key: cord-0758375-bze8odrh authors: Clare, Philip J.; Aiken, Alexandra; Yuen, Wing See; Upton, Emily; Kypri, Kypros; Degenhardt, Louisa; Bruno, Raimondo; McCambridge, Jim; McBride, Nyanda; Hutchinson, Delyse; Slade, Tim; Mattick, Richard; Peacock, Amy title: Alcohol use among young Australian adults in May‐June 2020 during the COVID‐19 pandemic: a prospective cohort study date: 2021-06-09 journal: Addiction DOI: 10.1111/add.15599 sha: 5eecc289b5c9054fd9b744a4252561234d8e52ce doc_id: 758375 cord_uid: bze8odrh AIMS: To estimate change in young people's alcohol consumption during COVID‐19 restrictions in Australia in early‐mid 2020, and test whether those changes were consistent by gender and level of consumption prior to the pandemic. DESIGN: Prospective longitudinal cohort SETTING: Secondary schools in New South Wales, Tasmania, and Western Australia. PARTICIPANTS: Subsample of a cohort (n=443) recruited in the first year of secondary school in 2010‐11. Analysis data included three waves collected in Sep 2017 – July 2018, Sep 2018 ‐ May 2019 and Aug 2019 ‐ Jan 2020), and in May‐June 2020. MEASUREMENTS: The primary predictors were time, gender, and level of consumption prior to the pandemic. Outcome variables, analysed by mixed‐effects models, included frequency and typical quantity of alcohol consumption, binge drinking, peak consumption, alcohol‐related harm, and drinking contexts. FINDINGS: Overall consumption (frequency x quantity) during the restrictions declined by 17% (incidence rate ratio [IRR] 0.83; 95% confidence interval [CI] 0.73, 0.93), compared to February 2020, and there was a 34% decline in the rate of alcohol‐related harms in the same period (IRR 0.66; 95% CI 0.55, 0.80). Changes in alcohol consumption were largely consistent by gender. CONCLUSIONS: From a survey of secondary school students in Australia, there is evidence for a reduction in overall consumption and related harms during the COVID‐19 restrictions. The COVID-19 pandemic has had consequences beyond disease morbidity and mortality, with social and travel restrictions imposed worldwide. There are questions regarding the impact of COVID-19 on alcohol. Restrictions on gatherings could result in declining consumption within social contexts (1) (2) (3) . Conversely, the pandemic and related restrictions have generated financial and psychological distress for many Australians (4): known risk factors for elevated alcohol consumption (5) . Further, some governments have temporarily relaxing liquor licensing restrictions, facilitating online purchasing and home delivery (1) . Financial analyses showed increased alcohol purchasing in Australia in March 2020 when restrictions were implemented (6) , although industry reported declines in total alcohol volume sold (1, 7) . Similarly, wastewater analysis showed a decline in alcohol consumption during initial COVID-19 restrictions in Adelaide, Australia (8) . By contrast, survey research from Australia and other countries suggest the majority of adults perceived their alcohol use as stable since COVID-19 restrictions (9, 10) (11) . Other Australian research indicated that self-reported alcohol consumption may be elevated relative to benchmark data collected before COVID-19 restrictions (12) . However, research to date largely does not account for pre-COVID-19 levels of alcohol use, highlighting the necessity for prospective, longitudinal data comparing alcohol consumption patterns before and after COVID-19 restrictions. Alcohol use is very common among young adults in Australia: 18-24-year-olds typically report the highest rates of drinking, binge drinking and alcohol-related harms of all Australians (13) . There is evidence of an ongoing decline in consumption overall among young Australians (13) , however, it is not clear whether this trend has been impacted by the COVID-19 pandemic. Young adults are typically less financially stable (14) than older adults, and overrepresented in service industries which have been particularly impacted by This article is protected by copyright. All rights reserved. COVID-19 restrictions (9) . Thus, restrictions may have disproportionately impacted younger people. Conversely, young people tend to consume more alcohol outside of the home (15) than older adults, so consumption may have declined with restrictions preventing gatherings. Emerging research also suggests gender differences in the response to the COVID-19 pandemic and associated restrictions. Men in the United States were more likely to drink following restrictions than women, but their drinking was largely unaffected by stress caused by the pandemic (16). Australian research (17) also found gender differences in alcohol use, mediated by other factors such as employment, exercise, diet, and sleep. However, that study occurred early in the pandemic when stress from isolation and economic factors may not have fully emerged. This study used data from the Australian Parental Supply of Alcohol Longitudinal Study (APSALS) to