems of their employees because of work-related injuries. Such selection bias would increase the association between "problem drinkers" and occupational injuries.
Table 2. Studies of alcohol and non-fatal occupational injuries
(BAC = blood alcohol concentration)
|Author (and year of study)||Place||Occupation||Number and/or type||Years covered by study||Study design||Alcohol measure||Outcomes||Effect estimates|
Cases: 25 percent
Controls: 28 per
Lost work days
Cases: 5.6 percent
Controls: 15.8 per
Odds ratio: 0.59 (1988)
Last 3 months:
last 3 months
|0.71 (0.41, 1.23)|
Last 24 hours:
last 24 hours
|Author (and year of study)||Place||Occupation||Number and/or type||Years covered by study||Study design||Alcohol measure||Outcomes||Effect estimates|
Odds ratio: 2.0
intake of five
or more drinks
Odds ratio: 2.0
of a large
Odds ratio: 1.65
Proportion of Pat-
ients with occupa-
tional injuries having
BAC z:0.08 per cent;
8.3 per cent
0.9 per cent
seven times as
likely to be
Alcoholics: 9 per
Odds ratio. 2.58
No relation with
binge. drinking or
eight diinks on
|Author (and year of study)||Place||Occupation||Number and/or type||Years covered by study||Study design||Alcohol measure||Outcomes||Effect estimates|
Share of patients
injuries having BAC
>:0.05 per cent:
4.9 per cent
Sources:P. B. Beaurnont and J. Hyman, "The work performance indicators of problem drinking: some British evidence",.Joumal of Occupational Behaviour (Chichester, Sussex), vol. 8, 1987, pp. 55-62; D. J. Buchanan, "Studies on blood alcohol in the workers of a Zambian copper mine", Alcohol and Alcoholism (New York), vol. 23, No. 3 (1988), pp. 239-242; R. B. Hertz and I- A. Emmett, 'Risk factors for occupational hand injury, Journal of Occupational Medicine (Baltimore), vol. 28, No. 1 (1986), pp. 36-41; R. W. Hingson, R. L. Lederman and D. C-Walsh, "Employee drinking patterns and accidental injury: a study of four New England states", Joumal of Studies on Alcohol (Piscataway, New Jersey), vol. 46, No. 4 (1985), pp. 298-303; M. A-Maxwell and others, "A study of absenteeism, accidents, and sicknees payments in problem drinkers in one industry, Quaterly Journal of Studies on Alcohol, vol. 20, 1959, pp. 302-312; A. W. Moll van Charante and G. H. Mulder, 'Perceptual acuity and the risk of industrial accidents", American Journal of Epidemiology (Baltimore), vol. 131, No. 4 (1990), pp. 652-663; L. Papoz and others, "Biological markers of alcohol intake among 4,796 subjects injured in accidents", British Medical Journal (London), vol. 292, 1986, pp. 1234-1237; R. B. Trent, 'Emergency room evidence on the role of alcohol intoxication in injury at work in the U.S., Saftey Science (Amsterdam), vol. 14,1991, pp. 241-252; H. M. Trice, Alcoholic employees: a comparison of psychotic neurotic, and 'normal' personnel", Journal of Occupational Medicine (Baltimore), vol. 7, No. 1 (1965), pp. 94-99; G. R. Webb and others, "The relationship between high-risk and problem drinking and the occurrence of work injuries and related absences', unpublished paper presented at the l8th Annual Alcohol Epidemiology Symposium, Toronto, 1-5 June 1990; H. Wechsler and others, 'Alcohol level and home accidents', Public Health Reports (Rockville, Maryland), vol. 84, No. 12 (1969), pp. 1043-1050.
A recent study  avoids the problem of selection bias by defining the high-risk populations using a questionnaire administered prospectively to all employees. That study used three separate definitions of high-risk workers: problem drinkers, as defined by the Mortimer- Filkins test; heavy alcohol consumption, as defined by a seven-day drinking diary; and binge drinking, as defined by eight or more drinks on more than one occasion. Of these three high-risk groups, only the problem drinkers showed a statistically significant association with occupational injuries. The odds ratio, a measure of the relative risk, was 2.58. However, there was no statistically significant association between average alcohol consumption over a seven-day period or a history of binge drinking with occupational injuries.
Other studies have shown similarly mixed results. A study of British municipal workers  showed no difference in the percentage of workers with reportable injuries or in the number of lost work days caused by occupational injuries when comparing "problem drinkers" with control workers. Similarly, a case-control study of hand injuries among Maryland municipal workers  showed no elevated risk of injury associated with drinking over the last three months or drinking within 24 hours preceding injury. That study was based on self -reported data on alcohol use, however, and might suffer from reporting bias because injured employees might have been reluctant to report their alcohol use even when guaranteed confidentiality.
