The impact of later trading hours for bars and clubs on alcohol-related ambulance call-outs and crimes in Scotland: a controlled interrupted time series study

Background Alcohol-related harms are prevalent late at night, especially on weekends, when high levels of intoxication contribute to increased rates of injury and violence. Reducing or increasing alcohol trading hours late at night in bars and clubs is generally associated with reduced and increased harms, respectively. This study evaluates the impact of later alcohol trading hours in the Scottish cities of Aberdeen and Glasgow on alcohol-related ambulance call-outs and crimes. Under local policy changes, 38 bars in Aberdeen had trading hours extended between 1 to 3 hours up to 3am, and 10 nightclubs in Glasgow had a 1-hour extension to 4am.

Following a natural experiment evaluation framework, we used a controlled interrupted time series design to compare outcomes before and after policy changes, from 2015 to 2021. The primary outcome was a count of total weekend night-time alcohol-related ambulance call- outs. Secondary outcomes included weekend night-time crimes.

In Aberdeen, the policy led to a significant relative increase of 17·4% (effect size=7·061; 95% CI=3·902 to 10·220; p<0·001) in alcohol-related ambulance call-outs, and 7·9% (effect size=3·187; 95% CI=0·564 to 5·810; p=0·017) in reported crimes (not robust across sensitivity analyses), at weekend night-times compared to Edinburgh (control). Findings were not significant and robust across analyses for Glasgow.

Later alcohol trading hours had a significant negative impact on alcohol-related ambulance call-outs in Aberdeen (where more premises had longer extensions) but not in Glasgow, . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 5 suggesting the number, capacity and type of premises moderated outcomes. This is important for the design of future national and local licensing policies and regulations.

later alcohol trading hours; alcohol-related ambulance call-outs; crimes; controlled interrupted time series; Scotland; KEY MESSAGES What is already known on this topic? Prior research has shown a strong association between alcohol availability, including late- night trading hours, and alcohol-related harms. Systematic reviews report that reducing hours of alcohol sales is linked to decreased alcohol-related ambulance-callouts, assaults, and road accidents, whereas later trading hours tend to intensify these alcohol-related harms. International studies, such as those conducted in Netherlands, Norway, and Australia, have provided empirical evidence that even modest reductions or extensions in trading hours can lead to significant decreases or increases in alcohol-related ambulance call-outs, crimes, and emergency department visits. What this study adds? This study provides novel and robust evidence on the impact of later trading hours for bars and clubs on alcohol-related ambulance call-outs and reported crimes in Scotland, using a controlled interrupted time series (CITS) design. By focusing on local policy changes in both Aberdeen and Glasgow, where certain venues of different types were permitted to extend their alcohol trading hours, this research leverages a natural experiment to provide novel insights. Our evaluation provides new evidence on the importance of local contexts on the . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 6 magnitude and timing of change in trading hours for informing future policymaking decisions and research. How this study might affect research, practice or policy? Our findings raise important questions for policymakers seeking to balance economic and community interests in late-night entertainment with the negative impacts of such premises on services including: what is a proportionate policy given our evidence? Is it legally possible for current licensing systems to contain later trading hours to a subset of premises? And, if so, how would those premises be selected? . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 7

Alcohol consumption led to approximately 1·8 million deaths globally in 2021 and is a risk factor for over 200 diseases [1,2]. Alcohol’s harms extend beyond those consuming alcohol, including in contributing to increased crime, violence, road accidents, and various forms of injury [3,4]. In Scotland, 1,277 deaths from causes wholly attributable to alcohol were recorded in 2023, averaging 24 deaths weekly [5]. Alcohol also poses a significant burden for the healthcare system in Scotland: in 2022/23, there were 31,206 alcohol-specific hospital admissions [6] and in 2019 it was estimated that 16% of ambulance call-outs were alcohol- related [7]. Generally, alcohol harms are associated with greater density of premises licensed to sell alcohol, as well as increased days and hours during which alcohol can legally be sold [8]. Acute alcohol harms are particularly prevalent late at night (i.e., 12am to 5am), especially on weekends, when higher levels of intoxication, can contribute to increased rates of injuries and violence [9]. In 2019, the highest concentration of alcohol-related ambulance call-outs in Scotland occurred between 9pm and 1am on weekends, whereas call-outs unrelated to alcohol were more frequent between 10am and 9pm [7]. Reducing late-night trading hours for alcohol on-premises outlets is associated with reductions in alcohol harms [10]. Considering extensions to permitted trading hours for alcohol in on-premises outlets, separate systematic reviews concluded that extensions generally increased the likelihood of assaults, injuries, and drink-driving incidents [11,12]. A 2009 review found evidence that extensions were associated with increased harm, while reductions was associated with a decrease [13]. The review emphasised the need for well- controlled research to further verify these associations. Another review based on a small number of studies concluded that extensions of two or more hours per day significantly increased alcohol harms, although no strong association was found for shorter spanning extensions due to limited evidence [14]. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 8 In the UK, three separate systems (one in England and Wales, one in Scotland and one in Northern Ireland) control the granting of licences required to legally sell alcohol. Both the systems in England/Wales and in Scotland were substantially reformed in the mid-2000s. The Licensing Act (Scotland) 2005 introduced a wholly reformed system empowering local government licensing boards to set local policy and make decisions on premises licensing, guided by five overarching objectives. These boards set local on-trade alcohol trading hours policy, but the legislation includes a presumption against 24-hour licensing. Under this system, late night alcohol trading hours have expanded in recent years [15], but their impact has never been evaluated in Scotland. The Licensing Act (2003) also established a completely new local government-led system to manage licensing in England and Wales that created a general presumption in favour of granting premises licences, with no presumption against 24-hour alcohol sales. The 2003 Act was underpinned by the view that the fixed closing times that predated the Act (most pubs closing at 11pm) contributed to overcrowding, violence, and disorder and that staggered dispersal may reduce these issues [16]. Studies on the impact of extended licensing in England and Wales have not found consistent evidence of their impact on violence [15], they used weaker before-and-after designs without considering a control series compared to those in other international studies [13,16]. Using a more robust interrupted time series (ITS) design, Humphreys and colleagues found no evidence of an overall increase in violent crime in the city of Manchester following the Act but a significant 36% increase between 3am and 6am [17]. Although more robust than earlier studies, wider health outcomes were not considered, further evidencing the need for additional research [17]. Another UK-based study reported small changes in the timing of self-reported drinking occasions due to the implementation of the Act but no evidence of changes in finish time variation, post-loading behaviour, late-night drinking, or overall alcohol consumption [18]. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 9 To our knowledge, no study in the UK has considered the impact of later trading hours on alcohol-related ambulance call-outs. Under the system established by the 2005 Act, local Licensing Boards (made up of elected members of local government) in Aberdeen and Glasgow have in recent years changed their local policies to allow premises in certain licence categories to apply for, and be granted, permission to sell alcohol later at night. This study aims to evaluate the impact of these later trading hours for bars and clubs on alcohol-related ambulance call-outs and reported crimes in both cities.

The policy changes: later trading hours for bars and clubs in Glasgow and Aberdeen In Glasgow, from 12th April 2019, 10 nightclubs, which met certain quality and operating standards, were granted permission by the Glasgow City Licensing Board (GLB) to remain open and sell alcohol for an additional hour to 4am (instead of 3am). This initial 12-month pilot became standard policy in 2023. In Aberdeen, a policy was introduced in January 2017 that allowed a wider range of alcohol premises to apply for later trading hours, which had previously only been available to nightclubs. From March 2017 to October 2020, 38 bars/pubs in Aberdeen were granted permission to sell alcohol between 1 and 2 hours later at night, up to 3am. Nightclubs remained closed throughout the COVID-19 restrictions and were therefore unable to use the additional hour, while bars and pubs were permitted to reopen earlier and to use extended hours during some periods. These recent policy changes in Aberdeen and Glasgow serve as the basis for this natural experiment evaluation. Study Design . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 10 We used a controlled interrupted time series (CITS) design to evaluate the later closing times, comparing outcomes before and after policy implementation (May 2015- July 2022) in Glasgow and Aberdeen with a control group. The policy change date, 12 April 2019 for Glasgow, and a gradual adoption of extended hours in Aberdeen beginning in early 2017, being the "interruption" in the time series. Data Sources We obtained all ambulance call-outs in Scotland from the Scottish Ambulance Service (SAS) and crime data from Police Scotland. Ambulance data included records such as call-out time, coordinates of location, patient demographics, and a flag indicating whether a call-out was alcohol-related (algorithmically generated from free-text fields in call-out records), as previously developed with SAS [7]. Crime data were gathered from Police Scotland using Scottish Government Justice Directorate (SGJD) crime codes. Two authors (NF & JL) scrutinised and selected crime codes they judged to be more likely to be associated with alcohol, in consultation with Police Scotland. These included assaults, drunkenness, driving under the influence of alcohol, and antisocial behaviour offences (see Table A4 in the appendix 2.1). Variables included time of recording, crime type, local council name, incident date, and coordinates of crime location. Outcomes Primary outcome: total weekend night-time alcohol-related ambulance call-outs (time between Fridays 20:00 to 23:59, Saturdays 00:00 to 05:59 and 20:00 to 23:59, and Sundays 00:00 to 05.59). Secondary outcomes: total weekend night-time ambulance call-outs and reported crimes. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 11 The outcomes were analysed as weekly counts. We also modelled population size-adjusted outcomes to check robustness of our analysis. Control Selection We developed a principled and structured framework for selecting the study control, shown in Figure A1 in the appendix 1 (with steps in the appendix 1.1 for alcohol-related ambulance call-outs and in the appendix 1.2 for reported crimes) and details in the published statistical analysis plan (SAP) [19]. Following this framework, among the potential 30 control candidates (Scottish council areas, apart from Glasgow City and Aberdeen City), the City of Edinburgh council area (Edinburgh) was identified as the most suitable control for Aberdeen and Glasgow. Measurement of Policy Exposure Variable Separate exposure variables were generated for Glasgow and Aberdeen. The dates on which premises were permitted to trade later were staggered over time in Aberdeen; in contrast, in Glasgow, the later trading hours came into effect from the same date for all premises. In Glasgow, a binary exposure variable taking the value “0”, for time periods before the implementation of the new trading hours (i.e., 12th April 2019), and “1” for time periods thereafter, was used. In Aberdeen, a numerical measure of exposure was developed based on the additional person-hours available weekly across all premises. This measure reflects the staggered granting of later licensed hours, taking values between 0 (no extended hours) and 1 (all additional person-hours had been adopted across participating premisses). The potential additional person-hours for a given venue in a week was calculated by using the formula [(additional hours) x (the number of nights additional hours are permitted to be used in a week) x (capacity of the venue)]. Full details can be found in the SAP [19]. Confounding Variables . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 12 We included data on per capita gross disposable household income, weather conditions (mean temperature, rainfall), and total number of on-premises alcohol outlets to account for potential time-varying confounding effects. Additional adjusting for COVID-19 lockdown (imposing closure or restricted hours for alcohol-premises), was done through dummy variables. We included dummy variables to account for public holidays, outliers and COVID- 19 restrictions which meant that the premises with later hours were not open at certain times (see appendix 2.2 for details). As noted earlier, the value of the COVID-19 dummy variable over time reflected that nightclubs were fully closed during COVID-19 restrictions, while bars and pubs remained open for some periods. Data sources for the confounding variables are provided in appendix 2.2.1. Statistical Analysis Data were analysed using autoregressive integrated moving average (ARIMA) models for both primary and secondary outcomes following the CITS design. We modelled the outcome differences between the interventions and control over time, for example, differences between alcohol-related ambulance call-outs in Aberdeen and Edinburgh. We followed Box-Jenkins model selection steps (identification, estimation, and diagnostic checking) to obtain the best ARIMA model that adequately captures the underlying patterns in the data and provides accurate estimation. First, we used an Augmented Dickey-Fuller test to establish that the stationarity (i.e., constant mean and variance over time) assumption was met. Then, we used autocorrelation and partial autocorrelation functions (ACF and PACF) to identify the terms of the autoregressive (AR) and moving average (MA) models. After that, from the candidate models, the best fit was chosen according to Bayesian (BIC) and Akaike (AIC) Information Criterions. We next tested suitability of the fitted model by performing diagnostic checks to ensure the residuals presented white noise characteristics. Policy exposure variables were added to the model as covariates to estimate intervention effect sizes. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 13 To check the robustness of our findings, we examined the effect size across ARIMA models adjusting for all covariates, and for statistically significant covariates only. Subgroup analyses to examine differential impacts by age, and sex, were conducted for the primary outcome only. Analyses were performed in Stata/MP 17.0 and RStudio (version 4.2.0) using the Synth package [20]. Sensitivity analysis We conducted sensitivity analyses by restricting the time series to pre-COVID-19 restriction periods in Scotland to assess the robustness of our estimates. Additionally, we modelled population size-adjusted outcome rates. To check robustness of our estimation, we additionally created two binary exposure variables specific to Aberdeen: one similar to Glasgow on the introduction of the new policy, and a second to indicate when the policy reached at least half its “strength”, taking a value of "0" when Aberdeen’s policy exposure variable is less than 0·5, and "1" otherwise. We also conducted Synthetic Control (SC) modelling. A separate SC for Aberdeen and Glasgow was created for the population-adjusted rates of alcohol-related ambulance call-outs and reported crimes, applying various specifications of pre-intervention outcomes [21]. Covariates such as per-capita gross disposable household income and population-adjusted number of on-premises alcohol outlets were incorporated in the construction of the SC. To assess the validity of our SC models, we split the pre-intervention period into training and validation phases and compared MSPEs (across validation periods) for the different model specifications. Appendices 3.1.1 and 3.2.1 present the model specifications and validity tests for alcohol-related ambulance call-outs in Aberdeen and Glasgow, respectively. Appendices 3.3.1 and 3.4.1 contain the model specifications and validity tests for reported crimes in Aberdeen and Glasgow, respectively. We performed in-space placebo tests and calculated . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 14 frequentist (frequency-based) p-values for the effect sizes generated through SC. Details of synthetic control analyses and sensitivity tests are provided in appendices 3.1.2 (Aberdeen) and 3.2.2 (Glasgow) for alcohol-related ambulance call-outs, and in appendices 3.3.2 (Aberdeen) and 3.4.2 (Glasgow) for reported crimes. Additionally, we performed time-series models (ARIMA) to the synthetic control series to check the robustness of the estimated effect size (see Tables 17 and 18 in the appendix 4). Finally, a falsification test was conducted, by generating a false policy exposure variable signalling exposure one year before the true intervention date.