investigate changes in alcohol consumption among young adults in May to June 2020 (i.e., after the implementation of restrictions and the first wave of the pandemic in Australia had peaked in late March), relative to consumption before March 2020. Specifically, we aimed to: 1) evaluate changes in alcohol consumption and harm, compared with pre-COVID-19; 2) test whether changes in alcohol consumption and harm were consistent in men and women; and 3) test whether changes in alcohol consumption and harm were consistent across levels of alcohol consumption prior to the COVID-19 pandemic. We also explored drivers for change in alcohol consumption. This article is protected by copyright. All rights reserved. We used the APSALS cohort (registered at ClinicalTrials.gov: NCT02280551) of 1927 families (18). Participants and a parent/guardian were recruited via an opt-in process in 2010 and 2011 from Grade 7 classes (mean age 12.9 years) in Australian schools across three jurisdictions (New South Wales, Tasmania and Western Australia), with signed consent obtained from parents. The target sample was N~1800 (18). Participants completed annual surveys, with parents/guardians surveyed until Wave 5. Participants were reimbursed for survey completion ($AUD20 for Wave 8; $AUD50 for Wave 9/10) and could enter a draw to win one of eight $AUD500 vouchers. Response rates for each wave are presented in Appendix B. Wave 10 data collection was underway when the first COVID-19 fatality was recorded in Five prospective outcome measures were used in addition to retrospective outcomes assessed in the COVID-19 survey. Further details of the measures are in Appendix C. Frequency of use. Participants were asked how often they consumed alcohol. In the COVID-19 survey, participants were asked first about the month of February, and then about the past month (Not at all, 1 day/month, 2-3 days/month, 1-2 days/week, 3-4 days/week, 5-6 days/week, or Daily). In the APSALS annual survey they were asked about the past year (Daily, 5-6 days/week, 3-4 days/week, 1-2 days/week, 2-3 days/month, 1 day/month, Less often, or Never). This was pro-rated to past month use (see Appendix C for details). Typical quantity consumed. Participants who reported any alcohol consumption were asked how many standard drinks (equivalent to 10g of alcohol) they typically consumed on a drinking occasion, with categorical responses in W8/9/10 (None; a sip or a taste; 1-2 drinks; 3-4 drinks; 5-6 drinks; 7-10 drinks; 11-12 drinks; or 13 or more drinks), and a continuous response in the COVID-19 survey, which was coded to the same categories as the W8/9/10 data. To limit impossibly high responses without making assumptions about maximum levels of consumption, the continuous response was capped at 100 drinks. Frequency of binge drinking. Participants who reported any alcohol consumption were asked how often they consumed ≥5 standard drinks, with the same response options as for frequency of use. Responses in the annual APSALS surveys were pro-rated to the past month (Appendix table C2 ). Overall consumption. We multiplied frequency of use and typical quantity consumed in each time point to create a proxy for overall consumption of alcohol in the month (19). survey were asked the maximum number of standard drinks they consumed on a single occasion in the past month, and in February 2020. Alcohol-related harms. The COVID-19 survey included a subset of the harms from the School Health and Alcohol Harm Reduction Project (SHAHRP) scale (20) (range 0-8; see Appendix C for list of items). Drinking contexts. The COVID-19 survey included questions about time spent drinking alone, physically with others, or virtually with others, coded as proportion of time drinking from 0 to 100. Use of delivery services. Participants were asked in the COVID-19 survey how often they had alcohol delivered to their home, coded as a binary (no/yes) variable of any use of alcohol delivery services. Self-reported change in alcohol use. Participants were asked to self-report whether their alcohol use had changed since the COVID-19 restrictions commenced in March. Those reporting increased or decreased consumption, were asked to select from a list of possible reasons why the change occurred (with multiple selections possible; Table 4 ). The primary predictor was time to assess changes before and during the COVID-19 restrictions. The primary time variable was a categorical variable with five time periods: analyses we coded the quarter and year in which the survey was completed (or to which data referred to). We also coded a binary variable of alcohol consumption in Wave 8 (above or below/equal to the median for consumption). Participants were asked about their experiences during the COVID-19 pandemic, including whether they had been tested for/diagnosed with the disease, and a range of other potential impacts such as being required to quarantine or losing employment. Further details are available in Appendix C. We included a range of