A limitation is that the proportion of the total sample in this study at risk of alcohol dependence (0·6%) may have been an underestimate (the 2014 Adult Psychiatric Morbidity Survey estimated 1·4% of the adult population in England10). Population survey research is limited by biases introduced through the exclusion of certain populations from both the sampling frame and through non-response bias. Due to the random location sampling used in these surveys, it is not appropriate to record response rates as the interviewers have a choice as to the properties approached, and therefore it is not possible to measure this non-response bias. The populations who are usually excluded, such as people who are homeless, tend to consume on average more alcohol than the general population and have a higher smoking prevalence.28 These biases can lead to the underestimation of alcohol consumption at the population level.29 This suggests that our study may have underestimated the prevalence of being at risk of alcohol dependence in the general population in England, and furthermore, smoking prevalence in this population. Furthermore, we relied on self-reported measures
