Representativeness
Despite achieving a reasonably high response rate to the questionnaire, non-response analyses were conducted to determine whether those who responded were representative of the total sample. Age, sex and baseline sample type (i.e., participants were originally classified based on the nature of the course of care they received at a public dental clinic at the time they were recruited, i.e., as either an emergency or general dental care patient) were examined between responders and non-responders. No significant differences were found between the proportions of responders and non-responders in terms of sex and baseline sample type (χ
2 test, P > 0.05). However, age differed significantly between those who responded and those who did not respond to the questionnaire, with responders being significantly older on average (55.9 years cf. 44.7 years; ANOVA, P < 0.0001). Thus the sample of responders was not totally representative of the age distribution of the total sample. However, in the regression model predicting behavioural intention where age was controlled for, age did not emerge as a significant independent predictor of intentions so in terms of the prediction of intention, it can be assumed that no significant biases were introduced in the sample. However, in the model predicting actual dental attendance, age was a significant independent predictor behaviour, so with respect to the assessment of behaviour, some bias may have been introduced the sample. That said however, in the National Survey of Adult Oral Health conducted in 2004–06 [
3], the average age of dentate adults eligible for public funded dental care was found to be 54.0 years in South Australia and 52.7 years nationally, so the age distribution of the sample used in this research does not differ very much from that in the National survey.
Patients' records were accessed electronically which enabled the collection of data relating to service use during the 3-year follow-up period and post-TPB questionnaire for the majority of participants in the sample. Databases across community dental clinics in South Australia were cross-matched with patient details to ensure that if a patient visited public dental clinics at different locations across the follow-up period, their data relating to visits and service provision would be identified and extracted. Therefore, most of, if not all, data related to public dental visitations was captured.
Whilst it is possible that some patients may have attended a private dental practice for dental care, data examining this group of patient's use of private dental services over the follow-up period was not collected. One of the main purposes behind following these patients was to determine their dental attendance behaviour based on a pattern of attendance in the public sector during the follow-up period. The questionnaire itself was designed to explain and predict dental visiting behaviour within the public sector, and so measures of intentions, attitudes, subjective norms and perceived behavioural control were derived from responses to questions relating to dental visiting within the public sector only. Consequently, by not incorporating private dental attendance into the measure of behaviour, and also because references to visiting the dentist privately were excluded from the derivation of the main components of the TPB, the resultant dataset strictly captured patients' perceptions of visiting the dentist within the public dental service and therefore eliminated as much as possible the confounding effects of visiting within the private sector, giving a clearer picture of the influential dental visiting factors specific to the public sector.
Dental visiting behaviour
This research sought to identify the motivational factors underlying dental visiting behaviour in a sample of users of public dental services. This study was conducted because there was limited evidence of the influence of psychosocial factors and health beliefs on public dental patients' patterns of public dental service use in SA, and more information was needed to help inform dental health policy and the delivery of services. It was found that public dental patients held fairly favourable attitudes toward visiting the dentist, perceived positive social pressure to do so and generally felt in control of visiting the dentist. The association of each of these factors with intention to visit the dentist varied, with correlational analyses showing self-efficacy to be more strongly associated with intention, followed by attitudes toward dental visiting and subjective norms. This research has highlighted the relative importance of the TPB constructs upon behavioural intention and subsequent behaviour. These relationships should be considered when designing educational programs to promote dental attendance. For instance, in order to increase people's motivation/intention to attend dentists, self-efficacy seems to be by far the most important factor to influence, followed in descending order by attitudes and subjective norm. In the behaviour model, both intention and self efficacy had a statistically significant association with dental attendance behaviour, calling for both a motivational and a structural educational approach. Furthermore, because perceived control was not statistically significantly associated with intention, the independent effect of self efficacy upon subsequent behaviour might reflect lack of confidence in ones ability to attend dentists (not perceived controllability or the extent to which attendance is up to the actor) and might call for reduction in structural barriers as a focus for intervention.
