Are the stages of change relevant for the development and implementation of a web-based tailored alcohol intervention? A cross-sectional study

This study is part of a large trial assessing the effects of adopting web-based computer tailoring on lifestyle behaviors and was approved by the Medical Ethics Committee of Maastricht University and the University Hospital Maastricht (NL27235.068.09/MEC 09-3-016).

Our present research was a cross-sectional, quantitative web-based questionnaire study, including written feedback messages. In spring 2008, the recruitment of respondents took place through advertisements and articles in different local newspapers (free sheets) in the Limburg province, which is located in the southern part of the Netherlands, and the distribution of flyers and posters in public places, such as supermarkets, universities, hospitals and airports, in both rural and urban areas. All people with a command of the Dutch language, at least 18 years old and with access to the Internet were invited to participate in this study. The purpose of the study was explained to the respondents and confidentiality was assured. Respondents did not receive any incentives for participation.

In total, 230 people visited our website. Of the 212 individuals (92.2%) who gave informed consent and started to fill out the questionnaire, 170 respondents (80.2%) completed the main part containing items about attitude, social influence, self-efficacy and stages of change, and were included in the analyses. The program was evaluated by 146 respondents (85.9%).

To deliver personalized advice, three inter-related elements were necessary: 1. a screening instrument (questionnaire) to measure alcohol consumption and influencing determinants; 2. a message source file containing all tailoring messages; and 3. a computer program that offered the opportunity to analyze the screening results and to select the correct messages from the message file [21, 46]. A web-based instrument was used to create and conduct an online questionnaire and to design and implement the tailored advice. By means of tailored formulas responses to questions yielded the tailored advice that respondents received on their screen and which could also be printed as a pdf file.

A questionnaire, based on earlier computer tailored studies [27], was used to measure demographics, alcohol consumption, attitude, social influence, self-efficacy and stages of change as well as to evaluate the computer-tailored program, particularly the tailored feedback messages.

The structure of the program was based on other computer-tailored programs which have already been proven to be effective in changing lifestyle behaviors, such as in increasing smoking cessation [30] and in increasing amounts of physical activity [48]. The I-Change model [49] was used as the theoretical framework for the tailored advice. The message file was used to generate tailored messages about adhering to the guideline for adequate alcohol intake, alcohol consumption, pros and cons of alcohol consumption, social influence and self-efficacy. Text-based messages were given in combination with suitable pictures. For example, we included tips and suggestions about how to decrease one's alcohol intake. In case respondents reported to drink in a healthy manner, maintenance-feedback was given which means that people were stimulated to also drink in a healthy manner in the future. With respect to social influence, feedback was given on how to deal with the presence of drinking others, like friends or family. Furthermore, we gave special personalized advice about social, emotional and routine risk situations. Graphic feedback was also used: A diagram was made which showed the relationship between the advantages and disadvantages of alcohol intake as experienced by the participants.

The data were analyzed with the statistical program SPSS, version 15.0. Descriptive statistics were used to describe the participants’ characteristics and their alcohol intake. Different groups (i.e. based on their actual stage of change) were compared with respect to demographics and drinking behavior by use of Chi-square tests for discrete variables and F-tests for continuous variables. In the case of significant main effects, variables were included in the analyses as covariates. Next, the groups were compared with regard to psychosocial determinants of drinking behavior and their program evaluation. Standardized T-scores were calculated with a mean of 50 and a standard deviation of 10. Differences between the groups were determined with the “Least Significant Difference” (LSD) method. For all tests, the significant value was set at p < .05.

Participants were aged from 18 to 74 years, with a mean age of 44.62 years (SD = 14.52). Slightly more women (n = 96; 56.5%) than men (n = 74; 43.5%) took part in this study. Nearly all respondents were of Dutch origin (n = 155; 91.2%). Two-thirds of the participants were classified as highly educated (n = 104; 61.2%). The majority were in a relationship (n = 123; 72.4%) and one quarter were living together with their children (n = 45; 26.5%).