Between 1 May 2015 and 31 July 2022, a total of 6,461 alcohol-related ambulance call-outs occurred in Aberdeen, 19,956 in Glasgow, and 13,781 in Edinburgh during weekend night- times (Table A1 in the appendix 2.1). A total of 5,927 crimes were reported in Aberdeen, 20,038 in Glasgow, and 12,268 in Edinburgh during weekend night-times, more than 80% of which in each city were ‘common assaults’ (Table A3 and A4 in the appendix 2.1). (Figure 1) Prior to the policy changes, trends in alcohol-related ambulance call-outs and reported crimes between Aberdeen/Glasgow and Edinburgh were broadly comparable, with seasonal variation (Figure 1). (Figure 2) In Aberdeen, alcohol-related ambulance call-outs peaked between 00:00–00:59 before the new policy but shifted to 01:00–01:59 after the policy change (Figure 2); such a shift was not observed in the control city (Edinburgh) and for reported crimes (Figure A2 in the appendix 2.1). . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 15 (Table 1) The estimated policy effect for Aberdeen using a staggered policy variable when the model included only statistically significant covariates, was a significant average relative increase of 7·061 extra call-outs per week in Aberdeen following the policy change compared with Edinburgh (corresponding to a 17·4% relative increase in Aberdeen compared to Edinburgh) (Table 1). Increasing policy effects were consistently observed for Aberdeen across various sensitivity analyses. However, the policy effect term was not significant when using the SC method. Subgroup analyses revealed that the policy effects in Aberdeen were comparatively higher for males (coefficient = 3·726, p-value <0·001, 95% CI = 1·946 to 5·507) and individuals under 45 years of age (coefficient = 7·050, p-value<0·001, 95% CI = 4·077 to 10·023) compared to their respective counterparts (Tables A9 to A12 in the appendix 2.3). Similarly, when only statistically significant covariates were included in the model, the effect was 3·187 (p-value = 0·017, 95% CI = 0·564 to 5·810) for reported crimes in Aberdeen (Table 2). This means an estimated average of 3·187 extra reported crimes per week in Aberdeen than Edinburgh, following the policy change (a 7·9% relative increase of reported crimes for Aberdeen compared to Edinburgh). Across our sensitivity analyses, the estimated policy effects show consistent increases after the policy change but are not consistently significant at 5% significance level (Table 2). (Table 1) In contrast, Glasgow showed a significant decreasing policy effect on alcohol-related ambulance call-outs (i.e., a reduction in 5·110 call-outs per week) based on ARIMA models (Table 1). However, Glasgow results varied substantially across the range of sensitivity analyses. Moreover, for reported crimes in Glasgow, no significant policy effects were observed (Table 2). . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 16 We also conducted falsification tests for both primary and secondary outcomes, results are provided in the appendix (Tables A14 to A16 in the appendix 2.4). (Figure 3) In Figure 3, we compared the rates of weekend night-time alcohol-related ambulance call- outs and reported crimes in Aberdeen and Glasgow with their respective synthetic counterparts (a counterfactual scenario that assumes no policy changes occurred). For alcohol-related ambulance call-out rates, the observed and synthetic trends in Aberdeen prior to the intervention show the best fit compared to Glasgow. After the policy change, the observed rates in Aberdeen mostly lie above the synthetic rates, suggesting an increase in call-outs due to the policy. In contrast, Glasgow shows considerable overlap between observed and synthetic rates after the policy change, indicating no clear effect. For reported crime rates, although some overlap exists between the observed and synthetic trends in Aberdeen after the policy change, the observed rates lie above the synthetic rates at most time points, suggesting a potential increase. In Glasgow, the observed crime rates also mostly lie above the synthetic rates following the policy change, implying an increase. However, the pre-intervention fit between observed and synthetic crime rates is weaker in Glasgow than in Aberdeen.

The results revealed clear differences in the effects of the policy changes in the two cities. In Aberdeen, the later trading hours policy applied to 38 bars and pubs led to a statistically significant 17·4% increase in alcohol-related ambulance call-outs and 7·9% increase in reported crimes relative to Edinburgh (control city) at weekend night-times. The inconsistency of the estimated effect size for reported crimes compared to alcohol-related ambulance call-outs in Aberdeen may be due to under reporting of night-time crime incidents . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 17 compared to the actual occurrences [22]. Furthermore, unlike ambulance call-outs, reported crimes do not have a specific alcohol-related flag, making it more challenging to assess their connection to alcohol intake and policy changes. Our findings in Aberdeen align with previous studies reporting an increase of alcohol harms after the implementation of extended alcohol premises hours policies [11,12]. Our findings also show a temporal shift in peak times for alcohol-related ambulance call-outs in Aberdeen, from 00:00–00:59 to 01:00–01:59, suggesting that the extended premises hours may have altered drinking patterns or increased alcohol availability, shifting the time at which individuals required ambulance services. Additionally, we found the impact of the policy change on alcohol-related ambulance call- outs in Aberdeen was significantly higher among men and individuals under 45 years of age. This is consistent with prior research showing that younger individuals, particularly men, are more likely to engage in risky drinking behaviours and are more at risk of alcohol-related ambulance call-outs [7]. In contrast, no significant association was observed due to the later trading hours policy for the 10 nightclubs in Glasgow in either alcohol-related ambulance call-outs or reported crimes when compared with our control, Edinburgh. Our Glasgow findings, of no consistent significant effect, conflict with one study that found a 1-hour extension to existing late-night licences was associated with a 34% more alcohol-related ambulance call-outs in Amsterdam, Netherlands [23]. Our findings are consistent with conclusions drawn from a systematic review (albeit based on a small number of studies), which suggested that extending alcohol premises hours by two or more hours is significantly associated with increased alcohol- related harms, while mixed results were found for extensions of less than two hours [14]. However, in separate qualitative interviews [24], frontline staff in both Glasgow and Aberdeen reported several harms from the extensions in trading hours including sometimes unmanageable pressures on police and support services due to greater numbers of vulnerable . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 18 or intoxicated individuals later at night. These incidents do not feature in the crime or ambulance data we analysed in this study. A key strength of our study is its robust design, which incorporated a structured pre-specified control selection process [19]. We triangulated our findings across modelling approaches to enhance reliability of the results. Unlike previous research that often overlooks staggered policy implementation, venue capacity, and premises type [16], our design incorporated a staggered policy covariate for Aberdeen based on the person-hours, calculated based on the number of permitted premises, their maximum capacities, and extended hours. While the study employs robust methodological approaches, including CITS and SC methods, limitations such as potential underreporting of crimes, incidents that are not recorded in our datasets (as above), and the observational nature of the study, including potential for unmeasured time-varying confounding factors should be acknowledged. We used the approval dates rather than implementation dates, since it is unclear whether all approved premises actually implemented extended hours; qualitative findings indicate that use of the extended hours was unpredictable across both cities due to mixed demand (paper forthcoming). For rate calculations, we used cities population as the denominator to represent those at risk, although visitors may also contribute to the night-time activities and related harms. Population data are recorded annually, therefore we used extrapolation technique. Both ARIMA and SC analyses showed similar effect direction in Aberdeen, but SC estimates were not statistically significant. This is maybe due to the frequentist (frequency based) nature of SC p-values, which are limited by number of control units in the donor pool [25]. Moreover, effect sizes are not directly comparable across models’ estimating counts versus rates, since ARIMA estimators are sensitive to scale/unit of analysis [26]. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 19 Our findings have important policy implications. This is the first UK study to find that extending trading hours for late-night licensed premises, may cause significant additional pressures on ambulance services. This ought to give pause to policymakers in both Ireland [27] and England and Wales [28,29] where further liberalisation of late-night opening hours has been proposed, especially given existing pressures on emergency services. Secondly, the findings illustrate some of the limitations of UK licensing systems [30]. If the volume and type of premises with later alcohol trading hours is an important factor dictating harms, then policymakers should have powers to control these factors. The current system in Scotland (and in England and Wales) does not distinguish between premises types [31]. Glasgow was only able to establish different standard trading hours for bars and nightclubs by including a robust definition of a nightclub in their licensing policy statement [32]. Further, local authorities cannot regulate the volume of premises eligible for later trading hours because each application has to be considered on its individual merits [33]. Although we studied the impact of the 4am extension in Glasgow’s pilot with 10 nightclubs, our findings suggest that granting the same hours to additional nightclubs would carry a further risk of negative outcomes, yet there is little to prevent multiple premises now being included. To protect public health and safety late at night, licensing systems may need to include powers to create type-specific caps on venue numbers with very late hours, for example per population or locality, similar to practice in several North American jurisdictions.

This study demonstrates the value of robust, multicomponent analysis of outcomes when assessing the impact of changes to licensing hours. While our findings confirm the general principle that widespread extending alcohol trading hours, especially late into the night, are . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 20 associated with some increased harms they also show that the intensity of these impacts is likely mediated by type and volume of outlets. It is therefore important that licensing systems can account for these factors. AUTHOR CONTRIBUTIONS NF led the overall design of the mixed-methods ELEPHANT (Evaluating Later and Expanded Premises Hours for Alcohol in the Night-time Economy) study. JL led the design of the statistical analyses and supervised NS and HH in conducting the analyses. JL, NF, CA, IHC, EM, HH, and NS contributed to the conceptualisation and methodology. NS was responsible for data curation, formal analysis, drafting the original manuscript. NS, JL, and FM contributed to validation of data curation and analysis. NS and CA were responsible for data visualisation. NF, EM, JN, ES, CE, RO, GM, IHC, MC, KM, AM, and IU contributed to interpretation. NF, JL, EM, CA, CE, ES, and AM are lead (NF) and co-applicants on the NIHR grant that funded ELEPHANT. All authors critically reviewed and edited subsequent versions of the manuscript and approved the final version of the manuscript for publication. FUNDING The ELEPHANT (Evaluating Later and Expanded Premises Hours for Alcohol in the Night- time Economy) study is funded by the NIHR Public Health Research programme (129885). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. DECLARATION OF INTERESTS All other authors declare no competing interests.

