A cross-sectional, nationally representative sample of 6,000 adults aged 18–88 years was randomly drawn from a web-panel of 30,000 participants (i.e., representative of age, sex, region of residence and size of municipality). Of these, 1,208 responded before recruitment was terminated. Response Analyse, a company specializing in survey research, performed the survey and the comparisons with national data on the time of the survey (December 2018). The gender balance of the responders was identical to the national population (49.7% women). There were fewer younger respondents than expected (18–24 years, 11.5% in the survey vs. 14.7% nationally) and slightly more responders in the oldest age group (65+ years, 20.8% in the survey vs. 19.4% nationally). For other age groups, the differences were less than 1%.

This study is part of an international project exploring attitudes towards and experiences with alcohol conversations in routine health care in Europe [13–15]. We aimed to recruit a minimum of 1000 respondents, based on population size (5.3 million inhabitants in 2018), and margins of error. Closing the survey was a manual procedure, and 1208 individuals had responded when the survey was closed after 2 weeks. Data was collected from the 7th to the December 20th, 2018 by means of a web-based questionnaire, accessed by the participants via a unique link provided in an e-mail invitation. No reminders were sent. Participants had previously consented to be part of the web panel of 30.000, and they consented to this specific study by actively accessing the link and answering the questionnaire.

We translated the questionnaire from English, and repeatedly tested and adjusted by comparing with the Swedish version. The questionnaire consisted of 16 questions regarding socio-demographic data, alcohol consumption, attitudes towards and experiences with alcohol conversations in routine health care. Education was reported in one of five categories (basic or secondary school; 1–3 years college/university; 4–5 years college/university; six or more years of college/university; PhD or specialist qualification from college/university). We collapsed the three highest categories into one (four or more years of college/university) in the analyses. Current occupational status was reported in one of eight categories (employed, student, unemployed, long-term sick leave, retired, parental leave or other leave, self-employed, other). Sociodemographic characteristics, health care visits and drinking status are reported in Table 1.

Alcohol consumption was measured using the three-item Alcohol Use Disorders Identification Test–Consumption (AUDIT-C), and three alcohol consumption categories were constructed: abstainers; low risk drinkers; and risky drinkers [23]. AUDIT-C consists of the first three questions in AUDIT, measuring alcohol consumption. The complete Alcohol Use Disorders Identification Test (AUDIT) consists of ten questions, assessing harm and addiction in addition to consumption, but with ten questions, it was less likely to be embedded into general history questionnaires [23]. Respondents reporting no alcohol consumption in the previous 12 months were classified as abstainers. Low risk drinking was defined as a weekly consumption of up to 14 standard drinks per week for men and up to nine drinks per week for women, and/or heavy episodic drinking (HED), on no more than a monthly basis. HED was defined as consuming five drinks or more per occasion for men or four drinks or more for women. Risky drinking was defined as consumption above these levels. One standard drink equated to 12 g of pure alcohol, as in previous research [15].

One item asked participants weather they had visited their health care centre in the past 12 months, with response options: “no”, “yes, once” or “yes, more than once”. Respondents who answered “no” did not receive further questions. The question concerning whether alcohol use was addressed in health care in the past 12 months was answered with “no”, “yes, once” or “yes, more than once”. Respondents who answered “no” did not receive further questions.

One item investigated the duration of alcohol conversations including four response options: less than 1 min, 1–4 min, 5–10 min, and more than 10 min. In Table 2, the categories 5–10 min and more than 10 min has been collapsed. Respondents with more than one conversation about alcohol were asked to report the average duration. To investigate perceived contents of the conversation, the respondents were asked to respond “yes” or “no” to each of five different items: information about health consequences; how much alcohol they usually drank; whether they wanted to cut down; advice on how to cut down; and written information about alcohol. Perceptions of the alcohol conversations were assessed by four items with a four-item Likert response scale, ranging from “do not agree” to “agree completely”. The four statements included: “it provided valuable knowledge”, “it was informative”, “it was routine”, “it felt awkward”, “it was irritating”, “it was offensive”, and “it felt judgmental”.

Perceived effects of alcohol conversations were assessed with six statements, using a similar four-item Likert scale. The following statements were used: “it had no effect at all”, “it made me think about my drinking”, “it gave me a better understanding of alcohol’s health risks”, “it led to a reduction in my drinking”, “it led to an increase in my drinking”, and “it made me think about a friend’s drinking”.

Descriptive statistics are presented as frequencies or as means and standard deviations. Differences in proportions between groups were examined with the chi-squared test or Fisher’s exact test. Logistic regressions were performed to examine associations between having had an alcohol conversation in the past 12 months and individual characteristics. The odds ratios (OR) of having had an alcohol conversation in the past 12 months were first estimated in a model adjusted for age and gender (model 1 in Table 3), then adjusted for age, gender, educational level, occupational status, marital status, drinking categories and number of healthcare visits in the past 12 months (model 2 in Table 3). Data were analysed using the statistical software IBM SPSS 25. Results were statistically significant at p ≤ 0.05 using two-tailed tests.

Eighty percent (79.4%) of the respondents had visited health care the past 12 months, and of these, one in six (16.8%) reported having had conversations about alcohol. Two or more visits in health care the past 12 months increased the likelihood of having had an alcohol conversation three-fold (OR 3.03, p < 0.001) in the multivariate model, compared to one visit. More women than men had visited health care in the past 12 months (84.9% of women compared to 72.8% of men), but a higher proportion of men reported a conversation about alcohol (18.2% of men, compared to 15.8% of women). In the multivariate logistic regression women were half as likely as men to have had an alcohol conversation (OR 0.56, p = 0.004) (see Table 3).

