OUP user menu

Self-Assessment of Drinking on the Internet—3-, 6- and 12-Month Follow-Ups

Anja Koski-Jännes, John Cunningham, Kari Tolonen
DOI: http://dx.doi.org/10.1093/alcalc/agn124 301-305 First published online: 16 January 2009


Aim: The aim of this work was to report on the results of a pilot study of a web-based self-assessment service (DHT) for Finnish drinkers (www.paihdelinkki.fi/testaa/juomatapatesti). Method: During the 7-month recruitment period in 2004 altogether 22,536 anonymous self-assessments were recorded in the database of this service. The study sample was recruited from the 1598 service users who also participated to a survey evaluating the DHT. Those who consented by providing required baseline data and their e-mail address (n = 343) were sent a message asking them to fill in the follow-up questions 3, 6 and 12 months later. Their self-reported use of alcohol and drinking-related problems served as the main outcome variables in this single-group follow-up study. Results: At 3, 6 and 12 months, 78%, 69% and 61% of the study participants, respectively, responded to the follow-up. The intention-to-treat (ITT) results revealed significant reductions (P < 0.001) in all the outcome measures. The reductions occurred during the first 3 months, after which the changes were non-significant. Conclusions: The results are in line with previous studies with mostly shorter follow-up periods suggesting that Internet-based self-assessment services can be useful tools in reducing excessive drinking. A randomized controlled trial would, however, increase our certainty about the causes of the observed changes.


During the last decade, the use of Internet has dramatically increased all over the world. Seeking and providing information about health-related issues is one common use of the Internet. Increasing numbers of web-based services have accordingly been developed also for substance abusers (e.g. Cunningham et al., 2000; Bewick et al., 2008; Kypri et al., 2008). This pilot study is a single group trial on the effects of an Internet-based self-assessment service for Finnish drinkers (see www.paihdelinkki.fi/testaa/juomatapatesti). Survey results on this service as well as 3-month per protocol follow-up data have previously been reported (Koski-Jännes et al., 2007). The study at hand reports the intention-to-treat (ITT) results of this intervention over the whole follow-up year.

When the data for this study were collected, 73% of Finns between the ages 16 and 74 years were using the Internet (Nordic Information Society Statistics, 2005, 28). Alcohol issues were a hot topic in the media due to the EU driven price reductions on alcohol in April 2004. As the consumption of alcohol was expected to grow, new means to facilitate self-regulation of drinking were clearly needed. Using Internet for this purpose offered an exciting new channel for this sort of work.

The self-assessment service with the name of Drinking Habit Test (DHT) was launched in 2003. It was modified from a previously developed Canadian service (Cunningham et al., 2000). DHT comprises a brief assessment and provides personalized feedback on one's drinking. It aims to raise consciousness of the use of alcohol and to reduce excess particularly among hazardous and harmful drinkers. Theoretically, it is based on the role of feedback and normative comparisons as tools of self-regulation (Shibutani, 1955; Carver and Scheier, 1998). According to West (2006), excessive behaviors often arise out of one-sided feedback on certain fast rewarding activities like drinking or using drugs. The failure of balancing input leads the person to give an unhealthy priority to these activities. A good way to provide balancing feedback without threatening the ‘face’ of a person (Goffman, 1967) is to do it anonymously through the Internet.

Previous research

Experiencing drinking-related problems raises interest in self-help materials and particularly in online tools for heavy drinkers (Koski-Jännes and Cunningham, 2001). In line with this observation, several pilot studies have been done on Internet sites providing self-help materials for people with unhealthy levels of drinking (e.g. Cunningham et al., 2000; Westrup et al., 2003; Linke et al., 2004; Saitz et al., 2004).

Some randomized studies have also been conducted with university students (e.g. Kypri et al., 2004; Chiauzzi et al., 2005; Walters et al., 2005; Kypri et al., 2008). Their results provide preliminary support for the efficacy of web-based assessment and feedback even up to 12 months. Cunningham et al. (2005) conducted a randomized study with general population participants using essentially the same web-based service as the study at hand. After receiving Internet-based personalized feedback on their drinking, people could sign up for the follow-up through their home computers. The volunteers (n = 58) were assigned either to no further treatment or to also receive a self-help book by regular mail. The combined approach produced better results. Drinking was reduced also under the Internet condition alone, but probably due to the small sample size the effect remained non-significant.

The Finnish DHT is also targeted to the general population. The first 3-month results of this pilot study displayed highly significant reductions (P < 0.001) in all the drinking-related per protocol outcome measures. The service appealed more to women than men, but there were no sex differences in drinking-related outcomes (Koski-Jännes et al., 2007). The dropout rate was only 22%, but it was expected to grow over time. To avoid possible bias by selective subject attrition, ITT results of the 3-, 6- and 12-month follow-up of the same subjects will be reported in this single group trial.