The mean daily alcohol intake among the whole sample was 2.69 standard drinks (SD = 3.04), with a statistically significant difference for males and females [t(107.93) = 4.98; p < .001]. Men had a mean daily alcohol intake of 4.00 glasses (SD = 3.60), whereas women had a mean daily alcohol intake of 1.68 glasses (SD = 2.02). Sixty-six respondents (38.8%) failed to meet the alcohol guideline.

When the 170 respondents were grouped according to motivation to change, 34.1% did not intend to change to healthier drinking habits within the next six months (PC: n = 58), 22.9% intended to drink in a healthy manner at some time in the near future (C: n = 15; P: n = 24), 4.1% had started to drink in a healthy manner during the past six months (A: n = 7) and 38.8% had already been drinking in a healthy manner for more than six months (M: n = 66). Because of the distribution of respondents among the different stages of change, three groups were formed. Precontemplators remained as one group (PC), contemplators and preparers were combined into one group (CP) and defined as those planning to reduce their alcohol intake in the near future and those in the action and maintenance stages were grouped together and defined as those currently adhering to the guideline (AM).

The distribution of respondents over the three groups differed significantly for gender, age and drinking behavior (Table1). Contemplators/preparers drank more alcohol than precontemplators. Females were more likely to be in the action or maintenance stage than males. Contemplators/preparers were significantly older in comparison with the other two groups. Owing to these differences, the analyses were corrected for gender and age.

To compare the three groups in terms of cognitive determinants, Figure2 represents the standardized T-score patterns on pros and cons, social modeling and support and social, emotional and routine self-efficacy. Table2 shows the standardized T-scores and the LSD contrasts in order to identify potential differences between the stages of change.

With regard to attitude, respondents in CP scored significantly higher on the pros than people in AM. The scores for cons did not differ significantly between the three groups. People in AM scored significantly higher on social modeling than those in PC and CP. Respondents in PC reported perceiving less social support than those in CP and AM. Respondents in PC and CP showed lower levels of self-efficacy in social, emotional and routine situations compared with those in AM whereas PC showed higher levels than CP.

Table3 represents the standardized T-scores and the LSD contrasts used to identify potential differences in the program evaluation data between the three groups. Respondents in CP reported more often than people in PC and AM that the advice would help to stimulate them to change their drinking behavior. Moreover, respondents in AM scored even higher than those in PC on the item “The advice stimulates me to improve my drinking behavior.” People in AM, on the other hand, stated that they found the advice more informative than people in PC. People in CP and AM agreed more with the advice than people in PC.

As a first goal, we investigated to what extent the Ø–pattern, as identified for smoking cessation, is applicable to the stage of change typology regarding intention to reduce the amount of alcohol intake.

Regarding attitude, we found as hypothesized an increase of cons in terms of alcohol between people in PC and CP, and a decrease of pros between people in CP and AM. The non-significant differences in the cons could be ascribed to a power problem, i.e. owing to a relatively small sample, significant differences may have failed to appear. The crossover of attitudes is partly in line with findings by previous studies [50, 51]. Our crossover, however, seemed to occur later in the behavioral change process: between the CP stage and the AM stage. Additionally, our pros did not clearly outweigh the cons in the early stages, nor did our cons outweigh the pros in the later stages of change. In general, people in AM scored lower on pros and cons than those in PC and CP, indicating that attitude may be less important in the last stages of change. Lower scores on these attitude beliefs items are in line with the Ø–pattern.

Concerning social influence, the findings are mostly consistent with the Ø–pattern, but, owing to the division in social support and modeling, a more detailed picture can be given. Concerning social support, a relatively large difference between people in PC and CP was found, indicating that social support seems to become more prevalent in CP. The analyses confirmed that people in the AM stage had more healthy drinking role models than people in the two other groups. Thus, in order to achieve transition from one stage to another, it is recommended to increase perceptions of social support and healthy role models among those in the earlier motivational stages.

