This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
The first UK COVID-19 lockdown had a polarizing impact on drinking behavior and may have impacted engagement with digital interventions to reduce alcohol consumption.
We examined the effect of lockdown on engagement, alcohol reduction, and the sociodemographic characteristics of users of the popular and widely available alcohol reduction app Drink Less.
This was a natural experiment. The study period spanned 468 days between March 24, 2019, and July 3, 2020, with the introduction of UK lockdown measures beginning on March 24, 2020. Users were 18 years or older, based in the United Kingdom, and interested in drinking less. Interrupted time series analyses using generalized additive mixed models (GAMMs) were conducted for each outcome variable (ie, sociodemographic characteristics, app downloads and engagement levels, alcohol consumption, and extent of alcohol reduction) for existing (downloaded the app prelockdown) and new (downloaded the app during the lockdown) users of the app.
Among existing users of the Drink Less app, there were increases in the time spent on the app per day (B=0.01,
Following the first UK lockdown, there was evidence of increases in engagement and alcohol consumption among new and existing users of the Drink Less app.
Alcohol consumption is a dose-dependent [
The first lockdown was introduced in the United Kingdom in response to the COVID-19 pandemic, in March 2020. Following initial government advice from March 16, 2020, to avoid group gatherings and to work from home, where possible, the first national lockdown was announced with behavioral restriction measures coming into force in the United Kingdom from the March 24, 2020, remaining in place until July 4, 2020 [
In addition to leading to increased alcohol consumption among some groups, the first lockdown also led to an increase in self-reported alcohol reduction attempts by increasing and higher-risk drinkers (28.5% during the lockdown vs 15.3% prelockdown) [
Digital interventions, such as websites and apps, are convenient and low cost [
Understanding how the initial lockdown affected engagement with digital alcohol interventions and subsequent alcohol reduction could inform the targeting of public health messaging and the provision of alcohol support in the future. It is also important to consider the consistency of these effects across the population. Drinkers who drink more heavily [
The aims of this paper are threefold. First, to understand how the use of Drink Less and drinking behavior may have changed during the lockdown, we examined whether the lockdown affected engagement and recorded drinking behavior among existing Drink Less users (ie, those who downloaded the app prelockdown). Second, we examined whether the lockdown led to a change in new app downloads. Third, to understand how the lockdown may have impacted the characteristics of users downloading the Drink Less app, we analyzed whether the sociodemographic characteristics of users, engagement with the app, and drinking behaviors recorded in the app differed between new users downloading the app prelockdown and during the lockdown.
This study addresses the following 3 research questions (RQs):
RQ1: Was the first UK lockdown associated with an immediate change in existing users of the Drink Less app in terms of:
The depth of use (percent of available screens viewed)
The amount of use (mean time spent on the app)
The frequency of use (number of sessions)
The number of alcohol units recorded each day
The number of alcohol-free days recorded each day
The number of heavy-drinking days recorded each day
RQ2: Was the first UK lockdown associated with a change in the number of new Drink Less downloads per day?
RQ3: Was the first UK lockdown associated with an immediate change in new users of the Drink Less app in terms of:
Sociodemographic and drinking characteristics at baseline and in the 28 days following app download
The depth of use (percentage of available screens viewed)
The amount of use (mean time spent on the app)
The frequency of use (number of sessions and number of days used)
The number of alcohol units recorded
The number of alcohol-free days recorded
The number of heavy-drinking days recorded
This was a natural experiment without active recruitment.