An anonymous telephone survey of New England workers  found no association between occupational injury and a history of drinking at work. It did, however, find an association between occupational injury and those workers who drank five or more drinks per day; the odds ratio was 2.0 (1.0, 4.1).
A case-control study of shipyard workers in the Netherlands  found that any regular alcohol consumption was associated with occupational injuries. The odds ratio was 1.65, even after controlling for a variety of other potential predictors of injury, such as noise at work, hearing loss and working on the dock.
Based on a review of the studies described above, the following conclusions can be drawn:
Acute alcohol impairment is present in about 10 per cent of fatal occupational injuries;
Acute alcohol impairment is present in about 5 per cent of non-transport, non-fatal, work-related injuries;
A history of alcohol abuse may be weakly associated with occupational injuries (odds ratios ranging from 1.0 to 2.58);
The wide variety of methodological difficulties in the various studies of the association of a history of alcohol abuse and occupational injuries should make the reader cautious in drawing conclusions from the literature.
Medical examiner data are not as useful in assessing the role of drugs in occupational fatalities as they are for alcohol. In the case of alcohol, identification of a BAC level greater than 0.08 per cent suggests strong evidence of impairment at the time of death, implying that alcohol contributed to the causation of the fatality.. For other drugs, such as marijuana and cocaine, the presence of blood or urine metabolites at death provides little evidence relating to, impairment. Thus, for illicit drugs, a control group is needed to compare with the fatally injured workers. If the percentage of fatally injured workers with drugs in their urine and blood is substantially greater than that of the control group, an argument could be made for an association of drug use and fatal occupational injuries. Unfortunately, appropriate controls were not included among the medical p examiner data summarized in table I above. In most of the studies, not even the results of toxicological screens on the patients were provided. In the study of Houston workers  , there were such low rates of drug positives as to raise concerns about the completeness of the toxicological studies. Parkinson and others  detected no illicit drugs in any of the fatally injured workers, but the cohort was too small to generate stable rates. In a larger study conducted in Alberta, Canada  , 459 fatalities were reviewed, but no illicit drugs were detected except marijuana, which was present in 8.5 per cent of the workers tested between 1983 and 1986. Those numbers might well be consistent with the prevalence of drug usage among injured Canadian workers at that time. In sum, these three mortality studies provide no evidence of an association between fatal occupational injuries and drug use but, at the same time, they do not rule out the possibility of there being such an association.
Few studies have carefully examined the relationship between non- fatal occupational injuries and drug usage. Two of them focused on postal workers in the United States. At Boston, 2,537 new postal employees were followed longitudinally  . In that prospective study, the employees were tested for marijuana, cocaine, opiates, phencyclidine (PCP), barbiturates and amphetamines in a double-blinded fashion; that is, neither the postal service nor the employees knew the results of the drug test. They were then followed up for an average of one year. Those whose urine tests were positive for marijuana were found to have a relative risk of injuries of occupational injuries of 1.85 per cent and a relative risk of accidents of 1.55 per cent. Those who were positive for cocaine were found to have a relative risk for occupational injuries of 1.85 per cent and a relative risk of accidents of 1.59 per cent. (All these results were statistically significant except the last, which had a 95 per cent confidence interval of 0.95, 2.67.)
In addition, the United States Postal Service carried out a multi-site study modelled on the Boston study  . , The multi-site study was a blind prospective study of 4,396 new postal employees in 21 sites nation- wide. Although the authors do not present details of their analysis, no increased risk for injuries or accidents was found. In an update of that longitudinal study  , no evidence was found of an association between injuries and accidents and marijuana and cocaine use. The conflicting results of the multi-site and Boston studies could both be consistent with a weak. association between injuries and accidents and cocaine and marijuana use in the postal service.
Finally, in a recent study  , the authors examined the relation- ship between drug testing and reportable accident and injury rates at 48 facilities in Wisconsin, United States of America, from 1984 to 1988. Ordinary least square regression showed that the 12 facilities that implemented drug- testing programmes during this period did not experience a reduction in those rates compared with the 36 facilities without drug-testing programme. A pooled time series regression analysis among the 12 facilities with drug testing suggested that post-accident testing was related to a decrease in accident and injury rates, but that was based on a sample of just three facilities. Reasonable cause testing was not related to any significant change in accident and injury rates, but, again, the result was based on only three facilities.