The study participants were recruited from the anonymous users of the DHT who responded to the voluntary survey on this service from March to October 2004 (see Fig. 1). During this period, 22,536 self-assessments were completed all of which were given an id-number and automatically recorded in a database set up by the service provider. Seven percent of them also agreed to evaluate the DHT by answering some survey questions. The subjects of this study were recruited from among these 1598 respondents by asking them in the end of the survey if they wanted to participate in the follow-up. One-fifth of them consented by providing their e-mail address (Koski-Jännes et al., 2007). Out of these 351 people, we excluded three persons who responded for somebody else and five persons due to incomplete data.

To compare the baseline data of the study subjects (n = 343) with the service users in general, we also drew a random sample of respondents (n = 538) from the general database of the DHT during the same data-collection period. Compared to this random sample, the study subjects were significantly more often females (61% versus 49%, P < 0.001), slightly older (32.6 versus 30.4, P = 0.007) with higher AUDIT scores (17.0 versus 14.1, P < 0.001) and had more negative consequences of drinking (2.7 versus 1.9, P < 0.001) (Koski-Jännes et al., 2007).


The DHT includes 10 AUDIT questions (Babor et al., 1989; Saunders et al., 1993), six psychosocial consequence questions (see e.g. Canada's Alcohol and Other Drugs Survey, 1994) as well as questions on the number of drinks consumed on each day of a typical week. Those with an irregular drinking pattern could fill in the past 30 days’ drinking diary. Some demographic data were also asked. The Finnish ‘drink’ contains 11.5 g of ethanol. When filling in the AUDIT, the respondents were informed what this equals to in terms of common beverage types consumed in Finland.

After entering these data, the respondents receive personalized feedback, which allows them to compare their drinking with the average of Finnish men and women (Metso et al., 2002). The feedback also summarizes the quantity, cost and caloric intake caused by their drinking as well as the time spent under the influence of alcohol over the last year. Feedback on their AUDIT scores is also given as well as information on the relationship of drinks to blood alcohol concentration and the likelihood of accidents at different levels of BAC.

Follow-up data collection

At 3, 6 and 12 months after the initial assessment, the subjects received an e-mail asking them to click a link to a website containing follow-up questions on their drinking, treatment or self-help group utilization and the perception of change in their drinking during the past 3 months. They were also asked if they had recommended this site to others and if the 2004 alcohol price reductions in Finland had affected their drinking. Non-responders were sent reminders 2 and 4 weeks later. No other incentive was offered to increase the follow-up rate.

Main outcome variables and data analysis

As all but the first three AUDIT questions have a 12-month reference, the full AUDIT is not well suited for shorter follow-up periods. The sum of scores for first three items without time reference (AUDIT-C) was, therefore, also reported as well as the number of drinks in a typical week, the number of drinks in the last occasion and the number of negative drinking consequences.

Due to markedly skewed distributions, the main outcome variables could not be consistently normalized. The analysis was, therefore, conducted both by non-parametric (Friedman test and Wilcoxon signed rank test) and parametric methods (repeated measures ANOVA) with basically similar results. Categorical data were analyzed with χ2-tests. ITT results are here presented on the main outcome variables. Those who dropped out after the initial assessment were regarded as unchanged, and their missing follow-up data were replaced by their corresponding baseline values.


Study participation and results on main outcome variables

At 3, 6 and 12 months, 78%, 69% and 61% of the study participants, respectively, responded to the follow-up. The only significant baseline difference between responders and non-responders appeared at 6 months: the non-responders had initially experienced more drinking problems than the responders (3.11 versus 2.57, Mann–Whitney U = 10,787.5, z = −2.187, P = 0.03). All the other baseline differences remained non-significant at the three follow-up points.

The results on the main outcome variables (see Table 1) show that the reduction in drinking and its harmful consequences mainly occurred during the first 3 months. The 6-month results are essentially similar to those at 3 months. Drinking appears as slightly increased by 12 months, but even here all the outcome measures remain significantly lower than those at the baseline (P < 0.001). All the changes after the first 3 months were non-significant.

View this table:
Table 1

ITT-results on drinking at the baseline and at the 3-, 6 and 12-month follow-up (n = 343)

Mean (SD) AUDIT17.0 (8.7)14.7 (8.4)14.9 (8.8)15.0 (9.0)<0.001a
Mean (SD) AUDIT-C7.6 (2.5)7.0 (2.6)7.0 (2.6)6.9 (2.7)<0.001a
Mean (SD) no. of drinks in a typical week19.0 (18.5)14.9 (15.6)14.9 (16.7)15.9 (17.8)<0.001a
Mean (SD) no. of drinks in the last occasion8.5 (6.2)7.5 (6.6)7.5 (6.3)7.3 (5.7)<0.001a
Mean (SD) no. of drinking consequences2.7 (2.1)2.3 (2.1)2.2 (2.2)2.4 (2.1)<0.001a
  • aFriedman test,

  • bWilcoxon signed rank test.

Drinking by gender and subjective experience of problem drinking

Figure 2 displays the number of drinks consumed in a typical week by gender at the four measuring points. The general tendency among both men and women displayed a remarkable drop in consumption during the first 3 months. The situation remains stable at 6 months but by 12 months drinking is slightly increased, more so among men than women.

