A Machine Learning Approach for Inferring Alcohol Intoxication Levels From Gait Data

250 adult volunteers who will each participate in a single laboratory-based visit. At the orientation, each participant will provide informed consent, be weighed, and undergo a medical history to confirm eligibility. A urine drug quick-screen will be given. Female participants will take a urine pregnancy test. Baseline questionnaires will be administered. After giving participants the phone on which the AlcoGait 2.0 app is installed, baseline assessment of gait will be performed with the participant walking a distance of 50 yards ten times.

Drinking will then commence. The total amount will be consumed over 30 minutes. Participants will have their BrAC assessed multiple times during and after drinking has commenced and finished. At BrAC levels of .02, .04, .06 and .08 g%, will perform the gait task. After the last gait task, participant data will be transmitted to a secure server. Participants will receive a meal, and will be escorted to a sitting area with a DVR and videos, and allowed to use their own electronic devices. Their BrAC will be periodically tested until it reaches .02 g% or below, then a taxi will be called to take them home.

Inclusion criteria

• ages 21-65
• drinking at least 5 drinks per occasion for men (4 for women) during a single drinking episode at least once in the last month to ensure, for safety reasons, that people are accustomed to consuming alcohol in amounts comparable to what we will administer
• willingness to drink beer (since only beer will be provided in the session to standardize drinks more easily).

Exclusion criteria

• past 6 months or current treatment for alcohol or drug abuse by self-report (for ethical reasons)
• Score > 5 on (the short version of the Michigan Alcohol Screening Test [SMAST]48) given during the screening interview
• positive family history of alcohol use disorder
• weight ≥ 230 lbs (to prevent excessive volume of beverage)
• current use of medications that are contraindicated for alcohol
• a health condition that contraindicates alcohol (e.g. peptic ulcer disease)
• a health condition that affects ability to walk
• if female, pregnant or nursing
• use of recreational drugs (e.g., marijuana, opioids, stimulants, benzodiazepines) within the last week (self-report plus urine screening cup)
• persons who report alcohol-induced flushing, an untoward reaction that is often accompanied by headache, nausea and other symptoms