What are the odds of causing a fatal crash by driving within a few hours of using cannabis?
Evidence quality 4.63/5
Eight-dimension review score against the quality rubric . Each dimension scored 1–5.
- D1 Source grounding
- 4/5
- D2 Source authority
- 5/5
- D3 Arithmetic
- 5/5
- D4 Uncertainty
- 4/5
- D5 Scope
- 5/5
- D6 Prose
- 5/5
- D7 Perception honesty
- 4/5
- D8 Caveat completeness
- 5/5
Lifetime probability · lifetime, activity-specific
1 in 53
1.9% lifetime chance
Most people underestimate this.
range 1 in 167 to 1 in 13
● your factors — click this risk ▾ to reveal
≈ As likely as
Perceived
Most regular cannabis users believe driving stoned is meaningfully safer than driving drunk, and a substantial minority believe it is safe or even improves their driving. Self-report surveys in legal-cannabis US states find that roughly 30-50% of past-month cannabis users have driven within two hours of use, and most of those describe themselves as cautious or unaffected. The subjective experience of acute cannabis impairment is qualitatively different from alcohol: drivers typically feel slower, more focused, and self-correctingly conservative — they drive more carefully on the straightaways and miss the high-attention edge cases (lane departures, sudden brake events, peripheral movement) where the per-trip crash risk actually lives.
Rough estimate: most regular cannabis users believe driving stoned is far safer than driving drunk
Source: editorial intuition, not polled
Actual
~4 per 100,000 trips result in a fatal crash for a driver within ~2 hours of cannabis use at ~5 ng/mL blood THC (≈2× the sober-driver rate)
US adult driver within ~2 hours of cannabis use at acute-effect concentrations (≥5 ng/mL whole-blood THC), per-trip crash involvement rate derived from Albrecht 2025 meta-regression applied to NHTSA per-trip baseline
Show derivation
The US population-average per-trip fatal-crash probability for a sober driver is approximately 1 in 50,000 (the baseline used in the driving-at-0.1pct-bac entry, derived from FARS and NHTSA per-trip estimates). Albrecht et al. 2025 (Drug Science, Policy and Law) pooled culpability studies in a dose-response meta-regression and found crash-culpability risk roughly doubles at ~5 ng/mL whole-blood THC and roughly quadruples at ~10 ng/mL; below ~1.5 ng/mL the increase is not distinguishable from baseline. Applying a 2× per-trip multiplier at acute-effect concentrations gives ~1 in 25,000 per cannabis-impaired trip. For a driver who operates within two hours of cannabis use roughly monthly (12 trips/year over 40 years ≈ 480 impaired trips), cumulative probability is 1 − (1 − 1/25000)^480 ≈ 0.019, or roughly 1 in 52. The uncertainty band reflects two main sources: the Rogeberg & Elvik 2016 pooled adjusted OR is only 1.36 (so weekly casual use yields a lower estimate than the Albrecht acute-dose point), while the European DRUID study found OR ≈ 6.6 at ≥5 ng/mL (closer to 0.15% BAC), which would push the high-end estimate to ~1 in 9.
Caveats: The cannabis-driving evidence base has unusually wide spread because the publish…
The cannabis-driving evidence base has unusually wide spread because the published estimates depend heavily on how the impaired population is defined. "THC-positive" includes drivers who used cannabis days earlier and retain detectable but non-impairing residual concentrations; "acute use" means within roughly 1-3 hours of inhalation or 2-4 hours of edible peak. Pooled "any positive" ORs (Rogeberg 1.36, Compton 1.00 adjusted) are dominated by the residual group and underestimate acute-use risk. Dose-response meta-regressions (Albrecht 2025, the DRUID studies) isolate the acute-use signal and find substantially higher per-trip risk at concentrations consistent with recent inhalation. The headline 1 in 52 estimate uses the acute- use Albrecht multiplier at ~5 ng/mL, which corresponds to typical post-inhalation peak concentrations but is conservative for high-dose edibles or concentrate use. The combination with even modest alcohol use is super-additive and not captured in the headline figure. Compared with the 0.10% BAC entry, the per-trip risk multiplier is smaller (~2× vs ~5.5×) but the per-event frequency for regular cannabis consumers can be higher (weekly is more common than weekly drunk driving), so the lifetime totals end up closer than the per-trip comparison would suggest.
