Step Away: Comparing a Chatbot-delivered Alcohol Intervention With a Smartphone App

The Timeline Followback (TLFB) will be used to gather information on alcohol and drug use. The TLFB method has demonstrated good test-retest reliability and high correlations with prospective daily recording methods for quantifying alcohol use. In the current study, a 30-day online version will be administered at baseline and the 4 month post-intervention follow-up assessment. Online versions of the TLFB have been shown to have high correlations with phone and in-person interviewer administered versions of the TLFB The TLFB will be used to calculate the following drinking variables: average (mean) drinks per week, days abstinent, and heavy drinking days. Heavy drinking days will be defined as days with 4 or more standard drinks for women and 5 or more standard drinks for men.

The Short Inventory of Problems - Revised (SIP-R) is a self-report inventory of adverse consequences associated with drug and alcohol use. It is comprised of 17 questions scored on a 4-point Likert scale from 1 (never) to 4 (daily or almost daily). Item responses are summed to produce a total score and five subscale scores (physical, intrapersonal, social responsibility, interpersonal, and impulse control). Higher scores indicate more life problems related to alcohol use. In populations of problem drinkers, the SIP-R has demonstrated good concurrent validity and internal consistency. The SIP-R will be administered at baseline and follow-up to examine between group changes over time. Scale ranges from 17 to 68.

The Timeline Followback (TLFB) will be used to gather information on alcohol and drug use. The TLFB method has demonstrated good test-retest reliability and high correlations with prospective daily recording methods for quantifying alcohol use. In the current study, a 30-day online version will be administered at baseline and the 4 month post-intervention follow-up assessment. Online versions of the TLFB have been shown to have high correlations with phone and in-person interviewer administered versions of the TLFB The TLFB will be used to calculate the following drinking variables: average (mean) drinks per week, days abstinent, and heavy drinking days. Heavy drinking days will be defined as days with 4 or more standard drinks for women and 5 or more standard drinks for men.

The Timeline Followback (TLFB) will be used to gather information on alcohol and drug use. The TLFB method has demonstrated good test-retest reliability and high correlations with prospective daily recording methods for quantifying alcohol use. In the current study, a 30-day online version will be administered at baseline and the 4 month post-intervention follow-up assessment. Online versions of the TLFB have been shown to have high correlations with phone and in-person interviewer administered versions of the TLFB The TLFB will be used to calculate the following drinking variables: average (mean) drinks per week, days abstinent, and heavy drinking days. Heavy drinking days will be defined as days with 4 or more standard drinks for women and 5 or more standard drinks for men.

Given that the main underlying theoretical framework of Step Away is motivational enhancement (ME), we want to examine readiness to change as a mediating variable. We will utilize the Readiness to Change Treatment Version (RCQTV) in this study to assess this construct. The RCQTV has been shown to have solid reliability and validity and has been modified for individuals contemplating or engaged in treatment for alcohol problems. Scale ranges from -8 to 8; higher scores represent higher motivation to change.

Intervention utilization data is used to calculate how frequently participants used the app or the chatbot during the study period. Means and standard deviations will be used to calculate this measure. Unit of measure is the number of times a user clicked on the app or the chabot during the study period.

Retrospective qualitative interviews conducted with 10 chatbot and 10 app users to understand how they used technology and their recommendations for improving the experience. Semi-structured interviews were conducted with participants who had completed follow-up assessments. Participants were purposefully selected to maximize variation in the app (n=10) or the chatbot (n=10) usage. Eleven core interview questions were asked of both app and chatbot users, plus an additional chatbot user question about their experience with and impressions of the chatbot's conversational quality. A random number generator was used to select the 5 high users and 5 low users of each intervention. Qualitative analysis software was used to identify themes related to perceived helpfulness, barriers, motivators, and improvements; key themes related to these topics are reported here. No quantitative data were collected during the interviews.