For all RQs, the interruption was conceptualized as the introduction of national lockdown measures in the United Kingdom on March 24, 2020. Due to differences in the way that the independent variables were operationalized (see the Analysis section), the time periods for RQ1, RQ2, and RQ3a differed from RQ3b-g. Specifically, for RQ1, each of the outcome variables was aggregated at a daily level by the number of active users in that week. For RQ2, the number of new downloads each day was attributed to the day of download. For RQ3a, baseline AUDIT scores and sociodemographic variables were measured once and attributed to the day of download. As such, data were collected for the 468 days between March 24, 2019, and July 3, 2020. This captured the period of time up until pubs in England reopened on July 4, 2020, which could have had a stepped effect on alcohol consumption. This period was divided into pre- (March 24, 2019-March 23, 2020; 366 days) and during-lockdown (March 24, 2020-July 3, 2020; 102 days) segments. For RQ3b-g, the depth, amount, and frequency of use, along with the number of alcohol units, alcohol-free days, and heavy-drinking days, recorded were aggregated over the 28-day period following app download by the number of users who downloaded the app on that day. Therefore, to limit potential confounding after pubs reopened, only respondents downloading the app a full 28 days prior to when pubs reopened (up to June 6, 2020) were included. Therefore, the study period for RQ3b-g was March 24, 2019-June 6, 2020 (441 days) and was divided into pre- (March 24, 2019-March 23, 2020; 366 days) and during-lockdown (March 24, 2020-June 6, 2020; 75 days) segments.
The sample was UK users who downloaded the Drink Less app from Apple App Store, where it is freely available. To be eligible for inclusion, users had to be aged 18 years or older, based in the United Kingdom, interested in drinking less (specified when downloading the app), and have agreed to the privacy policy and terms and conditions within the app, as well as completing the AUDIT. For RQ1, existing users were defined as all those who downloaded the app between March 24, 2019, and the March 23, 2020 (prelockdown). RQ1 focused on existing, regular users, defined as use of the app at least once a week for a minimum of 4 weeks. For RQ2 and RQ3a, new users were defined as those who downloaded the app between March 24, 2020, and July 3, 2020, with no limits on regularity of use for new or existing users. Finally, for RQ3b-g, new users were defined as those who downloaded the app between March 24, 2020, and June 6, 2020, with no limits on regularity of use for new or existing users.
Three sociodemographic characteristics were measured at download. These were age (in years, continuous), sex (percentage female), and employment type (percentage nonmanual). The AUDIT was asked of all users providing both an AUDIT score (continuous) and the percentage of increasing and higher-risk drinkers (AUDIT score>=8).
The number of new app downloads each day was recorded.
Three indicators of user engagement were derived from screen view records for each user: (1) number of sessions (where a new session is defined as a new screen view after 30 minutes of inactivity), (2) time spent on the app in minutes, and (3) percentage of available screens viewed. For RQ1, these measures were aggregated at a daily level across active users and attributed to the day of engagement. For RQ3b-g, each measure was aggregated over the 28-day period following app download for each user and was attributed to the date of download. Due to the differences in aggregation, the number of days used was also included as a measure of engagement for RQ3.
In the app, users were prompted to fill in a daily drinking calendar, where they either marked days as “alcohol free” or entered any alcoholic drinks they drank that day. Three drinking variables were calculated: (1) number of alcohol units (UK unit=10 mL of ethanol) consumed (aggregated daily), (2) number of alcohol-free days, and (3) number of heavy-drinking days (defined as >6 alcohol units). As described before, these measures were operationalized differently for RQ1 and RQ3.
All analyses were conducted in R Studio. The engagement measures were derived using Pandas, a Python framework, within a Jupyter Notebook, an open source web application.
Models 1a-1f (for RQ1) examined whether the lockdown was associated with the percentage of available screens viewed, the mean time spent on the app, the number of sessions on the app, the number of alcohol units, the number of alcohol-free days, and the number of heavy-drinking days among existing, regular users of the Drink Less app.
Model 2 (for RQ2) examined whether the lockdown was associated with a change in the number of new daily downloads of the Drink Less app.
Models 3a-3e (for RQ3a) examined whether the lockdown was associated with changes in age, the proportion of female users, the proportion of nonmanual users, the proportion of users who were at risk of alcohol dependence, and the AUDIT scores among new downloaders of the Drink Less app.
