To thine own self, be true: Examining change in self-reported alcohol measures over time as related to socially desirable responding bias among people with unhealthy alcohol use

Background: Due to a conscious or unconscious desire to be perceived favorably by others, some participants may under or overexaggerate when reporting sensitive behaviors or attitudes, including those related to addictions. This socially desirable responding bias should be considered when using self-reports in predictive models since it introduces error. Methods: A total of 1711 participants were recruited using Amazon’s Mechanical Turk crowdsourcing platform for two randomized controlled trials investigating the effectiveness of brief online interventions for hazardous alcohol use. At baseline, participants completed the Balanced Inventory of Desirable Responding (BIDR). Four measures of alcohol use were collected at baseline and follow-up: number of drinks consumed in a typical week, on one occasion, consequences experienced, and amount of perceived risk of injury or illness from alcohol use. Results: As expected, individuals scoring high on the BIDR subscales reported less alcohol use and related behaviors (p < 0.05); however, repeating the analyses for each gender showed no difference for females asked direct questions about the frequency of their alcohol use. Mixed-effects models investigating the interaction of socially desirable responding bias over time on alcohol-related measures showed some significant differences in the amount of change reported depending on BIDR scores. Participants with higher self-deceptive enhancement scores reported less change over time in their ratings of risk of illness or injury (p = 0.001) compared to lower-scoring participants. Likewise, high-scoring participants reported less change in the number of consequences experienced due to alcohol use over time on both BIDR subscales. Neither direct measure of alcohol use seemed affected by BIDR scores over time. A different pattern was found in males and females analyzed separately. Conclusions: These findings suggest that researchers should consider including measures of socially desirable responding bias in longitudinal studies involving self-reported alcohol use, particularly when modeling alcohol-related measures using rating scales across time. In addition, separate gender analyses may be appropriate. Trial registration: ClinicalTrials.gov NCT03008928. Registered 30 December 2016; ClinicalTrials.gov NCT03060135. Registered 17 February 2017.