We want to thanks to the Scottish Ambulance Service and Police Scotland. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 21 PATIENT AND PUBLIC INVOLVEMENT Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. DATA SHARING Time-series data on alcohol-related ambulance call-outs and reported crimes, used in this study can be accessible through formal application to the Scottish Ambulance Service and Police Scotland. Meta data used in this study have already been published online (Per capita gross disposable household income: https://www.webarchive.org.uk/wayback/archive/20180212144444/http://www.gov.scot/Topi cs/Statistics/Browse/Economy/QNA2017Q3; Temperature and rainfall: https://data.ceda.ac.uk/badc/ukmo-hadobs/data/insitu/MOHC/HadOBS/HadUK- Grid/v1.2.0.ceda/region; On-premises alcohol outlets: https://www.gov.scot/publications/scottish-liquor-licensing-statistics/; COVID-19 lockdown: https://spice-spotlight.scot/2023/05/10/timeline-of-coronavirus-covid-19-in-scotland/; and Public holidays: https://www.mygov.scot/scotland-bank-holidays) REFFERENCES 1 Brauer M, Roth GA, Aravkin AY, et al. Global burden and strength of evidence for 88 risk factors in 204 countries and 811 subnational locations, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2024;403:2162–203. 2 World Health Organization. Global status report on alcohol and health 2018. Geneva: World Health Organization 2018. https://doi.org/10.1037/cou0000248 3 Gakidou E, Afshin A, Abajobir AA, et al. Global, regional, and national comparative . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 22 risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet. 2017;390:1345–422. 4 Griswold MG, Fullman N, Hawley C, et al. Alcohol use and burden for 195 countries and territories, 1990-2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet. 2018;392:1015–35. 5 National Records of Scotland. Alcohol-specific deaths 2023. 2024. https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/vital- events/deaths/alcohol-deaths 6 Public Health Scotland. Alcohol Related Hospital Statistics Scotland 2022/23. 2023. 7 Manca F, Lewsey J, Waterson R, et al. Estimating the burden of alcohol on ambulance callouts through development and validation of an algorithm using electronic patient records. Int J Environ Res Public Health. 2021;18. doi: 10.3390/ijerph18126363 8 World Health Organization. The SAFER technical package: five areas of intervention at national and subnational levels. 2019. https://apps.who.int/iris/handle/10665/330053 9 Mäkelä P, Warpenius K. Night-time is the right time? Late-night drinking and assaults in Finnish public and private settings. Drug Alcohol Rev. 2020;39:321–9. 10 Wilkinson C, Livingston M, Room R. Impacts of changes to trading hours of liquor licences on alcohol-related harm: A systematic review 2005-2015. Public Heal Res Pract. 2016;26:1–7. 11 Nepal S, Kypri K, Tekelab T, et al. Effects of extensions and restrictions in alcohol trading hours on the incidence of assault and unintentional injury: Systematic review. J Stud Alcohol Drugs. 2020;81:5–23. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 23 12 Popova S, Giesbrecht N, Bekmuradov D, et al. Hours and days of sale and density of alcohol outlets: Impacts on alcohol consumption and damage: A systematic review. Alcohol Alcohol. 2009;44:500–16. 13 Stockwell T, Chikritzhs T. Do relaxed trading hours for bars and clubs mean more relaxed drinking ?A review of international research on the impacts of changes to permitted hours of drinking. Crime Prev Community Saf. 2009;11:153–70. 14 Robert A. Hahn, Kuzara JL, Elder R, et al. Effectiveness of Policies Restricting Hours of Alcohol Sales in Preventing Excessive Alcohol Consumption and Related Harms. Am J Prev Med. 2010;39:590–604. 15 Alcohol Focus Scotland. Review of statements of licensing policy 2018 to 2023. Published Online First: 2020. 16 Humphreys DK, Eisner MP. Evaluating a natural experiment in alcohol policy The Licensing Act (2003) and the requirement for attention to implementation. Criminol Public Policy. 2010;9:41–67. 17 Humphreys DK, Eisner MP, Wiebe DJ. Evaluating the Impact of Flexible Alcohol Trading Hours on Violence: An Interrupted Time Series Analysis. PLoS One. 2013;8. doi: 10.1371/journal.pone.0055581 18 Stevely AK, de Vocht F, Neves RB, et al. Evaluating the effects of the Licensing Act 2003 on the characteristics of drinking occasions in England and Wales: a theory of change-guided evaluation of a natural experiment. Addiction. 2021;116:2348–59. 19 Sheikh N, Henriques-Cadby I, Haghpanahan H, et al. Evaluating later or expanded premises hours for alcohol in the night-time economy (ELEPHANT): statistical analysis plan. medRxiv. Published Online First: 2024. doi: . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 24 https://doi.org/10.1101/2024.05.07.24306991 20 Abadie A, Diamond A, Hainmueller J. Synth: An R Package for Synthetic Control

in Comparative Case Studies. J Stat Softw. 2011;42:1–17. 21 Degli Esposti M, Spreckelsen T, Gasparrini A, et al. Can synthetic controls improve causal inference in interrupted time series evaluations of public health interventions? Int J Epidemiol. 2020;49:2010–20. 22 Justice Analytical Services. Bringing Together Scotland ’ s Crime Statistics. Edinburgh: Scotland 2014. 23 de Goeij MCM, Veldhuizen EM, Buster MCA, et al. The impact of extended closing times of alcohol outlets on alcohol-related injuries in the nightlife areas of Amsterdam: A controlled before-and-after evaluation. Addiction. 2015;110:955–64. 24 Mitchell et al. ‘It’s a bit of a mess’: The impact of later alcohol trading hours for bars and clubs in Scotland according to qualitative interviews with local stakeholders. 2025. 25 McClelland R, Mucciolo L. An update on the synthetic control method as a tool to understand state policy. 2022. https://www.taxpolicycenter.org/sites/default/files/publication/163919/an_update_on_t he_synthetic_control_method_as_a_tool_to_understand_state_policy.pdf 26 Huntington-Klein N. The Effect : An introduction to Research Design and Causality. 1st ed. Chapman & Hall 2022. 27 Department of Justice of Ireland. General Scheme: Sale of Alcohol Bill 2022 including press release. 2022. https://www.drugsandalcohol.ie/37347/ (accessed 29 October 2025) 28 Department for Business & Trade. Licensing policy sprint: joint industry and HM . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 25 government taskforce report. Policy paper. London 2025. https://www.gov.uk/government/publications/licensing-taskforce-report-and- government-response/licensing-policy-sprint-joint-industry-and-hm-government- taskforce-report 29 Department for Business & Trade. Licensing taskforce report: government response. Policy paper. London 2025. https://www.gov.uk/government/publications/licensing- taskforce-report-and-government-response/licensing-taskforce-report-government- response 30 Fitzgerald N, Egan M, O’Donnell R, et al. Public health engagement in alcohol licensing in England and Scotland: the ExILEnS mixed-method, natural experiment evaluation. Public Heal Res (Southampton, England). 2024;13:1–84. 31 Scottish Government. Licensing ( Scotland ) Act 2005. 2005. https://www.legislation.gov.uk/asp/2005/16/contents 32 City of Glasgow Licensing Board. Licensing Policy Statement: Licensing (Scotland) Act 2005. 2018. https://www.glasgow.gov.uk/CHttpHandler.ashx?id=17578&p=0 33 Home Office. Revised guidance issued under section 182 of Licensing Act 2003. Published Online First: 2025. 34 Scottish Government. Quarterly National Accounts Scotland, 2017 Quarter 3. 2018. https://www.webarchive.org.uk/wayback/archive/20180212144444/http://www.gov.sc ot/Topics/Statistics/Browse/Economy/QNA2017Q3 (accessed 22 November 2022) 35 Centre for Environmental Data Analysis. Met Office weather datasets. CEDA Arch. 2019. https://data.ceda.ac.uk/badc/ukmo-hadobs/data/insitu/MOHC/HadOBS/HadUK- Grid/v1.2.0.ceda/region (accessed 10 November 2023) . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 26 36 The Scottish Government. Scottish Liquor Licensing Statistics. 2022. https://www.gov.scot/publications/scottish-liquor-licensing-statistics/ 37 Scottish Parliament Information Centre (SPICe). Timeline of Coronavirus (COVID- 19) in Scotland. SPICe Spotlight. 2023. https://spice- spotlight.scot/2023/05/10/timeline-of-coronavirus-covid-19-in-scotland/ (accessed 11 March 2024) 38 The Scottish Government. Public and bank holidays. Scotl. bank holidays. 2022. https://www.mygov.scot/scotland-bank-holidays (accessed 12 March 2024) 39 Abadie A, Vives-i-Bastida J. Synthetic Controls in Action. arXiv Prepr. Published Online First: 2022. 40 Ferman B, Pinto C, Possebom V. Cherry Picking with Synthetic Controls. J Policy Anal Manag. 2020;39:510–32. 41 Kaul A, Klößner S, Pfeifer G, et al. Standard Synthetic Control Methods: The Case of Using All Preintervention Outcomes Together With Covariates. J Bus Econ Stat. 2022;40:1282–90. 42 Kreif N, Grieve R, Hangartner D, et al. Examination of the synthetic control method for evaluating health policies with multiple treated units. Health Econ. 2015;25:1514– 28. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 27 Table 1. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow Primary outcome, weekend night-time alcohol-related ambulance callouts Main analysis Sensitivity analysis Synthetic control Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 5·824 0·002 2·225, 9·423 4·641 0·037 0·279, 9·002 0·140 <0·001 0·071, 0·209 0·103 0·222 Policy effect (significant covariates) 7·061 <0·001 3·902, 10·220 4·643 0·036 0·292, 8·994 0·130 <0·001 0·061, 0·198 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 2·898 0·056 -0·078, 5·873 2·486 0·101 -0·483, 5·455 0·119 <0·001 0·052, 0·185 0·140 0·185 Policy effect (significant covariates) 2·970 0·051 -0·014, 5·955 2·565 0·098 -0·470, 5·599 0·126 <0·001 0·072, 0·180 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 3·145 0·041 0·136, 6·154 2·018 0·243 -1·369, 5·406 0·088 0·007 0·024, 0·152 - - . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 28 Policy effect (significant covariates) 4·473 0·001 1·846, 7·100 1·744 0·305 -1·585, 5·072 0·080 0·011 0·018, 0·142 - - Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -4·016 0·050 -8·027, -0·005 -2·307 0·329 -6·942, 2·329 -0·057 0·064 -0·116, 0·003 0·012 0·926 Policy effect (significant covariates) -5·110 <0·001 -7·491, -2·730 -2·951 0·096 -6·423, 0·520 -0·045 0·018 -0·081, -0·008 *Effect size for synthetic control adjusted by the covariates, number of on-premises alcohol outlets (population adjusted), and per capita gross disposable household income Table 2. Effect of policy changes on weekend recorded crimes in Aberdeen and Glasgow Secondary outcome, weekend night-time recorded crimes Main analysis Sensitivity analysis Synthetic control Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 5·770 0·072 -0·514, 12·055 4·596 0·218 -2·722, 11·914 -0·003 0·973 -0·194, 0·187 0·022 0·333 Policy effect (significant covariates) 3·187 0·017 0·564, 5·810 3·442 0·035 0·239, 6·645 0·057 0·092 -0·009, 0·124 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 29 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -1·551 0·388 -5·072, 1·969 -1·459 0·455 -5·286, 2·368 -0·036 0·593 -0·168, 0·096 0·028 0·370 Policy effect (significant covariates) 0·960 0·398 -1·269, 3·190 1·192 0·283 -0·985, 3·368 0·039 0·273 -0·031, 0·110 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 4·020 0·005 1·190, 6·850 3·205 0·029 0·332, 6·078 0·154 0·009 0·039, 0·270 - - Policy effect (significant covariates) 3·350 0·001 1·304, 5·395 3·245 0·003 1·128, 5·361 0·154 0·009 0·039, 0·270 - - Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -4·389 0·117 -9·877, 1·098 -0·229 0·873 -3·037, 2·579 -0·072 0·068 -0·149, 0·005 0·148 0·519 Policy effect (significant covariates) -2·734 0·273 -7·623, 2·155 -0·403 0·779 -3·224, 2·418 -0·072 0·067 -0·149, 0·005 *Effect size for synthetic control adjusted by the covariates, number of on-premises alcohol outlets (population adjusted), and per capita gross disposable household income . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 30 * Dots represent weekly data; solid lines represent 8-week moving averages; vertical dashed line represents start of intervention Figure 1: Number of weekend night-time alcohol-related ambulance call-outs and reported crimes over-time in the intervention and control cities . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 31 Figure 2. Distribution of weekend night-time alcohol-related ambulance call-out and reported crimes in Aberdeen and Glasgow * Dots represent weekly data; solid lines represent 8-week moving averages; vertical dashed line represents start of intervention Figure 3: Rate of weekend night-time alcohol-related ambulance call-outs and reported crimes in Aberdeen, Glasgow and synthetic Aberdeen and synthetic Glasgow. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 32 Appendix Contents Appendix 1: Control selection process ................................ ................................ ..... 33 Appendix 1.1. Alcohol-related ambulance call-outs ................................ ............... 34 Appendix 1.2. Reported crimes ................................ ................................ ............ 36 Appendix 2: Analysis ................................ ................................ .............................. 40 Appendix 2.1. Descriptive analysis ................................ ................................ ....... 40 Appendix 2.2. Description of confounding variables ................................ .............. 45 Appendix 2.2.1. Data sources for confounding variables................................ ....... 45 Appendix 2.3. Secondary and Sub-group analysis ................................ ................. 46 Appendix 2.4: Falsification test ................................ ................................ ............ 54 Appendix 3. Synthetic control ................................ ................................ ................. 59 Appendix 3.1. Alcohol-related ambulance call-outs (Aberdeen) .............................. 59 Appendix 3.1.1. Model specification and validation test ................................ ........ 59 Appendix 3.1.2. Main analysis and sensitivity tests ................................ ............... 59 Appendix 3.2. Alcohol-related ambulance call-outs (Glasgow)................................ 66 Appendix 3.2.1. Model specification and validation test ................................ ........ 66 Appendix 3.2.2. Main analysis and sensitivity tests ................................ ............... 66 Appendix 3.3. Reported crimes (Aberdeen) ................................ .......................... 72 Appendix 3.3.1. Model specification and validation test ................................ ........ 72 Appendix 3.3.2. Main analysis and sensitivity tests ................................ ............... 72 Appendix 3.4. Reported crimes (Glasgow) ................................ ............................ 79 Appendix 3.4.1. Model specification and validation test ................................ ........ 79 Appendix 3.4.2. Main analysis and sensitivity tests ................................ ............... 79 Appendix 4. Synthetic control and ARIMA ................................ .............................. 86 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 33 Appendix 1: Control selection process and study design Figure A1: Flow diagram for the control selection procedure . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 34 Appendix 1.1. Alcohol-related ambulance call-outs In step 1, we tested the parallel trend assumption both graphically and statistically for 30 potential Scottish cities/council areas, excluding Aberdeen and Glasgow. We estimated the slope coefficients for Aberdeen and Glasgow against each of the 30 potential control candidates, using data from the pre-intervention periods only. Any control candidate with a statistically significant slope coefficient, indicating a statistically significant difference in trends from Aberdeen or Glasgow, was excluded. As a result, Stirling was excluded as a control for Aberdeen, while Angus, Argyll and Bute, Dundee, East Ayrshire, Highland, Perth and Kinross, Scottish Borders, and Stirling were excluded as controls for Glasgow. In step 2, we reviewed the licensing policies of the remaining potential control cities/councils and excluded those with similar extended premises hours policies to Aberdeen or Glasgow. Additionally, we removed candidates whose population-adjusted density of on-premises alcohol outlets was more than one standard deviation above or below the densities of the intervention areas (Aberdeen and Glasgow). In step 3, we computed the mean squared prediction errors (MSPEs) for all remaining control candidates. The MSPE represents the mean squared differences between the intervention areas (Aberdeen/Glasgow) and the potential control candidates during the pre-intervention periods. For Aberdeen, the candidates with the lowest MSPEs were Edinburgh, Fife, and North Lanarkshire, while for Glasgow, Edinburgh had the lowest MSPE. Ultimately, Edinburgh was chosen as the control for both Aberdeen and Glasgow. For Aberdeen, this decision was informed by Edinburgh’s significant contribution in synthetic control analysis compared to Fife and North Lanarkshire, alongside our a priori decision to use Edinburgh as a control. Step 1: Testing parallel trend assumptions graphically and statistically Parallel trend assumption test: Ambulance call-outs (based on rate) Council name Aberdeen Glasgow Slope P-value Slope P-value Aberdeenshire -0·0009 0·268 -0·0001 0·763 Angus -0·0002 0·826 -0·0006 0·049 Argyll and Bute -0·0022 0·054 -0·0016 <0·001 City of Edinburgh -0·0017 0·061 0·0003 0·248 Clackmannanshire -0·0009 0·554 -0·0006 0·240 Dumfries and Galloway -0·0007 0·468 -0·0004 0·191 Dundee City -0·0007 0·514 -0·0009 0·010 East Ayrshire -0·0014 0·169 -0·0009 0·010 East Dunbartonshire -0·0011 0·258 0·0004 0·097 East Lothian 0·0003 0·760 0·0000 0·913 East Renfrewshire -0·0019 0·050 -0·0002 0·384 Falkirk 0·0003 0·803 -0·0003 0·319 Fife -0·0007 0·389 -0·0003 0·186 Highland 0·0006 0·469 -0·0006 0·007 Inverclyde 0·0013 0·321 0·0005 0·219 Midlothian -0·0005 0·639 -0·0001 0·847 Moray -0·0006 0·612 -0·0007 0·055 North Ayrshire 0·0004 0·757 0·0004 0·221 North Lanarkshire 0·0000 0·989 0·0003 0·286 Orkney Islands 0·0003 0·825 0·0005 0·307 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 35 Perth and Kinross -0·0014 0·169 0·0007 0·009 Renfrewshire 0·0001 0·925 0·0003 0·325 Scottish Borders -0·0018 0·095 0·0012 0·001 Shetland Islands -0·0008 0·587 -0·0002 0·718 South Ayrshire -0·0005 0·688 -0·0004 0·186 South Lanarkshire -0·0013 0·195 -0·0002 0·536 Stirling -0·0026 0·019 -0·0008 0·017 West Dunbartonshire 0·0014 0·246 -0·0003 0·477 West Lothian 0·0001 0·912 -0·0001 0·620 Western Isles - Eilean Siar -0·0001 0·945 -0·0008 0·090 Step 2: Assess qualitatively and quantitatively the councils/cities that passed parallel test assumption at Step 1 Potential control candidates that satisfied parallel trend assumption Aberdeen (no. of on sale licenses, 432; density per 1000 population, 1·90 (SD 0·87)) Glasgow (no. of on sale licenses, 1352; density per 1000 population, 2·13 (SD 0·85)) Council name Number of on sale licenses, 2022 Density per 1000 population Council name Number of on sale licenses, 2022 Density per 1000 population Aberdeenshire 370 1·40 Aberdeenshire** 370 1·40 Angus 246 2·12 City of Edinburgh 1467 2·75 Argyll and Bute** 451 5·35 Clackmannanshire* 76 1·48 City of Edinburgh 1467 2·75 Dumfries and Galloway** 456 3·10 Clackmannanshire* 76 1·48 East Dunbartonshire* 119 1·08 Dumfries and Galloway** 456 3·10 East Lothian 213 1·95 Dundee City 307 2·07 East Renfrewshire** 118 1·21 East Ayrshire 181 1·49 Falkirk** 199 1·22 East Dunbartonshire* 119 1·08 Fife 681 1·83 East Lothian 213 1·95 Inverclyde 125 1·64 East Renfrewshire 118 1·21 Midlothian 132 1·37 Falkirk 199 1·22 Moray 262 2·74 Fife 681 1·83 North Ayrshire 255 1·91 Highland** 922 3·90 North Lanarkshire** 394 1·16 Inverclyde 125 1·64 Orkney Islands 66 2·96 Midlothian 132 1·37 Renfrewshire 279 1·55 Moray 262 2·74 Shetland Islands** 104 4·53 North Ayrshire 255 1·91 South Ayrshire 271 2·42 North Lanarkshire 394 1·16 South Lanarkshire* 457 1·42 Orkney Islands** 66 2·96 West Dunbartonshire 140 1·58 Perth and Kinross** 436 2·86 West Lothian** 211 1·13 Renfrewshire 279 1·55 Western Isles - Eilean Siar 67 2·56 Scottish Borders** 343 2·96 Shetland Islands** 104 4·53 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 36 South Ayrshire 271 2·42 South Lanarkshire* 457 1·42 West Dunbartonshire 140 1·58 West Lothian 211 1·13 Western Isles - Eilean Siar 67 2·56 *Excluded due to extension of alcohol premises hours **Excluded if density of on-sale premises, ± 1 SD that of Aberdeen and Glasgow Step 3: Calculate mean square prediction error based on pre-intervention data to choose control candidate based on lowest value Potential control candidates satisfied parallel trend assumption and passed qualitative and quantitative assessment Intervention: Aberdeen Intervention: Glasgow Council name Mean squared prediction error (MSPE) Council name Mean squared prediction error (MSPE) Aberdeenshire 0·23 City of Edinburgh 0·07 Angus 0·14 East Lothian 0·29 City of Edinburgh 0·07 Fife 0·13 Dundee City 0·10 Inverclyde 0·16 East Ayrshire 0·11 Midlothian 0·20 East Lothian 0·16 Moray 0·24 East Renfrewshire 0·28 North Ayrshire 0·11 Falkirk 0·11 Orkney Islands 0·63 Fife 0·07 Renfrewshire 0·13 Inverclyde 0·14 South Ayrshire 0·16 Midlothian 0·12 West Dunbartonshire 0·16 Moray 0·16 Western Isles - Eilean Siar 0·45 North Ayrshire 0·12 North Lanarkshire 0·07 Renfrewshire 0·08 South Ayrshire 0·13 West Dunbartonshire 0·12 West Lothian 0·12 Western Isles - Eilean Siar 0·31 Appendix 1.2. Reported crimes In step 1, for reported crimes, Shetland Islands was excluded as a control for Aberdeen, while Aberdeenshire, East Dunbartonshire, East Lothian, East Renfrewshire, Highland, North Lanarkshire, South Ayrshire, and South Lanarkshire were excluded as controls for Glasgow. In step 2, Argyll and Bute, Clackmannanshire, Dumfries and Galloway, East Dunbartonshire, Highland, Orkney Islands, Perth and Kinross, Scottish Borders, South Lanarkshire, and Stirling . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 37 excluded as controls for Aberdeen. Similarly, for Glasgow, Argyll and Bute, Clackmannanshire, Dumfries and Galloway, Falkirk, Orkney Islands, Perth and Kinross, Scottish Borders, Shetland Islands, and West Lothian were excluded. In step 3, Edinburgh was chosen as the control for both Aberdeen and Glasgow for reported crimes based on the lowest MSPEs among the surviving control candidates. Step 1: Testing parallel trend assumptions graphically and statistically Parallel trend assumption test: Police call-outs (based on rate) Council name Aberdeen Glasgow Slope P-value Slope P-value Aberdeenshire 0·0000 0·973 -0·0008 0·012 Angus -0·0015 0·289 -0·0003 0·495 Argyll and Bute -0·0003 0·866 -0·0009 0·078 City of Edinburgh 0·0006 0·558 -0·0002 0·538 Clackmannanshire 0·0007 0·665 -0·0002 0·695 Dumfries and Galloway 0·0004 0·757 -0·0006 0·163 Dundee City -0·0002 0·899 0·0001 0·772 East Ayrshire -0·0019 0·168 0·0001 0·851 East Dunbartonshire -0·0008 0·514 -0·0011 0·001 East Lothian -0·0001 0·924 -0·0011 0·007 East Renfrewshire -0·0010 0·461 -0·0007 0·034 Falkirk 0·0012 0·414 -0·0007 0·061 Fife 0·0005 0·643 -0·0002 0·505 Highland 0·0024 0·081 0·0007 0·042 Inverclyde -0·0002 0·911 -0·0009 0·070 Midlothian 0·0011 0·492 -0·0005 0·210 Moray 0·0003 0·837 -0·0005 0·305 North Ayrshire 0·0008 0·512 0·0003 0·428 North Lanarkshire -0·0006 0·562 -0·0009 0·001 Orkney Islands 0·0024 0·369 -0·0007 0·349 Perth and Kinross 0·0002 0·878 -0·0005 0·118 Renfrewshire 0·0002 0·834 -0·0007 0·079 Scottish Borders 0·0004 0·758 -0·0006 0·075 Shetland Islands 0·0067 0·009 0·0012 0·101 South Ayrshire 0·0009 0·569 -0·0008 0·035 South Lanarkshire 0·0000 0·971 -0·0008 0·005 Stirling 0·0012 0·433 -0·0001 0·809 West Dunbartonshire 0·0022 0·207 -0·0002 0·687 West Lothian 0·0002 0·868 -0·0005 0·106 Western Isles - Eilean Siar 0·0036 0·063 -0·0006 0·325 Step 2: Assess qualitatively and quantitatively to the councils/cities passed parallel test assumption . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 38 Potential control candidates satisfied parallel trend assumption Aberdeen (no. of on sale licenses, 432; density per 1000 population, 1·90 (SD 0·87)) Glasgow (no. of on sale licenses, 1352; density per 1000 population, 2·13 (SD 0·85)) Council name Number of on sale licenses, 2022 Council name Number of on sale licenses, 2022 Council name Number of on sale licenses, 2022 Aberdeenshire 370 1·40 Angus 246 2·12 Angus 246 2·12 Argyll and Bute** 451 5·35 Argyll and Bute** 451 5·35 City of Edinburgh 1467 2·75 City of Edinburgh 1467 2·75 Clackmannanshire* 76 1·48 Clackmannanshire* 76 1·48 Dumfries and Galloway** 456 3·10 Dumfries and Galloway** 456 3·10 Dundee City 307 2·07 Dundee City 307 2·07 East Ayrshire 181 1·49 East Ayrshire 181 1·49 Falkirk** 199 1·22 East Dunbartonshire* 119 1·08 Fife 681 1·83 East Lothian 213 1·95 Inverclyde 125 1·64 East Renfrewshire 118 1·21 Midlothian 132 1·37 Falkirk 199 1·22 Moray 262 2·74 Fife 681 1·83 North Ayrshire 255 1·91 Highland** 922 3·90 Orkney Islands** 66 2·96 Inverclyde 125 1·64 Perth and Kinross* 436 2·86 Midlothian 132 1·37 Renfrewshire 279 1·55 Moray 262 2·74 Scottish Borders** 343 2·96 North Ayrshire 255 1·91 Shetland Islands** 104 4·53 North Lanarkshire 394 1·16 Stirling 273 2·84 Orkney Islands** 66 2·96 West Dunbartonshire 140 1·58 Perth and Kinross* 436 2·86 West Lothian** 211 1·13 Renfrewshire 279 1·55 Western Isles - Eilean Siar 67 2·56 Scottish Borders** 343 2·96 South Ayrshire 271 2·42 South Lanarkshire* 457 1·42 Stirling** 273 2·84 West Dunbartonshire 140 1·58 West Lothian 211 1·13 Western Isles - Eilean Siar 67 2·56 *Excluded due to extension of alcohol premises hours **Excluded if density of on-sale premises, ± 1 SD that of Aberdeen and Glasgow Step 3: Calculate mean square prediction error based on pre-intervention data to choose control candidate based on lowest value. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 39 Potential control candidates satisfied parallel trend assumption and passed qualitative and quantitative assessment Council name Mean squared prediction error (MSPE) Council name Mean squared prediction error (MSPE) Aberdeenshire 0·26 Angus 0·23 Angus 0·19 City of Edinburgh 0·09 City of Edinburgh 0·09 Dundee City 0·12 Dundee City 0·15 East Ayrshire 0·25 East Ayrshire 0·19 Fife 0·16 East Lothian 0·37 Inverclyde 0·34 East Renfrewshire 0·50 Midlothian 0·28 Falkirk 0·18 Moray 0·24 Fife 0·13 North Ayrshire 0·19 Inverclyde 0·31 Orkney Islands 0·63 Midlothian 0·26 Renfrewshire 0·22 Moray 0·20 Scottish Borders 0·36 North Ayrshire 0·15 Stirling 0·32 North Lanarkshire 0·15 West Dunbartonshire 0·23 Renfrewshire 0·16 Western Isles - Eilean Siar 0·67 South Ayrshire 0·26 West Dunbartonshire 0·27 West Lothian 0·13 Western Isles - Eilean Siar 0·57 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 40 Appendix 2: Analysis Appendix 2.1. Descriptive analysis Table A1. Total number of weekend night-time alcohol- related ambulance call-outs and reported crimes by Scottish cities from 1 May 2015 to 31 July 2022 Scottish cities Alcohol-related ambulance call- outs Reported crimes Aberdeen City 6,461 5,927 Aberdeenshire 3,399 3,655 Angus 2,294 2,358 Argyll & Bute 1,897 1,767 City of Edinburgh 13,781 12,268 Clackmannanshire 1,399 1,172 Dumfries & Galloway 3,104 3,636 Dundee City 4,483 4,595 East Ayrshire 2,609 2,269 East Dunbartonshire 1,271 1,047 East Lothian 1,792 1,434 East Renfrewshire 1,015 781 Eilean Siar 402 334 Falkirk 3,824 3,680 Fife 8,754 7,900 Glasgow City 19,956 20,038 Highland 5,026 4,665 Inverclyde 1,902 1,457 Midlothian 1,937 1,768 Moray 1,937 2,005 North Ayrshire 3,786 2,888 North Lanarkshire 8,216 6,906 Orkney Islands 252 287 Perth & Kinross 2,892 2,247 Renfrewshire 4,189 3,624 Scottish Borders 2,608 1,989 Shetland Islands 350 307 South Ayrshire 2,748 2,198 South Lanarkshire 6,932 5,580 Stirling 2,003 1,656 West Dunbartonshire 2,331 1,853 West Lothian 4,085 3,536 Total 127,635 115,827 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 41 Table A2. Age, sex and socio-economic