Portal de Periódicos da CAPES

THE TRANSTHEORETICAL MODEL OF BEHAVIOUR CHANGE AND THE SCIENTIFIC METHOD

2006; Wiley; Volume: 101; Issue: 6 Linguagem: Inglês

10.1111/j.1360-0443.2006.01502.x

1360-0443

Robert West,

Mental Health Research Topics

I am grateful to Prochaska for his response [1] to criticisms of the Transtheoretical Model (TTM) of behaviour change that I and others have levelled [2] and for the opportunity to clarify some important issues. It is worth reminding ourselves that for a scientific model to represent an advance it should do at least one of the following: provide a more complete or accurate description of the subject matter; provide a more coherent, complete or parsimonious explanation of observed phenomena; enable more accurate prediction; and foster an improved technology that enables things to be achieved that could not be achieved previously. My editorial argued that, after more than a decade of research, the TTM had not met any of these criteria adequately. Considering first the adequacy of the model as a description, I argued that there are no 'stages' in any meaningful sense and there is no 'cycle of change'. The simple common-sense model of behaviour change on which the TTM must improve is one which says that people experience varying degrees of desire for change; this desire shows a moderate degree of stability from occasion to occasion and, when the conditions are right, stimulates an attempt at change whose success depends on a range of contextual and personal factors including, for smoking, the level of nicotine dependence. In the case of smoking the desire to stop stems from a variety of factors, including worries about health consequences and dissatisfaction with the cost of smoking, and is undermined by enjoyment of smoking (for a related review see McCaul et al. [3]). This model applies to the whole population and not just those seeking treatment [4]. Thus Prochaska is incorrect in his assertion that the TTM model is the first to address the issues at a population level. Prochaska then defends the concept of 'stages' first by noting that many measurable entities, such as death, are discrete. However, this misses the point, because my argument was not against the existence or measurement of discrete entities—that would be nonsensical. Following on from this he notes that it is only the 'stage' concept in the TTM that is discrete. This again misses the point, because it is precisely the stage concept that is being criticized. Prochaska defends the use of 'somewhat arbitrary' criteria for turning continuous variables into categorical measures; he says that is common practice and he is right, but again this misses the point. Of course, there are cases where it is desirable or even necessary for practical reasons to classify entities using criteria applied to a continuous variable. However, the case for doing this must be made—sometimes it makes sense and sometimes it does not. The concept of 'stages' is used as though these were discrete and stable entities that classify individuals (e.g. pre-contemplators versus contemplators) in a meaningful way and form the basis for an orderly transition between them. It is that approach that I and others are criticizing. We can all agree that there are people who, at a given moment, are more or less interested in making a change in their behaviour, there are some who are in the process of trying to change and there are others who are more or less well established in their new behaviour pattern; that is simple common sense. What is at issue is whether it is useful to classify these people into stages of change as defined in the TTM. I pointed to reviews and analysis by others of the problems of doing this and also to evidence in the case of smoking that so-called 'stages' are unstable and incoherent [5]. There are now two studies finding that a substantial proportion of attempts to stop smoking are made without any pre-planning at all and that these unplanned quit attempts actually have a greater probability of success [6,7]. It seems that human motivation is much more dynamic and influenced by the immediate context than is implied by the TTM, with its labelling of individuals as 'pre-contemplators', 'contemplators', 'preparers', etc. This has important implications for practice because it implies that interventions to generate change should seek to use the moment of exposure to that intervention to generate maximum motivational tension, trigger the impulse to act and address the range of issues such as nicotine dependence in the case of smoking that frustrate or undermine those attempts. The technology of behaviour change involves finding ways of doing this that are both effective and acceptable (interested readers may wish to see West [8]) I also criticized the model for neglecting basic motivational processes that are not easily amenable to self-report. Prochaska points out that the TTM does include operant learning principles and not simply 'pros and cons' types of analysis. However, my reading of the model and its application is that it does focus greatly on communication strategies that assume that smokers have insight into relevant motivational processes. Moving on to the role of theory in prediction and improvements in technology, Prochaska's reply argues that the 'traditional models' which he says I support fall far short in terms of accounting