Models 3f-3l (for RQ3b-g) examined whether there were changes in the percentage of available screens viewed, the mean time spent on the app, the number of sessions on the app, the number of days the app was used, the number of units, the number of alcohol-free days, and the number of heavy-drinking days recorded among new users following the first UK lockdown.
To estimate the associations between the lockdown and each of the outcomes, we conducted separate interrupted time series analyses using generalized additive mixed models (GAMMs). Analyses were conducted at the daily aggregated level while controlling for day of the week and month of the year. Smoothing “splines” were fitted in order to account for seasonal nonlinear variations in, for example, drinking behavior. To account for differences in the trends prelockdown and during the lockdown, the regression models included terms for the baseline level for each outcome prelockdown, the trend in the prelockdown period, the level change in the outcome immediately after the lockdown, and the trend in the during-lockdown period.
Plots of the autocorrelation functions (ACFs) and partial autocorrelation functions (PACFs) were used to test for both autoregressive (AR) and moving average (MA) autocorrelation over time. The ACF plots were used to identify plausible values for AR and MA terms for the baseline model. Models with various plausible AR and MA terms were compared with our baseline model using the Akaike information criterion (AIC), where smaller values indicate a better model fit.
As little was known about how the trends in each of the outcome variables during the lockdown, secondary analyses assessed whether regression models with nonlinear trends (cubic and quadratic) provided a better fit to the data. Best-fitting models were selected with the AIC, and where appropriate, cubic and quadratic models were reported.
All continuous variables were normally distributed, but a negative binomial distribution was used for the number of new downloads (RQ2), as the outcome variable was operationalized as a discrete (rather than continuous) variable and overdispersion was present. Data and details for each model (ie, AR and MA terms, AICs) are available in the annotated R code and can be accessed through GitHub [
Preplanned sensitivity analyses (adjusting the date of the interruption to that on which social distancing measures were introduced, March 17, 2020, and controlling for potential confounders) were not conducted due to the complexity of the paper, the practical constraints associated with running additional analyses, and the robust model selection approach already taken.
Ethical approval was obtained from University College London’s Research Ethics Committee (CEHP/2016/556; CEHP/2020/579), and participants provided online consent to having their anonymous data used for scientific research purposes.
Descriptive statistics for the aggregated outcome variables of interest, stratified by period (prelockdown vs during the lockdown).
RQsa | Entire period (441 days) | Entire period (468 days) | Prelockdown (366 days) | During the lockdown (75 days)b | During the lockdown (102 days) | |
|
||||||
|
Mean percentage screens viewed | N/Ac | 4.66 (0.80) | 4.80 (0.83) | N/A | 4.16 (0.30) |
|
Mean time spent on app (minutes) | N/A | 1.58 (0.72) | 1.73 (0.73) | N/A | 1.05 (0.25) |
|
Mean number of logins | N/A | 0.95 (0.11) | 0.97 (0.11) | N/A | 0.89 (0.06) |
|
Mean alcohol units per day | N/A | 3.29 (1.52) | 3.38 (1.55) | N/A | 2.98 (1.36) |
|
Mean heavy-drinking days | N/A | 0.21 (0.10) | 0.21 (0.10) | N/A | 0.18 (0.09) |
|
Mean alcohol-free days | N/A | 0.46 (0.16) | 0.48 (0.16) | N/A | 0.42 (0.16) |
|
||||||
|
