The role of psychosocial factors and treatment need in dental service use and oral health among adults in Norway

Objectives: This study aimed to explore whether population characteristics were associated with the use of dental services, individual's personal oral health practices, dental caries and oral health-related impacts using the revised Andersen's behavioural model as the theoretical framework. Methods: This cross-sectional study included participants from a Norwegian general population (N = 1840; 20-79 years) included in the Tromstannen—Oral Health in Northern Norway (TOHNN) study. The variables included in the model were social structure (income, education, urbanization), sense of coherence (SOC), enabling resources (difficulties accessing the dentist, declined treatment, dental anxiety), treatment need, use of dental services, toothbrushing frequency, sugary soda drink consumption, decayed teeth and oral health-related impacts (OHIP-14). Structural equation modelling was used to test the direct and indirect effects within Andersen's


| INTRODUC TI ON
Regular dental attendance has been shown to have a positive impact on oral health. 1,2 Despite this, both access to oral health care and regular dental attendance have been shown to vary considerably within and between populations. 3,4 Why people do not regularly attend dental services is complex, and there is probably an interplay between psychosocial, material, cultural and behavioural aspects as well as the perceived need for treatment. Factors found to be associated with less use of dental health services are lower socioeconomic status (SES), 4-6 dental fear, 7-9 distant geographic location and distance to dental care services 10,11 and fatalistic beliefs. 12 Understanding why people do not seek regular dental care is an essential requirement for developing effective health-policy interventions to reduce inequalities in oral health outcomes in any population.
It is well known for example that dental attendance is related to dental caries. 1,13,14 Dental caries is also a result of a complex interplay of other factors including an individual's behaviour, nutrition, SES, genetics and local oral environment. 15 It is a diet-mediated disease and free sugars seems to be the primary determining factor. 15 In addition, SES explains variations in dental caries between people. In a systematic review by Costa et al 16 the authors found that an individual's educational level, income and/or occupation status were all associated with a greater prevalence of dental caries. In addition to SES indicators, psychological factors such as dental anxiety and sense of coherence are possible mediators of dental caries. 7,17 Sense of coherence (SOC) is a salutogenic construct relating to the way an individual makes sense of the world (comprehensibility), has the required resources to respond to life events (manageability) and feels that these responses are meaningful and make sense (meaningfulness). 18 A stronger SOC has been associated with greater dental attendance, more frequent toothbrushing and fewer oral health-related impacts. [19][20][21][22] The research to date would suggest that salutogenic factors-such as SOC-could play an important role in improving dental access and oral health-related behaviours and outcomes in adults.
To gain knowledge of the complex inter-relationship between the many determinants associated with the use of dental services, dental status and oral health impacts, biopsychosocial models have been shown to be useful tools to guide the selection of variables. One such theoretical model is Andersen's behavioural model for health services' use. 23, 24 Andersen 23 originally developed the model to predict and explain why and how people use health care services by integrating predisposing/social structure factors (eg education, physical environment), enabling resources (eg having the means to use available health services) and need for health care (eg how people view their need for care). These different population characteristics have been used to explain why some people are more likely to seek health care. The model has, during the last three decades, been further extended and developed. Personal health practices and health outcomes/status (both self-perceived and clinically evaluated) have been added ( Figure S1 in supporting information). 24 The model has been tested to explore which factors determine dental care use and self-reported oral health in two different general populations in the UK. 25 ing characteristics on treatment need and the use of dental services were mediated by enabling resources. That is, whilst there were individuals who were more predisposed to seek dental care, there had to be the means, meaning, the enabling resources for them to do so.
Furthermore, in both studies, they found that people with higher perceived need were less likely to attend regular dental appointments and reported increased oral health impacts.
Andersen's behavioural model was recently utilized to explore how individual characteristics (psychosocial factors, enabling resources and treatment need) were associated with dental attendance, smoking habits, toothbrushing frequency, periodontitis and oral health-related impacts in a Norwegian adult population. 27 In this study, SOC was also included in the model to examine key associations with adults' oral health as well as other key determinants. In line with the two previous UK studies, 25,26 the relationship between predisposing characteristics (income, education and urbanization) and use of dental services was mediated by enabling resources. In contrast, however, need was not associated with use of dental services or oral health impacts. Instead, SOC, predisposing characteristics and the severity of periodontitis were the main predictors of oral health-related impacts. Severity of periodontitis was linked to smoking and predisposing characteristics (eg income).
In order to further extend our understanding of the oral health of the Norwegian population, in particular, the association of key individual characteristics with the use of dental services and oral health impacts, the present study utilized Andersen's behavioural model with dental caries as the clinically evaluated outcome. In addition, the frequency of soda drink consumption-as a behavioural factor commonly associated with dental caries-was examined.
As discussed by Costa et al 16 in their recent systematic review, the majority of previous research on the socioeconomic factors related to dental caries have employed bivariate and/or multivariate analysis including a handful of factors whilst controlling for potential con- founders. Yet, given the complex range of clinical, psychological and social factors important in dental caries, it is necessary to study the inter-relationships between all of these potential factors at the same time. 28 Currently, one way to explore the inter-relationship between several contributing factors simultaneously is to apply theoretically driven structural equation modelling (SEM). The aim of the present study therefore was to employ SEM to examine how population

