Two major studies using the same U.S. crash databases (FARS and NRS) reached contradictory conclusions about marijuana and crash risk, primarily because of inconsistent drug testing practices and database limitations.
Readers trying to make sense of conflicting research on cannabis and driving safety.
Same databases, opposite conclusions: database limitations in drug testing are the primary reason
What the researchers found
The authors examined why two studies using similar databases and approaches produced opposite conclusions about marijuana's contribution to fatal crash risk. Li et al. (2013) found marijuana significantly increased crash risk, while Romano et al. (2014) found no significant increase after controlling for confounders.
The key source of divergence was limitations in the FARS (Fatality Analysis Reporting System) database. Drug testing of fatally injured drivers is not standardized across jurisdictions: different states test for different drugs at different rates, using different methods, and testing rates have changed dramatically over time. These inconsistencies mean that the pool of tested drivers is not representative of all drivers in fatal crashes.
The authors conclude that FARS should not be used to examine trends in drug use prevalence among crash-involved drivers or to obtain precise risk estimates. However, under certain conditions (data from jurisdictions that routinely test, with consistent procedures), it could provide relative risk comparisons.
Why it matters
Policy decisions about cannabis and driving are heavily influenced by epidemiological research. If the major databases used for this research produce contradictory results depending on analytic choices, policymakers need to understand these limitations. This paper serves as a critical methodological caution for all drugged driving research.
The numbers in context
Two studies compared: Li et al. 2013 (found significant crash risk from marijuana) vs. Romano et al. 2014 (found no significant risk). Both used FARS and NRS databases. Testing rates and methods varied widely across jurisdictions and over time.
How the study worked
Methodological comparison of two published case-control studies that used FARS and NRS databases. The authors identified specific analytic choices and database limitations that led to contradictory results.
What this study cannot tell us
This is a methodological critique, not new empirical research. The authors do not provide definitive answers about whether marijuana increases crash risk, instead highlighting why existing studies disagree. Their own prior work (Romano et al. 2014) is one of the two studies being compared, creating potential bias.
How to read the evidence
Moderate evidence: methodological critique illuminating limitations of the U.S. crash database infrastructure.
When this study was published
Published in 2017. Crash database limitations remain an ongoing challenge for drugged driving research.
The bigger picture
The cannabis-and-driving debate is complicated by the fundamental difficulty of measuring impairment. Unlike alcohol, where blood levels correlate reasonably well with impairment, THC blood levels do not reliably indicate whether someone is currently impaired. When the databases used to study the question also have significant limitations, reaching firm conclusions becomes extremely challenging.
Questions still open
- Can U.S. crash databases be improved to consistently test all fatally injured drivers for drugs? What is the best study design to determine marijuana's true contribution to crash risk? Should policy rely on laboratory impairment studies rather than epidemiological crash data?
Common questions
Does marijuana actually increase crash risk or not?
Why is it so hard to study marijuana and driving?
Read the original research
Marijuana and the Risk of Fatal Car Crashes: What Can We Learn from FARS and NRS Data?
The journal of primary prevention, 38(3), 315-328
Citation
Romano, Eduardo; Torres-Saavedra, Pedro; Voas, Robert B; Lacey, John H. (2017). Marijuana and the Risk of Fatal Car Crashes: What Can We Learn from FARS and NRS Data?. The journal of primary prevention, 38(3), 315-328. https://doi.org/10.1007/s10935-017-0478-3