Prediction of intention and behaviour
The use of regression models allowed for the valuation of the importance of each of the constructs of the TPB relative to dental visiting intention and behaviour. The four factors to emerge as important predictors of intention were attitudes, subjective norms, self-efficacy and perceived control. Users of public dental services were less likely to believe that the decision to visit the dentist was under their control (i.e., in terms of perceived difficulties in visiting the dentist), although those intending to visit were more likely to have positive attitudes toward visiting the dentist, perceive support from significant others to visit the dentist and be confident within themselves of their ability to make a visit. These results suggested that although one may hold positive beliefs and attitudes toward visiting the dentist, and despite feeling comfortable in going to the dentist, there are external influences that affect intentions to visit the dentist. As a result, the most effective interventions may be those that attempt to change structural/organisational influences. Direct attempts at encouraging dental visiting will need to involve initiatives that improve access to care, such as addressing the cost of dental care or the long waiting times that currently exist within the public dental system. Whilst this research suggests that these sorts of interventions seem appropriate, there are a number of issues raised by other researchers using this model that should also be considered. These issues relate to the measurement and conceptualisation of the perceived behavioural control construct. Whilst there is agreement about the multidimensionality of this construct (in terms of self-efficacy and perceived control), and indeed the findings from this study support, researchers have questioned whether perceived behavioural control when operationalised in terms of perceived difficulty is just another measure of attitudes, and whether perceived behavioural control when operationalised in terms of self-efficacy can really be discriminated from intentions [
39]. Bearing these issues in mind, further work should be carried out to explore the dimensional structure of perceived behavioural control in this context so that appropriate focuses for intervention can be determined.
In this research, a modest amount of the variability in intentions (i.e., 12.0%) could be explained by respective components of the model. The somewhat low predictive power is disappointing since intentions and its predictors were measured at the same time on the same questionnaire using similar items – conditions that should maximise predictive power. The upside however, was that the proximal determinants of intention (as specified by the TPB model) still explained a modest amount of the variation with all of the variables being significant predictors. Modest predictive power for intentions may have been the result of a lack of variation in responses to scales measuring intentions, attitudes, subjective norms and perceived behavioural control and may reflect a bias in the original sample selection. For example, those agreeing to participate in the study may have been those people with fairly, or very strong, positive attitudes. Even the addition of past behaviour as a potential predictor of intentions did not improve the predictive power of the model. Ajzen [
17] suggests that when individuals have ambivalent or uncertain attitudes and normative influences, the effect of prior experiences will more strongly affect intentions. In particular, when individuals have no clear plan of action, they are more likely to rely on their experiences to gauge their intentions as well (i.e. residual effects of prior experience can be powerful, particularly in situations where individuals have little certainty in terms of their attitudes, subjective norms, or their perceived behavioural control). As Ajzen [
17] suggests about the residual effects of past behaviour on behavioural intent, prior behavioural experience can affect behavioural intent; and yet prior experience may be mitigated by the intervening factors outlined within TPB as seems to be the case in this research. Participants in this study had strong intentions and quite favourable attitudes, subjective norms and perceptions of behavioural control.
The amount of variance explained in actual dental attendance by measures of intention and perceived behavioural control components was small, only 6.6%. The inclusion of past behaviour in the model strengthened the model's predictive power, with the amount of variance explained increasing to 14.1%. During the period of post-questionnaire follow-up only 35.4% of questionnaire respondents visited the dentist, so perhaps those who had yet to visit during the observed follow-up period did in fact intend to visit but simply had not yet done so. In fact, 77.4% reported that they did intend to visit the dentist, but only a small proportion actually did visit. Perhaps a longer follow-up period was needed to better capture the dental behaviour of study participants as longer time intervals allow more opportunities for the behaviour to be performed, increasing the intention-behaviour correspondence [
40].