Number of new downloads per day | N/A | 14.50 (13.00) | 18.00 (13.00) | N/A | 7.5 (6.00) |
|
||||||
|
Gender (proportion female) | N/A | 0.54 (0.16) | 0.53 (0.15) | N/A | 0.57 (0.21) |
|
Age (years) | N/A | 44.28 (4.07) | 44.16 (3.64) | N/A | 44.73 (5.34) |
|
Employment type (proportion nonmanual) | N/A | 0.71 (0.15) | 0.71 (0.13) | N/A | 0.69 (0.19) |
|
AUDIT score | N/A | 16.47 (2.41) | 16.40 (2.06) | N/A | 16.71 (3.40) |
|
At risk of alcohol dependence (proportion at risk) | N/A | 0.91 (0.09) | 0.91 (0.08) | N/A | 0.90 (0.11) |
|
||||||
|
Number of logins | 15.18 (6.60) | N/A | 14.80 (5.80) | 17.05 (9.42) | N/A |
|
Number of days used | 9.66 (3.41) | N/A | 9.46 (2.96) | 10.65 (4.97) | N/A |
|
Percentage screens viewed | 0.31 (0.04) | N/A | 0.30 (0.04) | 0.31 (0.06) | N/A |
|
Time spent on app (minutes) | 39.08 (21.64) | N/A | 38.73 (20.28) | 40.81 (27.51) | N/A |
|
Alcohol units | 72.21 (36.65) | N/A | 70.60 (31.51) | 80.15 (55.10) | N/A |
|
Heavy-drinking days | 4.44 (2.19) | N/A | 4.35 (1.63) | 4.90 (3.94) | N/A |
|
Alcohol-free days | 10.70 (5.42) | N/A | 10.74 (5.31) | 10.48 (5.93) | N/A |
aRQ: research question.
bShorter during-lockdown period as outcome variables 3b-e were aggregated over 28 days rather than at a monthly level.
cN/A: not applicable.
dMedian (IQR) presented here; a large variance could lead to a skewed mean.
There was an overall decline in the mean time spent on the app, the mean units recorded per day, and the mean heavy-drinking days recorded during the study period, with no step changes following the first COVID-19 lockdown. However, there was a change in slope during the lockdown, with a significant increase in the trajectory in the mean time spent on the app, the mean units recorded per day, and the mean heavy-drinking days recorded following the lockdown, though the magnitude of the change in these daily trends appeared small (
There was an overall decline in the mean percentage of screens viewed and the mean number of sessions during the study period, with no step change following the first COVID-19 lockdown in the United Kingdom. The declining trend plateaued during the lockdown, with no significant trend in the mean percentage of screens viewed or the mean number of sessions (
There was no significant trend in the mean alcohol-free days over the study period and no step change or change in slope following introduction of the first lockdown in the United Kingdom (
Results of the best-fitting model for each outcome variable for RQ1a (N=468 days, range 9-598 users per day).
Outcome variables | B (95% CI) | |||
|
||||
|
Trend | –0.0042 (–0.0074 to –0.0010) | .01 | |
|
Level | 0.0204 (–0.5541 to 0.5949) | .95 | |
|
Slope | 0.0038 (–0.0105 to 0.0181) | .60 | |
|
||||
|
Trend | –0.0061 (–0.0081 to –0.0041) | .00 | |
|
Level | 0.0604 (–0.3230 to 0.4438) | .76 | |
|
Slope | 0.0118 (0.0027-0.0209) | .01 | |
|
||||
|
Trend | –0.0005 (–0.0007 to –0.0003) | .00 | |
|
Level | –0.0247 (–0.1083 to 0.0589) | .56 | |
|
Slope | 0.0012 (–0.0001 to 0.0025) | .09 | |
|
||||
|
Trend | –0.0049 (–0.0066 to –0.0032) | .00 | |
|
Level | 0.0351 (–0.9218 to 0.9920) | .94 | |
|
Slope | –0.0297 (–0.1065 to 0.0471) | .45 | |
|
Slope2 | 0.0016 (–0.0001 to 0.0033) | .07 | |
|
Slope3 | 0.0000 (0.0000-0.0000) | .02 | |
|
||||
|
Trend | –0.0004 (–0.0005 to –0.0003) | .00 | |
|
Level | 0.0106 (–0.0498 to 0.0710) | .73 | |
|
Slope | –0.0019 (–0.0068 to 0.0030) | .44 | |
|
Slope2 | 0.0001 (0.0000-0.0002) | .06 | |
|
Slope3 | 0.0000 (0.0000-0.0000) | .02 | |
|
||||
|
Trend | –0.0002 (–0.0004 to 0.0000) | .10 | |
|
Level | –0.0183 (–0.1101 to 0.0735) | .70 | |
|
Slope | 0.0005 (–0.0009 to 0.0019) | .44 |
aRQ: research question.
bAdjusted for month of the year (cubic spline), day of the week (cubic spline), and autocorrelation.