A clearer view of the association between illicit drugs and alcohol and fatal injuries in the high-risk profession of heavy-truck driving can be obtained by comparing two separate studies. For a year, beginning 1 October 1987, the National Transportation Safety Board carried out a careful study of all accidents involving heavy trucks that were fatal to the drivers in eight states chosen as a representative sample of all trucking operations in the United States  . Toxicological testing was obtained for 168 (91 per cent) of the 185 drivers in the study.
Another careful study 1531 yielded estimates of drug and alcohol use among representative non-injured drivers of heavy trucks. That study was carried out in December 1986 at a truck-weighing station at Brownsville, Tennessee. Of the 349 representative drivers that were asked to participate in a health- related study, 317 (88 per cent) consented and were paid US$ 30 each. They each provided anonymous urine and blood samples for -toxicological analysis.
Table 3 shows the drug and alcohol use among the fatally injured drivers and uninjured controls. Alcohol was detected in one eighth of the fatally injured drivers, compared with 1 per cent of the controls. The contrast is more marked considering that the average BAC level among the fatally injured drivers was 0.15 per cent whereas the average'. BAC level among the controls was 0.02 per cent. The data suggest that alcohol impairment plays a major role in one eighth of accidents involving heavy trucks that are fatal to the drivers. For marijuana, the proportion of fatally injured drivers that tested positive (12.8 per cent) was similar to that of uninjured controls that tested positive (14.8 per cent). The data suggest that there is no association between marijuana-positive urine samples and crashes involving heavy trucks that are fatal to the drivers. For cocaine and amphetamines, about 8 per cent of fatally injured drivers were positive as opposed to 2.2 per cent of the uninjured controls. In reviewing the data, it must be kept in mind that it was not possible to control for :demographic differences between the two populations. Moreover, the controls were examined one year before the fatally injured drivers, so that it is possible that the results can be partially explained by temporal-or geographical variations.
The following urine drug-screening results obtained by the Federal Aviation Administration in 1991  illustrate to some extent the role of drug use in non -fatal accidents in that industry:
|Employees who tested positive|
|Number of employees tested||Number||Share (percentage)|
The share of drug-positive urine samples obtained by random testing pearly 170,000 aviation employees was compared with the share of drug- positive urine samples obtained from employees tested after accidents. The overwhelming majority of the post-accident tests did not involve fatalities. In both cases, about 0.75 per cent of the urine samples were
Table 3. Drug and alcohol use among fatally injured drivers of heavy trucks and uninjured controls
|Fatally injured drivers|
|Drivers who tested positive|
|Drivers who tested positive|
|Substance||Number of Completed tests||Number||Share (percentage)||Number of Completed tests||Number||Share (percentage)|
Source:A. K Lund and others, "Drug use by tractor trailer drivers", Journal of Forensic Sciences (Philadelphia), vol. 33, No. 3 (1988), pp. 648-661.
positive. Again, the results must be viewed with caution because of lack of adjustment for demographic differences between the two groups and for the nature of the job assignments of the various employees. Thus, the data neither provide evidence of an association between drug use and accidents in the aviation industry nor exclude the possibility of such an association.
In summary, a review of the data suggests that alcohol impairment is present in about 10 per cent of occupational fatalities. The evidence is especially strong among drivers of heavy trucks. Drivers with a clinical history of alcohol abuse may also have a slightly elevated relative risk of non -fatal occupational injuries compared with those without such a history. However, the data here are not entirely consistent. The relationship between fatal injuries and illicit drug use is more difficult to define. Medical examiner studies are not as useful because drug-positive urine is not necessarily associated with impairment. In the heavy trucking industry, the evidence suggests that drivers with marijuana-positive urine are not at increased risk of fatal injuries while those with urine that tests positive for amphetamines and cocaine may be. There is little evidence of an association between drug use and non-fatal injuries. The association has been studied in only one industry: a United States Postal Service study showed a weakly elevated relative risk for injury among cocaine- and marijuana-positive employees, but a second study showed no such increase. A review of Federal Aviation Administration data suggests that there is no association between accidents and drug-positive urine. However, those data are subject to methodological concerns.
