Background Under-reporting of alcohol consumption is widespread: surveys typically capture only 40-60% of alcohol sales. Little is known about the population distribution of under-reporting.

Methods This study uses mixed methods to identify social, demographic, and alcohol-related ‘risk factors’ for under-reporting, using the nationally-representative Health Survey for England (HSE) 2011. Two methods of measuring alcohol consumption were used: a retrospective computer-assisted personal interview (CAPI), and a seven-day drinking diary (n = 5,040 adults, 55% women). Predictors of an individual’s diary being greater than the CAPI were identified using multivariate linear regression. Three outcomes were investigated: drinking days in the week recorded, heaviest drinking day in the week recorded (in UK units), and weekly alcohol consumption (in UK units). Qualitative semi-structured interviews (n = 10) were conducted with HSE 2011 participants who completed the drinking diary in participants’ homes or workplaces. Emergent themes were identified to reveal additional ‘risk factors’ for under-reporting.

Results Number of drinking days was similar in the diary and the CAPI. The diary estimates were higher than those of the CAPI for heaviest drinking day in the week recorded (0.7 units greater among men, 0.8 units among women), and weekly alcohol consumption (2.9 units greater among men, 3.0 units among women). Unfortunately, the results from the quantitative study are embargoed until the HSE 2011 dataset is deposited in the UK Data Archive in spring 2013. I am happy to update this section once the embargo has been lifted. The qualitative interviews identified having a non-routine drinking pattern, perception of the self as a less frequent drinker, and usually tracking drinking using qualitative approaches as being linked to experiencing surprise at the diary.

Conclusion Heavier drinking and more ‘chaotic’ drinking patterns may be associated with under-reporting of alcohol consumption. Estimates of the prevalence of drinking above the recommended levels, hazardous levels, or even harmful levels that are based on self-reported consumption may be disproportionately under-estimated.