key: cord-0977871-4lkrc4j2
authors: Daly, M.; Robinson, E.
title: Problem drinking before and during the COVID-19 crisis in US and UK adults: Evidence from two population-based longitudinal studies
date: 2020-06-28
journal: nan
DOI: 10.1101/2020.06.25.20139022
sha: 6ce87a2ba954b11994a87d7e2e8dcf9bc7de6d9e
doc_id: 977871
cord_uid: 4lkrc4j2

Background The impact of the COVID-19 crisis on potentially harmful alcohol consumption is unclear. Aims To test whether the prevalence of problem drinking has changed from before to during the COVID-19 crisis in the US and UK. Design/Setting We examined nationally representative longitudinal data on how problem drinking has changed from pre-pandemic levels among adults in the US (N=7,327; Understanding America Study) and UK (N=12,594; UK Household Longitudinal Study). Methods In the US, we examined rates of consuming alcohol [≥] 4 times in the past week at baseline (March, 2020) and across four waves of follow-up (April-May, 2020). In the UK we assessed the prevalence of consuming alcohol [≥] 4 times per week and weekly heavy episodic drinking using the AUDIT-C at baseline (2017-2019) and during the COVID-19 lockdown (April, 2020). We also tested whether there were specific groups at greater risk of increased problem drinking during the pandemic. Results Among US adults, there was a statistically significant increase in the percentage of participants reporting drinking alcohol [≥] 4 times a week which rose significantly from 11.7% to 17.9% (53% increase, p < .001) as the COVID-19 crisis developed in the US. Among UK adults, the percentage of participants reporting drinking [≥] 4 times a week increased significantly from 14.2% to 23% (62% increase, p < .001) and heavy episodic drinking at least weekly increased significantly from 9.7% to 16.6% (71% increase, p < .001) when compared to pre-COVID-19 lockdown levels. Trends were similar across population demographics, although those aged under 50 years and higher income groups displayed the largest increases. Conclusions The COVID-19 crisis has been associated with substantial increases in problematic drinking in both US and UK adults.