K E Y W O R D S
Andersen's behavioural model, carbonated beverages, dental anxiety, epidemiology, oral hygiene, sense of coherence, socioeconomic status, structural equation modelling characteristics (predisposing factors; enabling resources; perceived need) were associated with the use of dental services, oral health practices, dental caries and self-reported oral health impacts using Andersen's behavioural model for health services' use as the theoretical framework.

| Study design and participants
Andersen's behavioural model was tested in a general adult population with data from the Tromstannen -Oral Health in Northern Norway (TOHNN) study utilizing dental caries (decayed teeth) as the clinical outcome. 29 The TOHNN-study was a cross-sectional study in an adult population in Troms County. 29 Details of the invitation procedure have been described in a previous publication. 29

| Measures
The variables in the study were chosen to reflect the constructs in Andersen's revised behavioural model 24  (a) <300 000 NOK, (b) 300 001-450 000 NOK, (c) 450 001-900 000 NOK and (d) >900 000 NOK. Urbanization was used as an indicator of the number of inhabitants and availability of dentists as a ratio of inhabitants per dentist. The municipality with the larger town had the highest dentist availability and was categorized as urban; two municipalities with smaller towns had the second highest availability and were categorized as suburban; and the remaining municipalities without towns and with lowest availability were classified as rural.
The second latent variable, SOC, was assessed with the Norwegian version of the previously validated 'Orientation to life questionnaire', which is comprised of 13 items. 30,31 Indicator variables were represented by the three SOC dimensions: comprehensibility (five items), manageability (four items) and meaningfulness (four items).
Enabling resources were measured with one latent variable. The three indicators for this latent variable were declined treatment due to costs, perceived difficulty accessing a dentist (each assessed with one question), and dental anxiety. Dental anxiety was assessed with the previously validated Norwegian version of Corah's Dental Anxiety Scale (DAS). 32,33 For analysis, the DAS-score was reversed so higher scores represented less dental anxiety.
Treatment need within the model was assessed with one observed variable: "If you saw a dentist tomorrow, do you think you would need treatment'? (yes/no)".
Oral health-related behaviours were represented by personal oral health practices and use of dental services. Personal oral health practices included frequency of toothbrushing and of sugary soda drink consumption. Each of these was measured with one item (observed variables). Use of dental services was measured as a latent variable with two indicators: attendance orientation and frequency of attendance. The response options for these, and all variables, can be seen in Table 1.
The clinically evaluated outcome, dental caries, was measured with one observed variable: number of decayed teeth. Decayed teeth was radiographically and clinically assessed for all tooth surfaces in all teeth, except third molars, using four bite-wing radiographs and a dental explorer (Hu-Friedy EXS9), mouth mirror and compressed air.
Decayed teeth were recorded on a five-grade diagnostic scale. 34 In this grading scale, caries Grades 1-2 are denoted as enamel caries and Grades 3-5 as dentine caries. In this study, teeth with only enamel lesions (Grades 1-2) were defined as healthy. Teeth with dentine caries (Grade 3-5) were defined as decayed, independently of severity. Two calibration tests were conducted during the study period with a 3-month interval. A set of bite-wing radiographs was examined by all examiners, and congruency towards the gold standard using proportion of agreement and Cohen's kappa (κ) was evaluated with an acceptable agreement (per cent agreement: 75%-100% and 81%-92%, respectively; median κ-values: 0.73 and 0.77, with quartile deviations between 0.5-0.85 and 0.74-0.79, respectively). 29 For more information on the distribution of caries and caries experience for this Norwegian population sample see Oscarson et al. 35 Subjective oral health-related impacts were measured as a latent

