Did self ‐ reported tobacco smoking, alcohol consumption, and physical activity change during the COVID-19 restrictions in Germany in spring 2020? Findings from a population survey (the DEBRA study)

Broad nationwide restrictions of social life and contacts were implemented in Germany on March 22nd, 2020, to reduce the spread of the severe acute respiratory syndrome coronavirus type 2 (SARS ‐ CoV ‐ 2). It is unclear how these restrictions affected peoples’ health behaviour. Objective To: i) examine changes in self-reported health behaviour of the German population regarding tobacco smoking, alcohol consumption, and physical activity during the restrictions compared with the time prior to these restrictions; ii) explore associations between potential changes and socioeconomic and sociodemographic characteristics.


Introduction
It is quite likely that these massive restrictions on personal liberty and the accompanying feelings of uncertainty (both job-related and health-related) also had an impact on the health behaviour of people [8].
The World Health Organization (WHO) already warned at the beginning of the pandemic that the risk of increasing alcohol consumption due to isolation is possible [9,10]. Besides, mental health problems, such as depression or anxiety disorders, which are also associated with increased alcohol and tobacco consumption, could be triggered or aggravated by these conditions [11][12][13][14].
Other health behaviours such as physical activity might have changed as well since sporting habits could not be pursued during this time for many areas of activity (e.g., tness studios, club sports, or active commuting [15]). This poses a risk, especially for people with pre-existing chronic diseases, such as cardiovascular diseases, diabetes or mental illness, for whom physical activity has positive effects [16][17][18][19][20][21][22][23][24][25]. On the other hand, the pandemic-related reduction of commuting and working hours (short-time work) might have offered more time to purse physical outdoor activities for some people.
In Germany, the number of cigarettes sold increased markedly during the rst comprehensive restrictions in March 2020 (number of cigarettes sold: February 2020 = 4,949,032,692; March 2020 = 8,174,948,907) [26]. For comparison: the relative increase in sales from February to March 2020 was 65%, whereas it was only 14% from February to March 2019.
Visiting bars or restaurants was severely or completely restricted during the aforementioned period, and these establishments were thus not eligible as a place for alcohol consumption. First data from other countries (e.g., United States [27], Belgium [28,29], Australia [30], England [31,32], Pakistan [33], Italy [34], Scotland [35]) report changes in tobacco smoking, alcohol consumption and/or physical activity during the rst national COVID-19 restrictions. Whereas data from United States, England, and Belgium indicate a substantial increase in alcohol consumption [27,28,32], further data from England and Australia showed mixed results with a proportion of respondents who reported increased alcohol consumption and roughly the same proportion who drank less during this time [30,31]. According to these studies certain person characteristics are associated with alternated consumer behaviour: women, people with a higher socioeconomic status, symptoms of anxiety, depression and stress were associated with increased alcohol consumption [30,31].
As for tobacco consumption, ndings are also mixed. While a study from Belgium (web-based survey) and a longitudinal study conducted in Pakistan found an increase [28,33] in tobacco smoking, webbased surveys from Australia and England, as well as a cross-sectional study from England reported no changes in smoking behaviour [30,32,36], or even a strong reduction in tobacco smoking during the COVID-19 restrictions [33].
Concerning physical activity, a decrease and therefore negative effect has been observed in the majority of the 66 articles (N = 86.981) included in a systematic review [37]. A representative study from Belgium, as well as further web-based studies from Italy and England showed a reduction in moderate to vigorous physical activity but an increase in strength training [29,34,38]. The representative (cross-sectional) population survey (N = 13,515) from Belgium showed that the change in physical activity behaviour seems to be dependent on the age of the respondents, their previous physical activity behaviour (before) and habits [29]. A reduction of the physical activity was particularly evident in people who previously had a high level of physical activity, were less educated, and were more used to training in groups or in a sports club [29].
The current evidence shows that speci c groups of the society are more likely to be affected by health behaviour changes during the rst COVID-19 restrictions. As the level of the restrictions differed substantially between countries, there is a need to obtain data from different countries to get an overview and a solid data basis on health behaviour changes related to COVID-19 restrictions and associations with personal characteristics. We are still in an exceptional situation worldwide. In Germany, only few data are available so far, and these data indicate an increasing trend in self-reported alcohol consumption as well as tobacco consumption during restrictions at the beginning of the pandemic [39]. These data are important for public health to understand the effects of these restrictions and to be able to counteract them with targeted health policy/care measures. In addition, these data help to measure the health economic relevance of these restrictions; and, if appropriate, to detect a worsening of the deprivation of people from more disadvantaged groups in terms of health behaviour.
The present study therefore aims to examine whether there was a change in self-reported health behaviour regarding tobacco smoking, alcohol consumption, and physical activity during the rst COVID-19 restrictions in Germany in spring 2020, compared with the time prior to these restrictions. Furthermore, this study aims to explore potential associations between changes in health behaviour, if any, and socioeconomic and sociodemographic characteristics. A sample of the nationwide, representative German Study on Tobacco Use (DEBRA) serves as data basis to answer these questions.