deprivation* distribution of patients in alcohol-related ambulance call-outs at the weekend night-time Patient characteristics Aberdeen, n(%) Glasgow, n(%) Edinburgh, n(%) Age group (years) < 25 1,786 (27·64) 4,609 (23·1) 3,954 (28·69) 25-34 1,136 (17·58) 3,531 (17·69) 2,366 (17·17) 35-44 849 (13·14) 2,827 (14·17) 1,671 (12·13) 45+ 2,190 (33·9) 7,588 (38·02) 4,828 (35·03) Unknown 500 (7·74) 1,401 (7·02) 962 (6·98) Gender Female 2,649 (41·00) 7,300 (36·58) 5,655 (41·03) Male 3,576 (55·35) 11,892 (59·59) 7,663 (55·61) Others - 3 (0·02) 4 (0·03) Unknown 236 (3·65) 761 (3·81) 459 (3·33) Socio-economic deprivation 1 (most deprived) 111 (1·72) 7,385 (37·01) 1,189 (8·63) 2 580 (8·98) 2,868 (14·37) 1,311 (9·51) 3 1,012 (15·66) 1,492 (7·48) 1,141 (8·28) 4 600 (9·29) 1,278 (6·40) 1,852 (13·44) 5 1,387 (21·47) 1,342 (6·72) 1,826 (13·25) 6 624 (9·66) 2,447 (12·26) 1,467 (10·65) 7 627 (9·70) 601 (3·01) 914 (6·63) 8 431 (6·67) 1,655 (8·29) 1,422 (10·32) 9 423 (6·55) 619 (3·1) 795 (5·77) 10 (less deprived) 666 (10·31) 269 (1·35) 1,864 (13·53) Total 6,461 19,956 13,781 *Socio-economic deprivation score is based on the postcode of the incident location Table A3. Summary of weekend night-time alcohol-related ambulance call-outs by cities City Year Minimum Maximum Mean SD Median IQR Total Interventions Aberdeen 2015 10 36 18 5 18 7 636 2016 4 26 15 5 16 7 800 2017 9 32 19 5 19 8 997 2018 9 36 20 5 20 6 1,042 2019 9 37 21 6 20 8 1,066 2020 4 26 15 5 14 6 763 2021 5 25 14 5 15 7 752 2022 1 25 14 5 14 6 405 Total 1 37 17 6 17 8 6,461 Glasgow 2015 42 89 59 9 58 12 2,069 2016 33 78 55 10 56 16 2,878 2017 35 85 60 10 60 13 3,143 2018 37 92 64 10 63 12 3,312 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 42 2019 38 78 59 9 58 13 3,063 2020 22 62 41 9 42 14 2,154 2021 21 62 40 11 40 14 2,067 2022 1 58 42 14 45 15 1,270 Total 1 92 53 14 54 17 19,956 Control Edinburgh 2015 25 61 40 10 38 16 1,413 2016 24 63 41 9 38 12 2,115 2017 27 57 41 7 40 10 2,124 2018 26 70 42 8 41 9 2,207 2019 20 59 41 8 42 9 2,151 2020 14 54 29 8 28 11 1,485 2021 11 48 27 8 28 13 1,423 2022 2 44 30 8 30 8 863 Total 2 70 37 10 36 13 13,781 Table A4· Total number of weekend night-time reported crimes according to crime types from 1 May 2015 to 31 July 2022 Type of reported crimes Aberdeen, n(%) Glasgow, n(%) Edinburgh, n(%) Common Assault 4,172 (70·39) 13,702 (68·38) 9,493 (77·38) Common assault of an emergency worker 765 (12·91) 3,091 (15·43) 1,135 (9·25) Consuming outwith permitted hours - - 4 (0·03) Culpable and reckless, causing injury 3 (0·05) 13 (0·06) 9 (0·07) Disorderly on licensed premises 104 (1·75) 200 (1) 60 (0·49) Drunk and Incapable 5 (0·08) 294 (1·47) 99 (0·81) Drunk and attempting to enter licensed premises 22 (0·37) 24 (0·12) 5 (0·04) Drunk in charge of a child 5 (0·08) 24 (0·12) 6 (0·05) Drunk in or attempting to enter designated sports ground 1 (0·02) 2 (0·01) - Licensed person, employee or agent drunk in licensed premise 1 (0·02) 3 (0·01) 6 (0·05) Licensed persons, other offences 2 (0·03) 9 (0·04) 5 (0·04) Liquor licensing laws, other offences - 8 (0·04) - Person under 18 buying excisable liquor in bar - - 2 (0·02) Permitting riotous behaviour in licensed premise - 7 (0·03) - Purchasing excise liquor for consumption by person under 18 1 (0·02) - - Refusing to quit a licensed premise 207 (3·49) 335 (1·67) 205 (1·67) Robbery 131 (2·21) 408 (2·04) 306 (2·49) Sale of drink to person under 18 - 7 (0·03) - Serious Assault 474 (8) 1,841 (9·19) 858 (6·99) Sports grounds offences possessing alcohol 1 (0·02) 9 (0·04) 2 (0·02) Threats and Extortion 33 (0·56) 61 (0·3) 73 (0·6) Total reported crimes 5,927 20,038 12,268 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 43 Table A5. Summary of weekend night-time reported crimes by cities City Year Minimum Maximum Mean SD Median IQR Total Interventions Aberdeen 2015 10 42 21 7 20 8 748 2016 7 30 19 7 19 13 984 2017 7 54 19 8 17 9 972 2018 4 37 14 6 12 6 722 2019 5 33 16 6 16 7 845 2020 1 37 10 6 10 6 512 2021 1 33 13 7 12 10 694 2022 7 32 15 6 13 9 450 Total 1 54 16 8 15 10 5,927 Glasgow 2015 43 108 66 13 64 16 2,305 2016 28 90 60 14 61 14 3,118 2017 37 174 62 19 57 16 3,201 2018 33 86 53 10 52 13 2,760 2019 29 78 54 10 55 11 2,810 2020 12 72 37 14 33 17.5 1,912 2021 23 77 45 15 42.5 22.5 2,321 2022 21 74 52 13 48 17 1,611 Total 12 174 53 16 54 20 20,038 Control Edinburgh 2015 23 83 44 12 42 13 1,536 2016 18 63 38 8 38 9 1,974 2017 24 107 38 12 36 12 1,992 2018 16 62 36 9 35 11 1,856 2019 18 53 32 7 32 9 1,660 2020 8 60 22 10 20 13 1,135 2021 10 52 25 10 24 15 1,292 2022 10 39 27 7 26 10 823 Total 8 107 32 12 33 14 12,268 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 44 Figure A2. Distribution of weekend night-time alcohol-related ambulance call-out and reported crimes in Edinburgh . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 45 Figure A3: Differences (intervention city minus (-) control city) of number of weekend night- time alcohol-related ambulance call-outs and reported crimes over-time between intervention and control cities Appendix 2.2. Description of confounding variables We included several confounding variables in our model, such as per capita gross disposable household income, weather conditions (mean temperature and rainfall), and the total number of on- premises alcohol outlets. These variables were incorporated as differences between the intervention and control areas (e.g., number of on-premises outlets in Aberdeen minus the number in Edinburgh). Additionally, we adjusted for dummy variables representing COVID-19 restrictions, public holidays, and outliers. These dummy variables were binary, taking the value “1” to indicate the presence of a restriction, holiday, or outlier, and “0” to indicate their absence. Public holiday data were extracted separately for each Scottish city. Outliers were defined as data points that fell more than 1.5 times the interquartile range above the upper quartile or below the lower quartile. To account for COVID-19 restrictions, we used city-specific lockdown policies based on the type of premises, as issued by the Scottish Government. While restrictions on bars, clubs, and nightclubs were implemented simultaneously across Scotland on 21st March 2020, those affecting nightclubs remained in place for a longer period compared to bars and pubs. We also included a time trend variable in the model to account for temporal effects, if it was statistically significant. Appendix 2.2.1. Data sources for confounding variables 1. Per capita gross disposable household income [34] 2. Weather conditions (mean temperature, rainfall) [35] 3. Number of on-premises alcohol outlets [36] 4. COVID-19 lockdown (imposing closure or restricted hours for alcohol-premises) [37] 5. Public holidays [38] . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 46 Appendix 2.3. Secondary and Sub-group analysis Table A6. Effect of policy changes on weekend restricted night-time (24:00 to 05:59) alcohol-related ambulance call-outs in Aberdeen and Glasgow Primary outcome, weekend restricted night-time (24:00 to 05:59) alcohol- related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 3·674 <0·001 1·796, 5·552 3·805 0·035 0·267, 7·343 0·139 0·179 -0·064, 0·343 Policy effect (significant covariates) 4·418 <0·001 2·868, 5·969 3·970 0·021 0·609, 7·331 0·071 0·024 0·009, 0·134 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 3·755 0·013 0·780, 6·731 2·950 0·070 -0·235, 6·136 0·099 0·104 -0·021, 0·219 Policy effect (significant covariates) 3·726 0·014 0·753, 6·700 2·140 0·316 -2·041, 6·322 0·080 0·003 0·027, 0·134 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 2·482 0·068 -0·179, 5·143 2·286 0·103 -0·459, 5·031 0·029 0·554 -0·067, 0·125 Policy effect (significant covariates) 3·261 <0·001 1·857, 4·666 2·352 0·068 -0·172, 4·875 0·046 0·106 -0·010, 0·101 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -0·996 0·453 -3·600, 1·608 1·865 0·221 -1·119, 4·849 -0·010 0·577 -0·043, 0·024 Policy effect (significant covariates) -1·269 0·139 -2·947, 0·410 2·183 0·070 -0·180, 4·546 -0·001 0·920 -0·024, 0·022 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 47 Table A7. Effect of policy changes on weekend restricted night-time (20:00 to 23:59) alcohol-related ambulance call-outs in Aberdeen and Glasgow Primary outcome, weekend restricted night-time (20:00 to 23:59) alcohol- related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 1·716 0·041 0·072, 3·360 0·497 0·631 -1·528, 2·521 0·117 0·044 0·003, 0·231 Policy effect (significant covariates) 2·408 0·001 1·000, 3·816 0·440 0·669 -1·573, 2·453 0·083 <0·001 0·050, 0·117 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -0·196 0·873 -2·606, 2·214 0·304 0·823 -2·365, 2·973 0·007 0·842 -0·065, 0·080 Policy effect (significant covariates) -0·334 0·751 -2·396, 1·728 -0·014 0·983 -1·365, 1·337 0·007 0·859 -0·066, 0·079 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) -2·118 0·038 -4·115, -0·120 0·093 0·935 -2·135, 2·322 -0·044 0·180 -0·108, 0·020 Policy effect (significant covariates) -2·040 0·113 -4·564, 0·484 -0·163 0·845 -1·792, 1·466 -0·046 0·156 -0·110, 0·018 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -2·642 0·067 -5·468, 0·183 -2·281 0·182 -5·630, 1·068 -0·030 0·144 -0·069, 0·010 Policy effect (significant covariates) -3·218 <0·001 -4·701, -1·734 -2·183 0·066 -4·513, 0·147 -0·040 0·002 -0·065, -0·014 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 48 Table A8. Effect of policy changes on weekend night-time all ambulance call-outs in Aberdeen and Glasgow Secondary outcome, weekend night- time all ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficien t P-value 95% CI Coefficien t P- value 95% CI Coefficien t P-value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 11·761 0·005 3·550, 19·972 6·373 0·139 -2·070, 14·815 0·328 <0·001 0·189, 0·467 Policy effect (significant covariates) 12·895 <0·001 6·003, 19·787 6·055 0·161 -2·406, 14·517 0·284 <0·001 0·145, 0·424 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 5·788 0·197 -3·011, 14·587 4·204 0·134 -1·297, 9·704 0·240 0·006 0·068, 0·411 Policy effect (significant covariates) 7·337 0·102 -1·469, 16·143 3·997 0·157 -1·543, 9·537 0·202 0·010 0·048, 0·357 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 8·816 0·010 2·119, 15·512 5·294 0·083 -0·683, 11·272 0·227 0·001 0·094, 0·360 Policy effect (significant covariates) 9·704 0·001 4·050, 15·358 4·593 0·122 -1·234, 10·419 0·208 0·001 0·084, 0·331 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -5·843 0·200 -14·787, 3·102 -3·437 0·468 -12·711, 5·836 -0·035 0·619 -0·173, 0·103 Policy effect (significant covariates) -5·990 0·084 -12·779, 0·798 -1·266 0·763 -9·509, 6·976 -0·021 0·674 -0·117, 0·075 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 49 Table A9. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow among female sub-group Primary outcome, weekend night- time alcohol-related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 2·481 0·006 0·724, 4·237 8·601 0·012 1·856, 15·346 0·080 <0·001 0·045, 0·114 Policy effect (significant covariates) 3·661 <0·001 2·285, 5·037 2·009 0·044 0·058, 3·959 0·081 <0·001 0·047, 0·115 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 0·988 0·162 -0·396, 2·372 0·928 0·168 -0·392, 2·248 0·065 <0·001 0·033, 0·097 Policy effect (significant covariates) 0·980 0·168 -0·413, 2·373 0·974 0·152 -0·358, 2·307 0·055 <0·001 0·027, 0·083 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 1·516 0·059 -0·057, 3·090 1·180 0·158 -0·457, 2·818 -0·002 0·940 -0·052, 0·048 Policy effect (significant covariates) 1·932 0·013 0·414, 3·450 1·149 0·148 -0·407, 2·705 -0·008 0·762 -0·060, 0·044 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -1·597 0·097 -3·485, 0·291 -2·060 0·107 -4·565, 0·444 -0·017 0·232 -0·044, 0·011 Policy effect (significant covariates) -2·104 <0·001 -3·070, -1·139 -2·491 0·009 -4·371, -0·611 -0·021 0·014 -0·037, -0·004 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 50 Table A10. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow among male sub-group Primary outcome, weekend night- time alcohol-related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficien t P- value 95% CI Coefficient P- value 95% CI Coefficien t P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 3·401 0·001 1·357, 5·444 4·045 0·00 1 1·552, 6·537 0·059 0·021 0·009, 0·109 Policy effect (significant covariates) 3·726 <0·001 1·946, 5·507 3·995 0·00 2 1·487, 6·502 0·055 0·025 0·007, 0·102 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 1·943 0·019 0·321, 3·566 2·336 0·00 4 0·734, 3·938 0·070 0·004 0·023, 0·117 Policy effect (significant covariates) 1·977 0·022 0·280, 3·673 2·261 0·01 7 0·404, 4·119 0·080 <0·001 0·037, 0·123 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 1·894 0·025 0·243, 3·545 2·114 0·04 0 0·098, 4·131 0·030 0·196 -0·015, 0·075 Policy effect (significant covariates) 2·315 0·002 0·815, 3·816 1·818 0·07 7 -0·200, 3·835 0·030 0·148 -0·011, 0·070 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) 2·392 0·300 -2·131, 6·916 -0·395 0·78 0 -3·162, 2·372 0·036 0·382 -0·044, 0·115 Policy effect (significant covariates) 1·463 0·500 -2·783, 5·709 -2·150 0·05 3 -4·330, 0·029 0·019 0·616 -0·054, 0·092 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 51 Table A11. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow among patients age less than 45 years Primary outcome, weekend night- time alcohol-related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P-value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 6·180 0·002 2·251, 10·109 15·626 <0·001 10·292, 20·959 0·210 0·002 0·074, 0·345 Policy effect (significant covariates) 7·050 <0·001 4·077, 10·023 14·904 <0·001 9·983, 19·825 0·105 <0·001 0·065, 0·145 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 1·527 0·648 -5·026, 8·080 1·897 0·406 -2·577, 6·371 0·066 0·123 - 0·018, 0·150 Policy effect (significant covariates) 1·355 0·705 -5·670, 8·380 1·860 0·390 -2·384, 6·104 0·088 <0·001 0·050, 0·127 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 1·500 0·873 -16·828, 19·827 2·821 0·245 -1·936, 7·577 0·025 0·423 - 0·037, 0·088 Policy effect (significant covariates) 3·732 0·562 -8·876, 16·340 2·595 0·179 -1·188, 6·378 0·068 <0·001 0·035, 0·101 Intervention area: Glasgow Dummy policy covariate . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 52 Policy effect (all covariates) -0·889 0·516 -3·575, 1·797 0·908 0·577 -2·282, 4·099 -0·011 0·633 - 0·054, 0·033 Policy effect (significant covariates) -2·543 0·014 -4·579, -0·506 2·250 0·110 -0·510, 5·010 -0·013 0·513 - 0·050, 0·025 Table A12. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow among patients age more than 45 years Primary outcome, weekend night- time alcohol-related ambulance call-outs Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 0·891 0·689 -3·478, 5·261 0·315 0·877 -3·664, 4·295 -0·010 0·870 -0·131, 0·111 Policy effect (significant covariates) 0·806 0·009 0·203, 1·408 0·357 0·681 -1·346, 2·060 0·033 0·054 -0·001, 0·066 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 0·925 0·288 -0·783, 2·633 0·886 0·355 -0·992, 2·764 0·018 0·544 -0·040, 0·076 Policy effect (significant covariates) 0·679 0·008 0·181, 1·178 0·499 0·365 -0·580, 1·578 0·046 0·004 0·015, 0·077 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 0·163 0·857 -1·616, 1·943 -0·341 0·697 -2·057, 1·376 0·008 0·761 -0·046, 0·062 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 53 Policy effect (significant covariates) 0·571 0·053 -0·008, 1·150 0·043 0·948 -1·243, 1·328 0·025 0·076 -0·003, 0·053 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -1·139 0·458 -4·144, 1·867 -4·629 <0·001 -6·691, -2·567 -0·055 <0·001 -0·084, - 0·027 Policy effect (significant covariates) -1·566 0·293 -4·484, 1·352 -3·452 <0·001 -4·713, -2·190 -0·066 <0·001 -0·090, - 0·042 Table A13. Effect of policy changes on weekend restricted night-time (24:00 to 05:59) recorded crimes in Aberdeen and Glasgow Secondary outcome, weekend restricted night-time (24:00 to 05:59) recorded crimes Main analysis Sensitivity analysis Outcome: number of incidents Restricted time series only before COVID-19 restrictions, (outcome: number of incidents) Outcome: population adjusted incident rates Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) 4·534 0·071 -0·385, 9·454 3·752 0·179 -1·721, 9·226 0·034 0·717 -0·152, 0·221 Policy effect (significant covariates) 2·853 0·002 1·073, 4·634 3·483 0·003 1·153, 5·814 0·023 0·497 -0·043, 0·090 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -0·060 0·967 -2·917, 2·797 -0·062 0·970 -3·239, 3·116 0·004 0·950 -0·110, 0·118 Policy effect (significant covariates) 1·572 0·028 0·174, 2·970 1·805 0·033 0·150, 3·460 0·031 0·325 -0·030, 0·091 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 54 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 2·757 0·007 0·759, 4·755 2·291 0·042 0·079, 4·503 0·076 0·023 0·011, 0·142 Policy effect (significant covariates) 2·642 <0·001 1·271, 4·013 2·821 0·001 1·171, 4·472 0·048 0·043 0·002, 0·094 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -6·301 0·003 -10·423, - 2·178 -1·869 0·280 -5·256, 1·518 -0·108 0·002 -0·175, -0·040 Policy effect (significant covariates) -5·863 0·003 -9·687, - 2·040 -0·625 0·654 -3·356, 2·106 -0·100 0·002 -0·164, -0·035 Appendix 2.4: Falsification test Table A14. Falsification test for weekend night-time alcohol-related ambulance call-outs and all ambulance call-outs in Aberdeen and Glasgow Primary outcome: weekend night- time alcohol-related ambulance call-outs Primary outcome: weekend restricted night-time (24:00 to 05:59) alcohol-related ambulance call-outs Primary outcome: weekend restricted night-time (20:00 to 23:59) alcohol-related ambulance call-outs Secondary outcome: weekend night- time all ambulance call-outs Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) -10·095 0·056 -20·464, 0·273 2·118 0·199 -1·115, 5·352 -6·320 0·066 -13·069, 0·429 -23·557 0·012 -41·914, -5·199 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 55 Policy effect (significant covariates) -6·997 0·066 -14·458, 0·464 2·135 0·187 -1·038, 5·308 -5·982 0·003 -9·945, -2·019 -13·751 0·050 -27·518, 0·016 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -4·234 0·071 -8·825, 0·356 -1·623 0·247 -4·372, 1·126 -1·112 0·401 -3·707, 1·483 -9·540 0·016 -17·266, -1·815 Policy effect (significant covariates) -3·887 0·099 -8·507, 0·732 0·260 0·880 -3·129, 3·650 -2·629 0·012 -4·685, -0·573 -9·176 0·018 -16·752, -1·601 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 3·104 0·019 0·500, 5·708 2·962 0·202 -1·590, 7·513 0·226 0·913 -3·851, 4·304 -4·256 0·535 -17·705, 9·194 Policy effect (significant covariates) 3·096 0·020 0·493, 5·699 1·233 0·252 -0·879, 3·345 -0·315 0·788 -2·617, 1·986 -1·864 0·651 -9·949, 6·221 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) 1·777 0·533 -3·804, 7·357 -1·301 0·493 -5·022, 2·419 2·178 0·204 -1·181, 5·536 5·260 0·301 -4·708, 15·228 Policy effect (significant covariates) 3·968 0·113 -0·945, 8·880 -0·793 0·378 -2·555, 0·969 0·089 0·921 -1·679, 1·857 4·305 0·377 -5·252, 13·862 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 56 Table A15. Falsification test for sub-group analysis for the weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow Sub-group: Female Sub-group: Male Sub-group: age less than 45 years Sub-group: age more than 45 years Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) -5·369 0·013 -9·607, -1·132 2·833 0·011 0·658, 5·007 -6·556 0·207 -16·745, 3·633 1·861 0·476 -3·259, 6·980 Policy effect (significant covariates) -4·478 0·010 -7·867, -1·089 -0·442 0·843 -4·828, 3·944 -4·454 0·294 -12·778, 3·871 0·813 0·013 0·173, 1·453 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -2·047 0·021 -3·789, -0·304 -1·349 0·313 -3·968, 1·271 -3·267 0·128 -7·479, 0·945 -0·268 0·761 -1·995, 1·459 Policy effect (significant covariates) -2·077 0·028 -3·932, -0·222 -1·169 0·377 -3·763, 1·425 -3·263 0·121 -7·390, 0·864 0·438 0·540 -0·961, 1·836 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) -3·043 0·061 -6·226, 0·141 1·753 0·391 -2·252, 5·757 2·223 0·504 -4·299, 8·746 0·489 0·758 -2·615, 3·592 Policy effect (significant covariates) -1·917 0·054 -3·871, 0·037 1·537 0·454 -2·484, 5·558 1·734 0·625 -5·222, 8·689 0·513 0·050 <0·001, 1·026 Intervention area: Glasgow . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 57 Dummy policy covariate Policy effect (all covariates) -2·127 0·103 -4·684, 0·430 3·219 0·098 -0·592, 7·031 1·680 0·387 -2·127, 5·487 -0·419 0·739 -2·881, 2·043 Policy effect (significant covariates) -1·965 <0·001 -2·989, -0·941 2·053 0·273 -1·617, 5·722 1·668 0·086 -0·234, 3·570 -1·086 0·347 -3·351, 1·179 Table A16. Falsification test for weekend night-time recorded crimes in Aberdeen and Glasgow Secondary outcome, weekend night-time recorded crimes Secondary outcome, weekend restricted night-time (24:00 to 05:59) recorded crimes Falsification test, 1 year before policy implementation, (outcome: number of incidents) Falsification test, 1 year before policy implementation, (outcome: number of incidents) Coefficient P-value 95% CI Coefficient P-value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates) -1·956 0·696 -11·761, 7·850 1·573 0·674 -5·767, 8·913 Policy effect (significant covariates) 2·248 0·117 -0·563, 5·059 2·474 0·007 0·684, 4·263 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -0·914 0·588 -4·227, 2·398 -0·003 0·998 -2·575, 2·569 Policy effect (significant covariates) 0·800 0·605 -2·227, 3·826 1·564 0·151 -0·570, 3·698 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) -3·758 0·245 -10·096, 2·579 -2·461 0·247 -6·633, 1·710 Policy effect (significant covariates) 1·608 0·693 -6·382, 9·598 -3·183 0·009 -5·554, -0·812 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 58 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates) -4·608 0·015 -8·331, -0·885 -5·240 0·002 -8·552, -1·929 Policy effect (significant covariates) -5·723 <0·001 -8·399, -3·046 -6·703 <0·001 -9·286, -3·919 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 59 Appendix 3. Synthetic control Appendix 3.1. Alcohol-related ambulance call-outs (Aberdeen) Appendix 3.1.1. Model specification and validation test The analysis of the synthetic control was conducted in two main steps. First, we assessed nine different model specifications derived from the existing literature on synthetic control methods, following the recommended validation procedure for synthetic control analysis (Table SC 1) [39,40]. The validation test involved calculating the mean squared prediction error (MSPE) by splitting the pre-intervention periods into training and validation sets. Since there are no recommended ratios for splitting the pre-intervention period, we tested three different split ratios—50:50, 70:30, and 80:20— and compared MSPEs across the different specifications. For the alcohol-related ambulance call-outs in Aberdeen, we identified three specifications (1, 2, and 3 (see below Table SC 1)) that produced the lowest MSPEs. However, literature recommends against using all pre-intervention outcome values as covariates, as this can reduce the predictive power of other variables that may significantly influence the construction of the counterfactual [40,41]. Therefore, we selected specification 3 for the remainder of the synthetic control analysis. This specification was chosen because it struck a balance between capturing the pre-intervention trends and allowing other important covariates to contribute to the counterfactual. Appendix 3.1.2. Main analysis and sensitivity tests For the analysis of alcohol-related ambulance call-outs, we observed that the predictor means for Aberdeen were closely aligned with those of the synthetic Aberdeen, as opposed to the average of all donor pool/control cities (Table SC 2). The number of on-premises alcohol outlets emerged as the most influential covariate in generating the synthetic control weights, alongside the pre-intervention outcome mean. Among the donor pools, Edinburgh was the most significant contributor to the creation of synthetic Aberdeen, followed by North Lanarkshire and East Renfrewshire (Figure SC 01). From the gap plot between Aberdeen and synthetic Aberdeen, a positive average treatment effect (ATT=0·103) appeared evident (Figure SC 02). However, a placebo test using permutation methods indicated that other council areas might also exhibit a positive ATT although no policy change occurred there (Figure SC 03a). As a sensitivity analysis, we also presented two graphs for the placebo test after excluding council areas with MSPE two times higher and one time higher than Aberdeen's (Figures SC 03b and SC 03c). This exclusion aimed to improve the robustness of the comparison. For Aberdeen, we calculated the ATT by accounting for the staggered implementation of the policy, adjusting the ATT accordingly using staggered policy weights. Additionally, we visualised the distribution of effect sizes (i.e., ATT) based on the placebo test, where the second figure only considered positive effect sizes, further narrowing the focus to more relevant comparisons (Figures SC 04a and SC 04b). This distribution provides a clear depiction of how extreme the ATT for Aberdeen is compared to the control/donor cities. From the figures, we observed that while some control units exhibited positive ATTs, the magnitude of Aberdeen’s ATT was comparatively larger. This suggests that the treatment effect in Aberdeen is more pronounced than in most control cities, strengthening the uniqueness of the observed impact in Aberdeen relative to the donor pool. Finally, we compared post/pre-intervention MSPE ratios for Aberdeen and the control cities and calculated a frequentist p-value. This was done by dividing the number of cities with post/pre MSPE ratios higher than Aberdeen's by the total number of cities. This resulted in a p-value of 0·222 (6/27) (Figure SC 05a). The second figure illustrating post/pre MSPE ratios was also presented, this time only considering cities with positive effect sizes, enhancing the interpretability of the results with a focus on potential treatment effects (Figure SC 05b). The estimated p-value in this analysis was 0·28, calculated as 5/18 (Figure SC 05b). It is important to note that the minimum possible p-value in . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 60 synthetic control studies depends on the number of control units included in the donor pool. For instance, with 27 control units, the minimum attainable p-value would be 1/27 (approximately 0·04). In contrast, if 18 control units are used, the minimum p-value would be 1/18, (approximately 0·06). Thus, the number of control units directly influences the lowest possible p-value that can be observed. In addition, we conducted ARIMA based on difference-in-difference analyses with the synthetic control where we modelled the differences of alcohol-related ambulance call-out rates between Aberdeen and Synthetic Aberdeen [42]. We estimated effect size of 0·143 with a p-value <0·001 (Table A17). Table SC 1. Validation test for synthetic control for alcohol-related ambulance call-outs in Aberdeen Specifications Pre-intervention periods split between training and validation Training/validation, 50/50 Training/validation, 70/30 Training/validation, 80/20 MSPE* (Validation) MSPE* (Validation) MSPE* (Validation) 1. (all pre-intervention outcome values and covariates) 0·0336 0·03522 0·0480 2. (all pre-intervention outcome values only) 0·0335 0·03524 0·0479 3. (pre-intervention outcome mean and covariates) 0·0412 0·0384 0·0435 4. (last pre-intervention outcome value and covariates) 0·0807 0·1267 0·1002 5. (pre-intervention outcome mean and last pre- intervention outcome value, and covariates) 0·0404 0·0389 0·0473 6. (first three-fourths of the pre-intervention outcome values and covariates) 0·0370 0·0411 0·0503 7. (last three-fourths of the pre-intervention outcome values and covariates) 0·0356 0·0386 0·0528 8. (odd pre-intervention outcome values and covariates) 0·0359 0·0388 0·0516 9. (even pre-intervention outcome values and covariates) 0·0367 0·0402 0·0458 *Mean Square Prediction Error Aberdeen- Synthetic Aberdeen, Specification 3: Table SC 2. Covariate balance in the pre-intervention periods Covariates Covariate weights Alcohol-related ambulance call- outs, predictor means Aberdeen Synthetic Aberdeen Average of other control cities Number of on-premises alcohol outlets (population size adjusted) 0·696 19·813 20·324 23·554 Per capita gross disposable household income 0·045 20,125 19,390 17,763 Pre-intervention outcome mean 0·259 0·724 0·712 0·538 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 61 Figure SC 01: Optimal unit weights Figure SC 02: Gaps of Aberdeen and synthetic Aberdeen 0.000 0.100 0.200 0.300 0.400 0.500 0.600 Midlothian East Renfrewshire North Lanarkshire City of Edinburgh . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 62 Figure SC 03a: Placebo test Figure SC 03b: Placebo test excluding control cities MSPE>2 times Aberdeen . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 63 Figure SC 03c: Placebo test excluding control cities if MSPE> Aberdeen Figure SC 04a: Effect size for placebo test . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 64 Figure SC 04b: Effect size placebo test considering only positive effect size Figure SC 05a: Post/Pre MSPE ratio across cities . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 65 Figure SC 05b: Post/Pre MSPE ratio across cities considering only positive effect size . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 66 Appendix 3.2. Alcohol-related ambulance call-outs (Glasgow) Appendix 3.2.1. Model specification and validation test For the analysis of alcohol-related ambulance call-outs in Glasgow, we identified two specifications (3 and 7) that yielded the lowest MSPEs (Table SC 3). Specification 3 produced the smallest MSPE on two occasions and had the lowest overall value. As a result, we selected specification 3 for the remainder of the analysis. Appendix 3.2.2. Main analysis and sensitivity tests We observed that the predictor means for Glasgow were closely aligned with those of the synthetic Glasgow, compared to the average of all donor pool/control cities (Table SC 4). Per capita gross disposable household income (63%) and the number of on-premises alcohol outlets (37%) emerged as the most influential covariates in generating the synthetic control weights. Among the donor pool cities, Dundee and North Ayrshire were the most significant contributors to the creation of synthetic Glasgow (Figure SC 06). For Glasgow, we estimated a positive average treatment effect (ATT) of 0·012. However, the effect sizes estimated using ARIMA models were negative, indicating a discrepancy between the synthetic control and ARIMA approaches. Additionally, the p-value from the placebo test was 0·926, indicating a high level of uncertainty (Figure SC 10a). When we further modelled ARIMA with synthetic control, we estimated negative and statistically significant effect sizes (Table A17). Despite the different nature of p-values (inferential for ARIMA and frequentist for Synthetic) the different results from different approaches in terms of direction and significance weakened our confidence in interpreting the overall impact of the policy change in Glasgow, making these results inconclusive. Table SC 3. Validation test for synthetic control for alcohol-related ambulance call-outs in Glasgow Specifications Pre-intervention periods split between training and validation Training/validation, 50/50 Training/validation, 70/30 Training/validation, 80/20 MSPE* (Validation) MSPE* (Validation) MSPE* (Validation) 1. (all pre-intervention outcome values and covariates) 0·0618 0·05474 0·0662 2. (all pre-intervention outcome values only) 0·0618 0·05474 0·0662 3. (pre-intervention outcome mean and covariates) 0·0604 0·0667 0·0489 4. (last pre-intervention outcome value and covariates) 0·0810 0·0706 0·0605 5. (pre-intervention outcome mean and last pre-intervention outcome value, and covariates) 0·0604 0·0532 0·0729 6. (first three-fourths of the pre-intervention outcome values and covariates) 0·0643 0·0606 0·0662 7. (last three-fourths of the pre-intervention outcome values and covariates) 0·0613 0·0531 0·0620 8. (odd pre-intervention outcome values and covariates) 0·0618 0·0547 0·0662 9. (even pre-intervention outcome values and covariates) 0·0618 0·0547 0·0662 Glasgow- Synthetic Glasgow, Specification 3: Table SC 4. Covariate balance in the pre-intervention periods . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 67 Covariates Covariate weights Alcohol-related ambulance call- outs, predictor means Glasgow Synthetic Glasgow Average of other control cities Number of on-premises alcohol outlets (population size adjusted) 0·370 21·416 21·446 23·524 Per capita gross disposable household income 0·628 15852 15851 18260 Pre-intervention outcome mean 0·002 0·964 0·835 0·595 Figure SC 06: Optimal unit weights 0 0.1 0.2 0.3 0.4 0.5 0.6 Falkirk West Lothian City of Edinburgh North Lanarkshire North Ayrshire Dundee City . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 68 Figure SC 07: Gaps of Glasgow and synthetic Glasgow Figure SC 08a: Placebo test . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 69 Figure SC 08b: Placebo test excluding control cities MSPE>2 times Glasgow Figure SC 08c: Placebo test excluding control cities if MSPE> Glasgow . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 70 Figure SC 09a: Effect size for placebo test Figure SC 09b: Effect size placebo test considering only positive effect size . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 71 Figure SC 10a: Post/Pre MSPE ratio across cities Figure SC 10b: Post/Pre MSPE ratio across cities considering only positive effect size . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 72 Appendix 3.3. Reported crimes (Aberdeen) Appendix 3.3.1. Model specification and validation test For the reported crimes in Aberdeen, we identified three specifications (6, 7, and 9) that generated the lowest MSPEs (Table SC 5). Specification 6 had the lowest MSPE value among them. As a result, we selected specification 6 for the remainder of the analysis. Appendix 3.3.2. Main analysis and sensitivity tests We observed that the predictor means for Aberdeen were closely aligned with those of the synthetic Aberdeen with some exceptions (Table SC 6). Among the donor pool cities, Dundee, Edinburgh, West Lothian, and Falkirk were the most significant contributors to the creation of synthetic Aberdeen for reported crimes (Figure SC 11). We estimated a positive average treatment effect (ATT) of 0·022, consistent with the direction of the effect sizes estimated using ARIMA based on difference-in-Differences approaches. However, the p- value from the placebo test for the synthetic control was 0·333 (Figure SC 15a). When we further modelled ARIMA with synthetic control, we also estimated a positive but statistically insignificant effect size of 0·016 (Table A18). Although the effect sizes across different approaches (ARIMA, synthetic control, and synthetic control + ARIMA) were homogeneous in terms of direction (positive), they were not always statistically significant. This underscores the uncertainty in the results and the need for caution in their interpretation. Table SC 5. Validation test for synthetic control for reported crimes in Aberdeen Specifications Pre-intervention periods split between training and validation Training/validation, 50/50 Training/validatio n, 70/30 Training/validatio n, 80/20 MSPE* (Validation) MSPE* (Validation) MSPE* (Validation) 1. (all pre-intervention outcome values and covariates) 0·0772 0·05798 0·0570 2. (all pre-intervention outcome values only) 0·0767 0·05798 0·0570 3. (pre-intervention outcome mean and covariates) 0·0724 0·0639 0·0665 4. (last pre-intervention outcome value and covariates) 0·1128 0·2336 0·0729 5. (pre-intervention outcome mean and last pre-intervention outcome value, and covariates) 0·0722 0·0640 0·0682 6. (first three-fourths of the pre-intervention outcome values and covariates) 0·0858 0·0714 0·0534 7. (last three-fourths of the pre-intervention outcome values and covariates) 0·0707 0·0640 0·0693 8. (odd pre-intervention outcome values and covariates) 0·0745 0·0647 0·0565 9. (even pre-intervention outcome values and covariates) 0·0772 0·0570 0·0556 Aberdeen- Synthetic Aberdeen, Specification 6: Table SC 6. Covariate balance in the pre-intervention periods . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 73 Covariates Covariate weights Reported crimes, predictor means Aberdeen Synthetic Aberdeen Average of other control cities Number of on-premises alcohol outlets (population size adjusted) 0·043 19·794 20·535 23·558 Per capita gross disposable household income 0·001 20125 17415 17773 Pre-intervention outcome, period-19 0·026 0·695 0·702 0·655 Pre-intervention outcome, period-20 0·000 0·825 0·976 0·534 Pre-intervention outcome, period-21 0·004 0·825 0·773 0·749 Pre-intervention outcome, period-22 0·103 1·086 0·895 0·677 Pre-intervention outcome, period-23 0·004 0·608 0·720 0·641 Pre-intervention outcome, period-24 0·000 0·825 0·926 0·695 Pre-intervention outcome, period-25 0·040 0·565 0·679 0·764 Pre-intervention outcome, period-26 0·019 0·825 0·902 0·644 Pre-intervention outcome, period-27 0·030 0·738 0·700 0·674 Pre-intervention outcome, period-28 0·001 0·434 0·733 0·574 Pre-intervention outcome, period-29 0·081 0·695 0·667 0·598 Pre-intervention outcome, period-30 0·001 0·739 0·884 0·642 Pre-intervention outcome, period-31 0·020 1·260 1·086 0·905 Pre-intervention outcome, period-32 0·017 0·565 0·966 0·765 Pre-intervention outcome, period-33 0·013 0·695 0·872 0·777 Pre-intervention outcome, period-34 0·047 0·826 0·783 0·595 Pre-intervention outcome, period-35 0·044 1·478 0·852 0·716 Pre-intervention outcome, period-36 0·062 0·826 0·691 0·642 Pre-intervention outcome, period-37 0·005 0·435 0·712 0·668 Pre-intervention outcome, period-38 0·047 0·869 1·028 0·631 Pre-intervention outcome, period-39 0·116 1·043 0·912 0·663 Pre-intervention outcome, period-40 0·078 0·696 0·677 0·545 Pre-intervention outcome, period-41 0·094 1·261 0·962 0·544 Pre-intervention outcome, period-42 0·097 0·870 0·725 0·724 Pre-intervention outcome, period-43 0·007 1·044 0·632 0·592 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 74 Figure SC 11: Optimal unit weights Figure SC 12: Gaps of Aberdeen and synthetic Aberdeen 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 Renfrewshire North Ayrshire Falkirk West Lothian City of Edinburgh Dundee City . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 75 Figure SC 13a: Placebo test Figure SC 13b: Placebo test excluding control cities MSPE>2 times Aberdeen . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 76 Figure SC 13c: Placebo test excluding control cities if MSPE> Aberdeen Figure SC 14a: Effect size for placebo test . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 77 Figure SC 14b: Effect size placebo test considering only positive effect size Figure SC 15a: Post/Pre MSPE ratio across cities . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 78 Figure SC 15b: Post/Pre MSPE ratio across cities considering only positive effect size . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 79 Appendix 3.4. Reported crimes (Glasgow) Appendix 3.4.1. Model specification and validation test For the reported crimes in Glasgow, we identified specifications 6, 7, and 9 that generated the lowest MSPEs (Table SC 7). Among them specification 7 a had the lowest value of MSPE. As a result, we selected specification 7 for the remainder of the analysis. Appendix 3.4.2. Main analysis and sensitivity tests We observed that the predictor means for Glasgow were closely aligned with those of the synthetic Glasgow with some exceptions (Table SC 8). Among the donor pool cities, Edinburgh, Dundee, and Falkirk were the most significant contributors to the creation of synthetic Glasgow for reported crimes (Figure SC 11). We estimated a positive average treatment effect (ATT) of 0·148. However, the effect sizes estimated using ARIMA based on difference-in-Differences models were negative and insignificant, indicating a discrepancy between the synthetic control and ARIMA approaches. Additionally, the p-value from the placebo test was 0·519, indicating a high level of uncertainty (Figure SC 20a). When we further modelled ARIMA with synthetic control, we estimated positive and statistically insignificant effect size (Table A18). Despite the different nature of p-values (inferential for ARIMA and frequentist for Synthetic) the different results from different approaches in terms of direction and significance weakened our confidence in interpreting the overall impact of the policy change in Glasgow, making these results unconclusive. Table SC 7. Validation test for synthetic control for reported crimes in Glasgow Specifications Pre-intervention periods split between training and validation Training/validation, 50/50 Training/validation, 70/30 Training/validation, 80/20 MSPE* (Validation) MSPE* (Validation) MSPE* (Validation) 1. (all pre-intervention outcome values and covariates) 0·0765 0·05166 0·0473 2. (all pre-intervention outcome values only) 0·0765 0·05133 0·0473 3. (pre-intervention outcome mean and covariates) 0·0853 0·0819 0·0632 4. (last pre-intervention outcome value and covariates) 0·1298 0·0670 0·0598 5. (pre-intervention outcome mean and last pre- intervention outcome value, and covariates) 0·0930 0·0643 0·0584 6. (first three-fourths of the pre-intervention outcome values and covariates) 0·0745 0·0513 0·0473 7. (last three-fourths of the pre-intervention outcome values and covariates) 0·0724 0·0513 0·0473 8. (odd pre-intervention outcome values and covariates) 0·0766 0·0517 0·0473 9. (even pre-intervention outcome values and covariates) 0·0724 0·0513 0·0475 Glasgow- Synthetic Glasgow, Specification 7: Table SC 8. Covariates balance in the pre-intervention periods Covariates Covariate weights Reported crimes, predictor means . