Purpose The market for smartphone-paired breathalysers has rapidly expanded, potentially influencing decisions to drive after drinking, but these devices have not been validated.

Methods/approach We enrolled 10 healthy participants 21–39 years old in a laboratory validation study of the accuracy of smartphone-paired breathalysers available for iPhone and Android. Participants were given weight-based doses of vodka with a target blood alcohol level of 0.100. The breath alcohol content was measured every 20 min using a police-grade breathalyser (Intoxilyzer 240) and 3 smartphone breathalysers: Alcohoot AHT 101, BACtrack Mobile Pro, and DRIVESAFE Evoc until the participants breath alcohol declined to 0.02 on the Intoxilyzer 240. The order of the 4 breathalysers was randomised for each test. A peak blood alcohol level was also drawn 20 min after the third dose of alcohol. We measured the mean paired differences of: 1) breath alcohol measurements relative to the blood alcohol level; and 2) smartphone breath alcohol measurements relative to the police breathalyser.

Results The mean peak blood alcohol level was 0.107. The mean paired differences between the breathalysers and peak blood alcohol were: police (−0.015; 95% CI: - 0.025,–0.006), Alcohoot (−0.015; 95% CI: −0.022,–0.008), BACtrack (−0.010; 95% CI: −0.019,–0.001), and DRIVESAFE (−0.037; 95% CI: −0.048,–0.027). The mean paired differences between the smartphone breathalysers and the police breathalyser were: Alcohoot (−0.001; 95% CI: −0.002, 0.001), BACtrack (0.005; 95% CI: 0.004, 0.007), and Evoc (−0.015; 95% CI: −0.016,–0.013).

Conclusions Two of the three smartphone breathalysers tested (Alcohoot and BACtrack) were accurate relative to a police grade breathalyser with differences in breath alcohol levels+/-0.01).

Significance and Contributions to Injury and Violence Prevention Science Smartphone breathalysers are variable in accuracy, with some fit for use with injury prevention interventions, while others dangerously underestimate levels of intoxication.