sociodemographic characteristics of the sample, with means and standard deviations for continuous measures and ns and percentages for categorical measures. We compared the characteristics of participants who completed the COVID-19 survey to all participants who completed ASPALS W8, with t-tests and chi-square tests to assess differences between responders and non-responders (including those not invited to the COVID-19 survey). We reported descriptive statistics on alcohol consumption of participants over time. We compared the subjective report of change in alcohol use to a measure of change derived from change in overall consumption between February 2020 and May-June 2020. We conducted analyses of change in alcohol use using mixed effects models to control for repeated observations for each participant. Primary analyses were conducted using negative binomial regression, with results reported as incidence-rate ratios (IRRs). Because the primary difference of interest was with alcohol use immediately prior to the pandemic, we used February 2020 (or Quarter 1 2020 for analyses using calendar time) as the referent in all analyses. Although participants who completed Wave 10 up to April 30 were invited to participate in the COVID-19 survey, to maintain separation between the Wave 10 data and the COVID-19 restrictions, only Wave 10 responses to the end of January 2020 were included. We analysed alcohol outcomes that were only asked in the COVID-19 survey with mixedeffects models across two time periods (February 2020 and May-June 2020), using linear regression (maximum consumption and drinking contexts), negative binomial regression (alcohol-related harms), and logistic regression (use of delivery services).Results were reported as changes in mean, IRRs, and ORs, respectively. To examine whether changes over time differed by gender, we reconducted all models with a gender by time interaction effect. Two participants identified as a gender other than male or female and were excluded from all analyses. To examine whether changes differed as a function of the amount of alcohol consumed, we conducted secondary analyses using an interaction between time and Wave 8 consumption. As a post-hoc sensitivity analysis, we conducted an analysis based on Australian government guidelines (21), with participants classified based on reported binge drinking more than once in the month, or heavy drinking (10+ drinks a week) at Wave 8. We conducted sensitivity analyses with time considered as continuous, grouped into Years/Quarters (e.g., Q1 2020) rather than split categorically into waves. Another sensitivity analysis was conducted using the same continuous time measure, but excluding the retrospective data collected about February 2020. For both sensitivity analyses, all Wave 10 data up to April 30 2020 were included. All regression models controlled for confounding, using Wave 1 and Wave 8 variables. The confounding variables included: child gender and age (22, 23), socio-economic status of area of residence (24), family history of alcohol problems (25), having older siblings (22), and This article is protected by copyright. All rights reserved. peer substance use and disapproval of alcohol and tobacco use (22, 23, 25) . More details on the confounding variables are included in Appendix C. Analysis was conducted using the 'mi' commands and mixed effects models (mixed, meologit, menbreg) in Stata 16.1 (26). We used a critical p-value of p<0.05, and all results of inferential statistics are presented with 95% confidence intervals. Analysis code is available at https://www.philipclare.com/code/apsals/. The specific analyses in this study were not preregistered and should be considered exploratory. To reduce the chance of bias due to missing data, we used multiple imputation, implemented using the R package 'mice' (27) and combined using the package 'Amelia' (28) in R 4.0.3 (29). Further detail of missing data and the imputation is included in Appendix D. Participants who completed the COVID-19 survey (n=443) largely recruited from Tasmania (37.5%), Western Australia (26.9%) and New South Wales (25.7%) ( Table E1 ). The COVID-19 sample was broadly similar to the APSALS cohort although there was a higher proportion of women (60.8% compared with 49.3%, respectively; Table 1 ). The COVID-19 sample also came from slightly more educated (44.0% of parents with university education vs 36.3%) and religious (34.2% of parents religious vs 29.5%) backgrounds (Table 1) . Six percent of COVID-19 survey respondents reported having been tested for SARS-CoV-2, and one participant reported having received a positive test result. Three quarters (74%) reported voluntary home isolation (i.e., only leaving their primary residence for governmentspecified essential reasons, including grocery shopping and medical appointments), and one in ten (9%) reported they had been in quarantine (i.e. at risk of contracting COVID-19 and required to quarantine for 14 days). Further details of