Approaches to Characterizing Drinking Episodes in College Students from Wearable Alcohol Sensors

Introduction: Heavy episodic drinking episodes are prevalent amongst college students and are typically characterized using self-reports about the number of drinks consumed during the episode. These self-reports, however, are challenging to collect in real life and subject to biases that can produce inaccuracies in measuring intoxication (e.g., heavy alcohol use may compromise the compliance or accuracy of drinking reports). Wearable devices, such as the Secure Continuous Remote Alcohol Monitoring (SCRAM) Device, provide a way to more objectively measure intoxication through ongoing measurement of transdermal alcohol concentration (TAC), a direct analogue to blood alcohol concentration (BAC). These repeated measurements of TAC yield a time series of intoxication readings, allowing nuanced characterization of drinking episodes based on intoxication levels (max, mean, area under the curve [AUC]) and intoxication dynamics (trends and fluctuations). In this poster, we characterize drinking episodes measured via the SCRAM device using these approaches and discuss the advantages and disadvantages of different characteristics.

Methods: Young adults (N = 100, M_age = 21, 98% undergraduates) were recruited on or near a large northeastern university. Participants were eligible if they reported engaging in heavy drinking (4+ drinks for females and 5+ drinks for males) at least once a week over the past calendar year or engaging in heavy drinking at least once in a typical week during the academic year. During a 1-hour baseline session participants completed questionnaires and received training on the SCRAM device. Following the baseline session, participants wore the device in their day-to-day lives for 5 consecutive days (Wednesday through Monday) and completed ecological momentary assessments three times per day and every 30 minutes during drinking episodes.

Discussion: Examination of drinking episode plots of TAC revealed non-uniformity in the shape of intoxication functions; for example, some episodes exhibit multiple peaks and valleys. This reveals the difficulty of making comparisons of episodes based on typical metrics of intoxication. We provide exemplars of drinking episodes and the appropriateness/capacity of different characterizations to capture the nuances of real-life drinking experiences. For example, participants similar on some characteristics (e.g., AUC) may have meaningfully different episodes – with such differences able to be captured by complementary metrics (e.g., mean TAC and length of episode). We argue that effective characterization of drinking episodes will require the use of multiple metrics, and careful consideration of the features that truly differentiate potentially harmful drinking episodes from those that may be more typical.