Until the results of the Women's Health Initiative trial [1] were known, hormone therapy (HT) was promoted for preventing osteoporosis and was thought to protect from cardiovascular diseases [2]. Although physicians agreed that menopause was not a disease, they considered it a serious health problem [3]. Hot flashes, night sweats and osteoporosis were reasons for gynaecologists and general practitioners (GPs) to prescribe HT [4]. Notwithstanding these benefits, breast cancer was a major concern [5, 6]. Moreover, many women objected to adopting HT at menopause from fear of medicalization of an otherwise natural process [7]. Regardless of these conflicting evidences, in Canada, at the time this study was conducted, Premarin^®, a known estrogenic compound, was the most prescribed drug by gynaecologists and the third most prescribed drug by GPs [8].

Following the Women's Health Initiative trial [1], there was a decrease in the number of prescriptions of standard-dose HT but a slight increase in the number of prescriptions of low-dose HT in the United States [9]. In fact, HT remains the most effective option to alleviate severe perimenopausal symptoms such as hot flashes [10]. Moreover, although HT is not a first-line treatment recommended for preventing osteoporosis [11], its use is associated with a decrease in fractures in current users [12]. Based on the results of the Women's Health Initiative trial, HT use for 1 year in 10 000 healthy postmenopausal women is associated with 7 more cardiovascular disease events, 8 more invasive breast cancers, 8 more strokes, 8 more pulmonary emboli, 6 fewer colorectal cancers and 5 fewer hip fractures [1]. Weighing the benefits with the risks associated with HT reminds women and their doctor that deciding about HT requires a careful and individualised assessment of their personal situation.

It is in this context that decisions about HT is said to be representative of clinical decision-making in the face of scientific uncertainty [13, 14]. Thus, it is likely that many factors play a role in a doctor decision-making process leading to a prescription of HT. Attitude towards HT is associated with the medical specialty [15, 16], and age [17] as well as gender of the provider [15, 18]. The country in which a doctor practices medicine also appeared to play a role [19].

Conceptual underpinnings of this study

In studies of doctors' decision-making and related prescription behaviours, more attention needs to be given to the use and combination of different theories [20]. The lack of use of theories restrains effective implementation of change in patients' care because it restrains our understanding of the pathways through which a given implementation strategy is effective. The theory of planned behaviour [21] is well known through its previous applications to the study of doctors' behaviours [22–25]. This theory provides a theoretical account of the way in which attitude, subjective norm and perceived behavioural control combine to predict a given behavioural intention (decision-making) [26] and in turn, a given behaviour [27]. Thus, it provides direction for elaborating effective strategies that will influence the decision-making process leading to change in doctors' behaviours.

This theory postulates that under a controlled situation, intention is the immediate determinant of behaviour [21]. In turn, this intention is under the influence of three main factors: attitude, subjective norm and perceived behavioural control. Each can be assessed directly or indirectly (beliefs-based measure). Attitude is conceptualized as a personal evaluation of the action. It is the product of a set of salient beliefs about the consequences of performing the behaviour, each weighted by an evaluation of the importance of the respective consequences. Subjective norm refers to a perceived social pressure to perform the behaviour in question. The belief-based measure requires that the individual's normative beliefs be multiplied by the individual's motivation to comply with these socially-normative referents. Perceived behavioural control is a measure of the amount of control the individual has over the behaviour in question. It refers to the individual's perception of barriers or facilitating factors likely to influence the adoption of the behaviour. It can also be measured directly or indirectly. For the indirect measurements, the individual's control beliefs must be weighted by the corresponding perceived evaluation of how much each of these control beliefs will impact on the adoption of the behaviour. According to the authors of this theory, sociodemographics and other variables will influence behaviour through their influence on the attitude, the subjective norm and the perceived behavioural control [21]. Successful behavioural change will occur only if the underlying determinants of intention change.

Although the theory of planned behaviour has proven useful when studying health related behaviours, two systematic reviews found that components of this theory explain on average 41% of the variance in intention and 28% to 31% of the variance in behaviour thus, suggesting that other variables must play a direct role on the behavioural intention and possibly, on the behaviour itself [28, 29]. Consequently, some authors have expanded the theory of planned behaviour to include other relevant psychosocial constructs [28]. For example, the measure of moral norm [30] was found to be useful in understanding women's intention to use HT [31, 32]. Moral norm takes into account feelings of personal responsibility regarding adoption of a specific behaviour, that is, the individual's perception of the moral correctness or incorrectness of performing the behaviour. Because of the potential influence of the medical specialty, in this study, role belief was also taken into consideration [30]. Role belief refers to the perception by the individual that members of a specific group would perform the behaviour under study. The initial theoretical model adopted as a basis for examining the psychosocial determinants associated with doctors' intention to prescribe HT is presented in Figure 1.