restrictions during the COVID-19 pandemic in Australia are included in Appendix A(30). There was relatively little concern about COVID-19 itself, with 17% reporting they were at least moderately worried about contracting the disease. Alcohol use Between 76.6% and 85.4% of respondents reported consuming alcohol in each period from September 2017 to June 2020 (Table 2) , while 50.0% to 64.7% reported binge drinking. There was evidence of increased frequency of consumption, which did not abate during the pandemic, with frequency of alcohol consumption in May-June 2020 31% higher than in Wave 8 ( Figure 1a , Table E3 ), but not different to February 2020 (IRR 1.08; 95% CI 0.98, 1.19). In contrast, typical quantity consumed per occasion in May-June 2020 declined by 27% from February 2020 (Figure 1b) , continuing a downward trend since Wave 8 (41% lower than Wave 8). Binge drinking was stable across the period up to February 2020, but then declined by 28% in May-June 2020 (IRR 0.72; 95% CI 0.63, 0.83; Figure 1c ). Overall consumption (frequency x quantity) declined 17% from February 2020 (IRR 0.83; 95% CI 0.73, 0.95; Figure 1d ). Respondents reported a lower maximum number of drinks consumed in May-June 2020 compared with February (coef -1.58; 95% CI -1.99, -1.16; Table 3 and Table E3 ). Over a third (36.8%) reported experiencing an alcohol-related harm in February 2020, which declined to 26.2% in May-June 2020 (Table 4) . Similarly, there was around a 30% decline in the number of harms reported (IRR 0.65; 95% CI 0.54, 0.79; Table 3 ). In May-June 2020, compared to February 2020, there were increases in the proportion of drinking alone reported by participants (coef 2.15; 95% CI 0.50, 3.81) and in drinking with others virtually (coef 7.80; 95% CI 5.57, 10.02), and decreases in drinking with others in person (coef -13.80; 95% CI -17.56, -10.04) and in the use of alcohol delivery services (OR 0.68; 95% CI 0.43, 1.08). Changes in alcohol consumption were relatively consistent by gender, although men showed higher incidence rates of drinking across the period than women ( Figure E1 ). However, while there were significant declines in typical quantity consumed and binge drinking for both men and women ( Figures E1b and E1c) , there was a decline in overall consumption only among women (IRR 0.72; 95% CI 0.55, 0.94; Figure E1d and Table E4 ). Both men and women showed declines in the maximum amount of alcohol they consumed on an occasion, with mean declines of 1.74 drinks among men (95% CI: -2.41, -1.07) and 1.47 among women (95% CI: -1.99, -0.95; Figure E2a and Table E5 ). Patterns of alcohol-related harms were consistent across men and women ( Figure E2b ). Men drank alone more in May-June than February ( Figure E2c and Table E6 ), but women did not. Otherwise, the results for drinking context were similar to the overall results. There was no change in the use of alcohol delivery services by gender ( Figure E2f Those in the upper half of drinkers drank more frequently, and in greater quantity, than the lower half of drinkers. Both groups showed similar trends over the course of the study, with no change in frequency, but declines in typical quantity consumed and frequency of binge drinking. However, only those in the upper half of drinkers showed a decline in overall consumption ( Figure E3 and Table E7 ). Similarly, both the upper half and lower half of drinkers showed declines in the maximum number of drinks consumed ( Figure E4a and Table E8 ) and the number of alcohol-related harms experienced ( Figure E4b ). Neither group reported changes in use of alcohol delivery services ( Figure E4f and Table E9 ). Sensitivity analysis split by 'high risk' drinking based on binge drinking or consuming >10 drinks a week were very similar to the primary analysis ( Figures E5-E6 , and Tables E10-E12). Around one in five participants (19.4%) perceived their alcohol use had increased since the COVID-19 restrictions, while around a quarter (23.5%) said their use had decreased. Compared to a measure of change derived from the amount of alcohol reported consumed in in February 2020 and May-June 2020, there was a relatively high degree of discrepancy, with 61% in the same category (alcohol use decreased on both, alcohol use stayed the same on both, or alcohol use increased on both measures; Table E13 ). Of those who reported increased consumption, the most common reasons endorsed related to boredom and having fewer things to occupy them (78.0%) and having more time to drink (55.7%; Table 5 ). The most common reason for decreased consumption was having fewer opportunities to go out or be with people (91.2%). Results of the sensitivity analyses using continuous rather than categorical time were broadly consistent with the primary analyses, with increases in the frequency of consumption, but decreases in typical quantity consumed ( Figure E7 ). These results were similar when split by gender ( Figure E8 ) and overall drinking in Wave 8 ( Figure E9 ). These