ADRF RESEARCH REPORT

Australian Dental Journal 2007;52:(4):271-275

Geographic location and indirect costs as a barrier to dental treatment: a patient perspective
B Curtis,* RW Evans,* A Sbaraini, E Schwarz

Abstract Background: The recently published National Survey of Adult Oral Health 200406 indicated that tooth loss, mean decayed and number of DMF teeth were all higher outside capital city locations. In addition, dental attendance patterns were worse in terms of frequency, reason for visit, and continuity in rural and remote locations, but there was no difference by geographical location in terms of financial barriers to dental care. The objective of this research was to identify, quantify and analyse some of the non-treatment costs associated with dental treatment from the perspective of the patient and to determine whether the perceived impact of those costs may limit access to dental care. Methods: This cohort study was nested within a clinical trial. Patients had been allocated to treatment arms within clusters dependent on the randomization status of the dental practice they usually attended, classified as major city, regional or remote. A questionnaire was developed from a series of focus groups in which patients were asked to identify the domains of non-treatment costs associated with a dental visit that were important to them and to quantify those costs. Factor analysis was used to reduce these items to four core scales. These scales were assessed for reliability and validity. Regression and ANCOVA was used to explore differences in DMFS scores between the three groups and a predictive model developed to adjust for potential confounders. Results: Two core scales were identified as key drivers on the perceived impact of indirect costs associated with dental visits; travel impact and family impact. Patients living in remote locations incurred significantly higher indirect costs associated with dental treatment and higher mean DMFS scores. Conclusions: Patient perception of the impact of travel costs and impact on family life are major drivers restricting access to dental services for people living in remote locations in New South Wales. Further research using outcomes directly related to

access is required to validate the claim that patients living in regional and remote locations suffer both perceived and real financial barriers to dental care.
Key words: Dentistry, indirect cost, prevention, patient perspective. Abbreviations and acronyms: ARCPOH = Australian Research Centre for Population Oral Health; DMF = decayed, missing, filled; DMFS = decayed, missing and filled tooth surfaces; GLM = generalized linear model; ICC = intra-class correlation coefficient. (Accepted for publication 18 July 2007.)

*Community Oral Health and Epidemiology, Faculty of Dentistry, The University of Sydney. National Health and Medical Research Council Research Assistant, Faculty of Dentistry, The University of Sydney. Professor and Dean, Faculty of Dentistry, The University of Sydney.
Australian Dental Journal 2007;52:4.

INTRODUCTION Growing interest in economic evaluation has been fuelled by decision makers seeking data that will assist in making resource allocation decisions to achieve efficiency in an environment of increasing healthcare costs. Research has tended to focus on the benefit or effectiveness of treatment, especially indicators based on the patients own assessment of changes to their oral health.1-5 One aspect of the evaluation that is less understood or studied is the ability to define and interpret the nontreatment or indirect costs incurred by the patient in obtaining that treatment. This is particularly relevant in a nation such as Australia, where large geographical areas are classified as remote (using road distance as a proxy for remoteness) and under-serviced (using population size as a proxy for availability of services) in terms of access to dental care.6 It has previously been suggested that those living in remote/rural areas may routinely access services in major city locations, and as a result experience a similar standard of perceived oral health to urban residents.7 Access may depend significantly on factors such as availability of transportation, access to childcare and the opportunity to take time away from work. More recently, little variation was found between perceived treatment needs defined by socio-demographic characteristics.8 This clinical trial is being undertaken to establish the cost and effectiveness of a structured caries prevention
271