The COVID-19 pandemic has resulted in governments introducing drastic measures to reduce 76 viral transmission. Many governments have introduced 'social lockdown' orders, which have 77 had severe effects on the economy and far reaching interpersonal consequences on working 78 life, childcare, travel and social contact. Although social lockdown orders will have reduced 79 the number of deaths caused by COVID-19, as of June 2020, in the UK alone there have been 80 more than 30,000 deaths attributed to 2) . There is also emerging evidence on 81 the indirect effects the COVID-19 pandemic has had on population health. For example, 82 initial findings from both the UK and the US indicate that the COVID-19 pandemic is likely 83 to have impacted mental health, with substantial increases in the prevalence of mental health 84 problems and depression estimated from nationally representative studies (3, 4) . 85 The extent to which alcohol use has changed as a result of COVID-19 crisis is 86 unclear. There is already a considerable public health burden caused by problematic drinking 87 (5) and alcohol misuse could increase risk of mortality from COVID-19 because of immune 88 function related health effects (6). For these reasons, it is crucial to understand how patterns 89 of problematic drinking have changed since the emergence of the COVID-19 crisis. In the US 90 and UK, there have been mass closures of non-essential businesses, including pubs, bars and 91 restaurants, which may have reduced the amount of alcohol that the population are drinking. 92 However, this has also coincided with a sharp rise in alcohol sales in supermarkets (7). There 93 are also concerns that COVID-19 social lockdown measures may result in a spike in alcohol 94 misuse, particularly among groups that are already at risk for problematic drinking patterns 95 (8, 9) . 96 Prior research has shown that exposure to traumatic events such as Hurricane Katrina 97 (10) and the 9/11 terrorist attacks (11) predicts alcohol misuse and drinking to alleviate 98 distress and worry related to the event. Yet, research studies examining problematic drinking 99 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020 . . https://doi.org/10.1101 /2020 Pg.5 during the COVID-19 crisis are limited. In two non-representative cross-sectional studies 100 relying on retrospective recall of alcohol drinking prior to the COVID-19 pandemic, 101 approximately one quarter of Australian adults and one third of Chinese adults reported that 102 their alcohol consumption had increased as a result of 13) . In a 103 repeated cross-sectional survey, there was an increase in the prevalence of high-risk drinking 104 of approximately 50% among 1700 UK adults (14) when comparing drinking before and after 105 COVID-19 social lockdown. Although these studies are suggestive of changes in problematic 106 drinking, findings may be explained by the use of retrospective recall and/or differences 107 between participants sampled before vs. during the COVID-19 crisis. Therefore, there is a 108 need for research that allows for examination of longitudinal changes in person-by-person 109 problematic drinking behavior before and after the development of the COVID-19 pandemic. 110 In the present research we examine changes in problematic drinking among US and 111 UK adults before and during the development of the COVID-19 pandemic. We make use of 112 two longitudinal studies with well characterized sampling frames and sampling weights that 113 provide a correction for selection probabilities and attrition bias enabling population 114 inferences to be generated. We examined problematic drinking patterns among US adults by 115 making use of data collected as part of the Understanding America Study. In this study, 116 drinking behavior was reported on early in the COVID-19 pandemic and before social 117 lockdown restrictions had been widely introduced in the US (March, 2020) vs. during 118 lockdown restrictions (April, 2020) and after the easing of restrictions (May, 2020). We also 119 examined drinking patterns among UK adults by making use of data collected as part of the 120 UK Household Longitudinal study in 2017-2019 and again in April, 2020 one month after the 121 introduction of UK-wide lockdown restrictions. To understand whether trends in problematic 122 drinking were socially patterned, we also examined changes in problem drinking based on 123 demographic sub-groups (age, sex, ethnicity, marital status, and income).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 28, 2020. were not completed on the assigned day but were completed within two weeks of the 147 assigned date. Sampling weights were applied in all analyses to adjust for non-response and 148 generate nationally representative estimates. In the UAS survey-wave specific sampling 149 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg.7 weights are generated using an adaptive sampling algorithm described elsewhere (19) . The 150 weights account for unequal probabilities of selection into the UAS and ensure each wave of 151 the study is aligned with the distribution of sociodemographic characteristics of the US 152 population.

The UKHLS is a longitudinal study that collects extensive information annually on CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. were asked "Out of the past 7 days, what is your best estimate of the number of days that you 177 did each of the following activities?" and were asked to complete the number of days they 178 "Consumed alcohol" alongside other health behaviors. To identify potentially problematic 179 drinking and enable comparisons with the UKHLS we dichotomized responses to this 180 question into those who drank more or less frequently than 4 times in the past week. (1=Never, 2=2-4 times per month, 3=2-3 times per week, 4=4+ times per week). In order to 185 capture drinking levels during the pandemic lockdown in the COVID-19 study the reference 186 period for this question was changed from "past 12 months" to "past 4 weeks" and response 187 scales were as follows: 1=Never, 2=Once, 3=2-4 times in total, 4=2-3 times per week, 5=4-6 188 times per week, and 6=Daily. In both waves, we characterized problematic drinking as 189 consuming alcohol 4 or more times per week.

The AUDIT-C also includes a question on the frequency of heavy episodic alcohol 191 use, defined for women/men as drinking 6/8 or more units on a single occasion: "How often period was "the past year" and response options were: 1=Never, 2=Less than monthly, 196 3=Monthly, 4=Weekly, 5=Daily or almost daily. Once again in April, 2020 the reference 197 period referred to was adapted from "the past year" to "the past 4 weeks" and heavy episodic 198 drinking was assessed with the response options: 1=Never, 2=Once, 3=Weekly, 4=Daily or 199 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg.9 almost daily. In both 2017-2019 and April, 2020 those indicating that they drank heavily on a 200 'Weekly' or 'Daily or almost daily' basis were classified as engaging in problem drinking.