results were largely unaffected by the omission of the retrospective data for February 2020 ( Figures E10-12 ). We examined the impact of the COVID-19 pandemic and associated restrictions on alcohol consumption among a cohort of young adults, finding declines in typical quantity consumed, overall consumption, and frequency of binge drinking. The use of alcohol is the leading risk for disability adjusted life years in 10-24-year-olds (31), as well as being linked to a range of other health burdens (32, 33). As such, altered patterns of alcohol consumption during the pandemic, with young people less likely to drink in large quantities, is of considerable interest. The number of acute alcohol-related harms (which are more strongly associated with quantity consumed per drinking occasion than frequency of drinking) (34) saw a similar decline. This decline may be driven by the fact that drinking was more likely to occur alone or 'virtually' with others due to the need to isolate, which reduces the risk of harms such as fighting with strangers, and traffic accidents. Our findings broadly align with research showing that alcohol consumption in Australia declined due to restrictions early in the pandemic (35), although other research has shown an increase in quantity of alcohol consumed during the pandemic in Australia and internationally (36). In contrast, past research has shown greater increases in alcohol consumption among women than men during pandemics (37), while this study found relatively similar changes in both men and women, with pre-existing gender differences remaining. The changes observed in drinking context also align with other research showing young people consume more alcohol outside of the home (15). With restrictions preventing people from going out, 'drinking with others' was expected to decline. However, this was matched largely by increases in drinking 'virtually' with others. That is, it appears that alcohol consumption in this demographic remains a social activity, with only the medium changing. There were differences based on level of alcohol consumption pre-COVID-19, with declines in overall consumption among those drinking at heavier levels before COVID-19 but not those drinking at lower levels. This suggests studies examining population-level changes may miss important nuances in individual-level changes in consumption. While some of this may be attributable to regression to the mean, the changes in the higher consumption group mirrored the overall changes in the sample, while the lower consumption group was more stable, suggesting that declines in overall consumption were driven by reduced consumption by heavier drinkers. Importantly, the sample reported relatively little concern about the COVID-19 disease itself, and only one respondent had been diagnosed with the disease. Rather, changes in consumption appear to be driven by the COVID-19 restrictions, although given the range of resulting social and economic impacts, the precise cause remains unclear. In addition, while restrictions have softened since these data were collected, the pandemic is ongoing, and some restrictions will likely remain for some time. As such, it is likely that other impacts may emerge that were not present at the time of this survey and may have implications for alcohol consumption. A major strength of our study is that we used prospective longitudinal data, and thus do not rely on delayed recall of alcohol consumption prior to the pandemic, an issue underlined by the discrepancy between self-reported change and changed derived from prospective data. However, not all items were included in our prospective survey, and thus some analyses rely on recall about February 2020, a period three-four months prior to the survey completion. This article is protected by copyright. All rights reserved. Nevertheless, sensitivity analyses were similar regardless of whether they included the data for February 2020, providing some reassurance to the reliability of the data collected. In addition, the period covered by the COVID-19 survey was shorter than the annual APSALS surveysthus, seasonal differences may affect the COVID-19 survey that are less present in the main data. We also note the APSALS cohort were recruited using an opt-in process and thus results may not generalise to the general population of young adults. However, APSALS has similar levels of alcohol use and a comparable demographic profile to the Australian population, although families of lower socioeconomic status are underrepresented (18). While the APSALS sample was broadly representative, the COVID-19 survey sub-sample was more likely to be female. However, as the aim of the study was to investigate trends rather than prevalence, this should not impact the reliability of the findings. We present evidence indicating decreased overall alcohol consumption, binge drinking, and alcohol-related harms among young Australians during the COVID-19 restrictions, although this differed by levels of alcohol consumption prior to the pandemic. Callinan S, Livingston M, Room R, Dietze P. Drinking Contexts and Alcohol Consumption: How Much Alcohol Is Consumed in Different Australian Locations? J Stud Alcohol Drugs. 