| Data analysis
In order to assure the estimated minimum sample size for the struc- showed that missing values appeared to be at random and at a low rate (1.9% of the total values used in the present analysis). Missing variables were replaced as follows: All one-item variables were replaced with the median. Individuals with more than three missing SOC items were excluded from analysis. If three or fewer items were missing, they were replaced by the median value of the remaining SOC items for that individual. 38 Individuals with more than two

TA B L E 1 (Continued)
missing OHIP-items were excluded from analysis. When two or less items were missing, they were replaced with the sample median of the relevant OHIP-item. 39 Individuals with more than one missing item in the DAS-scale were excluded from analysis. Single missing items were replaced with the median value of the remaining DAS items for that individual. Re-analysis of data excluding individuals with any missing items did not change mean scores by more than two decimal places or frequency distributions by more than one percentage point, except for income and soda drink consumption that changed 2.4 and 3.4 percentage points, respectively (not reported).
The excluded individuals did not differ significantly on key outcomes (decayed teeth and OHIP-14 sum mean score) compared with those that remained in the analysis. In Step 1 of the analysis, to identify whether the indicators chosen to measure the five latent constructs were acceptable, confirmatory factor analysis (CFA) was used. CFA is the first in the two-stage process of SEM (the measurement model). 40 (Table 2). In this model, 53%, 27%, 58%, 12% and 48% of the bootstrapped variance was accounted for in enabling resources, need, use of dental services, number of decayed teeth and oral health-related impacts, respectively (Figure 1).
The direct effects can be seen in Table 2 and Figure 1. linked to less use of dental services, more frequent toothbrushing, less frequent sugary soda drink consumption, fewer decayed teeth and fewer oral health-related impacts. As hypothesized, more perceived treatment need was linked to more decayed teeth.
There were sixteen significant indirect paths (Table 2 and Figure 1). More social structures were associated with less treatment need, more use of dental services, more frequent toothbrushing, less sugary soda drink consumption and fewer decayed teeth. A stronger SOC was associated with less treatment need, more use of dental services, more frequent toothbrushing, less sugary soda drink consumption and fewer decayed teeth. More enabling resources were linked to more frequent toothbrushing, less sugary soda drink consumption and fewer decayed teeth. More treatment need was linked to more sugary soda drink consumption and more decayed teeth. Finally, more use of dental services was linked to fewer decayed teeth. These are total indirect paths and can consist of a single potential effect or a multitude of potential effects. For example, greater social structures were linked to a greater perceived treatment need via a single effect: more enabling resources. On the other hand, more social structures were linked to fewer decayed teeth via two effects: more enabling resources and greater use of dental services.