Methods
We used data from the German Study on Tobacco Use (DEBRA: "Deutsche Befragung zum Rauchverhalten", www.debra-study.info), an ongoing representative household survey [40]. The DEBRA study collects bi-monthly data by means of computer-assisted, face-to-face household interviews in a new sample of about 2000 people aged 14+ each survey wave. People are interviewed about their use of tobacco and of alternative nicotine delivery systems, as well as on sociodemographic characteristics. Additional questions on self-reported health behaviour changes in tobacco smoking, alcohol consumption, and physical activity during the nationwide COVID-19 restrictions in spring 2020 (March 22 nd to June 5 th ) were added to two waves of the DEBRA study: wave 24 (June/July 2020) and wave 25 (July/August 2020). For the present analysis, data on both waves were aggregated resulting in a total sample of 4078 respondents (N wave 24: 2038 respondents, N wave 25: 2040 respondents).
Up to January 2020, respondents of the DEBRA study were selected through multistage, multi-strati ed random probability sampling. In January 2020, the market research institute Kantar, which conducts the survey, switched to a dual frame design of data collection: random strati ed sampling (50% of the sample) and quota sampling (50% of the sample). The COVID-19 pandemic substantially affected the possibility and willingness of the general public to participate in face-to-face household surveys, leading to lower response rates during the pandemic than before. As consequence, the proportion of the quota sampling had to be increased up to 100% during the study waves 24 and 25 to further balance expected non-response effects. Details on these changes in the sampling design of DEBRA have been published elsewhere: https://osf.io/s2wxc/.
The study was approved by the ethics committee of the Heinrich-Heine-University Duesseldorf (HHU 5386R) and was registered at the German Clinical Trials Register (DRKS00011322 and DRKS00017157).

Measuring sociodemographic characteristics
We collected self-reported data on age, sex, education, and on monthly net household income of respondents. The level of education was categorised into three groups: low (no quali cation or junior high school equivalent ("Hauptschulabschluss")), middle (secondary school equivalent ("Realschulabschluss")), and high level of education (advanced technical college equivalent ("Fachhhochschulreife") or high school equivalent ("Allgemeine Hochschulreife")). We used an equalisation technique provided by the Organisation for Economic Co-operation and Development (OECDmodi ed equivalence scale) to adjust the net household income for household size and composition. In order to have the distribution of income in the German population re ected in our data in the best possible way, we assigned the monthly net income to the following categories: low=approximately <20 th income percentile, middle=approximately 20 th to 80 th income percentiles, and high= approximately >80 th income percentile (details: https://osf.io/387fg/).

Self-reported changes in health behaviour
Behaviour changes in the German population during the time of the rst COVID-19 restrictions in spring 2020 were collected for three health behaviours of interest by using the following answer categories: Tobacco smoking: "During the time of corona restrictions, I … 1. smoked much more than before 2. smoked somewhat more than before 3. smoked the same amount than before 4. smoked somewhat less than before 5. smoked much less than before . don't smoke tobacco at all 7. no answer" Alcohol consumption: "During the time of corona restrictions, I … 1. drank much more alcohol than before 2. drank somewhat more alcohol than before 3. drank the same amount of alcohol than before 4. drank somewhat less alcohol than before 5. drank much less alcohol than before . I don't drink alcohol at all 7. no answer" Physical activity: "During the time of corona restrictions, I was 1. much more active than before 2. somewhat more active than before 3. just as active as before 4. slightly less active than before 5. much less active than before . I can't move around 7. no answer" Answers were classi ed into "increase" (answer options 1 and 2), "no change" (answer option 3) and "decrease" (answer options 4 and 5). People who reported that they do not smoke, drink or cannot move around (answer option 6), and those who did not provide an answer (answer option 7) were excluded for the analyses.