2016;77(4):612-9. Rodriguez LM, Litt DM, Stewart SH. Drinking to cope with the pandemic: The unique associations of COVID-19-related perceived threat and psychological distress to drinking behaviors in American men and women. Addictive behaviors. 2020;110:106532-. Neill E, Meyer D, Toh WL, van Rheenen TE, Phillipou A, Tan EJ, et al. Alcohol use in Australia during the early days of the COVID-19 pandemic: Initial results from the COLLATE project. Psychiatry and Clinical Neurosciences. 2020;n/a(n/a). Aiken Note: Outcomes were analysed using negative binomial mixed effects models, with Feb 2020 as the reference category, and results presented as exponented coefficients corresponding to incidence-rate ratios (IRRs), with bounds of 95% confidence intervals shown in dashed lines. Models are adjusted for covariates; full results are included in Supplementary Table E2 . Results are based on multiple imputation with estimates combined using Rubin's rules. 0.68 (0.43, 1.08) * Analysed using linear mixed effects models, with results presented as a coefficient equating to the difference in the mean maximum alcohol consumed. # Analysed using negative binomial mixed effects models, with results presented as an incidence-rate ratio (IRRs). † Analysed using linear mixed effects models with results presented as coefficients equating to the difference in the mean proportion of consumption in each setting. ‡ Analysed using logistic mixed effects models, and presented as an odds ratio (OR).Models are adjusted for covariates; full results are included in Supplementary Table E4 . Results are based on multiple imputation with estimates combined using Rubin's rules. The first official COVID-19 diagnosis was reported in Australia on 25 January 2020, leading to an increase in cases through February and March, with a peak of 469 new cases on March 28. This was followed by a period with relatively low rates of infection (<20 cases per day), until a '2 nd wave' from late June, largely in Victoria (1, 2) . Each state implemented its own restrictions on movement and gathering, with the exact rules varying by jurisdiction. However, restrictions on gatherings were implemented in all states from early March, and by the end of March Australia entered a period of lockdown, where residents were only permitted to leave their residence for essential reasons such as shopping for essentials (e.g., groceries) and medicines, accessing medical and health care services, exercise (limited in time and distance), and work (if unable to do so from home). These restrictions were reduced from mid-June, although significant restrictions were later enforced in Victoria from July. A summary of the outcome measures, including the waves for which each was asked and included, can be seen in Table C1 . Participants were asked how often they consumed alcohol. In Wave 9 and 10 of the APSALS survey, participants were asked how often they consumed alcohol in the past year, with the options being: Never; less often; about 1 day a month; 2-3 days a month; 1-2 days a week; 3-4 days a week; 5-6 days a week; and every day. In the COVID-19 survey they were asked how often they consumed alcohol in the past month, or the month of February, with the options: Not at all; once in the month; 2-3 days in the month; 1-2 days a week; 3-4 days a week; 5-6 days a week; and every day. W9 and W10 data was then prorated to past month consumption, with categories matching the COVID-19 survey (table C2) Typical quantity consumed. Participants who reported any alcohol consumption were asked how many standard drinks (where a standard drink was 10g of alcohol) they typically consumed on a drinking occasion. In the Wave 9 and Wave 10 surveys, participants were asked to select from the options: None; a sip or a taste; 1-2 drinks; 3-4 drinks; 5-6 drinks; 7-10 drinks; 11-12 drinks; or 13 or more drinks. In the COVID-19 survey, participants were asked to provide a number of drinks, which could range from 0 to 100. The COVID-19 survey responses were then coded to the same response levels as the Wave 9 and 10 surveys. Frequency of binge drinking. Participants who reported any alcohol consumption were also asked how often they consumed ≥5 standard drinks, with the same frequency options as for frequency of alcohol consumption. As with frequency of use, responses in the main APSALS survey were pro-rated to the past month (Table C2) . Overall consumption. Frequency of use and typical quantity consumed were multiplied together to create a proxy for overall consumption of alcohol in the month, in each of the four time periods. Maximum consumption. Those who reported any alcohol consumption in the COVID-19 survey were also asked what their maximum number of standard drinks on a single occasion in the past