Risks at similar odds
Other risks with roughly the same likelihood — useful for calibration.
Driving on sedating meds
What are the odds of causing a fatal crash by driving while on a sedating prescription medication?
Hand-held phone call + driving
What are the odds of a crash from holding a phone to your ear while driving?
Video watching + driving
What are the odds of a crash from watching video on a phone while driving?
Spaceflight (astronaut)
What are the odds of dying as an astronaut on a spaceflight mission?
Cruise ship norovirus
What are the odds of being on a cruise ship voyage that has a norovirus outbreak?
Recently viewed on this device
Stored locally — clear anytime.
Pick challenger
The cleanest recent estimate of acute cannabis-impaired driving risk comes from Albrecht et al.’s 2025 dose-response meta-regression in Drug Science, Policy and Law, which pooled culpability studies that reported blood THC concentrations rather than just THC-positive/negative status. They found a clear dose-response curve: below roughly 1.5 ng/mL whole-blood THC, crash- culpability risk is statistically indistinguishable from baseline; risk doubles around 5 ng/mL and quadruples around 10 ng/mL. The 5 ng/mL point corresponds approximately to the peak concentration produced by typical inhaled cannabis use within the first hour. Pooled “any THC positive” odds ratios — Rogeberg & Elvik’s 2016 adjusted 1.36, the 2015 NHTSA Virginia Beach study’s adjusted 1.00 — appear much lower because the positive group is dominated by drivers with residual THC from prior days who are not currently impaired.
The lifetime framing is similar in shape to the alcohol-impaired case but with a smaller per-trip multiplier and (for regular consumers) a higher per-event frequency. For a driver who operates within two hours of cannabis use roughly monthly — 480 acute-use trips over 40 years — applying the Albrecht 2× per-trip multiplier to the sober per-trip baseline of ~1 in 50,000 gives a cumulative lifetime probability of roughly 1 in 52 of being involved in a fatal crash specifically caused by their cannabis impairment. A weekly cannabis-driving pattern pushes the estimate toward 1 in 13; daily use with high-dose edibles or concentrates pushes it further still. The uncertainty band is wide because the European DRUID studies found OR ≈ 6.6 at ≥5 ng/mL (comparable to 0.15% BAC), which would land the headline closer to 1 in 9 lifetime — a meaningful contrast with the Albrecht meta-regression’s 2×.
Two structural features of cannabis impairment make the lifetime calculation harder than the alcohol case. First, the impairment is short-lived but the detection window is long: detectable THC can persist for days after impairment has resolved, which contaminates every observational study that uses positive/negative status as the exposure variable. Second, the subjective experience does not match the objective deficit. Drivers under acute cannabis influence typically drive more slowly and leave larger following distances, but their reaction times to unexpected events lengthen and lane-keeping degrades — the deficit shows up in the high-attention edge cases, not the routine straightaway. The combination with even modest alcohol is super-additive and not captured in the headline figure: a 0.04% BAC plus acute cannabis use roughly multiplies risks rather than adding them, which is the pattern most regular cannabis-using social drinkers encounter most often.
Claim ledger
Every number below is what each source reported, with the verbatim quote we relied on and how we arrived at our figure. Click any link to verify directly.