Although safety concerns have loomed large in discussions of work- place drug screening, it has also been argued that drug testing could increase productivity  . It has been suggested that this could be accomplished by reducing absenteeism, containing medical costs, restricting disciplinary procedures,- and containing EAP costs. Both the Boston and multi-site postal studies showed an association between increased rates of absenteeism and positive pre-employment drug screens.. The initial results of the Boston postal study after 13 months of follow-up showed that, compared with persons with drug-negative urine, the relative risk of persons with marijuana-positive urine being in the higher absenteeism groups was 1.56 per cent and that the relative risk of those with cocaine-positive urine being in the higher absenteeism groups was 2.37 per cent  . The mean absenteeism rates were 7.1 per cent for the marijuana-positives, 9.8 per cent for the cocaine-positives and 4.0 per cent for the negatives. After 24 months of follow-up, the relative risk of being in the high absenteeism groups had decreased to 1.31 for persons with marijuana-positive urine and was no longer statistically significant; for persons with cocaine-positive urine, the relative risk of being in the high absenteeism groups had increased to 2.65  . After two years, the mean absence rates were 9.1 per cent for persons with marijuana- positive urine, 13.8 per cent for those with cocaine-positive urine and 6.0 per cent for those with drug-negative urine. The multi-site postal study did not distinguish between persons with marijuana-positive and those with cocaine-positive urine. It did find, however, after 8.2 months of follow-up, that those with drug-positive urine averaged 4.35 per cent absenteeism whereas those with drug - negative urine averaged 3.0 per cent absenteeism  . After3.3 years of follow-up, the multi-site postal study found that persons with drug-positive urine averaged 11.39 per cent absenteeism whereas those with drug-negative urine averaged 6.85 per cent absenteeism  .
Both postal studies also showed increased risks of disciplinary action associated with positive pre-employment drug screens. In the Boston study, persons with marijuana-positive urine had a 1.48 per cent relative risk of formal discipline compared with those with drug-negative urine  ; and those with cocaine-positive urine had a 1.72 per cent relative risk compared with those with drug-negative urine. In the multi-site study, the odds ratio for discipline among persons with marijuana- positive urine was 1.87; among those with cocaine-positive urine it was 5.52  . The multi-site postal study provides further insight into the nature of these disciplinary infractions: of the 149 drug-positive subjects who had been disciplined, 118 had been disciplined for poor attendance. Surprisingly, those who tested positive for drugs did not have an increased risk of being disciplined for poor work performance  .
There are limited and conflicting data concerning medical benefits used by employees who tested positive for illicit drugs. The multi-site postal study  analysed medical claims only for the 19 per cent of the study group who subscribed to Blue Cross/Blue Shield Insurance. It was found that the median dollar amount of claims of persons with drug- positive urine (US$ 486.65) was 83 per cent higher than that of persons with drug-negative urine (US$ 265.81). However, no data were provided for 80 per cent of the sample. In a study conducted at a power and light company in Utah  , employees who had tested positive in a random drug test were compared with a random sample of company employees. Annual medical expenditure per drug-positive employee was US$ 504.00 whereas annual medical expenditure in the control group averaged US$ 719.00. However, the study included only 12 employees with drug-positive urine.
In addition, the multi-site postal study  suggested that persons who had tested drug-positive in pre-employment drug screens were more likely to be referred to EAP in the course of employment. For those with marijuana-positive urine, the odds ratio was 1.91 per cent; for those with cocaine-positive urine, it was 6.27 per cent. Twenty-nine (10.7 per cent) of the 270 persons with marijuana-positive urine were eventually referred to EAP. Twenty-six (28.3 per cent) of those with cocaine-positive urine were eventually referred to EAP. Most of the overall referrals (54 per cent) were for alcohol-related problems.
Finally, in three recent, studies  ,  ,  the relation between productivity, as measured by wages, and drug use among young adults have been examined. All have analysed data from the National Longitudinal Survey of Youth using different methods of analysis. Kaestner  , using a two -stage least squares analysis, found that use of marijuana and cocaine was associated -with higher wages, suggesting increased productivity. Gill  found that, after allowing for self-selection effects (the possibility that unobservable factors simultaneously influence both wages and the decision to use drugs), drug users had higher wages than non-users. Finally, Register  found that long-term or on-the-job marijuana use was associated with decreased wages, but that the net productivity for all marijuana users together was increased. No association between cocaine use and productivity was found. Taken together, the studies raised serious doubts about previous assertions about drug use being a cause of decreased productivity in the United States  .
In summary, there is good evidence, at least in a single industry, that persons who test -positive in pre-employment drug screens will have higher rates of absenteeism, which will lead to higher rates of disciplinary action. The evidence on the association between positive results in drug screens and medical costs is weak and conflicting.
Advocates of workplace drug testing often cite the global costs of drug and alcohol abuse to society. They talk about the large number of substance abusers in the workplace and conclude by advocating workplace drug and alcohol testing as a solution to the problem. However, this quick transition from a global problem to a specific solution skips over several important steps. First, it must be shown that the specific solution works, that is, that workplace drug testing can be expected to decrease advers employment outcomes such as injuries and absenteeism. Sections V and VI present reviews of the literature on the subject. Secondly, it must be shown that the specific intervention is cost-effective, that is, that it achieves favourable outcomes without incurring disproportionate costs. The present section deals with the cost-benefit data.