month, and the month of February, which could range from 0 to 100. This information was not collected in the main APSALS survey, so only change from February 2020 to May/June 2020 was examined. Alcohol-related harms. The COVID-19 survey included a subset of the harms from the School Health and Alcohol Harm Reduction Project (SHAHRP) scale (1). We chose items based on examination of the scale in the main APSALS data via factor analysis (excluding items that did not load >.40 on a single factor), as well as one additional item to cover accidents/injuries due to alcohol. The items included in the scale can be seen in Table C3 . The reduced scale is strongly correlated with the full scale (r>0.9 in all main waves of the APSALS survey). Harm items were queried for the past month and February in the COVID-19 survey. Alcohol-related harms were also asked in the APSALS main surveys, however because of the way the questions are phrased it was not possible to pro-rate to 'monthly' experience of harm, so we only conducted retrospective analysis of February and May-June 2020. Were you sick after drinking? Did you have a hangover after drinking? Have you been unable to remember what had happened while you had been drinking? Did you get in a physical fight with someone because you were affected by alcohol? Did you damage something because you were affected by alcohol? Did you get into trouble with your friends because of alcohol (your friends got annoyed with you)? When affected by alcohol, did you have sex that you later regretted? Did you have an accident, injury or fall due to alcohol? Drinking contexts. In the COVID-19 survey, participants were asked "what percentage of your drinking did you do…" with separate responses for: alone; physically with others; and virtually with others. This was asked of both past month, and the month of February. Each drinking context was analysed separate, coded as proportion of time drinking, ranging from 0 to 100. Use of delivery services. Participants were asked how often (if at all) they had alcohol delivered to their home in the past month, and in the month of February, as part of the COVID-19 survey. The possible response options were: Never; once; 2-4 times; 5-7 times; or 8 or more times. However, less than 2% of the sample reported using a delivery service more than once in either month, so responses were recoded into a binary (no/yes) variable of any use of alcohol delivery services. This article is protected by copyright. All rights reserved. Money to buy alcohol. Participants were also asked whether, in the past month or the month of February, they "usually have money to buy alcohol if you wanted to buy it?". This was recorded as a binary (no/yes) variable for each wave. Self-reported change in alcohol use. Participants were asked to self-report whether their alcohol use had increased, decreased or remained the same since the COVID-19 restrictions commenced in March as compared with February 2020. Those who reported their consumption had increased or decreased were then asked to select from a number of reasons why the change had occurred (with multiple selections possible). A full list of the possible response options can be seen in Table 3 . Covariates were chosen based on the literature on adolescent drinking and measured at each wave. The variables included were: child gender and age (2, 3), socio-economic status of area of residence (4), family history of alcohol problems (5), having older siblings (2), and their peer's substance use and disapproval of alcohol and tobacco use (2, 3, 5). Child gender. Participants were given the opportunity to identify as male, female, or another gender. Because only n=2 participants identified as other than male or female, they were excluded from the analysis. Worry about COVID-19. Participants were asked how worried they are about contracting the disease with response options ranging from 'Not at all worried' to 'Extremely worried'. The data contained a number of missing data points, both because not all participants completed all three surveys used in the analysis, and also because participants could refuse to answer individual questions in the surveys. Data was confirmed to be not missing completely at random via Little's test. As such, we have assumed the data to be missing at random. While there was relatively little missing data, because missingness can introduce bias when there is missingness in both the outcome and exposure variables (1), the analyses were conducted using multiple imputation. Based on past research (2), imputation conducted via the 'just another variable' approach, in which data is imputed in 'wide' form, with one record per individual, and each repeated measurement of the same variable imputed as a separate variable. We conducted the imputation in the R package 'mice' (3) . Just under a quarter of cases had at least some missing data (23%). Frequencies and the most common patterns of missing data are reported in Table D1 . To be