programme compared to the standard care provided by a range of major city and regional dental practices in New South Wales (NSW) and the Australian Capital Territory (ACT). The methodology of recruitment and the standardization of practices and techniques have been reported previously.9 Although not a primary end-point of our costeffectiveness study, which is being undertaken from the perspective of the dental practitioner, we have had the opportunity to develop a valuable research relationship with a large cohort of patients and feel that the patient perspective on the cost side of the equation is worthy of further investigation. Accordingly, this study details the development of a questionnaire designed to quantify the non-treatment costs of dental care and how this impacts on perceived access to care from the perspective of the patient. We were particularly interested in investigating whether patients living in rural or remote locations found these costs to be an impediment to access. In addition, we sought to assess the validity and reliability of the questionnaire. MATERIALS AND METHODS Ethics approval for the project was obtained from the Human Research Ethics Committee at The University of Sydney in compliance with guidelines issued by the Declaration of Helsinki. A questionnaire was developed from a series of focus groups (n=20) in which patients were asked to identify the drivers of non-treatment costs associated with a dental visit that were important to them. The number of participants in each focus group ranged between 5 and 10 individuals. Participants were questioned on a range of background demographic information (age, gender, education), more specific dental information (dental insurance status, fluoride history, access to fluoridated water, toothbrushing habits), transportation arrangements for dental appointments (mode, distance travelled, time taken), family (need for spouse or other family member to be present at appointment, childcare needs) and work commitments (ability to take time off, time taken, lost pay). These questions were specified to be relative to the previous dental visit of the patient. The results of the focus groups were used to generate the items, choose the administration method, and the recall period of the questionnaire. Baseline DMFS (number of decayed, missing and filled tooth surfaces as a result of dental caries) was calculated from the patients chart at the recruitment appointment. Accordingly, the questionnaire was designed to be self-administered and mailed to each study participant within one month of recruitment (into the clinical trial). Detailed instructions and a stamped, self-addressed envelope were included in the package. As the overall trial design used a cluster randomization technique, the allocation of each patient to treatment or control group was, in fact, determined by the allocation of the dental practice which they attended.
272

All dental practices that were recruited into the trial fell within the major city or regional classification of remoteness,6 although adjustment was made for those patients who had travelled significant distances to attend the surgery. For the purposes of this analysis, patients were classified according to their residential address rather than the dental practice they attended for treatment. Item discrimination was estimated with the Pearson item-total correlation between the item and its scale score. Item-total correlations are often reported as part of an internal consistency analysis, but are also an aspect of validity as they estimate the degree to which the item is correlated with the overall measure of the construct of interest. A correlation matrix was produced and those variables that did not show significant correlation were dropped from the subsequent factor analysis. Extreme multicollinearity and singularity were also assessed. Factor analysis was undertaken in an attempt to explain the variability within each of the items. Extraction using the principle components method was performed; varimax rotation with the maximum iterations for convergence was set at 30. Eigen values over 1 were extracted and regressed with communalities indicating the amount of variance within each variable accounted for. The scree plot was the primary method for determining the number of factors to extract, but was supplemented with an inspection of the magnitude of the variance explained by each factor. Despite its subjective nature the scree plot performs as well or better in simulation studies than the K1 rule.10 When more than one solution appeared reasonable from the scree plot we inspected several solutions and determined the most valid solution based on the magnitude of factor loadings. It has been suggested that a minimum sample size for effective factor analysis to be 300 subjects with communalities after extraction set at greater than 0.5.11 In this case the Kaiser-Meyer-Olkin measure of sampling adequacy was assessed.12 We imputed missing values (with the mean of available scale items) when at least 70 per cent of the items on the scale were not missing. Using this method, 20 items were factored into four clinically meaningful core scales: impact of travel, impact on family, previous dental history and socio-demographic. The perception of the impact of non-treatment costs on access to care was assessed for each of the core scales identified from the factor analysis using a sevenpoint Likert scale. The values of the scale ranged from 1 no impact at all to 7 a major impact. Patients were also asked to quantify that cost if able. The raw scores were summed and as such the maximum score for the perception of non-treatment cost scale was 28. A higher score represented a greater perceived impact of non-treatment costs on access to care. This draft instrument was pilot-tested on 878 patients and assessed for reliability, validity,
Australian Dental Journal 2007;52:4.

Table 1. Patient characteristics stratified by location

Major city No. subjects (%) 532 (65%) Gender (% = female) 57% Age (SD) 43.5 (19.2) Mode of transportation (% = motor vehicle) 76% Travel time (SD) min 30.1 (31.8) Distance (SD) km 16.0 (25.2) Average number of children (SD) 1.5 (1.7) Childcare required (% = yes) 23.1% Accompanied to visit (% = yes) 22.2% Taken time off work (% = yes) 61% Baseline DMFS (SD) 21.1 (22.9) Insured for dental care (% = yes) 68% Using fluoride toothpaste (% = yes) 91% Frequency of brush (SD) (each day) 1.7 (0.7) Born in Australia (% = yes) 71% Perception of indirect costs on total cost. A higher score indicates greater perceived impact max. = 28 (SD) 13.6 (7.3) Practice location Regional 248 (30%) 63% 36.1 (20.9) Remote 43 (5%) 47% 45.8 (17.4)

96% 100% 30.0 (29.1) 238.1 (115.1) 25.3 (31.9) 291.3 (148.2) 2.5 (1.8) 12.1% 41.1% 61% 13.0 (16.4) 65% 90% 1.7 (0.7) 92% 2.5 (1.5) 49% 97.7% 28% 29.9 (23.0) 47% 86% 1.8 (0.9) 90%

Fig 1. Diagram of the conceptual framework.