Respondents in the AM stage had higher self-efficacy expectations regarding healthy drinking in difficult situations than respondents in the PC and CP stages. This finding applies to all three kinds of self-efficacy expectations. It can be assumed that self-efficacy is strongly influenced by performance of a desired behavior [52] and people in the AM stage are by definition performing healthy drinking behavior successfully. Furthermore, people in the CP stage appeared to perceive lower social, emotional and routine self-efficacy than people in the PC stage. People in PC may make unrealistic and optimistic self-efficacy estimations. This finding is in line with a previous study [53], which found that unmotivated people often overestimate their ability to behave in a healhy manner. Another explanation for this finding may be the amount of alcohol reported. Contemplators/preparers reported drinking more than precontemplators. It can be imagined that people who drink less feel more able to abstain in certain situations.

Regarding our second goal – a comparison of the three groups concerning their evaluation of the program – we found differences between motivational groups regarding the evaluation of the computer-tailored advice. Although all groups reported a favorable evaluation of the program, the results also suggest that respondents who were motivated to change their alcohol consumption (CP) reported more often than those who were not motivated or those in action/maintenance that the tailored advice contributed by helping and stimulating them to change their drinking behavior. These findings imply that further research is needed to analyze how the tailored advice might be adapted to increase its attractiveness and stimulating effects for people less motivated to change. Unmotivated people (PC) agreed less with the advice and perceived the advice as less informative, which may suggest that respondents in PC may need additional information targeting, for instance, awareness of their unhealthy behavior and their personal risks. The motivated group (CP) scored, for example, somewhat higher on the two items “The advice was relevant for me” and “The information in my advice was interesting,” which may indicate that these people perceived the advice as more relevant and more interesting than did individuals in precontemplation. In sum, a positive finding is that mostly no differences between the program evaluation scores between the three groups were found, suggesting that the program also addressed most items satisfactorily for unmotivated respondents.

There are several reasons why the current study is important. First, previous research failed to examine differences in cognitions data across distinct stages of reducing alcohol consumption (“the Ø–pattern”). Second, this is the first study to investigate the relationship between the stages of change and the appreciation of a (computer-tailored) intervention (program evaluation). Third, in comparison with studies about the stages of change model and other health-damaging behaviors, such as smoking [52, 54] and lack of physical activity [34], this study included the distinction of social influence and self-efficacy expectations according to the factor analysis when testing the Ø–pattern. The results of the factor analysis indicated that attitude could be subdivided into pros and cons, that social influence could be subdivided into support and modeling and that self-efficacy could be subdivided into social, emotional and routine self-efficacy, all consisting of reliable scales. These findings are consistent with assumptions of the I-Change model [49].

Our findings should be regarded with a certain amount of caution owing to a number of shortcomings of the study. First, the design was cross-sectional, precluding causal inferences. Longitudinal studies could give additional information about the causality between the stages of change and psychosocial determinants as well as evaluation data. Second, the sample group was small, which could have led to lack of statistical power to show significant differences, and as a result of the small sample we had to combine the different groups, based on their stage of change, into three groups. Third, the group which completed the questionnaire differed from the partial completers. Partial completers scored significantly higher on the baseline assessment of alcohol consumption (selection bias). Fourth, the Dutch alcohol consumption guideline changed in 2009. Females are now recommended to drink no more than one alcoholic drink per day and males no more than two alcoholic drinks per day [55]. In future studies, intention to comply with the new guideline should be measured. Additionally, it is recommended to have at least two alcohol-free days a week in order to prevent health problems, such as liver diseases and addiction [56]. This additional recommendation was not yet included in our feedback, but is recommended for future programs. Finally, only self-reports could be used, which may have resulted in recall difficulties, recall bias, or both. Respondents could have either overestimated or underreported their amount of alcohol intake, feelings or expectations. This can result in misclassifications [57]. Our results revealed that this was indeed the case for a small group, but reanalysis on the basis of actual consumption showed that the overall results did not change. Moreover, the possibility of participants giving socially desirable answers may have been induced by reliance on self-reports.