for variance in behaviour change, because they deal only with the small proportion of smokers who seek help with stopping. There are two separate points being confused here, and in both cases Prochaska misstates my position: my own view and indeed the so-called traditional model are concerned as much with population-level changes in behaviour as change in people willing to receive an intervention to help them. The questions are: (1) has the TTM generated better prediction of attempts to quit and success of those attempts either in population samples or special populations than asking about desire/intention to quit and assessing level of dependence; and (2) has the TTM generated more effective population-level or clinical interventions? With regard to prediction, I was not able to find evidence that TTM-based measures were superior to measures of desire/intention and dependence in predicting smoking cessation. I was careful not to say that TTM constructs had no predictive power—it would be surprising if this were the case, as they capture some aspect of intention and behaviour; I wanted to convey the idea that there was a need for their predictive power to represent a genuine advance on simpler accounts. Prochaska considers in some detail several of the papers I cited to support the view that the TTM has performed no better or, in some cases worse, than pre-existing accounts. Abrams et al. [9] found that a simple addiction measure derived from two variables already known or suspected to relate to smoking abstinence predicted long-term abstinence independently of stage of change but not vice versa. Prochaska argues that if stage of change had been entered before the addiction measure in the key regression analysis it might have emerged as a predictor and knocked out the addiction variable. However, from my reading of the paper the variables were entered simultaneously (p. 228) so that argument does not apply. Similarly, Farkas et al. [10] clearly demonstrated strong discrimination between future smokers and ex-smokers using variables already known to relate to cessation and a failure of TTM variables to improve upon or match this (see Figs 1 and 2, pp. 1275 and 1276). Prochaska argues that I was selective in citing studies that fail to show the TTM model to be an improvement on other models, but of the two studies that he argues support his case, one is the Abrams study discussed above and the other is a study that does not attempt to predict smoking cessation [11]. On the other hand, Prochaska does not address the finding cited in my editorial that a simple rating of desire to stop smoking in a large study strongly predicted cessation in 1264 smokers undergoing an intervention, while the pre-cessation stage defined in terms of not planning to quit, planning to quit within 6 months and planning to quit within 1 month was unsuccessful [12]. There are other studies I might have cited that have not supported the predictive value of the stage variable in other behaviours such as dietary choices (e.g. Resnicow et al. [13]), and there are more recent studies on smoking that have not found key TTM concepts to be predictive (e.g. Callaghan & Herzog [14]). The key point remains that there does not appear to be evidence that the TTM has improved our ability to predict smoking cessation (the area in which it has been tested most extensively) beyond assessment of desire/intention to stop and dependence. When it comes to interventions, the approach that the TTM needs to improve upon is one in which a range of common-sense techniques are used to increase the desire to stop at a population level, trigger attempts to stop and provide treatment to assist those who are unable to stop by themselves. This is set out clearly in the English smoking cessation guidelines that I co-authored [15]. It is incorrect for Prochaska to assert that the model I espouse is directed solely at those wishing to stop smoking. At the population level the strategies supported by evidence include the use of policies such as tax increases, anti-tobacco communications and advocacy and smoking restrictions to increase desire to stop and reduce temptations to smoke in those trying to stop. Within the health-care system there is clear evidence that brief advice from a physician to all smokers triggers quit attempts in some of them, while nicotine replacement therapy, bupropion and psychological support increase the chances of success in those willing to use them [15]. There appears to be no evidence that tailoring brief opportunistic advice to stop smoking to stage of change is more effective than simply advising all-comers to stop and offering them treatment to help. I cited a review that concluded that TTM-tailored interventions had not been found to improve on non-tailored ones [16]. In reply, Prochaska draws attention to a comprehensive review of the application of the TTM to smoking which, he claims, supports the improved efficacy of stage-matched interventions over conventional ones [17]. However, careful examination of that review (Table 19, pp. 47–53) shows that where TTM-tailored interventions were compared with non-tailored interventions it was not possible to detect a reliable effect on the outcome measure that matters—which is actually stopping smoking. There were claims of improvements in terms of 'stage progression' but, as I pointed out in my editorial, that