Engagement indicators among existing, regular users of the Drink Less app over the study period (RQ1a-f). The red line indicates fitted values, the gray area indicates the 95% CI, and the dashed blue line indicates the interruption (ie, the first national UK lockdown). RQ: research question. Higher-resolution version of this figure is available in
There was an overall declining trend in the number of downloads per day over the full study period, with no step change detected following the first COVID-19 lockdown in the United Kingdom. However, the declining trend plateaued during the lockdown, with no significant trend in downloads per day (
Results of the best-fitting model for the number of downloads (N=468 days; range 0-288 downloads per day).
Number of downloadsa, cubic model | Incidence rate ratio (95% CI) | |
Trend | 0.9962 (0.9945-0.9979) | .00 |
Level | 0.5365 (0.2332-1.2342) | .14 |
Slope | 1.0161 (0.9495-1.0875) | .64 |
Slope2 | 1.0006 (0.9991-1.0021) | .45 |
Slope3 | 1.0000 (1.0000-1.0000) | .18 |
aAdjusted for month of the year (cubic spline), day of the week (cubic spline), and autocorrelation.
Number of new Drink Less downloads per day over the study period (RQ2). The red line indicates fitted values, the gray area indicates the 95% CI, and the dashed blue line indicates the interruption. RQ: research question.
There was no significant overall trend detected in the proportion of female users over the study period or the proportion of nonmanual (vs manual) workers, with no step change following the lockdown. However, there was a change in slope following the introduction of the first UK lockdown to a significant upward trajectory in the proportion of female users and a significant negative trajectory in the proportion of nonmanual workers (
There was no significant overall trend in the proportion of new users who were at risk of alcohol dependence over the study period. However, there was a step decrease during the lockdown, followed by a change in slope to an upward trend, whereby the proportion of new users who were at risk of alcohol dependence increased after the first lockdown in the United Kingdom, though the magnitude of this trend appeared small (
We did not detect a significant trend over the study period or a step change or change in slope following the introduction of the first UK lockdown in age or AUDIT scores of new app users (
Results of the best-fitting model for each outcome variable for RQ3aa (N=440 days; range 1-245 users per day).
Outcome variables | B (95% CI) | ||
|
|||
|
Trend | –0.0001 (–0.0003 to 0.0001) | .10 |
|
Level | 0.0109 (–0.0692 to 0.0910) | .79 |
|
Slope | 0.0013 (0.0001-0.0025) | .04 |
|
|||
|
Trend | 0.0006 (–0.0036 to 0.0048) | .77 |
|
Level | 0.3648 (–1.5329 to 2.2625) | .71 |
|
Slope | 0.0015 (–0.0272 to 0.0302) | .92 |
|
|||
|
Trend | 0.0000 (–0.0001 to 0.0001) | .65 |
|
Level | 0.0367 (–0.0271 to 0.1005) | .26 |
|
Slope | –0.0010 (–0.0020 to 0.0000) | .04 |
|
|||
|
Trend | 0.0020 (–0.0003 to 0.0043) | .09 |
|
Level | –0.3994 (–1.4571 to 0.6583) | .46 |
|
Slope | 0.0044 (–0.0116 to 0.0204) | .59 |
|
|||
|
Trend | 0.0000 (–0.0001 to 0.0001) | .73 |
|
Level | –0.0596 (–0.1060 to –0.0132) | .01 |
|
Slope | 0.0028 (0.0009-0.0047) | .004 |
|
Slope2 | 0.0000 (0.0000-0.0000) | .01 |
aRQ: research question.
bAdjusted for month of the year (cubic spline) and day of the week (cubic spline).
cAdjusted for autocorrelation.
dAUDIT: Alcohol Use Disorders Identification Test.
