It's an experience that many of us have had at one time or another: You're hanging out with friends at a bar (or, these days, on a Zoom call) and tossing back deceptively sweet drinks until suddenly the room starts to rock like the hull of a ship.
While social lubrication can be an effective way to unwind from the stress of your day or reconnect with friends, it can also pose a serious personal and public health risk if abused. Keeping track of your intoxication is important but judging by the number of drinks consumed alone isn't always a great way to gauge, especially because individuals have different tolerances.
A team of emergency medicine researchers recently investigated how your smartphone's accelerometer (similar to sensors used to track your steps) can predict intoxication based on how you walk. In a sample of 17 participants, the researchers were able to predict intoxication with 92 percent accuracy.
In the study, published Tuesday in the Journal of Studies on Alcohol and Drugs, researchers from Stanford University and the University of Pittsburgh say that their research is complementary to previous studies that identified a varied gait (e.g., swaying side to side when you walk) as a sign of intoxication. While past research focused on connections between the number of drinks consumed and a person's gait, this study zoomed in on blood alcohol concentration (BAC) and breath alcohol concentration (BrAC).
Because an individual's tolerance can vary based on a number of variables, including body size, the researchers behind this new study believe BAC and BrAC may be a more accurate way to measure intoxication.
This proof-of-concept incorporation of smartphone accelerometer data has the potential to be scaled up because of the accessibility of smartphones, the researchers argue.
To get to that future, the team rounded-up a few willing volunteers to get drunk (or, more specifically, to reach a .08 BrAC level) to test out how well this system actually works in the wild.
What're you having — In total, the team recruited 22 participants and measured their height and weight to determine what level of alcohol consumption would be necessary to have them reach a truly intoxicated .08 BrAC.
Brian Suffoletto, the study's lead researcher and an associate professor at the Stanford University School of Medicine's Department of Emergency Medicine, tells Inverse that one shot of a spirit (vodka, in this case) will on average raise a man's BAC by 0.02 (for men) woman's BAC by 0.03. At the time of this research, Suffoletto was a researcher at the University of Pittsburgh School of Medicine.
"Of course, for an individual, many other factors can affect how much a standard drink raises BAC," says Suffoletto.
The participants were served their alcohol in the form of a vodka gimlet and required to drink it within an hour. After that, participants completed walking trials (10-steps forward and 10-steps backward) every hour for seven-hours with their smartphone strapped to their lower backs. Their BAC and BrAC were also measured periodically throughout.
These trials gave the researchers 17 gaits to analyze (12 male and 5 female) and, subsequently, the team found that they were able to accurately predict whether a participant was at or above a 0.08 BrAC based on their gait with 92 percent accuracy.
The smartphone accelerometers collected multi-axis data on the participants' movement and the researchers found that lateral movement (aka swaying back and forth) was a key sign of intoxication.
"The biggest challenge in my mind is how to get someone who is already impaired with alcohol to respond to supportive messaging."
And while the average drinker isn't likely to strap their phone to the lower backs, Suffoletto tells Inverse that they're in the process of analyzing how this set-up will work with participants' phones in their pockets or hands. Suffoletto hypothesizes that phone's swaying in a participant's hand may disrupt their prediction but that keeping a phone in their pocket should not.
What's next — In order to ensure this approach will work outside the meticulously controlled environment of a lab, Suffoletto says a next step will be testing out the set-up in more realistic environments, like crowded bar hallways. The authors also note in the study that expanding their sample size will be important for showing the generalization of these results.
From there, the challenge will be how to communicate these intoxication signals to someone effectively, Suffoletto says.
"I have spent the past 10-plus years designing and testing communication-based strategies to help individuals make better choices related to alcohol consumption," he explains. "The biggest challenge in my mind is how to get someone who is already impaired with alcohol to respond to supportive messaging."
If all goes according to plan, Suffoletto predicts we might start seeing apps with this capability in the next year.
Suffoletto doesn't quite imagine a sci-fi future where our cars will lock us out based on an intoxication analysis beamed over from our phones, but he says perhaps this form of monitoring could be used as a way to drive down car insurance prices, similar to Progressive's safe driver "Snapshot" in-car monitoring device.
"Time will tell, but I expect in the US that we will not find it acceptable for our cars to tell us when we can or cannot safely drive them," Suffoletto says. "We might however in circumstances where our car insurance is $50 less per month if we agree to this sort of monitoring."
Abstract: Objective: Sensing the effects of alcohol consumption in real time could offer numerous opportunities to reduce related harms. This study sought to explore accuracy of gait-related features measured by smartphone accelerometer sensors on detecting alcohol intoxication (breath alcohol concentration [BrAC] > .08%). Method: In a controlled laboratory study, participants (N = 17; 12 male) were asked to walk 10 steps in a straight line, turn, and walk 10 steps back before drinking and each hour, for up to 7 hours after drinking a weight-based dose of alcohol to reach a BrAC of .20%. Smartphones were placed on the lumbar region and 3-axis accelerometer data was recorded at a rate of 100 Hz. Accelerometer data were segmented into task segments (i.e., walk forward, walk backward). Features were generated for each overlapping 1-second windows, and the data set was split into training and testingdata sets. Logistic regression models were used to estimate accuracy for classifying BrAC ≤ .08% from BrAC > .08% for each subject. Results: Across participants, BrAC > .08% was predicted with a mean accuracy of 92.5% using logistic regression, an improvement from a naive model accuracy of 88.2% (mean sensitivity = .89; specificity = .92; positive predictive value = .77; and negative predictive value = .97). The two most informative accelerometer features were mean signal amplitude and variance of the signal in the x-axis (i.e., gait sway). Conclusions: We found preliminary evidence supporting use of gait-related features measured by smartphone accelerometer sensors to detect alcohol intoxication. Future research should determine whether these findings replicate in situ.