Covariates. In both the UAS and UKHLS participants reported their age, sex, 202 ethnicity (grouped into white, non-white due to a low proportion of Black, Asian, and 203 minority participants in the UKHLS), marital status (married, not married) and household 204 income levels. Participants were grouped into four approximately even sized age groups (18- 

To do this, we first estimated the predicted probability of each problem drinking 220 outcome at each survey wave in logistic regression models that adjusted for differences in 221 participant age, sex, ethnicity (white, non-white), marital status (married, not married), and 222 household income tertiles. The Stata 'margins' command was then used to estimate 223 percentage-point changes in the binary outcomes of interest from the first survey wave / 224 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. 

Understanding America Study 239 The sample for our analyses included 7,327 participants assessed over five waves (Ns from 240 5,395-6,819). Participants were aged 48.5 (95% CI[48-49.1]) and 53.6% were female and 241 66% white, as shown in Table 1 . The characteristics of participants in each survey wave are 242 outlined in Table S1 which shows that the weighted sample composition was very similar in 243 each wave. The prevalence of drinking four or more times in the past week was 11.7% in the 244 first survey wave conducted in March, 2020 and increased to 17.8% in early April and 245 remained elevated at 16.5% by late May, 2020 (see Table 1 ). The increase in frequent 246 drinking was largest in magnitude amongst those aged under 50 years, whites, those who 247 were not married, and those living in households earning $40,000 or more per year. 248 249 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. 19.2%]) in early April, 2020 a statistically significant increase of 6.2% (95% CI[5.0%-7.5%]) 253 (p <.001), as shown in Table 2 . We also examined problem drinking among participants who 254 completed their baseline survey before lockdown measures were enacted (completed 255 assessments on 10 th -19 th March). We compared the prevalence of drinking ≥4 times per week 256 in this group with the high frequency drinking levels of the same group of participants as 257 averaged across assessment waves from April 15 th -May 13 th . As can be seen in Table S2 May (see Table 2 ).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. Changes in frequent drinking and heavy episodic drinking were largest in magnitude amongst 290 those aged 35-49, females, whites, and those on middle or high incomes (see Table 4 ). Table 5 . Similarly, the prevalence of heavy episodic drinking at least once a 297 week rose significantly from 9.7% (95% CI[9.0%-10.0%]) to 16.6% (95% CI[15.8%-17.4%]) 298 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg.13 over this period, a significant change of 6.9% (95% CI[6.1%-7.7%]). The increases in both 299 frequent drinking and heavy episodic drinking were statistically significant at the p <.001 300 level for all population subgroups examined (see Table 5 ) with the exception of non-whites 301 who did not show an increase in drinking four or more times per week.

The largest increase in frequent drinking (4 times or more a week) was observed 303 among those aged 35-49 years (11.7%, 95% CI[10.0%-13.4%]), followed by those aged 18-304 34 (10.0%, 95% CI[7.8%-12.2%]). Both groups showed increases that were significantly 305 larger than the increase identified among those aged 65 years and above, as shown in Table 6 .

Large rises in frequent drinking were also identified among middle income (10.1%, 95% 307 CI[8.7%-11.5%]) and high income (10.2%, 95% CI[8.7%-11.7%]) groups, and these 308 increases were significantly larger than the increase observed in the low income group (see 309   Table 6 ). The increase in frequent drinking among whites was also significantly larger than 310 the change among non-whites (by 7.4%, 95% CI[4.9%-10.0%]), as shown in Table 6 .

