The evidence for drug testing in Alberta workplaces

Last week, I appeared in a CBC call in show about the implementation of random drug testing among Suncor employees as part of the Drug and Alcohol Risk Reduction Pilot Project (DARRPP) in Alberta’s energy and construction sectors. 

The basic argument made by DARRPP for random testing is a safety one: violating employee privacy rights is warranted because random testing will make workplaces safer. Preparing for the interview, I read a number of studies that (to my surprise) indicate (1) random drug testing (as distinct from other kinds of drug testing) is not generally associated with a reduction in injuries and (2) drug use (as distinct from alcohol use) on the job or off is not associated with an increase injuries. This seems to undercut the safety argument almost entirely, which was the message I delivered on the show.

Since then, I’ve had a fair number of emails and phone calls with people asking about the evidence. This is a complicated request: (1) there are many studies, and (2) designing and executing a study often entails significant tradeoffs so sometimes findings require some expertise to understand. There is also the matter that it is not possible to prove a negative (i.e., that drugs never cause injury rates to go up and that random drug testing never causes injury rates to go down). Rather, we typically require those who assert a relationship (i.e., DARRP) to show evidence that it exists, so let’s start there.

The DARPP website cites two studies. The first indicates random alcohol testing of drivers in the US transport industry reduced fatal crashes by 23% among large trucks. DAARPP doesn’t tell us which study that is and I didn’t run across it. But the question at hand is whether illicit drug testing (not alcohol-testing) reduces injuries, so this study has no real bearing on the issue.

The DARRPP website cites another US study that shows mandatory alcohol and drug testing reduces positive tests by 50%. Again, I couldn’t find the study but let’s assume this is correct. The question is whether a reduction in positive test results is associated with any difference in injury outcomes? On this, we have no information. As drug tests measure prior use (but not impairment), positive tests might have little to do with workplace injury rates. Or the reduction might indicate gaming behaviour (“Psst. Buddy, can I borrow some urine?”). If so, a reduction in positive tests is largely meaningless.

DARRPP also kindly sent me a press release about a 2001 study they thought I should look at: Gerber, J. and Yacoubian, G. (2001). Evaluation of drug testing in the workplace: A study of the construction industry. Journal of construction engineering and management. 127(6): 438-444. While this study found that firms that implemented drug testing saw a 51% reduction in injury incident rates, there was no indication that these results stemmed from random drug testing (as opposed to pre-employment, reasonable cause and post-incident testing—none of which are the subject at hand). This study also has a small sample and poor response rate. Overall, this study really tells us nothing useful about the impact of random drug testing on workplace injuries.

I looked back through my notes and browser history and started to pull together what I reviewed for the CBC interview. Rather than attempt a half-assed systematic review, what I’ve done is selected those articles which present such reviews themselves. This should allow anyone who is keen to do so, the ability to go back and review each article on their own.

Institute of Medicine (1994) Under the Influence? Drugs and the American Work Force. Washington: National Academy Press. This book-length review is one of the better reads, despite the relative age. It found no good studies on the effectiveness of random drug testing (pp. 226-227). The effect of a drug in a laboratory varies and can be either positive and negative (depending on dose and testing). It is difficult to transfer this knowledge to determine the impact of drug use in the workplace. There was too little data to indicate that drug use was associated with increase injury and it is unclear if there is a causal relationship (i.e., some other factor may be at play).

A slightly more recent review is Macdonald, S. (1997). Work-place alcohol and other drug testing: A review of the scientific evidence. Drug and alcohol review. 16(3): 251-259. This article also indicates that there was little evidence for a linkage between drug use and workplace injury and no evidence drug testing significantly reduces workplace injury.

A study by three Alberta researchers for the government of Alberta is Beach, J., Ford, G. and Cherry, N. (2006). Final report: A literature review of the role of alcohol and drugs in contributing to work-related injury. Edmonton: University of Alberta, Department of Public Health Sciences.  On the issue of “is personal drug use associated with workplace injury”, this study is worth quoting at some length:

Overall this literature is difficult to evaluate, and probably inadequate to reach a definitive conclusion. If drug use is associated with work-related injury it appears to be more apparent among males, possibly more likely to occur with cocaine use, and possibly only occurs in certain situations such as when driving or in other ‘high risk’ work. Further, while a slim majority of studies appear to show an association of some sort, even within these studies this does not necessarily constitute causation. (pp. 13-14)

The authors found the evidence also too slim to know if drug use at work or immediately before a workplace injury increased the risk of injury. On the question of whether drug testing of any type affected workplace injuries, the studies were too weak to say anything definitive.

Moving onto actual studies, one of the larger studies is Hoffman, J. and Larison, C. (1999) Drug use, workplace accidents and employee turnover. Journal of drug issues. 29(2): 341–364. This study used nationally representative US data to find that drug use in general is not associated with a higher risk of workplace injury, with little indication of any such relationship in higher risk occupations such as construction, transportation and skilled labour.

An interesting study is Macdonald, S. (1995). The role of drugs in workplace injuries: Is drug testing appropriate? Journal of drug issues, 25(4): 703–23. This study uses self-reports to conclude illicit drug use is not a major cause of job injuries. This study seeks to control for (and understand) the impact of other variables on the relationship between drugs use and injuries—a failing of many other studies.

One of the studies that is sometimes cited as indicating random testing is effective is Miller, T., Zaloshnja, E. and Spicer, R. (2007). Effectiveness and benefit-cost of peer-based workplace substance abuse prevention coupled with random testing. Accident analysis and prevention. 39(3): 565-573. A careful read shows that the interactive effect of peer-based interventions and random drug testing was associated with a one-third drop in workplace injuries. The authors note that is not possible to disentangle the effect of the two interventions (peer intervention and random drug testing). The time series data shows the vast majority of the reduction followed the peer intervention; injury data is effectively flat after random testing starts. It may be the threat of testing may have contributed to willingness to participate in peer intervention, but we don’t know that.

Another construction-specific survey is Olbina, S., Hinzo, J. and Arduengo, C. (2011). Drug testing practices in the US construction industry in 2008. Construction management and economics. 29(10): 1043-1057. This study reports a statistically significant relationship between positive random drug tests and higher injury rates, suggesting drug use can increase the change of injury. There were a number of limitations noted in the study: 14% response rate, with only two-third using random testing, small sample size (53 firms), and a bias towards larger firms. An average of 2.23% of random tests were positive, suggesting illicit drug use is minimal.

The question I’m left is this: if there is no evidence that random drug testing reduces workplace injury rates, why would employers spend money on it?

There are several possible answers. They may have been sold a bill a good by companies that do the random testing. The notion that testing will reduce injuries accords with most people’s (incorrect) common-sense reaction that testing should reduce injury rates so that would be an easy enough sell.

Or there may be some other agenda at play. Testing may dissuade drug users from working at places with testing. If drug-use is associated with other undesirable behaviours (e.g., propensity to quit), then perhaps random testing is really about winnowing out these workers.

Testing also implicitly blames workers for workplace injuries, taking the heat off employers for organizing work unsafely and off the government for doing a poor job of enforcing the rules. And it makes it look like employers are doing something about the problem of workplace injuries. There is also a strong moral dimension to the issue of drug testing that can sometimes cloud people’s judgment.

Or perhaps these companies just believe (contrary to the evidence) that testing works. They are certainly entitled to believe that. Except that testing is a huge invasion of workers’ privacy and is demeaning, so ought to occur only when there is some evidence of its efficacy at increasing safety. Absent such evidence, the case for random testing looks pretty weak.

-- Bob Barnetson