Statistical analyses
A protocol of the present analyses, including a pre-speci ed analysis plan, was published prior to the statistical analyses (https://osf.io/emaq2/). Before we aggregated the data from the two waves, we carried out a preparatory analysis to test if the data from both survey waves differ signi cantly regarding reported health behaviour changes (see Supplemental Table I).
Prevalence data on self-reported health behaviour changes (increase, no change and decrease) were analysed using descriptive statistics and presented as percentages together with 95% con dence intervals (CI). Data were weighted (reported as: "w" ) to be representative of the German population accounting for personal and household characteristics. Details on the weighting technique are described in the study protocol [40].
Three separate multinomial logistic regression analyses (three categories: increase, no change, decrease) were used to analyse associations between each health behaviour and the following socioeconomic and sociodemographic characteristics: age, sex, monthly net household income, and education level of the respondents. All regression analyses used unweighted data.
Age in years was used as a continuous variable for the regression analyses, and as categorical variable (14-17, 18-24, 25-39, 40-64, 65+ years) for descriptive statistics. Monthly net household income was used as a continuous variable in € among over 18-year-olds (€0-€7000 or more) for the regression analyses, and as categorical variable (low, middle, high income) for descriptive statistics. Education was analysed as a categorical variable (low, middle, high level of education) for all analyses. Data were analysed using IBM SPSS version, 25.
Regarding the total sample (N=4078), 69.6% (n=2838) of the respondents reported not to smoke tobacco, 25.4% (n=1035), not to drink alcohol, and 1.4% (n=56) reported not to move around at all. Of the remaining samples for each health behaviour, missing data on health behaviour change was ≤1% (smoking: 1.0% (n=39); alcohol consumption: 0.3% (n=12); physical activity: 0.2% (n=7)). These persons were excluded from the regression analyses. Missing values of predictor and outcome variables were sparse (because this was a face-to-face survey) and assumed to be "missing at random".

Results
The preparatory analysis showed that survey waves 24 and 25 did not differ signi cantly with regard to reported changes in smoking and drinking behaviour. However, a small difference was observed in selfreported changes in physical activity (Chi square test (4)=13.173, p=.010, n=4015; Cramers V=0.57).
Sociodemographic and socioeconomic characteristics of the total sample of 4078 respondents are presented in Table 1 (unweighted data). The mean age of the sample was 49.3 years (SD (standard deviation) = ±18.6 years) and 51.1% (n=2084) of the respondents were female. Regarding changes in alcohol consumption 12.9% w (95%CI=11.7-14.1; n w =384) of the respondents reported an increased consumption, 67.3% w (95%CI=65.6-69.0; n w =2006) no change, and around one out of ve persons (19.9% w , 95%CI=18.4-21.3; n w =592) reported to drink less alcohol than during the time before the rst strict COVID-19 restriction in spring 2020.
The results of the three multinomial ordinal regression analyses are presented in Table 2.

Tobacco smoking
We found that increasing age was associated with reduced odds of increasing or decreasing tobacco consumption (increase: OR=0.98, 95%CI=0.97-0.99; decrease: OR=0.99, 95%CI=0.98-1.00). Respondents with lower educational level also had lower odds of reporting a decrease in tobacco smoking (middle educational level: OR=0.61, 95%CI=0.40-0.95; low educational level: OR=0.49, 95%CI=0.28-0.85) compared to people with a high level of education. As income rises, the odds of reporting a change in tobacco consumption rises in the direction of increase (OR=1.30, 95%CI=1.08-1.58).

Alcohol consumption:
Similar associations were found when comparing increase, respectively decrease, of alcohol consumption. The associations were more pronounced here, especially between age and low educational level and a decrease in alcohol consumption. As age rises, the odds of a change in alcohol consumption also rises in the direction of increase (OR=0.98, 95%CI=0.97-0.98) or decrease (OR=0.99, 95%CI=0.98-0.99) compared to no change.
Physical activity: The associations regarding physical activity showed a quite similar pattern. Men have lower odds than women to change their behaviour regarding physical activity towards decrease (OR=0.80, 95%CI=0.70-0.93). Again, the analyses show that people with lower educational level also had lower odds of changing their physical activity during the strict restrictions, and this applies to both decrease (middle educational level: OR=0.83, 95%CI=0.70-0.99; low educational level: OR=0.71, 95%CI=0.58-0.87) and increase (middle educational level: OR=0.63, 95%CI=0.51-0.77; low educational level: OR=0.40, 95%CI=0.30-0.52) compared to people with a high level of education.