Results concerning health care visits and alcohol conversations by drinking categories are presented in Table 2. There were no significant differences between drinking categories with regard to health care visits, nor any significant differences between drinking categories on having had an alcohol conversation. However, risky drinkers reported considerably longer alcohol conversations than low risk drinkers and abstainers. Fifty percent of risky drinkers who had an alcohol conversation reported that it was longer than 1 min, compared to 22% of low risk drinkers and 28.6% of abstainers (p = 0.007). The difference was largest for the longest conversations.

Table 3 presents the results of logistic regressions of having had an alcohol conversation, adjusted for age and sex, and a multivariate model. The odds ratio of such a conversation shows a gradual decline with increasing age compared to the youngest age group, statistically significant for all age groups older than 40 years in the multivariate model. In this model, the odds ratio increased more than eight-fold (OR 8.62, p = 0.001) for having a conversation about alcohol for respondents on parental leave or other leave.

Table 2 shows experiences, and the perceived effects, of alcohol conversations by drinking status. Of the risky drinkers, 66.1% (n.s.) compared to low risk drinkers (75.8%) and abstainers (78.6%) reported that questions about alcohol consumption was part of the conversation, while 10.7% of risky drinkers and 2.2% of low risk drinkers (p = 0.047) reported questions regarding their willingness to reduce consumption. Furthermore, 8.9% of risky drinkers and 1.1% of low risk drinkers (p = 0.038) had received advice on how to reduce consumption, and 10.7% of risky drinkers and 1.1% of low risk drinkers (p = 0.031) had received written information about alcohol.

A considerably higher proportion of risky drinkers found the alcohol conversations awkward, compared to low risk drinkers and abstainers (14.3 vs. 2.2% and 7.1%, p = 0.019). The same pattern is evident for participants perceiving the conversations as judgmental (19.6 vs. 6.6% and 7.1%, p = 0.045). Almost a quarter of risky drinkers (23.4%) reported that the alcohol conversations made them think about their drinking, compared to around 7% for low risk drinkers and abstainers (p = 0.020). However, 17.9% of risky drinkers reported that the conversation led to an increase in their drinking, compared to 4.4% of low risk drinkers and 7.1% of abstainers (p = 0.023). Other experiences and perceived effects were not significant by drinking status.

Additional analyses were performed of having had an alcohol conversation and experiencing a positive effect (“it made me think about my drinking”, “it gave me a better understanding of alcohol’s health risks”, “it led to a reduction in my drinking”). A logistic regression applying the same definition of positive effect as in Table 2 (1 = “agree completely” or 2 = “agree to a large extent” vs. 3 = “agree to some extent” or 4 = “do not agree”) did not yield significant results. In a new logistic regression where positive effect was defined as either 1, 2 or 3 (“agree completely”, “agree to a large extent”, “agree to some extent”) was then performed. In the multivariate model, women had a lower odds ratio of having had a conversation in healthcare and having reported a positive effect, compared to men (OR 0.26, p = 0.002). Risky drinkers were more likely to report a positive effect of their alcohol conversation compared to abstainers (OR 5.49, p = 0.035). These results are not presented in a table because of the wider definition of positive effect compared to Table 2, but a supplementary table is available from the corresponding author, upon request.

This study is based on a large, representative panel in all age groups, using a validated screening instrument (AUDIT-C) to assess drinking status, and a questionnaire previously employed in comparable national surveys in Europe, enabling comparisons with international studies. These studies provide new knowledge on how alcohol conversations are perceived, important for improving quality and reach of brief alcohol interventions in health care.

The study also has important limitations. We do not know what happens in the conversations or afterwards, and we do not know who initiated the conversation. All data are self-reported, and the cross-sectional design does not allow causal inferences. While the panels are representative on demographic variables, we do not know how representative they are on drinking habits and experiences with alcohol conversations. Additionally, non-responders might differ from responders. Our sample was relatively more educated than the general population in Norway; approximately 70% of our sample were educated to university-level, which is twice the national rate [41]. This suggests that the sample is skewed towards higher educational level, and probably higher alcohol consumption, although we lack relevant population data to confirm this [26]. The high proportion having visited health care indicates that our sample it is not healthier, which supports the representativeness. However, we lack data on health status of the respondents.

We need a better understanding of how people drink and how they think about drinking, and of how risky drinkers experience alcohol conversations in health care. Alcohol should be addressed as part of the regular clinical work, and exploration of willingness to change and advice on how to cut down, should be an integral part of alcohol conversations for all risky drinkers. Although the majority of risky drinkers did not find the conversation awkward or judgmental, clinicians need to know that a substantial proportion of risky drinkers may perceive the conversations as awkward and judgmental, and strive to find ways to overcome this. Future research and quality improvement should be based upon an improved understanding of patient’s experiences and perspectives, to enable more targeted and effective interventions [42].

Our study provides important insight in the different aspects of risky drinker’s perceptions of alcohol conversations. The proportion experiencing an alcohol conversation in routine health care was low, also among risky drinkers, and risky drinkers more frequently experienced the conversations as judgmental. Older adults had a significant lower likelihood of experiencing an alcohol conversation. Addressing alcohol should be an integrated task in health care, and especially for groups more at risk for alcohol-related health problems, due to increased consumption and/or increased vulnerability for alcohol. This is particularly relevant for older adults, as physiological changes, health problems and medications increase their vulnerability for alcohol. To increase the normalisation of addressing alcohol in health care, more research is needed on the development and implementation of sensitive and acceptable interventions targeting at-risk patients. Future research should address both implementation factors and acceptability of brief alcohol interventions.