Drinking pattern is more strongly associated with under-reporting of alcohol consumption than socio-demographic factors: evidence from a mixed-methods study.

Background: Under-reporting of alcohol consumption is widespread; surveys typically capture 40-60% of alcohol sales. However the population distribution of under-reporting is not well understood.
Methods: Mixed-methods study to identify factors associated with under-reporting, using the nationally-representative Health Survey for England (HSE) 2011 (overall response rate 66%). Comparison of retrospective computer-assisted personal interview and seven-day drinking diary (n = 3,774 adults 18+, 50% women, diary response rate 69%) to identify factors associated with diary responses exceeding those of the interview using multivariable linear regression for three outcomes: drinking days in the week recorded, volume consumed on heaviest drinking day in the week recorded, and weekly alcohol consumption. Qualitative semi-structured interviews (n = 10) explored reasons for under-reporting in further detail.
Results: Number of drinking days was slightly greater in the diary than the interview (P < 0.001). Reported consumption was higher in the diary than in the interview for heaviest drinking day in the week recorded (0.7 units greater among men, 1.2 units among women, P < 0.001), and weekly alcohol consumption in women only (1.1 units among women, P = 0.003). Participants who drank more frequently, more heavily, and had a more varied drinking pattern with respect to the types of drink consumed or choice of drinking venues had a larger difference between their diary week and their interview week.The qualitative interviews identified having a non-routine drinking pattern, self-perception as a non-frequent drinker, and usually tracking drinking using experiential approaches as linked to more drinking being reported in the diary than the retrospective interview.
Conclusions: Heavy drinking and non-routine drinking patterns may be associated with greater under-reporting of alcohol consumption. Estimates of drinking above recommended levels are likely to be disproportionately under-estimated.