In 1986, in an editorial in the Journal of the American Medical Association, it was pointed out that a great deal of pre-employment drug screening was being done without one proper cost- benefit analysis of this process in any peer review journal"  . Since that time, two cost-benefit analyses of pre-employment drug screening have appeared in the literature  ,  . Before presenting a review of those analyses, some of the requisites for a convincing cost-benefit analysis should be discussed  ,  . A good cost-benefit analysis must include the following-
A clear and accurate description of the intervention to be evaluated;
Evidence of the effectiveness of the intervention;
A fair and comprehensive assessment of the costs and consequences of the programme;
Adjustment of the costs and consequences according to when they occur (discounting);
A sensitivity analysis to evaluate the effect of changing the underlying assumptions.
The first cost-benefit analysis of pre-employment drug screening was carried out by the United States Postal Service based upon its multi-site study  . Assuming that the United States Postal Service would need to test approximately 180,000 Applicants in order to hire 61,588 new employees per year and assuming that 9 per cent of them would test positive in drug screens, it was concluded that for each yearly cohort of new employees, the postal service would save nearly US$ 52.8 million over the 10-year average postal career. Comparing this analysis with the above-described reveals some problems. First, the analysis assumes that the increased cost of absenteeism will continue unchanged for 10 years even though the study is based on an average follow-up of only 1.3 years. It would have been more appropriate to present the cost and benefits after a single year of employment. Secondly, the. study does not discount the value of future savings; although economists differ as to the appropriate rate of discounting, almost all would agree that some discount percentage should be included in cost-benefit calculations  . Thirdly, the analysis assumes the cost of testing is only the laboratory fee of US$ 11, ignoring the costs of collection, storage and transportation of the urine and of seeking replacements for applicants screened out by the tests. A recent report by the government accounting office  suggests that the average cost per urine drug screening by the United States Government is US$ 73.46. The report also acknowledges that this estimate is low because it includes only the direct costs; it does not include the costs associated with delayed hiring, and recruiting replacement workers for those screened out. Finally, the multi-site study provided no sensitivity analysis to show how changes in its underlying assumptions would affect the cost-benefit analysis. These assumptions are somewhat subjective and could vary significantly. A reanalysis illustrates the importance of sensitivity analysis by repeating the multi-site cost-benefit analysis limited to the first year of employment and using more reasonable assumptions  . Under these new assumptions, the costs of the programme, US$ 8.87 million, exceed the benefits of the programme, US$ 5.41 million, by US$ 3.46 million.
A second cost- benefit analysis based upon the Boston postal service study  found that drug testing would save the Boston postal service US$ 163 per newly hired employee. But more important, the study found that these results were sensitive to the assumptions used in the calculations. The most important assumption was the prevalence of drugs in the populations screened. For prevalences under 1 per cent, the programme would lose money. For prevalences above 10 per cent, the programme would save money. For prevalences between 1 and 10 per cent, the exact nature of the other cost assumptions determined whether the programme would save or lose money. it is likely that the prevalence of drugs in the population screened would be the driving force in other cost-benefit analyses of pre- employment drug screening. When drug use is rare, many applicants must be screened to identify a drug user. Thus, unless the costs associated with a single person's drug use are high, the programme will lose money. Since these costs, as well as the prevalence of drug use, are likely to vary from industry to industry and from company to company, it is important that a careful cost-benefit analysis be carried out before instituting drug screening in any specific situation. Moreover, as shown in section I above, the prevalence of drug use can change substantially over the years. Thus, it would be prudent for a company to repeat its cost-benefit analysis periodically.
Finally, the costs and benefits described above are only those accruing to the single enterprise. They do not include external costs  . Thus, the social costs of the invasion of privacy to obtain the urine, of the unemployment insurance paid to applicants screening positive, and of the damaged reputations of those falsely identified as positive did not enter into those calculations.
Over the last six years, there has been a great increase in the amount of workplace drug testing in the United States, in large part in response to federal regulation. As described above, an entire industry has grown up around these regulations. Thus, there are strong economic interests that can be expected to continue to advocate expansion of workplace drug testing, both in the United States and elsewhere. Recent data, however, suggest that drug usage is beginning to decline. Should this continue, public support for workplace drug testing might begin to weaken. Until recently, most workplace drug screening had been focused on illicit drugs. However, the passage of the Omnibus Transportation Employee Testing Act in 1991 in the United States and the recent publication of proposed regulations by the United States Department of Transportation suggest that alcohol testing in the workplace is likely to increase in that country in the years ahead.
