Workplace Safety and Randomized Controlled Trials: Another Weapon of Delay?
A fundamental principle of modern workplace safety laws holds that if scientific evidence suggests the health and safety of the public is at risk, the federal government should step in and take action, even if no conclusive proof has yet been generated. The principle is that the government should adopt public protections based on the “best available evidence” rather than wait indefinitely for “proof” of a hazard while workers suffer harm.
In practice, however, this has meant that industries and interests with an economic stake in the status quo often fight against federal health and safety improvements by trying to invalidate the scientific evidence showing that certain processes or products are dangerous to human health. The most flagrant examples are the money the tobacco industry invested in “scientific studies” questioning the finding that smoking causes cancer and the funding oil and gas companies provided to studies questioning the impact of carbon emissions on the earth’s climate. Other examples include industry-sponsored studies questioning the relationship between chrysotile asbestos and mesothelioma (a type of cancer) and formaldehyde and cancer. The deliberate production of studies designed to undermine the scientific consensus about various health risks has delayed the establishment and enforcement of new standards that could have saved the health and lives of hundreds of thousands of Americans.
Now a new hurdle is emerging. The Office of Management and Budget’s (OMB) recent draft report to Congress on costs and benefits of federal regulations states that randomized controlled trials (RCTs) might be “appropriate and useful” as a “form of advance testing of regulatory alternatives.” This is an alarming development for several reasons.
A randomized controlled trial is a type of epidemiology study, extensively used to test the effectiveness of new drugs seeking Food and Drug Administration approval. One group of people with an illness is given a new drug, and a second group is given the existing treatment (or nothing), and the results are compared over a limited time period, under the careful supervision of medical professionals. Over the past few decades, some social scientists have adapted these methods to try to test the impact of policy interventions. For example, one group of high school dropouts participates in a training program and another demographically similar group does not; employment rates at the end of the training period are compared to determine if the training had the desired effect of boosting employment rates. In these kinds of studies, there are multiple “external” factors that can affect outcomes, so large numbers of participants are required to achieve statistically reliable outcomes.
A number of characteristics of such controlled random experiments make them an inappropriate method for determining occupational or environmental risks. First, as noted above, large numbers of participants are required for statistical validity. To achieve the necessary numbers, one study might need to involve multiple worksites, which would dramatically increase the amount of time required to set up and operate a study and reduce the likelihood that all “control” and “treatment” groups experience the same conditions.
Second, these kinds of studies are extremely costly to conduct. When controlled experiments were used to evaluate job training programs during the 1980s and 1990s, each experiment cost millions of dollars to carry out.
Third, the longer the “experiment” lasts, the less reliable the method becomes because individuals in both control and treatment groups drop out of the study, and it is impossible to know if there is a pattern to the dropouts that would bias the results. Since it often takes decades before workplace health hazards make people sick, random assignment models that last only a year or two cannot accurately identify workplace hazards.
Most importantly, the use of RCTs to measure the effectiveness of workplace safety standards raises moral and legal questions. If there is enough evidence of harm to warrant a multi-year, multi-site, multi-million dollar research project, then surely there is enough evidence to require better working conditions for workers in the facilities. Failing to act is relegating workers in the control group to long-term, and in some cases, irreversible health problems. Some might argue that conducting a five-year study to finally determine that universal standards are needed is worth the sacrifice of the health of people in the control group. Those in the control group would probably disagree.
Despite these clear problems, opponents of workplace standards on repetitive motion disorders, for example, have been pushing RCTs as the gold standard for scientific evidence. Opponents of Occupational Safety and Health Administration (OSHA) ergonomics regulation have argued that repetitive motion injuries do not really exist – despite hundreds of epidemiology studies showing that workers who perform repetitive tasks have significantly higher incidences of tendonitis, carpal tunnel syndrome, and other injuries than those who do not. Having failed to discredit the evidence of a causal link, the industry is now asking for a random assignment experiment showing that regulations that reduce repetitive tasks would reduce workers’ injuries.
From a legal perspective, reliance on RCTs would move agencies away from relying on the “best available evidence” as a basis for acting on public health risks to a higher, much more difficult standard to meet – “best conceivable evidence.” Public health laws now recognize that government should issue protections when the cumulative scientific evidence suggests serious harm to the public, even if there is no “incontrovertible proof of harm.” It was the “best available evidence” standard that allowed federal agencies to caution the public about the hazards of secondhand smoke, even when the tobacco industry protested that there was not enough proof of harm; this standard allowed OSHA to regulate vinyl chloride, even though the plastics industry protested there was not enough proof of harm; and it was this standard that allowed EPA to regulate diesel particulate matter even when manufacturers protested there was not enough proof to do so.
The standards of “best available evidence” and “cumulative scientific evidence” have served the public well for four decades. Promoting the use of random assignment evaluations is just another way for industry to slow or shut down new standard setting by federal agencies. This is not an idea OMB should promote.