Drivers who test positive for drugs — prescription drugs or illegal substances — often fail to report the fact that they’ve taken them within 24 hours of getting behind the wheel, according to a recent study from the Insurance Institute for Highway Safety (IIHS).

For example, an analysis of two of the most recent national roadside surveys found that fewer than one in five respondents who tested positive for cocaine reported taking it within the past 24 hours in their answers to the questionnaire.

The overall findings from IIHS’ new study underscores the importance of conducting blood or saliva tests to understand the extent of the drug-impaired driving problem, noted IIHS.

In short, self-reported drug use is proving to be an ineffective measure for monitoring trends in drug use and road use.

Since 1973, the national roadside survey of alcohol and drug use by drivers has been conducted five times. In the most recent two surveys, workers asked about drug use too. They also collected saliva and blood samples in return for a small payment.

In the 2007 and 2013-14 surveys, workers collected anonymous data from 300 locations across the continental U.S., gathering both biological and self-reported information from more than 7,000 drivers during each survey. In this new study, researchers looked specifically at the results related to cannabis, opioids, cocaine, antidepressants and benzodiazepines, which include such drugs as Valium and Xanax.

While only 20% of cocaine users reported they got behind the wheel within 24 hours of using, the portion of marijuana-positive drivers who said they’d taken the drug within the past 24 hours increased from 25% to nearly 40% between 2007 and 2013-14. This may be due to broader social acceptance of marijuana following its legalization in several states.

For prescription drugs, the reporting rate was somewhat higher. Over 40% admitted to using opioids and more than 70% to using antidepressants in both surveys. More than 50% of those surveyed in 2013-14 said they use benzodiazepines. These higher rates may reflect lower social stigma associated with these drugs or less fear of legal consequences.

Although roadside surveys are useful, trends like the increasing acceptance of marijuana and the opioid crisis make it ever more important to develop a better understanding of drug-impaired driving, noted IIHS.

Roadway crashes have increased as much as 6% following the start of retail sales of recreational marijuana in Colorado, Nevada, Oregon and Washington, according to studies conducted by IIHS and the Highway Loss Data Institute.