The most substantial rise in episodic heavy drinking at least once per week was also 312 among those aged 35-49 years (11.1%, 95% CI[9.3%-12.8%]) and was 7.5% (95% CI[5.3%-313 9.6%]) greater than the increase among those aged 65 year and over, as shown in Table 6 . All 314 age groups experienced increases in episodic heavy drinking that were significantly larger 315 than the increase in the 65+ years age group (see Table 6 ). Whites also showed a greater 316 increase in episodic heavy drinking compared to non-whites (by 3.8%, 95% CI[1.8%-6.0%]) 317 and those on high incomes showed a larger increase in episodic heavy drinking than those on CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. In the present research we examined changes in problematic drinking among US and UK 325 adults following the development of the COVID-19 crisis. Among US adults, we first 326 examined problem drinking in March 2020 (baseline) when the US death toll of COVID-19 327 was relatively low (~5,000) and few states had enacted social lockdown restrictions. We There are a number of plausible mechanisms that may explain population-wide 345 increases in problem drinking. The COVID-19 crisis is thought to have had a considerable 346 burden on population level mental health and this may have resulted in an increase in people 347 using alcohol to cope with stress and negative affect (24, 25) . In line with this, a cross-348 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020 . . https://doi.org/10.1101 /2020 Pg.15 sectional study of alcohol use in COVID-19 social lockdown found that using alcohol to cope 349 was associated with increased drinking in lockdown (26) and physical social engagement, which for many may have resulted in increased boredom.

Boredom is thought to have a range of effects on behavior and boredom proneness is linked 355 to higher alcohol consumption (28, 29), which may in part explain why alcohol use has 356 increased alongside the introduction of social lockdown measures.

Across both UK and US samples, higher income participants experienced the largest 358 increases in problem drinking. Associations between socioeconomic status and alcohol use 359 are complex, but higher income tends to be associated with more frequent binge drinking (30, 360 31). Historical data also suggests that alcohol related harm during times of economic crisis is 361 disproportionately large among the wealthy (32) and more educated (33). During the 362 COVID-19 crisis, it may be the case that existing tendencies towards problem drinking and 363 available material wealth make higher incomes groups more likely to respond to boredom 364 and/or stress by drinking heavily. In a similar vein, increases in problem drinking being larger 365 in white vs. other ethnic groups among UK participants may reflect that abstinence is more 366 common in non-white ethnic groups (34) and such groups would be less likely to use alcohol 367 to cope in times of stress. We also found that among UK (but not US) participants, those 368 under the age of 65 showed the smallest increases in problem drinking. It is plausible that 369 older adults may not be experiencing some of the stressors that younger age groups (e.g. job 370 insecurity due to already being retired, childcare and homeschooling arrangements) will be 371 having to cope with as a result of the COVID-19 crisis.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020 . . https://doi.org/10.1101 /2020 Pg.16

There are a number of strengths and limitations of the present research. We were able CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg.17 differences in reported alcohol use at baseline vs. follow-up. However, we note that in the US 398 sample the same reporting time frame was used and similar sized increases in problem 399 drinking were observed. As is the case with any longitudinal study there was some level of 400 attrition in both the UK and US samples. In the context of the COVID-19 crisis it is difficult 401 to predict how this may affect estimates of problem drinking. It is conceivable that 402 participants who have developed more substantial alcohol use problems may be more likely 403 to be lost at follow-up and this would underestimate size of change in problem drinking. The COVID-19 crisis has been associated with substantial increases in problematic drinking 407 in both US and UK adults. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020 . . https://doi.org/10.1101 /2020 Pg.19 is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020 . . https://doi.org/10.1101 /2020 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg.23 Table 1 .

Sample characteristics and the prevalence of drinking alcohol 4+ times in the past week before (March, 2020) and during the COVID-19 pandemic (April, 2020) Note: Estimates are derived from weighted data.

a Households earning less than $40,000 a year classified as low income, those earning $40,000 -$100,000 middle income, and those above this threshold as high-income households.

Pg. 25 Note: Estimates are derived from weighted data. 23 a Age groups are based on age reported in April, 2020. 24 b Households earning less than £2,500 a month (net) are classified as low income, those earning 25 £2,500-£4,000 middle income, and those above this threshold as high-income households. 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 

Note: Estimates are from marginal effects calculated after a logistic regression clustered by the individual participant 4 identifier and controlling for all characteristics presented

Note: Estimates are from marginal effects calculated after a logistic regression clustered by the individual participant identifier 16 and controlling for all characteristics presented. Coefficients indicate the difference in the percentage point increase in 17 drinking problems across survey ways between the groups examined

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020 Pg. 31 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted June 28, 2020. . https://doi.org/10. 1101 /2020