Fig. 2

Mean, SEM and confidence intervals of weekly drinks by gender. ITT results of the 343 study subjects.

The best predictors of reduced weekly drinking at the 3-month follow-up were the AUDIT scores and the subjective assessment of ones drinking as problematic. The effect of these factors on weekly drinking were, therefore, checked also at the later follow-up points. Figure 3 displays the ITT results on weekly drinking in three equal sized AUDIT score groups (0–11/12–20/21–40). It shows that the subjects with >20 AUDIT scores reduced initially the most, but they tended to increase their drinking from the 3- and 6-month follow-up to the 12-month follow-up, while those with lower scores maintained the change for the whole year. The single question ‘Has your drinking been problematic during the past year?’ (no/yes, partly/yes, clearly) produced a very similar picture.

Fig. 3

Mean, SEM and confidence intervals of drinks per week by AUDIT scores. ITT results of the 343 study subjects.

Additional questions

The subjects were also asked in the follow-up about their perception of change in drinking during the previous 3 months. At 3, 6 and 12 months, 36%, 39% and 44% of respondents, respectively, claimed having reduced, whereas 10%, 11% and 11% said that they had increased their drinking. At 12 months, men and women differed significantly in this respect (χ2 = 6.59, df = 2, P = 0.037). Fifty percent of women versus 33% of men claimed drinking less now, whereas 8% of women versus 17% of men claimed having increased their drinking during the past 3 months.

Women were more eager than men to recommend the test to others. At 3 months, 34% of women versus 18% of men had done it (χ2 = 7.57, df = 1, P = 0.006). The respective proportions were 36% versus 17% (χ2 = 10.99, df = 1, P = 0.001) at 6 months and 36% versus 22% (χ2 = 4.56, df = 1, P = 0.033) at 12 months.

Only 10 persons (4%) reported having participated in treatment or self-help groups during the follow-up year. The use of this self-assessment site thus did not seem to increase interest in face-to-face treatment services.

The subjects were also asked whether they felt that the alcohol price reductions of 2004 had affected their drinking. The great majority (69.5%) claimed that it had not had an effect, 12.6% stated that it had caused them to increase their drinking and 17.8% said that they had reduced their drinking despite of it. There was no significant difference between males and females in this respect.


The ITT results of this web-based intervention confirmed the previous per protocol 3-month results and extended the follow-up to 6 and 12 months. Drinking and its negative consequences significantly decreased during the first 3 months after the initial assessment and feedback, and these gains were mostly maintained at the later follow-up points. The results are in line with those of Kypri et al. (2008) with equally long follow-up. This finding is encouraging particularly in relation to the general growth of consumption that took place in Finland after the large price reductions on alcohol in 2004 (Österberg 2005).

The sample means in this study were rather high compared to previous data from other countries indicating that most of our study participants seemed to be relatively heavy drinkers (see e.g. Cunningham et al., 2006). This implies that Internet-based self-assessment and feedback may have an effect even among some dependent drinkers. In fact, the consumption of alcohol seemed to initially decrease most in the highest AUDIT score group, but by 12 months their drinking had started to increase again suggesting that this group may need more support to maintain the change. A clinically useful observation was that the single question on the subjective experience of drinking problems (no problem/some problems/clear problems) produced a very similar picture as in Fig. 3 with three levels of AUDIT scores.

Previous web-based self-assessment studies have usually been targeted to problem drinkers with AUDIT scores >7. Sixteen percent (n = 55) of our subjects were below that criterion. We included this group because there are basically no limits to using this Internet service. Moreover, attention to one's drinking could also cause light drinkers to increase their use of alcohol, which did not seem to occur in this group. For the sake of comparison we also analyzed the data by excluding these subjects, but the results remained essentially similar. Both including light drinkers and using ITT results reduced our chances of finding large changes in the outcome data. Despite these cautions, the observed reductions in drinking remained significant.

A pleasing feature of this study was the unusually high response rate with a general population sample—almost two-thirds responded even 1 year later. In most Internet studies, response rates have been much lower even with shorter follow-ups (Eysenbach, 2005). Other merits are the long follow-up and the naturalistic setting.

A clear limitation of this study is the lack of a randomized control group. Therefore, we cannot be sure that the observed improvements were, in fact, caused by the intervention. Regression to the mean could explain some of the observed reduction in drinking, particularly because study participants were initially heavier drinkers than service users on average and could have reduced their drinking because of chance fluctuation or other events in their lives causing them to change even without exposure to any intervention (Cunningham, 2006.). Thus, a randomized study would increase our certainty about causal connections. When designing such a study, it should be noted that assessment per se already has some effect on drinking as noted by Kypri et al. (2007). So the only additional element that can be varied is the personalized feedback, which could be compared to providing general information on possible risks of excessive drinking.


This study was supported by a grant from the Finnish Foundation for Alcohol Studies.


View Abstract