Sociodemographic and drinking characteristics of new users of the Drink Less app over the study period (RQ3a). The red line indicates fitted values, the gray area indicates the 95% CI, and the dashed blue line indicates interruption. AUDIT: Alcohol Use Disorders Identification Test; RQ: research question. Higher-resolution version of this figure is available in
There was no significant trend in the number of days used by new users across the study period and no change in slope. However, there was a step increase in the number of days used by new users immediately following the introduction of the first UK lockdown (
There was an overall upward trend in terms of the percentage of available screens viewed across the study period by new users. There was a step decrease immediately following the introduction of the first UK lockdown, and the upward trend stabilized, with no significant trend during the lockdown (
There were no significant trends across the whole study period for mean alcohol units or heavy-drinking days reported by new users. However, there was a step increase for both following the introduction of the first UK lockdown but no significant change in slope.
There was an overall upward trend in alcohol-free days reported over the whole study period by new users, with no significant step change. The upward trend stabilized, with no significant trend during the lockdown.
There was no overall trend in the number of logins or time spent on the app in minutes across the whole study period for new users, with no significant step change and no significant change in slope following the introduction of the first UK lockdown.
Results of the best-fitting model for each outcome variable for RQ3b-ga (N=440 days; range 1-245 users per day).
Outcome variables | B (95% CI) | |||
|
||||
|
Trend | –0.0063 (–0.0126 to 0.0000) | 0.05 | |
|
Level | 2.6500 (–0.6376 to 5.9376) | 0.12 | |
|
Slope | 0.0262 (–0.0437 to 0.0961) | 0.46 | |
|
||||
|
Trend | –0.0022 (–0.0055 to 0.0011) | 0.19 | |
|
Level | 2.0484 (0.3449-3.7519) | 0.02 | |
|
Slope | –0.0099 (–0.0461 to 0.0263) | 0.59 | |
|
||||
|
Trend | 0.0001 (0.0000-0.0002) | 0.00 | |
|
Level | –0.0265 (–0.0511 to –0.0019) | 0.04 | |
|
Slope | 0.0000 (–0.0005 to 0.0005) | 0.87 | |
|
||||
|
Trend | –0.0188 (–0.0408 to 0.0032) | 0.09 | |
|
Level | 0.9802 (–10.4248 to 12.3852) | 0.87 | |
|
Slope | 0.1412 (–0.1013 to 0.3837) | 0.25 | |
|
||||
|
Trend | –0.0314 (–0.0668 to 0.0040) | 0.08 | |
|
Level | 20.1247 (1.7605-38.4889) | 0.03 | |
|
Slope | –0.0975 (–0.4880 to 0.2930) | 0.62 | |
|
||||
|
Trend | –0.0019 (–0.0039 to 0.0001) | 0.06 | |
|
Level | 1.3845 (0.3318-2.4372) | 0.01 | |
|
Slope | –0.0105 (–0.0329 to 0.0119) | 0.36 | |
|
||||
|
Trend | 0.0083 (0.0028-0.0138) | 0.003 | |
|
Level | –1.7637 (–4.3836 to 0.8562) | 0.19 | |
|
Slope | –0.0144 (–0.0692 to 0.0404) | 0.61 |
aRQ: research question.
bAdjusted for month of the year (cubic spline) and day of the week (cubic spline).
cAdjusted for autocorrelation.