As described in sections V and VI above, the empirical basis for workplace drug screening is weak. More importantly, no strong relation- ship between drugs and alcohol and occupational injuries has been established. Two types of studies would be especially useful in addressing these issues. First, case-control studies looking at drug and alcohol use among injured employees and non-injured controls could well shed further light on these questions. In the United States transportation industry, hundreds of thousands of random and post-accident drug screens are being performed. If this was done while collecting the appropriate demographic data in a centralized fashion, it would facilitate an examination of the association between drug and alcohol use and accidents in that industry. Secondly, prospective longitudinal studies following cohorts of workers after drug testing would be useful. The two postal studies  ,  mentioned above show that the efficacy of pre- employment drug screens is less than previously supposed. It would be valuable to see if those results applied to other industries. This type of study is especially important because it forms the basis for cost-benefit analyses of drug screening.
For the policy maker, the present review of the literature suggests three lessons:
Humility in the face of sparse empirical data. Relatively little is known about the role of alcohol and drugs in the workplace. In making decisions, policy makers should avail themselves of findings from existing studies, paying special attention to the methodologies used;
Attention to variation among different industries: The contradictory findings in different studies suggest that substance abuse may well play different roles in different occupational and cultural settings. Caution should be exercised in transposing results from one setting to another;
The importance of the prevalence of drug use in cost-benefit analyses: Any economic analysis of workplace drug screening is likely to be greatly influenced by the prevalence of drug use in the population screened.
A.M Nicholi, "The nontherapeutic use of psychoactive drugs: a modern epidemic", New England Journal of Medicine (Waltham, Massachussetts, vol. 308, 1983, pp. 925-933.02
D. L. Ackerman, "History of drug testing", Drug Testing: Issues and Options, R. H. Coombs and L. J. West, eds. (New York, Oxford University Press, 1991).03
National Institute an Drug Abuse, National Household Survey on Drug Abuse (Rockville, Maryland, National Institute on Drug Abuse, 1991).04
L. D. Johnston, P. M. O'Malley and J. G. Bachman Drug Use among American High School Seniors, College Students and Young Adults, 1975-1990: Vol 1; High School Seniors (Rockville, Maryland, National Institute on Drug Abuse, 1991).05
M. R. Burt and others, Worldwide Highlights from the Worldwide Survey of Nonmedical Drug Use and Alcohol Use among Military Personnel ( Bethesda, Maryland, Burt Associates, 1980).06
D.Catlin, A Guide to Urine Testing for Drugs of Abuse (Washington, D.C., Special Action Office' for Drug Abuse Prevention 1973).07
United States of America, President's Commission on Organized Crime, Report to the President-and Attorney General: ' America's Habit; Drug Abuse, Drug Trafficking, and Organized Crime (Washington, D.C., United States Government Printing Office, 1986).08
R. Reagan, 'Executive order 12564", Federal Register,vol. 51, No. 180 (17 September 1986).09
J. M. Walsh and J. G. Trumble, "The politics of drug testing', Drug Testing: Issues and Options, R. H. Coombs and L. J. West, eds. (New York, Oxford University Press, 1991).10
United States General Accounting Office, Employee Drug Testing: Estimating Costs to Test All Executive Branch Employees and New Hires; Report to the Chairman, Subcommittee on the Civil Service, Committee on Post Office and Civil Service, House of Representatives (Washington, D.C., United States General Accounting Office,. 1992).11
C. Skrzycki, "Drug testing industry shows its wear", Washington Post, 17 October 1990.12
L. Zimmer and J. Jacobs, "The business of drug testing technological innovation and social control", unpublished manuscript, 1 February 1992.13
M. Freudenheim, "Booming business: drug use tests", New York Times, 3 January 1990.14
"1992 business retrospective", Medical Review Officer's Alert, vol. 4, No. 11 (1993).15
A. H. Malcolm "Affluent addicts, road back begins getting past denial", New York Times, 2 October 1989.16