To our knowledge, at the time this study was planned, we were not aware of any studies that had applied the theory of planned behaviour to study variations of doctors' intention to prescribe HT across countries. Physicians in France and the province of Quebec, Canada, share the same language (French) but use it in very different geographical and cultural settings. Therefore, studies of their intention to prescribe HT offered a unique opportunity to provide more insight in the cultural variation in the psychosocial determinants associated with this behaviour without the difficulties associated with having questionnaires in two different languages. Therefore, this study is an application of the theory of planned behaviour with additional variables. It was designed to assess which variables were associated with the intention to prescribe HT. Hypotheses refer mainly to the assumption underlying the theoretical framework. It could nonetheless be added that we hypothesized that variations in the intention to prescribe HT between gynaecologists and GPs would be observed as well as variations in the intention to prescribe HT between French and French Canadians physicians would be observed.

To our knowledge, this is the first study that adopted the theory of planned behaviour as a theoretical basis for assessing and comparing the psychosocial determinants of GPs and gynaecologists' intention to prescribe HT in two countries, France and in Quebec. To our knowledge, this is also one of the few studies to adopt the theory of planned behaviour for identifying factors that influence doctors' behavioural intention to prescribe a medication in the context of preference-sensitive care.

The results of this study indicated that doctors' intention to prescribe HT varied according to their specialty. In both countries, gynaecologists showed higher intention to prescribe HT than GPs. This is in line with the results reported by Hovi and colleagues who assessed Estonian and Finnish GPs and gynaecologist opinions and prescribing practices in HT [19]. However, this intention did not vary significantly between GPs in France and those in Quebec, nor did it between gynaecologists in France and those in Quebec.

In the overall model (for all physicians in both countries), perceived barriers to prescription of HT was found to be the most important determinant of the intention to prescribe HT. This means that the more confident a doctor was regarding his/her ability to counter the perceived barriers that were assessed in this study, the more likely he/she was to express a high intention to prescribe HT. This is congruent with the observation made previously that a significant proportion of physicians considered menopause to be a serious health problem that needed to be treated [3]. In other words, even in the absence of a clear indication to prescribe HT (for example, severe hot flashes) or a clear indication by a woman that she would like to use HT, at that the time this study was conducted, some physicians had high intention to do so.

Moral norm was found to be the second most important determinant. In this study, moral norm was assessed with three items referring to the physician's personal convictions, sense of guilt and ethics. In the past, moral norm was found to be the strongest predictor of women's intention to adopt HT over a period of one year [32]. However, to our knowledge, the present study is one of the first to document that moral norm is associated with a doctor's prescription behaviour of HT. This means that, in this context, doctors' most profound personal values might overcome those of the women. As reported before by Hoffmann and colleagues, this suggests that doctors might be inclined to promote their own decision [38]. On one hand, this is in line with the literature that points to the contribution of doctors' preferences on the unexplained practice variation observed in the medical community [39]. On the other hand, this is contrary to the growing social trends in regards with shared decision-making [40]. Shared decision-making is grounded in ethics in which a deliberative model of decision-making is advocated [41]. A deliberative model of decision-making recognizes that both protagonists are competent individuals. It refers to the concepts of autonomy and self determination which underline the obligation of health care providers to provide information [41]. Shared decision-making emphasizes the importance of agreement on the treatment option by the patient and the physicians as a "the test of a shared decision" because "mutual acceptance ... remains a necessary prerequisite for shared decision-making" [42]. Shared decision-making rests on the best evidence of the risks and benefits of all the available options and takes into account the establishment of a context in which the values and preferences of the patient are sought and his opinions valued [40]. Therefore, these results suggest a need for interventions that will facilitate the inclusion of the specific characteristics and values of the patient in the decision. In the overall model, the intention to prescribe HT was also under the influence of role belief. This means that intention to prescribe HT was determined by a sense of belonging to a specific group and how one ought to act accordingly. This is different from the influence of subjective norm, which refers to perceived social pressure to perform a specific behaviour. In other words, doctors' intention to prescribe HT was dependent on what they believe someone like them should do. Doctors attach a high value to their clinical autonomy, which is viewed by certain critics as a major barrier to changing doctors' behaviour [43]. The positive influence of role belief in predicting doctors' behaviour had been reported before in the area of technology of information [24]. It appears that in the context of HT prescription, doctors might not be influenced by what others would want them to do (including a specific patient), but rather by what they believe a doctor like themselves ought to do.

