A smartphone app to assess alcohol consumption behaviours: development, validity, compliance and reactivity

Background: Although research into problem drinking often relies on retrospective measures to assess alcohol intake behaviour, such methods have been found to distort actual consumption levels and patterns. Consequently, real-time electronic protocols are increasingly advocated. However, there are limitations – related to cost, participant burden, and missing data – associated with first generation electronic methods of assessing drinking behaviour. This potentially diminishes some of the advantages – namely, enhanced sample size and diversity – typically attributed to real-time research. Smartphone apps participants source and download to their own phones might preserve some of these advantages. To date, there is limited research exploring issues pertaining to the development, validity, compliance, and reactivity of using such apps in the experimental arena.

Methods: An iterative process that included elements of agile software design guided the development of the CNLab-A alcohol consumption assessment app. Healthy individuals (N = 671, M age 23.12) completed demographic questions plus the Alcohol Use Questionnaire and a 21-day Timeline Followback before using CNLab-A for 21 days. Submissions were either event- or notification-contingent. We considered the size and diversity of the sample; compared data reported via retrospective measures with that captured using CNLab-A; and, assessed the data for evidence of compliance and reactivity as a function of hazard versus non-hazard drinker status.

Results: CNLab-A yielded a large and diverse sample. On average, participants submitted data on 20.27 out of 21 days (96.5%). Compared to Timeline Followback, a significantly greater percentage of drinking days (24.79% vs. 26.44%) and significantly higher total intake (20.30 vs. 24.26 standard drinks) was recorded via the app. CNLab-A captured a substantially greater number of high intake occasions than Timeline Followback at all levels from 8 or more drinks. Relative to the Alcohol Use Questionnaire, a significantly faster rate of consumption was recorded via the app. Both hazard and non-hazard drinkers were highly compliant with app protocols. There were no differences between groups in terms of number of days of app use (p = .490) or average number of app responses (p = .540). Linear growth analyses revealed hazardous drinkers decreased their alcohol intake by 0.80 standard drinks over the 21-day experimental period. There was no change to the drinking of non-hazard individuals. Both hazard and non-hazard drinkers showed a slight decrease in responding (“yes”) to drinking behavior over the same period.

Conclusions: Smartphone apps sourced by participants and downloaded to their own phones appear an effective and methodologically sound means of obtaining alcohol consumption information. CNLab-A provided more nuanced information regarding quantity and pattern of alcohol intake than the retrospective measures. In particular, it revealed higher levels of drinking than retrospective reporting. Such apps might, in future, allow for a more thorough examination of the antecedents and consequences of drinking behaviour.