Discussion
Overall, the majority of the German population (around 52%-67%) did not report changes in their health behaviour with regard to smoking, alcohol consumption, and physical activity during the rst COVID-19 restrictions in spring 2020 compared to the time immediately before. Nevertheless, signi cantly more respondents reported having smoked more than having smoked less. A comparable negative effect was seen concerning physical activity: more people reported having exercised less than more. About alcohol consumption, the result is different: here, slightly more respondents said they had reduced their alcohol consumption instead of increasing it. It seems that theses behavioural changes are associated with socioeconomic and sociodemographic characteristics such as education, age, gender and income of the respondents.
Our results are comparable with the current gures of the Federal Statistical O ce: in the "corona year" 2020, the per capita consumption in Germany for various alcoholic beverages decreased (-5.4% for beer, -2.1% for sparkling wine and − 0.9% for spirits). As possible reasons, the experts cite a lack of drinking opportunities due to closed gastronomy and the absence of many festivities [41]. On the other hand, a total of €28.8 billion worth of tobacco products were taxed, 5.0% more than in the previous year (2019). Differentiated by the various tobacco products (cigarettes, cigars, pipe tobacco, ne cut), only for conventional cigarettes a small decrease of -1.1% was observed. The strongest increase was observed in pipe (+ 44.3%) and ne cut tobacco (+ 10.6%). The strong increase regarding ne cut is assumed to be due to the lack of availability of alternative low-priced cigarettes from other countries and the fact that people roll their own cigarettes [42].
We found associations between changes in all analysed health behaviours (smoking, alcohol drinking, and physical activity) and speci c person characteristics, such as higher age, lower education level, and higher net household income. Comparable associations have also been found by other international research groups [28,29,31,43]. As previous studies suggested, aspects such as stress, boredom and anxiety seem to negatively in uence health behaviour change during country-speci c COVID-19 restrictions [27,28,30,31,43].
Current studies report heterogeneous results about the change in smoking, drinking behaviour, and physical activity [27][28][29][30][31][32][33][34][43][44][45]. Many of these studies were conducted as online surveys, and data were reported as being not representative for the general population of the respective countries. The objective was often to get an initial overview on trends in pandemic-related health behaviour changes. Besides, the duration of the restrictions, as well as the severity of the measures (e.g., curfews) in the speci c countries, could in uence the change of health behaviour.
Regarding tobacco consumption, our data showed that signi cantly more people increased (24%) than decreased (12%) their smoking behaviour during strict restrictions. Whereas in our study nearly 36% of the respondents changed their smoking behaviour, only 10.3% (increase 6.9%, decrease 3.4%) of the respondents in an Australian study (online survey, N = 1491) reported this [30]. Similar was reported by a group from Belgium: increase 7.4%, decrease 2.5% [28]. A different pattern was reported from Pakistan: of around 6000 interviewed smokers, 86% changed their smoking behaviour (increase 18%, decrease 68%) [33]. In contrast, a high increase (43%) in tobacco smoking was reported in a different study from Germany (N = 558) and was particularly observed among people with lower educational level or pandemic-related changes in the type of employment (leave of absence/home o ce) [39].
Positive health effects regarding alcohol consumption were reported during the COVID-19 restrictions: more people reported drinking less alcohol than more during this time. Findings from an Australian online survey are in line with our results (increase 26% (Australia) vs.13% (our study), decrease 18% vs. 20%) [30]. A large study from 21 European countries (N = 40064) also showed a decrease in alcohol consumption during the period of restrictions (April 24th -July 22nd 2020). In Germany, however, this decrease was less pronounced than in other European countries. The authors see possible reasons for this in the increase in alcohol consumption among women and people with risky consumption patterns [45]. However, other studies reported an ampli ed increase in alcohol consumption during the countryspeci c restrictions [28, 31,39,46]: an international online survey carried out from May to June 2020 (languages: Danish, Dutch, English, Finnish, French, German, Hungarian, Italian, Portuguese, and Spanish; N = 55811) revealed major changes in alcohol consumption: around 36% consumed (slightly) more alcohol, and 22% (slightly) less [46]. The motivational reasons for increased consumption include: boredom, lack of social contact and the lack of a daily structure [28], but also the increase in anxious, depressive behaviour and symptoms of stress [30].
In our study more people reported having exercised less than more during COVID-19 restrictions and can possibly be explained by a limited range of sport opportunities. Similar pattern was also found in further studies [30]. Compared with the results of our study, the proportion of less active people in Australis is lower (29.4% vs.48,9%). Studies from England [32], Belgium [29] and Italy [34] showed that the type of training (strength training, endurance training), the sport location (indoor-/outdoor-sport) the level of activity before restrictions (active vs. not active) [44], the type of sport (team sport vs. individual sport) [29], and the offer of online courses changed the behaviour of physical activity during this time. Reasons given were to have less time, to sit more and to lack the sporty character (competitive element, team sport) [29]. Maybe certain groups of people now have more time available due to the new circumstances or that online courses are increasingly offered and used [47] and can be an important aspect to maintained or even increased the level of physical activity during the restricted sport offers. A web-based survey conducted on May 2020 (12-29th ) among 5021 students of four German universities (mean age 24.4 years) stated, that 30.6% of the respondents reported an increase in vigorous physical activity and 19.3% a decrease [43]. Characteristics associated with a change in health behaviour were being female, younger age, being bored and having depression symptoms.
Our data help to identify and analyse the effects of the COVID-19 restrictions on the consumption of tobacco cigarettes, alcoholic beverages, and physical activity in the German population. Besides, our data complement the data from many other countries to generate as much knowledge as possible around the world. We were able to indicate that the restrictions in spring 2020 has consequences for the health behaviour of people living in Germany. It is positive that some people have used this time to improve their health behaviour, but for many (especially smokers) there are signi cant negative health consequences. Regarding physical activity, it would be interesting to differentiate between physical activities (strength training, team sports, other activities, etc.) to nd out whether it would be possible and useful to switch to new formats (online courses) to counteract the decrease here.
The present study has some limitations. First, the survey period (June-August 2020) is a few months away from the interested period of our study (March-May 2020, as well as the time immediately before). It is possible that the respondents do not remember the past events exactly (recall bias). Second, data were assessed by self-reports. People may answer in a socially desirable way in such surveys, for example by reporting a lower alcohol and tobacco consumption, or higher physical activity. Third, we did not collect the precise level of behaviour change. Knowledge on the increase in daily cigarettes or alcoholic beverages, for example, may allow a deeper understanding on relevant changes. Another aspect is that no causal relationship between a change in health behaviour and the pandemic-related restrictions can be proven with the cross-sectional study design.
However, our study has strengths: besides a large and representative sample, another strength is the analysis of associations of health behaviour change with socioeconomic and sociodemographic characteristics.