When specific regression models were built by medical specialty and country, power of control beliefs was kept in all models. Moral norm was kept in three out of the four models: GPs in Quebec and gynaecologists in France and in Quebec. Role belief was kept in both models from France: GPs and gynaecologists, but not in the models of doctors from Quebec (GPs and gynaecologists). This suggests that, notwithstanding the medical specialty, there might be differences between doctors from France and those from Quebec regarding this particular behaviour.

In three out of the four models, variables that were external to the initial theoretical model made a statistically significant contribution. All of these variables pertained to diagnostic testing. For example, in each of the regression models pertaining to GPs' behavioural intention, variables associated with diagnosing menopause were found. This suggests that GPs in Quebec and in France might need reassurance in the final objective diagnosis of menopause before embarking on a prescription of HT.

Interestingly, attitude and attitudinal beliefs were not retained in any of the regression models. This means that the intention of doctors to prescribe HT was not under the influence of those doctors' personal evaluation of this action (for example, the consequences of prescribing HT on the woman's health outcomes). This suggests that, when controlled for other psychosocial variables, the prescription of HT was not under the influence of what doctors believe HT would have had as potential benefits or risks to this woman.

Another interesting finding of the present study was the U-shaped distribution of the doctors' intention to prescribe HT. Indeed, given the recommendation regarding HT by some of the professional societies at the time this study was conducted [44–47], one could have expected a skewed distribution towards a more positive physicians' intention to prescribe HT. This U-shaped distribution of the doctors' intention to prescribe HT is in line with the results of a previous study published in 1986, long before HT was thought to be benefiting the cardiovascular system. Interestingly, in this study published in 1986, doctors were shown to act decisively when facing decisions about HT [14]. Facing twelve clinical vignettes, most of the fifty doctors enrolled in this study tended to act decisively (prescribe HT or not) although the decision analysis model had recommended a "toss-up" strategy (uncertain decision) 60% of the time. The authors concluded that "doctors ought to carefully re-examine their practice patterns in light of their beliefs and opinions." [14]. This conclusion would still hold true two decades later. It suggests that it is possible that physicians' intention to prescribe HT and its determinants could be stable overtime.

This U-shaped distribution of the doctors' behavioural intention regarding HT was in sharp contrast with the results from two other studies that applied the theory of planned behaviour to predict middle-aged women intention to use HT at menopause in Québec. These two studies were conducted in the same period as the present study was. In both these studies, the distribution of the intention to use HT was normal. In other words, while most women appeared to be undecided about adopting HT, physicians from the same geographic area appeared to be decided about either prescribing HT or not. Again, these results emphasise the need for future interventions that will facilitate the active participation of women in the decision-making leading to a prescription of HT.

Despite its limitations, the present study still has clinical and research implications. Geographic variations in health care delivery remain a challenge [39]. This study provides empirical data on the significant contribution of moral norm in the decision-making leading to a prescription of HT by doctors and possibly in other contexts of preference-sensitive care [39]. This is preoccupying because it is contrary to on-going efforts to incorporate the best available scientific evidences as well as the patients' views in health [53], one of the cornerstones of improving the quality of health care [54]. Therefore, it would be important in this type of clinical situation to provide guidance to doctors on how to support active participation of patients in the decision-making process.

This study should be useful to those interested in the application of social cognitive models in studying doctors' behaviour. Given the lack of theorization in the area of health care professional practice [20, 55], this study results improve the knowledge base in translation of evidence as well as the knowledge base of physicians' decision-making in context of preference-sensitive care [39]. It provides insight about the contribution of moral norm and role beliefs above variables provided by the theory of planned behaviour to explain doctors' behaviour and perhaps, variation in health care delivery. In future studies applying the theory of planned behaviour, these constructs should be tested with other type of doctors' behaviours. This study should also be useful in improving our understanding of cross-cultural variations in medical practices by focusing on theory-based psychosocial variables. Lastly, this study points to possible pathways through which a given implementation strategy is likely to be effective in modifying doctors' prescription of HT. Researchers interested in elaborating interventions to modify this prescription behaviour could use this study results accordingly.