| D ISCUSS I ON
In the present study, we found support for Andersen's behavioural model of access and health outcomes when applied to the oral health of a Norwegian adult general population sample. These findings therefore lend further support to the use of this as a conceptual framework for understanding the key determinants for dental health service use, self-reported oral health impacts and tooth decay in adults.
The findings regarding use of dental services were in line with our previous analysis testing Andersen's behavioural model with periodontitis as the clinically evaluated outcome. 27 In relation to both tooth decay and periodontitis, enabling resources seemed to be a key factor associated with regular dental attendance. Those people without dental anxiety, who perceived no or few economic or practical difficulties accessing dental care were more likely to visit the dentist regularly regardless of educational level, household income or indeed the availability of dentists (ie social structures). In the present analysis, we found that these social structures were mediated through enabling resources suggesting that having a lower socioeconomic status does not necessarily mean that people do not attend dental services regularly. This is in contrast to previous studies 4-6 but F I G U R E 1 Significant pathways in a model of dental service use and oral health outcomes based on the revised Andersen's behavioural model for health services use. Solid lines = direct effect; dashed lines = indirect effect As this was the first study to use dental caries as the clinically evaluated outcome in Andersen's behavioural model, comparison to other studies was not possible. The relatively small amount of variance in the decayed teeth outcome measure explained by the model could be expected since other well-known factors associated with dental caries (eg fluoride supplement, diet) were not included in the current analysis. Unexpectedly, however, social structures (education, income, urbanization) were not directly associated with decayed teeth as has been previously reported in many cross-sectional studies. 16 Rather, in a population with a relatively high educational level, socioeconomic factors appeared to be linked to tooth decay through people's enabling resources (dental anxiety, treatment costs, access to dental services) or increased toothbrushing frequency. One interesting difference between the present analysis and our previous research 20 was that here regular dental attendance was associated with less disease-that is fewer decayed teeth whilst in relation to periodontitis, more frequent use of dental services was associated with higher likelihood of having periodontitis. As expected, less frequent toothbrushing and more sugary soda drink consumption were associated with decayed teeth. Furthermore, there seemed not only to be a direct relationship between SOC and number of decayed teeth, SOC also appeared to be linked to oral health-related behaviour, which in turn predicted dental caries, as has been suggested in previous studies. 19,47 Number of decayed teeth was not associated with impaired oral health-related quality of life, in contrast to findings from other studies. [48][49][50] The studies are however, not directly comparable due to age differences between study populations. Also, in the present study several factors were analysed simultaneously in an a priori theoretical model and variables that were correlated in bivariate analysis were no longer associated when all other variables were added. When all variables were tested together, lower education, lower income and low availability to dentists were associated with impaired oral health-related quality of life. In our study, SOC was the factor with the strongest association to self-reported oral impacts.
This adds to existing literature on the important role of SOC, as a key psychosocial factor, associated with oral health practices and oral health outcomes. 19,22 Interestingly, a strong SOC was directly associated with less use of dental services but to more use when of dental services, having decayed teeth and oral health-related impacts but also how different factors are related by assessing these with complex statistical methods that allow for testing of both direct and indirect effects. Interestingly, the proportion of irregular dental attendees was higher compared to previous reports of dental attendance in Norway, 44,55 indicating that people who do not use dental health services regularly is not underrepresented.
This can be considered a strength of the study when exploring use of dental services in the Norwegian population.
In conclusion, the findings suggests that, in addition to focusing on reducing socioeconomic inequalities in relation to oral health in the Norwegian population, it is also important to consider how people perceive their own resources (eg financial, psychological, social) as well as their access to dental care in order to support regular dental attendance and potentially, in turn, enhance oral health.

ACK N OWLED G EM ENTS
Thanks to Dr Nils Oscarson and Dr Anders Tillberg who together with the first and second author carried out the TOHNN-study.
Thanks to all dental teams at the public dental clinics for helping with the data collection. The TOHNN study was funded by the Troms County Council and The Norwegian Directorate of Health.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

CO N FLI C T O F I NTE R E S T S
The authors have stated explicitly that there are no conflicts of interest in connection with this article.