Development of an accelerometer-based wearable sensor approach for alcohol consumption detection

Abstract

Background: Alcohol is a commonly used substance associated with significant public health consequences. Treatment is often stigmatized and limited with regard to both access and affordability, demonstrating the need for innovations in alcohol treatment. Accelerometer sensors can detect drinking without user input and are widely incorporated into wearable devices, increasing accessibility and affordability.

Methods: We compared a distributional and random forest classification approach to detect and evaluate sensor-based drinking data. Data were collected at a local state fair (n = 194), where participants drank water at specified intervals interspersed with confounding behaviors (e.g., touching nose, rubbing forehead, or yawning) while wearing an Android-based smartwatch for 10 min. Participants were randomized to receive one of three drinking container shapes: pint, martini, or wine.

Results: The random forest model achieved an overall testing accuracy of 93% (sensitivity = 0.32; specificity = 0.99; positive predictive value = 0.74). The distributional algorithm achieved an overall accuracy of 95% (sensitivity = 0.76; specificity = 0.97; positive predictive value = 0.72). The distributional algorithm had a significantly greater accuracy (t(193) = 7.73, p < 0.001, d = 0.56) and sensitivity (t(193) = 24.5, p < 0.001, d = 1.76). Equivalency testing demonstrated significant equivalency to the ground truth for sip duration (t_lower(193) = 16.92, p < 0.001; t_upper(193) = -9.85, p < 0.001) and between-sip interval (t_lower(193) = 1.72, p = 0.044; t_higher(193) = -3.96, p < 0.001). However, the random forest did not have significant equivalency to the ground truth for between-sip interval (t_lower(193) = 1.98, p = 0.025; t_higher(193) = 0.160, p = 0.564).

Conclusions: Overall, the results indicated that consumer-grade smartwatches can be utilized to detect and measure alcohol use behavior using machine learning and distributional algorithms. This work provides the methodological foundation for future research to analyze the behavioral pharmacology of alcohol use and develop accessible just-in-time clinical interventions.

Keywords: accelerometer; alcohol; classification; machine learning.

FIGURE 1

Scatter plots of accelerometer values for defined behaviors. (A–C) Scatter plots of representative accelerometer values during each behavior. (D). Scatter plot of y‐axis accelerometer values during a sip behavior (black) overlaid with a sinusoidal regression line (red). Dotted lines represent the manually coded start and end times of the sip behavior.

FIGURE 2

Results of paired t‐tests between ground truth and classification algorithms. (A) Results of the paired t‐test between the manually coded ground truth and the classification models on average sip duration. (B) Paired t‐test between the manually coded ground truth and classification models on average sip interval. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 3

Results of the paired t‐test between the manually coded ground truth and the classification models on average number of sips. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 4

Results of the ANOVA post hoc tests for effect of container on F₁‐score demonstrating the significant post hoc results of the effect of container shape on the random forest classification metrics. *p < 0.05, **p < 0.01, ***p < 0.001.