conservative, we used M=50 imputations (4) . Imputation models included all variables included in the analysis, as well as auxiliary variables related to mental health, use of services, use of other substances, and personal situation (e.g. employment status, living situation). Analyses were then conducted on each imputed dataset, and combined using Rubin's rules. This article is protected by copyright. All rights reserved. Drinking frequency X 17 Typical quantity X 13 Binge frequency X 15 Maximum consumption X 16 Drinking alone X 13 Drinking with others in person X 13 Drinking with others virtually X 13 Use of delivery services X 16 Money for alcohol X 15 Alcohol-related harms 0 February 2020 Drinking frequency X 15 Typical quantity X 14 Binge frequency X 14 Maximum consumption X 13 Drinking alone X 11 Drinking with others in person X 11 Drinking with others virtually X 11 Use of delivery services X 13 Money for alcohol X 12 Alcohol-related harms 0 Wave 10 Drinking frequency X X 22 Typical quantity X X 24 Binge frequency X X 22 Money for alcohol X 12 Wave 9 Drinking frequency X X 18 Typical quantity X X 19 Binge frequency X X 18 Money for alcohol 15 Feb 2020 May -June 2020 This article is protected by copyright. All rights reserved. Feb 2020 May -June 2020 This article is protected by copyright. All rights reserved. This article is protected by copyright. All rights reserved. Table E13 Comparison of self-reported perceived change in alcohol consumption and derived change in consumption based on report of overall consumption. Derived change in drinking based on change in overall consumption between February 2020 and May-June 2020 n (%) Stayed the same Decreased COVID-19 and alcohol in Australia: Industry changes and public health impacts Illicit drug use and harms in Australia in the context of COVID-19 and associated restrictions: Anticipated consequences and initial responses Mitigating and learning from the impact of COVID-19 infection on addictive disorders Taking the Pulse of the Nation Stressful life experiences, alcohol consumption, and alcohol use disorders: the epidemiologic evidence for four main types of stressors Global Economic & Markets Research report: CBA Card Spend -week ending Impact of COVID-19 on the drinks industry: Alcohol Beverages Australia Changes in alcohol consumption associated with social distancing and self-isolation policies triggered by COVID-19 in South Australia: a wastewater analysis study Household Impacts of COVID-19 Survey, cat Depression, Anxiety and Stress during COVID-19: Associations with Changes in Physical Activity, Sleep, Tobacco and Alcohol Use in Australian Adults Alcohol Consumption Reported during the COVID-19 Pandemic: The Initial Stage Alcohol consumption during the COVID-19 period Australian Institute of Health and Welfare. National Drug Strategy Household Survey Patterns of financial attributes and behaviors of emerging adults in the United States COVID-19: a chronology of state and territory government announcements (up until Australian Government Department of Health. Coronavirus (COVID-19) current situation and case numbers 2021 COVID-19) in Australia Oxford COVID-19 Government Response Tracker The school health & alcohol harm reduction project: details of intervention development & research procedures. The School Health & Alcohol Harm Reduction Project: Details of intervention development & research procedures Predictors of initiation of alcohol use among US adolescents: findings from a prospective cohort study Use and abuse of alcohol and illicit drugs in US adolescents: results of the national comorbidity surveyadolescent supplement Information paper: an introduction to socio-economic indexes for areas (SEIFA) A model to determine the likely age of an adolescent's first drink of alcohol Socio-economic indexes for areas (SEIFA), postal area (POA) Index of Relative Socio-Economic Advantage and Disadvantage, data cube Young adult sequelae of adolescent cannabis use: an integrative analysis. The Lancet Psychiatry Monitoring the Future: Questionnaire Responses from the Nation's High School Seniors Accounting for missing data in statistical analyses: multiple imputation is not always the answer A comparison of multiple imputation methods for missing data in longitudinal studies mice: multivariate imputation by chained equations in R How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prevention Science Upper half of drinkers Upper half of drinkers Upper half of drinkers Upper half of drinkers Upper half of drinkers Low-moderate risk drinkers High risk drinkers High risk drinkers High risk drinkers High risk drinkers High risk drinkers Low-moderate risk drinkers High risk drinkers High risk drinkers Family history of alcohol No REF REF REF REF This article is protected by copyright This research includes computations conducted using the computational cluster Katana supported by Research Technology Services at UNSW Sydney. This article is protected by copyright. All rights reserved.