In order to examine relationships between alcohol consumption and levels of inebriation in patrons of UK nightlife, a cross-sectional survey of drinkers' %BAC and drinking behaviours was undertaken in three UK cities. Individuals who were already severely inebriated [21, 31] were excluded. This may have resulted in conservative estimates of drunkenness and alcohol consumption. However, our methods included many individuals who assessed themselves as being drunk and who would become drunk later in the night. Moreover, this did not affect within-individual comparisons of drinking behaviours with observed, self-assessed and biologically measured (BrAT) drunkenness. Consequently, GLMs were employed to calculate the %BAC of participants at their point of departure from nightlife centres, frequently after they would have drunk substantially more alcohol (Additional file 2). The explanatory variables in this observational study were restricted to demographic and self-reported behavioural information, from which GLMs explained 40% of the variation in %BAC. However, the impact of the unexplained variance is likely to be mitigated by comparisons and conclusions being based on mean %BACs rather than calculations for any individual. Further, this methodology: achieved good compliance; is likely to have allowed better recall than interviewing more inebriated people at the end of the night ; and resulted in reported consumption consistent with other studies of similar populations . Unlike in other studies, our final model did not improve from the inclusion of sex or body size and even forced inclusion of these measures did not improve the variance explained [32–34]. However, the model was developed using data from the same individuals to which it was subsequently applied and was bespoke to real drinking environments rather than based on laboratory like conditions. Thus, while other generic models for calculating %BAC from consumption are available , when used in comparable drinking environments they have resulted in poorer predictions of %BAC (e.g. only around 20% of variance explained ). Importantly, when using the model predictions of %BACs at home time, the vast majority of home time (estimated) values for hours drinking (97.1%), drinking rates (100%), hours since at a meal (97.1%) and %BAC (99.5%) were within the observed range at interview. Therefore the model was not extrapolating substantially beyond its training values. Our %BAC measurements also relied on the appropriate use of the BrATs. In particular, a period of 20 minutes is recommended between last drink and breath analysis . Consequently, our study was designed to maximise time between drinking and BrAT, with participants approached outside of drinking establishments, and then introduced to the study and interviewed before %BAC was measured. Finally, there was no control of whether, post-interview, individuals would consume another meal before home time. However, individuals typically seek takeaway meals or other food at the end of the evening and food consumption appears to have made only a relatively small difference to %BAC (table 1) compared with other factors in our model.
Even at interview and with study criteria excluding those showing strong signs of drunkenness, 49.53% of respondents assessed themselves as being drunk. At least in our sample, drunkenness was a typical part of nights out rather than, as sometimes suggested, limited to a few individuals . At home time, modelling suggests %BACs will be considerably higher (Additional file 2d) and drunkenness will be the rule rather than the exception. Given the patterns of alcohol consumption identified this is not surprising. By home time, 10.53% of individuals (15.38% males; 4.35% females; Additional file 2a) intended to have drunk more than 40 units. In fact, even at interview 20.00% of males and 21.28% of females had drank more than the weekly alcohol limits (21 units, males; 14 units, female ) recommended by the UK government prior to the introduction of daily recommended limits, and at home time these figures were estimated to rise to 60.68% and 44.57% respectively. Excess alcohol was often consumed over long drinking periods. For males 21.18% of individuals were expecting to have been drinking for more than 12 hours before returning home. Time drinking, total units consumed and, to a lesser extent, units consumed per hour were positively related to preloading and resulted in preloaders having a significantly higher predicted %BAC at home time.
As well as effects relating to preloading, our results are consistent with extensions to licensing hours contributing to a higher prevalence of drunkenness (Figure 1). In 2005, licensing regulations in England and Wales changed to allow alcohol to be sold 24 hours a day . Although most on- and off-licence establishments have not adopted 24 hour opening, many have extended their opening hours . Here, those individuals intending to utilise the later hours were also most likely to have the highest %BACs (Figure 1). While our sample could not be considered representative of all UK nightlife users, results would at least support the hypothesis that later opening hours can increase inebriation and our methodology provides a mechanism for subsequent tests of this relationship. Already, police and health resources are stretched into early morning hours to allow drunkenness to progress in relative safety and to respond when incidents occur [12, 40]. Importantly, high visibility policing and easily accessible emergency health care may actually encourage individuals to get drunk in the knowledge that the immediate risks associated with drunkenness are substantially reduced [41, 42].
As technologies such as those employed in this study to measure %BAC (BrAT) become more accessible and affordable they may create an additional pressure to consume more alcohol. Only 3.55% of individuals said they would reduce their drinking once informed of their %BAC while nearly one in four individuals thought they would drink more. While this phenomenon needs more study, individuals in the UK can feel that an important feature of a night out drinking is to become drunk. A measured %BAC close to, or even under, the legal driving limit may appear to some drinkers as inconsistent with such an objective and consequently provide an incentive to drink more. Commercial use of BrATs to encourage individuals to drink more has already been attempted in some bars in the UK  and similar problems have been seen elsewhere . Such tests also pose a danger for drink-driving. In our sample, of those below the legal maximum %BAC for driving in the UK (0.08%BAC), 18.31% considered themselves drunk (Additional file 3) but on BrAT realised that they could still legally drive a car. Consequently, easy access to BrAT in night time environments may increase the risk of those who feel drunk at lower blood alcohol levels attempting to drive home in the knowledge they are under the legal blood alcohol limit. Self-reported drunkenness was less common (10%, Additional file 3) in those within the typical European driving limit (up to 0.05%BAC) than in those between 0.05%BAC and the UK limit, where 29.03% considered themselves drunk (Additional file 3). Thus, moves to a lower legal %BAC limit for driving in the UK [44, 45] could help prevent drunks driving, with less people who feel drunk identifying (e.g. through BrAT) that they are legally allowed to drive.
In our sample, half (51.16%) of those who considered themselves drunk at interview intended to consume more alcohol that night. In the UK and elsewhere it is illegal to sell alcohol to those who are drunk. However, research suggests that such laws are often ignored through, for example, commercial pressures to sell alcohol, low awareness and responsibility among bar servers, and difficulties recognising and refusing service to drunks [7–9]. Importantly, those breaking the law are rarely identified and penalised; in 2007 available data show just one individual (out of just seven proceeded against) was found guilty of selling alcohol to a drunk person in England and Wales, with 81 penalty notices for disorder (PNDs) issued for the same offence  (PNDs can be issued by police for certain alcohol-related offences, carrying a fine of £50-£80 to the offender). Findings here identify that a series of relatively simple behavioural observations are strongly correlated with drunkenness and %BAC (Additional file 3). Of those both unsteady on their feet and with a %BAC over 0.20, 83.33% self-assessed as drunk. Our results, and those of others, suggest that simple diagnostic observations of drunkenness (with or without BrAT measures) could be developed for nightlife settings and implemented by trained door and bar staff . Along with such training, measures would have to be implemented to ensure staff feel confident about their own safety when refusing entry to a premises or service to a drunk and potentially aggressive individual. Critically however, such measures run counter to commercial interests. Thus, much of the alcohol sold in night time environments is consumed by those who are already drunk and, to a large extent, the economic viability of many late night businesses in the UK relies on patronage by drunks. Consequently, measures to reduce alcohol sales to drunk individuals are unlikely to be adopted unless made mandatory. However, properly implemented such measures could substantially reduce the number of drunk individuals who continue to access alcohol in nightlife environments and result in fewer highly inebriated individuals leaving nightlife environments early in the morning or requiring health and judicial attention.