P. B. Bensinger, "Drugs in the workplace", Psychiatric Letters, vol. 4, No. 8 (1986), pp. 39-44.17
"Establishing a drug free workplace", Federal Personnel Manual FPM Letter 792.16 (Washington, D.C., United States Office of Personnel-Management, 1986).18
P. B.,Bensinger, 'Drugs. in the workplace", Harvard Business Review (Boston), vol. 60, 1982, pp. 48-53.19
J. D. Quayle, "American productivity: the devastating effect of alcoholism and- drug abuse", American Psychologist (Washington, D.C.), vol. 38, April 1983, pp. 454-458.20
United States of America, Office of National Drug Control Policy, Building a Drug Free Workforce (Washington, D.C., 1990). :,21
United States General Accounting Office, Employee Drug Testing: Information on Private Sector Programs (Washington, D.C., United States Government Printing Office, March 1988).22
H. V. Hayghe, "Antidrug programs in the workplace: are they here to stay?", Monthly Labor Review (Washington, D.C.), April 1991, pp. 26-29.23
United States of America, Bureau of Labor Statistics, Survey of Employer Antidrug Programs, Report No. 760 (Washington, D.C., Bureau of Labor Statistics, 1989).24
International Labour Office, Washington Tripartite Symposium on Drug and Alcohol Abuse Prevention and Assistance Programmes at the Workplace: Proceedings (Geneva, International Labour Office, 1991).25
H. H. Axel, Corporate Experiences with Drug Testing Programs, Conference Board Report No. 941 (New York, Conference Board, 1990).26
A. Hoffman, Steal this Urine Test: Fighting Drug Hysteria in America (New York, Penguin Books, 1987).27
P. J. Mulloy, "Winning the war on drugs in the military", Drug Testing: Issues and Options, R. H. Coombs and L. J. West, eds. (New York, Oxford University Press, 1991).28
R. M. Bray and others, Worldwide Survey of Alcohol and Nonmedical Drug Use among Military Personnel (Research Triangle Park, North Carolina, Research Triangle Institute, 1983).29
R. M. Bray and others, 1985 Worldwide Survey of Alcohol and Nonmedical Drug Use among Military Personnel (Research Triangle Park, North Carolina, Research Triangle Institute, 1986).30
R. M. Bray and others, 1988 Worldwide Survey of Substance Abuse and Health Behaviors among Military Personnel (Research Triangle Park, North Carolina, Research Triangle Institute, 1989).31
R. M. Bray, M. E. Marsden and S. C. Weeless, MilitarylCivilian Comparisons of Alcohol, Drug and Tobacco Use (Research Triangle Park, North Carolina, Research Triangle Institute, 1990).32
R. L. Berkelman and others, "Fatal injuries and alcohol', American Journal of Preventive Medicine (New York), vol. 1, No. 6 (1985), pp. 21-28.33
S. P. Baker and others, Injury Fact Book, 2nd ed. (New York, Oxford University Press, 1992).34
G. S. Smith and J. F. Kraus, "Alcohol in residential, recreational and the occupational injuries: a review of the epidemiologic evidence", Annual Review of Public Health (Palo Alto, California), vol. 9, 1988, pp. 99-122.35
H. Wechsler and others, "Alcohol level and home accidents", Public Health Reports (Rockville, Maryland), vol. 84, No. 12 (1969), pp. 1043-1050.36
L. Papoz and others, "Biological markers of alcohol intake among 4,796 subjects injured in accidents", British Medical Journal (London), vol. 292, 1986, pp. 1234-1237.37
R. B. Trent, "Emergency room evidence on the role of alcohol intoxication in injury at work in the U.S.", Safety Science (Amsterdam), vol. 14, 1991, pp. 241-252.38
P. B. Beaumont and J. Hyman, "The work performance indicators of problem drinking: some British evidence", Journal of Occupational Behaviour (Chichester, Sussex), vol. 8, 1987, pp. 55-62.39
M. A. Maxwell and others, "A study of absenteeism, accidents, and sickness payments in problem drinkers in one industry', Quarterly Journal of Studies on Alcohol, vol. 20, 1959, pp. 302-312.40
H. M. Trice, "Alcoholic employees: a comparison of psychotic neurotics and 'normal' personnel", Journal of Occupational Medicine (Baltimore), vol. 7, No. 1 (1965), pp. 94-99.41
G. R. Webb and others, "The relationship between high-risk and problem drinking and the occurrence of work injuries and related absences", unpublished paper presented at the 18th Annual Alcohol Epidemiology Symposium, Toronto, 1-5 June 1990.42
R. B. Hertz and E. A. Emmett, "Risk factors for occupational hand injury", Journal of Occupational Medicine (Baltimore), vol. 28, No. 1 (1986), pp. 36-41.43