Interestingly, those with a past pattern of emergency dental visiting had significantly greater odds of visiting the dentist post-questionnaire. Perhaps familiarity with visiting the dentist as an emergency dental patient in the past subdued the effect of their perceptions of difficulty associated with visiting the dentist since perceived control failed to be a significant predictor of visiting the dentist. This finding supports Ajzen's [
12] assertions about one's perceptions of control being altered once they believe that they are able to perform a particular behaviour.
Further explanations are offered for the modest predictive power for intentions and behaviour. Firstly, a lack of variation in responses to scales directly measuring intention, attitude, subjective norm and perceived behavioural control may have been a contributing factor. This may reflect a bias in the original sample selection. For example, those agreeing to participate in the study may have been those people with fairly, or very strong, positive attitudes and beliefs. Secondly, there was a lack of correspondence between the intention scale, which had 7 points, and the behaviour measure, which had 2 points. It is impossible to have a linear relationship, let alone a perfect linear relationship, and explain all of the variation when there are unequal numbers of scale categories [
40]. However, even with equal categories for the two measures, if the distributions are not the same, then it is not possible to obtain a perfect correlation between the two measures [
40]. For the dichotomous measures of intention (i.e., intend to visit or do not intend to visit) and behaviour (i.e., visited or did not visit) there was a 23%/77% No/Yes split on intention but a 65%/35% No/Yes split on behaviour. Because the two distributions were not concordant, only a small proportion of the variance in behaviour will be accounted for by intention, explaining why intention, although significant, had low predictive power of actual behaviour in the behaviour model. Thirdly, a high intentions-behaviour association is likely to be obtained if intentions remain stable. However, intentions may change. Sutton [
40] explains that the longer the time period between the measurement of intention and behaviour, the greater the probability that unexpected events will occur, leading to changes in intention. This may certainly be the case when measuring one's intention to visit the dentist at a public dental clinic. For instance, an individual may not intend to ever visit the dentist but may develop an unexpected dental problem causing them to make an emergency dental visit. Or perhaps personal circumstances change and they become ineligible to receive public dental care and therefore do not make a dental visit despite their reported intention to do so. Also, Sutton [
40] comments that some participants may not be 'engaging in real decision making' when they are completing the questionnaire. If intentions are measured before they have been formed the relationship between intention and behaviour will not be as strong [
40].
In the context of the TPB, health behavioural change is the result of reciprocal relationships between the environment, personal factors, and attributes of the behaviour itself. People's perceived control over the opportunities, resources, and skills needed to perform a behaviour affect behavioural intentions and actual performance of the behaviour. In this research, regression analyses highlighted the predictive strength of the self-efficacy and perceived control construct for dental visiting intention and the self-efficacy construct for actual dental visiting behaviour. Based on these analyses, interventions should target individuals' perceptions of behavioural control when seeking to increase dental visiting intentions and promote dental attendance, particularly preventive dental attendance. An approach to enhancing an individual's control over visiting the dentist would be to make changes or intervene at an environmental level. This may involve measures that increase the availability and accessibility of public dental facilities.
The results from this study can also be used in patient- and community-centred health education by identifying and enhancing the psychological features (such as self-efficacy) that characterise dental visiting behaviours. Perceptions of self-efficacy can be used to explain behavioural changes, to predict effects of interventions, and to improve dental health behaviour. In relation to dental health behaviour, self-efficacy determines whether a given behaviour is initiated (as demonstrated by the behaviour models) and for how long the behaviour may continue against any obstacles that are encountered. This is because self-efficacy beliefs provide 'the foundation for human motivation, well-being, and personal accomplishment' [
41]. There is little incentive for individuals to carry out a behaviour, or to persevere in the face of difficulties and barriers, if they believe that the behaviour will not lead to outcomes they desire [
41]. In addition, the results suggest that attitudinal considerations and beliefs regarding other people's support of the behaviour also have a role to play in dental visiting intention.
Effective interventions for behavioural change must therefore influence multiple levels because, as this research has demonstrated, dental health behaviour is shaped by many environmental subsystems, including family, community, beliefs, organisation of dental services, and the physical and social environments in which people live.