Aggregated engagement indicators among new users of the Drink Less app over the study period (RQ3b-e). The red line indicates fitted values, the gray area indicates the 95% CI, and the dashed blue line indicates the interruption. RQ: research question. Higher-resolution version of this figure is available in
Following the first COVID-19 lockdown in the United Kingdom, there was a significant increase in the time spent on the app and in the mean alcohol units per day and the number of heavy-drinking days recorded by existing, regular users of the Drink Less app, although no change was detected in the percentage of screens viewed, the number of sessions logged, or in the number of alcohol-free days recorded. There was no increase in downloads per day following the lockdown, although the overall negative trend in new daily app downloads plateaued following the introduction of the first UK lockdown. Among the new users of Drink Less, there were increases in the proportion of female users, manual workers, and those at risk of alcohol dependence following the first UK lockdown. With regard to changes in engagement indicators, there was a step increase in the number of days the app was used but a step decrease in the percentage of available screens viewed within Drink Less, suggesting that users engaged with the app for a longer period but with less of the available content. In terms of drinking characteristics, new users reported step increases in the mean number of alcohol units and heavy-drinking days aggregated over 28 days after app download following the first UK lockdown.
A strength of this study is that it was a natural experiment based on longitudinal data exploring self-motivated engagement with a freely available alcohol reduction app in the real world in a large sample against the backdrop of a global pandemic. However, there are also limitations associated with this approach. The period here reflects the immediate effects of COVID-19, which may not have been consistent over a longer period. These findings were also isolated to the United Kingdom. The app is reliant on self-reported alcohol consumption data. It is possible that changes in drinking contexts during the lockdowns could have affected the accuracy of self-report data. People were likely to be drinking in smaller groups in private, rather than public, settings, where they may have been more willing or less likely to forget to log drinks. Conversely, logged drinks during the lockdown might be more likely to be underestimated as individuals pouring their own drinks at home may be less likely to use standard measures than in on-trade settings. Finally, although the GAMMs incorporated cyclic cubic terms for day and month, additional seasonality terms may have further improved the model fit. This should be explored in future research involving app data. A recent study showed that COVID-19 had different effects on health behaviors in different countries, increases in alcohol consumption during the early months of the pandemic were recorded in the United Kingdom and Ireland, and decreases in consumption were recorded among 20 other European countries within the same period [
This research focused on 1 digital intervention, and it would be of interest to attempt to triangulate these findings across other forms of digital support available in the United Kingdom and internationally. This would aid in building a comprehensive overview of the effect of the COVID-19 pandemic on the use of digital alcohol reduction support and drinking patterns. The longer-term impact of the ongoing pandemic on engagement with digital interventions is also of interest. Furthermore, although increased engagement is positive, it is also necessary to examine the success of alcohol reduction attempts and whether they can be better supported.
This study indicated increases in units of alcohol consumed and heavy-drinking days among both existing and new users of the Drink Less app following the first UK lockdown. This is in line with other research outlining the polarizing impact of the first UK lockdown on alcohol consumption [
There were also increases in the proportion of female users, manual workers, and those at risk of alcohol dependence following the first UK lockdown. Shifts in engagement with the Drink Less app may be linked to more dramatic changes in lifestyles throughout the lockdown. There is evidence that women were disproportionately affected by additional caring responsibilities during the lockdowns [
Following the first COVID-19 lockdown in the United Kingdom, there is some evidence of increased engagement with the alcohol reduction app Drink Less
Higher resolution versions of
autocorrelation function
Akaike information criterion
autoregressive
Alcohol Use Disorders Identification Test
generalized additive mixed model
moving average
partial autocorrelation function
research question
MO, GL, and CG are funded by the National Institute for Health Research (NIHR) Public Health Research Programme (project reference NIHR127651). CG and OP are funded by Cancer Research UK (CRUK; C1417/A22962). JB is a member of the SPECTRUM consortium. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) license to any author-accepted manuscript version arising.
MO and CG are paid scientific consultants for the behavior change and lifestyle organization One Year No Beer. JB has received unrestricted research funding to study smoking cessation from companies who manufacture smoking cessation medications (Pfizer and J&J). OP and GL declare no conflicts of interest.