The intervention effect of local alcohol licensing policies on hospital admission and crime: a natural experiment using a novel Bayesian synthetictime-series method.

Background: Control of alcohol licensing at local government level is a key component of alcohol policy in England. There is, however, only weak evidence of any public health improvement. We used a novel natural experiment design to estimate the impact of new local alcohol licensing policies on hospital admissions and crime.

Methods: We used Home Office licensing data (2007-2012) to identify (1) interventions: local areas where both a cumulative impact zone and increased licensing enforcement were introduced in 2011; and (2) controls: local areas with neither. Outcomes were 2009-2015 alcohol-related hospital admissions, violent and sexual crimes, and antisocial behaviour. Bayesian structural time series were used to create postintervention synthetic time series (counterfactuals) based on weighted time series in control areas. Intervention effects were calculated from differences between measured and expected trends. Validation analyses were conducted using randomly selected controls.

Results: 5 intervention and 86 control areas were identified. Intervention was associated with an average reduction in alcohol-related hospital admissions of 6.3% (95% credible intervals (CI) -12.8% to 0.2%) and to lesser extent with a reduced in violent crimes, especially up to 2013 (-4.6%, 95% CI -10.7% to 1.4%). There was weak evidence of an effect on sexual crimes up 2013 (-8.4%, 95% CI -21.4% to 4.6%) and insufficient evidence of an effect on antisocial behaviour as a result of a change in reporting.

Conclusion: Moderate reductions in alcohol-related hospital admissions and violent and sexual crimes were associated with introduction of local alcohol licensing policies. This novel methodology holds promise for use in other natural experiments in public health.