Conclusion
The rst national COVID-19 restrictions in 2020 in uenced the health-related behaviour of around 40% of the people living in Germany. People with a lower level of education, higher income and younger age seems to be particularly more vulnerable of changing their health behaviour negatively. During the ongoing pandemic, monitoring of health behaviour changes will remain important. Further health policies also need to be developed to counteract negative changes and support positive changes in health behaviour. Particularly affected groups of the population should be targeted.  Data are presented as percentages (absolute numbers) within column unless otherwise noted. Differences when calculating the total percentages in column can be explained by missing data. § German equivalents to education levels listed from lowest to highest: low=no qualification or junior high school equivalent ("Hauptschulabschluss"), middle=secondary school equivalent ("Realschulabschluss"), high=advanced technical college equivalent ("Fachhhochschulreife") or high school equivalent ("Allgemeine Hochschulreife"). ''Income is listed from low to high: income in three categories: low (=approximately < 20 th income percentile), middle (=approx. 20 th to 80 th income percentiles), and high (=approx.>80 th income percentile).The sample of all available DEBRA waves is roughly comparable with the income distribution in the German population.  Prevalence data on self-reported health behaviour changes during the rst COVID-19 restrictions (March to May 2020) compared to the time immediately before these restrictions were implemented (with 95% con dence interval, weighted data).