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 80 Glasgow Synthetic Glasgow Average of other control cities Number of on-premises alcohol outlets (population size adjusted) 0·016 21·371 23·402 23·529 Per capita gross disposable household income 0·008 15852 18900 18246 Pre-intervention outcome, period-169 0·019 0·733 0·835 0·748 Pre-intervention outcome, period-170 0·017 0·701 0·720 0·456 Pre-intervention outcome, period-171 0·012 0·637 0·701 0·413 Pre-intervention outcome, period-172 0·011 0·987 1·085 0·539 Pre-intervention outcome, period-173 0·017 1·002 0·779 0·543 Pre-intervention outcome, period-174 0·015 1·114 0·676 0·519 Pre-intervention outcome, period-175 0·005 0·891 0·411 0·421 Pre-intervention outcome, period-176 0·026 0·954 0·811 0·516 Pre-intervention outcome, period-177 0·016 0·922 0·964 0·567 Pre-intervention outcome, period-178 0·014 0·763 0·398 0·501 Pre-intervention outcome, period-179 0·021 0·938 0·681 0·495 Pre-intervention outcome, period-180 0·020 0·826 0·589 0·470 Pre-intervention outcome, period-181 0·014 0·826 0·618 0·468 Pre-intervention outcome, period-182 0·014 1·080 1·178 0·795 Pre-intervention outcome, period-183 0·006 0·889 0·545 0·525 Pre-intervention outcome, period-184 0·017 0·826 0·798 0·556 Pre-intervention outcome, period-185 0·020 0·730 0·636 0·458 Pre-intervention outcome, period-186 0·015 0·746 0·855 0·562 Pre-intervention outcome, period-187 0·016 0·793 0·717 0·505 Pre-intervention outcome, period-188 0·014 0·809 0·622 0·523 Pre-intervention outcome, period-189 0·017 0·682 0·920 0·565 Pre-intervention outcome, period-190 0·006 0·682 0·857 0·528 Pre-intervention outcome, period-191 0·024 1·094 0·723 0·597 Pre-intervention outcome, period-192 0·018 0·824 0·690 0·585 Pre-intervention outcome, period-193 0·020 0·713 0·535 0·401 Pre-intervention outcome, period-194 0·018 0·681 0·573 0·419 Pre-intervention outcome, period-195 0·016 0·776 0·804 0·539 Pre-intervention outcome, period-196 0·023 0·792 0·969 0·543 Pre-intervention outcome, period-197 0·017 0·586 0·516 0·566 Pre-intervention outcome, period-198 0·027 0·886 0·575 0·629 Pre-intervention outcome, period-199 0·020 0·918 0·950 0·591 Pre-intervention outcome, period-200 0·018 1·013 0·962 0·528 Pre-intervention outcome, period-201 0·017 0·823 0·822 0·475 Pre-intervention outcome, period-202 0·013 0·854 0·694 0·552 Pre-intervention outcome, period-203 0·018 0·522 0·468 0·370 Pre-intervention outcome, period-204 0·028 1·360 0·913 0·616 Pre-intervention outcome, period-205 0·022 0·711 0·666 0·459 Pre-intervention outcome, period-206 0·013 0·616 0·794 0·563 Pre-intervention outcome, period-207 0·020 1·043 0·720 0·595 Pre-intervention outcome, period-208 0·018 0·964 0·702 0·594 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 81 Pre-intervention outcome, period-209 0·020 0·521 0·469 0·267 Pre-intervention outcome, period-210 0·021 0·805 0·447 0·344 Pre-intervention outcome, period-211 0·036 0·758 0·527 0·341 Pre-intervention outcome, period-212 0·013 1·074 0·562 0·413 Pre-intervention outcome, period-213 0·016 0·932 0·950 0·529 Pre-intervention outcome, period-214 0·016 0·774 0·622 0·557 Pre-intervention outcome, period-215 0·011 0·821 0·563 0·410 Pre-intervention outcome, period-216 0·022 0·868 0·726 0·476 Pre-intervention outcome, period-217 0·012 1·026 0·608 0·496 Pre-intervention outcome, period-218 0·020 0·884 0·617 0·559 Pre-intervention outcome, period-219 0·021 0·758 0·512 0·483 Pre-intervention outcome, period-220 0·018 0·852 0·630 0·480 Pre-intervention outcome, period-221 0·018 0·947 0·900 0·507 Pre-intervention outcome, period-222 0·019 0·458 0·597 0·521 Pre-intervention outcome, period-223 0·014 0·647 0·422 0·401 Pre-intervention outcome, period-224 0·017 0·883 0·706 0·422 Figure SC 16: Optimal unit weights 0 0.1 0.2 0.3 0.4 0.5 0.6 Falkrik Dundee City City of Edinburgh . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 82 Figure SC 17: Gaps of Glasgow and synthetic Glasgow . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 83 Figure SC 18a: Placebo test Figure SC 18b: Placebo test excluding control cities MSPE>2 times Glasgow . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 84 Figure SC 18c: Placebo test excluding control cities if MSPE> Glasgow Figure SC 19a: Effect size for placebo test . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 85 Figure SC 19b: Effect size placebo test considering only positive effect size Figure SC 20a: Post/Pre MSPE ratio across cities . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 86 Figure SC 20b: Post/Pre MSPE ratio across cities considering only positive effect size Appendix 4. Synthetic control and ARIMA Table A17. Effect of policy changes on weekend night-time alcohol-related ambulance call-outs in Aberdeen and Glasgow Primary outcome, weekend night-time alcohol-related ambulance call-outs Synthetic control + ARIMA Outcome: population adjusted incident rates Coefficient P-value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates)* 0·143 <0·001 0·069, 0·218 Policy effect (significant covariates) 0·143 <0·001 0·069, 0·218 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) 0·132 <0·001 0·072, 0·192 Policy effect (significant covariates) 0·132 <0·001 0·072, 0·192 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 0·083 0·031 0·008, 0·159 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 87 Policy effect (significant covariates) 0·084 0·029 0·009, 0·159 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates)* -0·095 <0·001 -0·146, -0·043 Policy effect (significant covariates) -0·095 <0·001 -0·147, -0·044 *Models are adjusted by covariates- Covid-19, public holidays, and outliers Table A18. Effect of policy changes on weekend night-time recorded crimes in Aberdeen and Glasgow Secondary outcome, weekend night-time recorded crimes Synthetic control + ARIMA Outcome: population adjusted incident rates Coefficient P- value 95% CI Intervention area: Aberdeen Staggered policy covariate Policy effect (all covariates)* -0·029 0·599 -0·136, 0·079 Policy effect (significant covariates) -0·032 0·488 -0·123, 0·059 Dummy policy covariate, sensitivity analysis Policy effect (all covariates) -0·032 0·444 -0·116, 0·051 Policy effect (significant covariates) -0·035 0·393 -0·115, 0·045 Dummy policy covariate with policy implemented with at least half strength, sensitivity analysis Policy effect (all covariates) 0·007 0·876 -0·081, 0·094 Policy effect (significant covariates) -0·001 0·985 -0·076, 0·075 Intervention area: Glasgow Dummy policy covariate Policy effect (all covariates)* 0·029 0·268 -0·22, 0·080 Policy effect (significant covariates) 0·029 0·269 -0·022, 0·080 *Models are adjusted by covariates- Covid-19, public holidays, and outliers . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 88

1 Brauer M, Roth GA, Aravkin AY, et al. Global burden and strength of evidence for 88 risk factors in 204 countries and 811 subnational locations, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2024;403:2162–203. 2 World Health Organization. Global status report on alcohol and health 2018. Geneva: World Health Organization 2018. https://doi.org/10.1037/cou0000248 3 Gakidou E, Afshin A, Abajobir AA, et al. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet. 2017;390:1345–422. 4 Griswold MG, Fullman N, Hawley C, et al. Alcohol use and burden for 195 countries and territories, 1990-2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet. 2018;392:1015–35. 5 National Records of Scotland. Alcohol-specific deaths 2023. 2024. https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/vital- events/deaths/alcohol-deaths 6 Public Health Scotland. Alcohol Related Hospital Statistics Scotland 2022/23. 2023. 7 Manca F, Lewsey J, Waterson R, et al. Estimating the burden of alcohol on ambulance callouts through development and validation of an algorithm using electronic patient records. Int J Environ Res Public Health. 2021;18. doi: 10.3390/ijerph18126363 8 World Health Organization. The SAFER technical package: five areas of intervention at national and subnational levels. 2019. https://apps.who.int/iris/handle/10665/330053 9 Mäkelä P, Warpenius K. Night-time is the right time? Late-night drinking and assaults in Finnish public and private settings. Drug Alcohol Rev. 2020;39:321–9. 10 Wilkinson C, Livingston M, Room R. Impacts of changes to trading hours of liquor licences on alcohol-related harm: A systematic review 2005-2015. Public Heal Res Pract. 2016;26:1–7. 11 Nepal S, Kypri K, Tekelab T, et al. Effects of extensions and restrictions in alcohol trading hours on the incidence of assault and unintentional injury: Systematic review. J Stud Alcohol Drugs. 2020;81:5–23. 12 Popova S, Giesbrecht N, Bekmuradov D, et al. Hours and days of sale and density of alcohol outlets: Impacts on alcohol consumption and damage: A systematic review. Alcohol Alcohol. 2009;44:500–16. 13 Stockwell T, Chikritzhs T. Do relaxed trading hours for bars and clubs mean more relaxed drinking ?A review of international research on the impacts of changes to permitted hours of drinking. Crime Prev Community Saf. 2009;11:153–70. 14 Robert A. Hahn, Kuzara JL, Elder R, et al. Effectiveness of Policies Restricting Hours of Alcohol Sales in Preventing Excessive Alcohol Consumption and Related Harms. Am J Prev Med. 2010;39:590–604. 15 Alcohol Focus Scotland. Review of statements of licensing policy 2018 to 2023. Published Online First: 2020. 16 Humphreys DK, Eisner MP. Evaluating a natural experiment in alcohol policy The Licensing Act (2003) and the requirement for attention to implementation. Criminol Public Policy. 2010;9:41–67. 17 Humphreys DK, Eisner MP, Wiebe DJ. Evaluating the Impact of Flexible Alcohol Trading . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 89 Hours on Violence: An Interrupted Time Series Analysis. PLoS One. 2013;8. doi: 10.1371/journal.pone.0055581 18 Stevely AK, de Vocht F, Neves RB, et al. Evaluating the effects of the Licensing Act 2003 on the characteristics of drinking occasions in England and Wales: a theory of change-guided evaluation of a natural experiment. Addiction. 2021;116:2348–59. 19 Sheikh N, Henriques-Cadby I, Haghpanahan H, et al. Evaluating later or expanded premises hours for alcohol in the night-time economy (ELEPHANT): statistical analysis plan. medRxiv. Published Online First: 2024. doi: https://doi.org/10.1101/2024.05.07.24306991 20 Abadie A, Diamond A, Hainmueller J. Synth: An R Package for Synthetic Control Methods in Comparative Case Studies. J Stat Softw. 2011;42:1–17. 21 Degli Esposti M, Spreckelsen T, Gasparrini A, et al. Can synthetic controls improve causal inference in interrupted time series evaluations of public health interventions? Int J Epidemiol. 2020;49:2010–20. 22 Justice Analytical Services. Bringing Together Scotland ’ s Crime Statistics. Edinburgh: Scotland 2014. 23 de Goeij MCM, Veldhuizen EM, Buster MCA, et al. The impact of extended closing times of alcohol outlets on alcohol-related injuries in the nightlife areas of Amsterdam: A controlled before-and-after evaluation. Addiction. 2015;110:955–64. 24 Mitchell et al. ‘It’s a bit of a mess’: The impact of later alcohol trading hours for bars and clubs in Scotland according to qualitative interviews with local stakeholders. 2025. 25 McClelland R, Mucciolo L. An update on the synthetic control method as a tool to understand state policy. 2022. https://www.taxpolicycenter.org/sites/default/files/publication/163919/an_update_on_the_synt hetic_control_method_as_a_tool_to_understand_state_policy.pdf 26 Huntington-Klein N. The Effect : An introduction to Research Design and Causality. 1st ed. Chapman & Hall 2022. 27 Department of Justice of Ireland. General Scheme: Sale of Alcohol Bill 2022 including press release. 2022. https://www.drugsandalcohol.ie/37347/ (accessed 29 October 2025) 28 Department for Business & Trade. Licensing policy sprint: joint industry and HM government taskforce report. Policy paper. London 2025. https://www.gov.uk/government/publications/licensing-taskforce-report-and-government- response/licensing-policy-sprint-joint-industry-and-hm-government-taskforce-report 29 Department for Business & Trade. Licensing taskforce report: government response. Policy paper. London 2025. https://www.gov.uk/government/publications/licensing-taskforce-report- and-government-response/licensing-taskforce-report-government-response 30 Fitzgerald N, Egan M, O’Donnell R, et al. Public health engagement in alcohol licensing in England and Scotland: the ExILEnS mixed-method, natural experiment evaluation. Public Heal Res (Southampton, England). 2024;13:1–84. 31 Scottish Government. Licensing ( Scotland ) Act 2005. 2005. https://www.legislation.gov.uk/asp/2005/16/contents 32 City of Glasgow Licensing Board. Licensing Policy Statement: Licensing (Scotland) Act 2005. 2018. https://www.glasgow.gov.uk/CHttpHandler.ashx?id=17578&p=0 33 Home Office. Revised guidance issued under section 182 of Licensing Act 2003. Published Online First: 2025. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint 90 34 Scottish Government. Quarterly National Accounts Scotland, 2017 Quarter 3. 2018. https://www.webarchive.org.uk/wayback/archive/20180212144444/http://www.gov.scot/Topic s/Statistics/Browse/Economy/QNA2017Q3 (accessed 22 November 2022) 35 Centre for Environmental Data Analysis. Met Office weather datasets. CEDA Arch. 2019. https://data.ceda.ac.uk/badc/ukmo-hadobs/data/insitu/MOHC/HadOBS/HadUK- Grid/v1.2.0.ceda/region (accessed 10 November 2023) 36 The Scottish Government. Scottish Liquor Licensing Statistics. 2022. https://www.gov.scot/publications/scottish-liquor-licensing-statistics/ 37 Scottish Parliament Information Centre (SPICe). Timeline of Coronavirus (COVID-19) in Scotland. SPICe Spotlight. 2023. https://spice-spotlight.scot/2023/05/10/timeline-of- coronavirus-covid-19-in-scotland/ (accessed 11 March 2024) 38 The Scottish Government. Public and bank holidays. Scotl. bank holidays. 2022. https://www.mygov.scot/scotland-bank-holidays (accessed 12 March 2024) 39 Abadie A, Vives-i-Bastida J. Synthetic Controls in Action. arXiv Prepr. Published Online First: 2022. 40 Ferman B, Pinto C, Possebom V. Cherry Picking with Synthetic Controls. J Policy Anal Manag. 2020;39:510–32. 41 Kaul A, Klößner S, Pfeifer G, et al. Standard Synthetic Control Methods: The Case of Using All Preintervention Outcomes Together With Covariates. J Bus Econ Stat. 2022;40:1282–90. 42 Kreif N, Grieve R, Hangartner D, et al. Examination of the synthetic control method for evaluating health policies with multiple treated units. Health Econ. 2015;25:1514–28. . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted November 1, 2025. ; https://doi.org/10.1101/2025.10.30.25339133doi: medRxiv preprint