R. W. Hingson, R. L. Lederman and D. C. Walsh, "Employee drinking patterns and accidental injury: a study of four New England states", Journal of Studies on Alcohol (Piscataway, New Jersey), vol. 46, No. 4 (1985), pp. 298-303.44
A. W. Moll van Charante and G. H. Mulder, "Perceptual acuity and the risk of industrial accidents", American Journal of Epidemiology (Baltimore), vol. 131, No. 4 (1990), pp. 652-663.45
R. J. Lewis and S. P. Cooper "Alcohol, other Drugs and fatal work related injuries", J. of Occ. Medicine (Baltimore) vol.31, No.1(1989), pp. 23-28.46
D. K. Parkinson, W. F. Gauss and E. S. Perper, "Traumatic workplace deaths in Allegheny County, Pennsylvania, 1983 and 1984", Journal of Occupational Medicine (Baltimore), vol. 28, No. 2 (1986), pp. 100-102.47
E. C. Alleyne, P. Stewart and R. Copes, 'Alcohol and other drug use in occupational fatalities", Journal of Occupational, Medicine (Baltimore), vol. 33, No. 4 (1991), pp. 496-500.48
C. Zwerling, J. Ryan and E. J. Orav, "The efficacy of pre-employment drug screening for marijuana, cocaine in predicting employment outcome", Journal of the American Medical Association (Chicago), vol. 264, No. 20 (1990), pp. 2639-2643.49
J. Normand and S. D. Salyards, Applicant Drug Testing: Update of a Longitudinal Study (Washington, D.C., United States Postal Service, 1990).50
J. Normand, S. D. Salyards and J. J. Mahoney, "An evaluation of pre-employment drug testing", Journal of Applied Psychology (Washington, D.C.), vol. 75, No. 6 (1990), pp. 629-639.51
D. M.. Feinau and S. J. Havlovic, "Drug testing as a strategy to reduce occupational accidents: a longitudinal analysis, Journal of Safety Research (Tarrytown, New York), vol. 24, No. 1 (1993), pp. 1-7.52
National Transportation Safety Board, Safety Study: Fatigue, Alcohol, Other Drugs, and Medical Factors Fatal to the Driver in Heavy Truck Crashes (vol. 1) (Springfield, Virginia, National Transportation Safety Board, 1990).53
A. K. Lund and others, "Drug use by tractor trailer drivers", Journal of Forensic Sciences (Philadelphia), vol. 33, No. 3 (1988), pp.648-661.54
"FAA 1991 aviation drug test results", MRO Alert, Vol. 3, No. 7, pp. 7-8.55
J. Ryan, C. Zwerling and M. Jones, "The effectiveness of pre-employment drug screening in the -prediction of employment outcome", Journal of Occupational Medicine (Baltimore), vol. 34, No. 11 (1992), pp. 1057-1063.56
S. D. Salyards, Utility Analysis of Pre-employment Drug Screening as a Selection Device (Washington, D.C., United States Postal Service, 1991).57
S. D. Salyards, Pre-employment Drug Testing: Associations with -EAP, Disciplinary, and Medical Claims Information (Washington, D.C., United States Postal Service, 1992).58
D. J. Crouch and others, "A critical evaluation of the Utah Power and Light Company's substance abuse management program: absenteeism, accidents and costs", Drugs in the Workplace: Research and Evaluation Data, National Institute on Drug Abuse Research Monograph Series, No. 91 (Rockville, Maryland, National Institute on Drug Abuse, 1989), pp. 169-193.59
R. Kaestner, "The effect of illicit drug use on the wages of young adults", Journal of Labor Economics (Chicago), vol. 9, No. 4 (1991), pp. 381-412.60
A. M. Gill and R. J. Michaels, "Does drug use lower wages?", Industrial and Labor Relations Review (Ithaca, New York), vol. 45, No. 3 (1992), pp. 419-434.61
C. A. Register and D. R. Williams, 'Labor market effects of marijuana and cocaine use among young men", Industrial and Labor Relations Review (Ithaca, New York), vol. 45, No. 3 (1992), pp. 435-451.62
G. Lundberg, "Mandatory unindicated urine and drug screening: still chemical McCarthyism", Journal of the American Medical Association (Chicago), vol. 256, 1986, pp. 3003-3005.63
C. Zwerling, J. Ryan and E. J. Orav, "Cost and benefits of pre-employment drug screening", Journal of the American Medical Association (Chicago), vol. 267, No. 1 (1992), pp. 91-93.64
K. E. Warner and B. R. Luce, Cost-Benefit and Cost-Effectiveness Analysis and Health Care: Principles, Practice, and Potential (Ann Arbor, Michigan, Health Administration Press, 1982).65
M. F. Drummond, G. L. Stoddart and G. W. Torrance, Methods for the Economic Evaluation of Health Care Programmes (Oxford, Oxford University Press, 1987).66
C. Zwerling and J. Ryan, "Pre-employment drug screening: the epidemiologic issues", Journal of Occupational Medicine (Baltimore), vol. 34, No. 6 (1992), pp. 595-599.67
T. R. Sexton, "On the wisdom of mandatory drug testing", Journal of Policy Analysis and Management (New York), vol. 7, 1988, pp. 542-547.