-
[1] Albrecht, Hasan, Kekez, Zhou — Drug Science, Policy and Law — Dose-response relationship between blood concentrations of THC and crash culpability risk: An updated meta-regression of culpability studies
Dose-response relationship between blood concentrations of THC and crash culpability risk: An updated meta-regression of culpability studies- Statistic
Crash culpability risk increases with rising whole-blood THC concentration, with an inflection around 1.5-3.0 ng/mL where risk begins to rise above baseline; risk approximately doubles at ~5 ng/mL and approximately quadruples at ~10 ng/mL. Below ~1.5 ng/mL culpability is statistically indistinguishable from baseline (<30% increase).- Excerpt
“"Crash culpability risk increases with increasing THC concentration, with an inflection around 1.5-3.0 ng/ml where risk begins to increase... a doubling of culpability risk around 5 ng/ml and a potential quadrupling of risk around 10 ng/ml." ”
- Source data from
- 2025-03-01
- Accessed
- 2026-05-25 · archived copy
- Calculation
- Albrecht 2025 is the most recent dose-response meta-regression and the cleanest source for translating a blood THC concentration into a per-trip crash-culpability multiplier. The doubling at ~5 ng/mL whole blood is used here as the headline per-trip risk multiplier for "driving within ~2 hours of typical inhaled cannabis use" (which produces peak THC concentrations in roughly that range). The quadrupling at ~10 ng/mL and Rogeberg 2016 pooled 1.36 OR bracket the uncertainty band.
-
[2] Rogeberg, O. & Elvik, R. — Addiction 111(8):1348-1359 — The effects of cannabis intoxication on motor vehicle collision revisited and revised
The effects of cannabis intoxication on motor vehicle collision revisited and revised- Statistic
Pooled odds ratio across 28 estimates from 21 observational studies for acute cannabis use and motor-vehicle collision involvement was 1.36 (95% CI 1.15-1.61); roughly half of earlier higher estimates from the Asbridge 2012 BMJ meta-analysis disappear after correcting for confounding and methodological inconsistencies.- Excerpt
“"Our updated meta-analysis suggests that the increase in crash risk caused by cannabis intoxication is moderate, around 20-30%, much smaller than that of drink-driving and similar in magnitude to that of driving with a blood alcohol concentration between 0.01% and 0.05%." ”
- Source data from
- 2016-08-01
- Accessed
- 2026-05-25 · archived copy
- Calculation
- Rogeberg & Elvik 2016 (DOI 10.1111/add.13347, PMID 26878835) is the most methodologically careful pooled estimate available. It used 28 estimates from 21 observational studies (case-control and culpability designs) and explicitly corrected for the confounding by age, sex, and time-of-day that the Asbridge 2012 BMJ meta-analysis partially missed. The 1.36 odds ratio is the low-end pooled multiplier used in the uncertainty band; it represents the "any cannabis-positive driver" average, dominated by drivers with residual THC from prior days rather than acute peak concentrations. Albrecht's 5 ng/mL doubling is the appropriate reference for acute post-use trips and anchors the headline.
-
[3] Compton, R.P. & Berning, A. — National Highway Traffic Safety Administration — Drug and Alcohol Crash Risk (Research Note, DOT HS 812 117)
Drug and Alcohol Crash Risk (Research Note, DOT HS 812 117)- Statistic
The unadjusted odds ratio for crash involvement among THC-positive drivers in the Virginia Beach case-control study was 1.25; after adjustment for age, sex, race/ethnicity, and alcohol concentration the odds ratio dropped to 1.00 (95% CI 0.77-1.31), indicating no statistically significant elevation. The same study found 0.08% BAC associated with adjusted OR of 3.93 and 0.15% BAC with adjusted OR of 12.04.- Excerpt
“"After adjustment for age, gender, race/ethnicity, and alcohol use, the odds of being a crash-involved driver was not statistically different from that of a control driver who tested positive for THC (OR=1.00, 95% CI 0.77-1.31)." ”
- Source data from
- 2015-02-01
- Accessed
- 2026-05-25 · archived copy
- Calculation
- The 2015 NHTSA Virginia Beach study is the largest US case-control study of drug-positive driving and is frequently cited as evidence that cannabis-impaired driving carries no elevated risk after controlling for confounders. Methodologists (Rogeberg, Albrecht, Compton) note that the null result reflects the dominance of residual-THC-positive drivers in the cannabis-positive group; separating acute-use from residual-positive cases (which the blood-concentration meta-regressions do) recovers a measurable dose-response signal. Used here as the lower bound on the "any THC positive" risk multiplier and as evidence that the headline number depends strongly on how the impaired population is defined.