is not a sufficiently robust outcome measure and would not be considered acceptable in any technology assessment by bodies such as the UK's National Institute of Health and Clinical Excellence, which produces guidance on cost-effective health-care interventions [18]. In places, Prochaska appears to concede that matching on stage alone would be unlikely to result in a significant improvement in effectiveness, but argues that where full tailoring was used, five of seven studies in the Riemsma [16] and Spencer [17] reviews showed significant effects and the two ineffective ones were with adolescents. However, these studies need to demonstrate an effect on smoking cessation that is greater than in a non-tailored control condition and they need to do this using appropriate outcome criteria. I had difficulty in identifying any studies in the Spencer review that would meet this standard, so it is difficult to see how Prochaska arrives at his conclusions. Perhaps there is a disagreement in our calculations of success rates. In the Velicer et al. study, for example [19], the authors claim that there was in the order of a 5% difference at the 18-month follow-up between an expert system-based tailored intervention and a simple stage-matched intervention. However, the percentages were based on only those participants who were followed-up, and the rate at which participants in the expert system condition refused to provide follow-up data at 18 months was 15.1% versus 10.3% in the simpler stage-matched control condition—a difference of 5%. Thus, if refusers are counted as smokers, which would be normal practice in smoking cessation trials and which Prochaska accepts as good practice in later papers, there would have been no effect detected. Prochaska cites some papers that have appeared since the Spencer and Riemsma reviews in support of the TTM approach to interventions. Taking them in order, Hollis et al.'s [20] paper was not a comparison with a non-TTM approach; Lawrence et al.'s study involving pregnant smokers was not a comparison with a non-TTM approach and, moreover, produced what the authors called a 'marginal effect' of 'doubtful value'[21]—on an intent-to-treat analysis with biochemical verification of abstinence, sustained abstinence at 30 weeks' gestation was 1.4% in the control condition, 2.6% in a condition involving midwife advice plus TTM-based materials and 3.1% in a condition involving midwife advice and a TTM-based computer programme, the differences not being statistically significant (p. 173). Prochaska cites a third recent paper in support of the effectiveness of TTM-based interventions [22], but this is a review paper. It appears that the study being described is in another paper which did not have a generic intervention control [23]. Prochaska also challenges critics such as myself who 'prefer to substitute an addiction model' to explain how this would be of help with 'populations of addicts [sic] with multiple behaviour risks' that have been treated effectively by TTM tailoring. I hope it is clear by now that I do not propose the kind of addiction model Prochaska ascribes to me. Of the papers used to support the evidence for TTM-tailoring one I have already considered, and the others are not published at the time of writing, so it is impossible to comment. One can point, however, to a recent large-scale trial of a stage-matched intervention with smokers who failed to show any benefit, leading the authors to question whether stage-matching does indeed lead to weaker interventions in so-called pre-contemplators [24]. A recent review of stage-matched interventions to promote physical activity has also failed to find any benefit [25]. When considering the above it is very difficult to see how Prochaska can refute the statement that TTM interventions often go for 'soft' outcomes. The Spencer review (Table 19) shows this very clearly in the case of smoking: there is constant reference to effects on 'stage transition' when there are no demonstrated effects on actual cessation of smoking. A more recent example of going for the 'soft' option comes from a study in which a stage-matched motivational enhancement programme for nicotine dependence in methadone-maintained pregnant smokers had no effect on cessation, but was claimed to have moved smokers forward on a stage-based measure [26]. In another paper, the authors claim effectiveness for a hospital-based counselling intervention that yielded a significant effect on 'stage-transition' but not on continuous biochemically verified abstinence [27]. This has been a detailed response to the detailed reply by Prochaska. To end on a positive note, it is good news that Prochaska is interested in moving forward with models of behaviour change. In so doing, I would urge that he and others in the field adopt what is commonly accepted practice in scientific discourse, which is to consider new theoretical propositions as hypotheses until and unless there is good evidence that they represent a genuine advance in the terms described at the start of this piece; then if the evidence fails to support the hypotheses it is important to be willing to revise or discard them. The author's post is funded by a grant from Cancer Research UK. The author undertakes research and consultancy for manufacturers and developers of smoking cessation medications and has written a book on the theory of addiction.