We investigated the association of three behavior change constructs (readiness, importance and confidence) with drinking and smoking behaviors. In this prospective cohort sample, it appears that changes in alcohol use are far more frequent than changes in smoking; while 29% of the subjects with baseline unhealthy alcohol use were no longer drinking unhealthy amounts at follow-up, only 13% of the baseline smokers no longer smoked at least one cigarette per day.
These results demonstrate that high confidence levels were associated with subsequent changes in drinking and smoking risk status. The magnitudes of association were similar for both behaviors, i.e. subjects who had high confidence in their ability to change were about three times more likely to no longer report the target behavior (unhealthy alcohol use, daily smoking) than were subjects with low confidence levels. Confidence appears to be a stable predictor of subsequent reductions in both alcohol and tobacco use. These results were confirmed in secondary analyses of alcohol use; subjects with high confidence levels reported fewer drinks per week and fewer binge episodes per month than did subjects with low confidence levels. Results also suggest that there is a dose–response relationship between confidence and drinking outcomes. For smoking, results of the primary outcome were not confirmed in the secondary analyses. Nevertheless, the measure of effect suggests a dose–response relationship between confidence and number of cigarettes smoked per smoking day, even if the association failed to reach statistical significance.
Our findings are consistent with other reports that point out the potential role of confidence in ability to change as a good predictor of subsequent change [20
]. They can be linked to other studies conducted in other populations (such as adults or females) that show the impact of self-efficacy on relapse and abstinence for both smoking and drinking [32
The importance of changing results were mixed. For primary outcomes, there was an association of high importance with favorable changes in drinking, but not with smoking. There were no associations found for secondary outcomes.
Readiness to change did not seem to be associated with changes in either drinking or smoking, and is seemingly at odds with current behavior change theories. Past research among adolescents has shown an association between readiness to change (measured with a ruler) and actual changes in alcohol and tobacco use [22
]. Unlike those in our sample, the Maisto and Chung subjects were recruited at substance use treatment facilities, so it is possible that the research setting influenced the results. In addition, the authors used a slightly different ruler than was used herein. It should be noted that the confidence scale in the Chung study is comparable to ours, and yielded similar results. The psychometric properties of the readiness rulers may be influenced by differences in wording, e.g. “10” in the Maisto study corresponds to “trying hard to change”, whereas “10” in ours corresponds to “ready to change”. However, other studies also failed to show an association between readiness to change and behavior change [20
]. As opposed to confidence, readiness may reflect severity of use rather than a dimension associated with the ability to enact changes [19
], but there is no definitive answer with respect to the predictive value of readiness. Heterogeneity in the measures across various studies is a likely explanation. Similarly, readiness may operate differently within different populations, e.g. treatment/non-treatment or specialized treatment milieus, etc.). In addition, most of the research conducted on readiness to change refers to the Transtheoretical Model (TTM), whose value has been questioned [27
]. However, our study was not intended to be a validation of TTM. Even though measures we used relate more or less to some of the processes described in TTM, we did not assess the experiential and behavioral components comprising TTM. We evaluated whether clinical measures were associated with subsequent substance use, so our results yield useful information for clinical and predictive utility. Since the component constructs of motivation to change (readiness, importance, and confidence) are not highly correlated and predict independently in a combined regression model, they could be measured and used separately in research and practice. A consistent finding of ours across the two substances studied is that individuals with higher levels of confidence were more likely to modify their substance use than were those reporting lower levels of confidence; this may help identify individuals who need more support in order to change their alcohol and tobacco use. Our results question the utility of relying on readiness to change as a predictor of substance use.
There are several limitations to consider in this study. First, our subjects agreed to participate in research allowing them to receive a brief motivational intervention, and thus might have been predisposed to changing. In addition, secondary analyses of randomized trial data can present methodological challenges, and sample size was determined for the randomized trial, not for the present analyses (note: at alpha level 0.05, the power to detect differences observed for the main outcomes of unhealthy alcohol use and smoking was 100%). However, unlike secondary analyses in other cohort designs, the intervention is well-specified and its recipients are in an identified group where all of the analyses can be controlled for intervention delivery. The fact that all analyses were adjusted for the receipt of brief intervention and they showed only a small, non-significant intervention effect [39
], makes it unlikely that our results are the consequence of receiving a brief intervention. Furthermore, efforts were made in this randomized trial within this particular population to ensure an acceptable follow-up rate of 80%. Nevertheless, the sample was small, and the study needs replication using larger samples. An additional limitation is the use of self-reported measures of alcohol and tobacco use. There is a long-standing debate about the reliability of self-reports of substance use [40
]. Despite its limitations, self-report remains an affordable, acceptable and often used method of assessing substance use. Self-report may be subject to social desirability and recall biases (especially for substances like alcohol that may affect memory); consequently, substance use may be under-reported, especially in situations involving social pressure or disapproval [44
]. In the present study, efforts were made to assure confidentiality, and a non-judgmental approach to discussing substance use was stressed in order to limit the risk of bias, but we cannot rule out the potential for inaccurate reporting among participants.
Since our study included only young men, our results should not be seen as generalizable to women or older adults, whose drinking and smoking habits may differ and for whom readiness, importance and confidence may play a different role. Generalizability is limited to young men agreeing to discuss their substance use, since subjects who agree to participate in a study where they have to talk about their substance use are likely to differ on motivation and substance use from those who do not. Even though recruited at a site where virtually all Swiss young men could be contacted, those willingly participating differed slightly from the total population. As reported for the randomized controlled trial, 22.1% of those available for participation were interested in receiving a brief motivational intervention. Since 98.6% of non-participants in that trial completed a short screening questionnaire, we were able to compare them to participants, who had a higher prevalence of smoking (54.4% vs. 47.7%). The prevalence of drinking >21 drinks per week was similar (9.6%) between the two groups, but the prevalence of binge drinking was higher (55.3 vs. 49.8%) among participants compared to non-participants [39
Our study also has several noteworthy strengths. We used a non-clinical sample of young men at the Lausanne army recruitment center that processes all French-speaking Swiss males in order to assess eligibility for military service. This procedure is mandatory in Switzerland and is a unique opportunity to contact a population-based sample. Typically, individuals in this setting do not seek treatment and seldom access primary care services. Other population-based studies evaluating behavior change constructs are relatively scarce.