Almost a quarter of the population aged 18 years and older reports binge drinking in the last month. Excessive alcohol consumption is associated with significant health and psychosocial consequences. Despite these high rates of problematic drinking, the majority (85%-90%) of individuals believe they can control drinking behavior on their own. Self-monitoring of drinking is key to behavior change, yet recent approaches utilizing mobile technology to collect real-time data have been limited by participant burden, poor compliance, and inaccuracy. Gait analysis, is one of the most reliable way to determine intoxication in humans. We developed AlcoGait, a smartphone-based alcohol tracking technology that passively infers level of drinking from gait impairment. In the current study, we will continue to refine AlcoGait's machine learning algorithms to passively analyze gait and trunk sway data in order to accurately determine gait impairment associated with elevated BrAC levels, particularly .08 g%. While the use of smartphone technology to conduct gait analysis has been employed with various patient populations, it has yet to be utilized for assessing gait in the context of alcohol intoxication. We will administer alcohol to 250 participants in a laboratory setting. We will collect gait and trunk sway data from the smartphone's accelerometer and gyroscope at preset breath alcohol concentration (BrAC) levels up to the legal limit of .08g%. In Stage 1 (technology preparation), we will program the AlcoGait 2.0 smartphone application to passively gather and export both accelerometer and gyroscope data (gait information) based on the first 10 participants. In Stage 2 (data gathering, analysis and machine learning), the next 190 participants will serve as a ?training set,? providing input data for AlcoGait 2.0; these data will allow for the comparison of 4 classifiers types (machine-learning methodologies) that infer BrAC levels associated with impairment, particularly with .08 g%, using gait and trunk sway data. In Stage 3, the best performing classifier as determined using the training set will be programmed into the final AlcoGait 2.0 app and validated with an additional 50 participants completing the alcohol administration laboratory study (?validation set?). AlcoGait 2.0's ability to accurately associate gait impairment with legal intoxication would represent a great advance over other smartphone applications that require the user to manually record and track amount, time, and types of alcohol consumption. With our app's focus on motor impairment, we see our primary users as those who, despite subjective perceptions or counting drinks, are unable to estimate their BrAC accurately but would want to know if they are too impaired to drive. This technology could be used preventively to alert drinkers of impairment and intoxication in real time, mitigating negative consequences. Passive smartphone-based sensing technologies present minimal burden to users and can have ubiquitous impact as smartphones are now widely owned across all ages, ethnicities, and socioeconomic groups.
Almost a quarter of the population aged 18 years and older reports binge drinking in the last month. Despite these high rates of problematic drinking, the majority of individuals do not receive formal alcohol treatment, often because they believe they can control drinking behavior on their own. Self-monitoring of drinking is key to behavior change, yet recent approaches utilizing mobile technology to collect real-time data have been limited by participant burden, poor compliance, and inaccuracy. Smartphone technology, leveraging the long-studied connection between alcohol intoxication and gait impairment, allows alcohol use self-monitoring to enter a new era. Such technology, as pioneered in our phone app Alcogait, where gait is passively recorded during usual activities, could be used preventively to alert drinkers when they are passing various blood alcohol concentrations, and when they are legally intoxicated in real time based on gait impairment.