13.3 (7.6)

22.8 (7.3)

administrative and respondent burden, and ability to detect a meaningful difference in scores. A random subsample of participants was retested one month later (n=120). A diagram of the conceptual framework is displayed in Fig 1. Internal consistency reliability was estimated with Cronbachs alpha. The primary source of measurement error assessed by coefficient alpha is content sampling error.13,14 The test-retest reliability was estimated using the intra-class correlation coefficient (ICC) estimated from a two-way random effects model in which both subjects and occasions are random.15 This yields an ICC that assesses agreement, the day-today fluctuation within individuals being the primary source of measurement error.14 Due to the longer time interval between applications, we also computed the ICC that measures consistency as the true scores may have changed slightly.15 Demographic characteristics of the participants were analysed using chi-square tests for proportions, t-test for continuous outcomes (or ANOVA for group comparisons), and logistic and linear regression for predictors of outcome. Review of the literature10 and a power calculation assuming a minimum detectable difference of 0.5 (SD=2.0) in the perception scale, undertaken prior to the study, indicated an approximate power to detect a statistically significant difference of 0.86. All subjects were asked to estimate the amount of time required to complete the questionnaire. Data analysis was undertaken using SPSS v15 and SAS v8.2. The alpha for statistical significance was set at 0.05.
Australian Dental Journal 2007;52:4.

RESULTS At the time this study was undertaken 878 patients had been recruited into the clinical trial. In all, 823 fully completed questionnaires were returned (94 per cent response rate). The mean completion time for the questionnaire was 6.9 (3.2) minutes. Less than 20 per cent of respondents provided information on income and that variable was not analysed further. The characteristics of the patient sample stratified by geographic location are displayed (Table 1). There was no significant difference in gender between the groups (p=0.08). The regional group was significantly younger than both the major city (p<0.01) and remote groups (p<0.01), who in turn were not significantly different to each other. A Bonferroni adjustment was performed due to the post hoc nature of this analysis.16 As expected, the major mode of transportation was the motor vehicle and not surprisingly 100 per cent of patients living in remote locations relied on its use. Patients living in remote locations travelled significantly further distances (p<0.001) and incurred significantly more travel time (p<0.001) than their counterparts in regional areas and major cities. In light of this fact and combined with a greater need to be accompanied on such visits by a spouse or significant other, a statistically significantly greater proportion of childcare was required in this group (p<0.001), adding
273

an average of $60.91 (SD $29.10) to the cost of the visit. Surprisingly, in light of the above, and the fact that a higher proportion of subjects within the group were male, a significantly fewer percentage of patients from remote areas required time off work. All geographic group baseline DMFS scores were significantly different to each other: major city*regional (p<0.001); major city*remote (p=0.03); and regional*remote (p<0.001). In an attempt to explain these unexpected variations we constructed a forward stepwise linear regression model. That model identified five confounders within this dataset: age, gender, travel distance, childcare needs and tank water usage. This model accounted for 33 per cent of the variability in the baseline DMFS scores (R2=32.98; all p<0.05), the largest driver being age with a partial R2=0.30. These variables were then added to a generalized linear model (GLM). However, unexplained differences were still apparent between the mean baseline DMFS scores after this analysis of covariance was performed. The adjusted major city score was 19.8 (95% CI 18.321.3), the regional score 16.2 (95% CI 13.918.4) and remote 27.1 (95% CI 21.832.5). Oral hygiene habits (including frequency of brushing, flossing, fluoride toothpaste use, use of fluoride supplements and self-reported previous exposure to fluoride) were homogenous within the group. Factor analysis of the perception of cost questionnaire resulted in 20 items being factored into four clinically meaningful core scales: travel impact, family impact, previous dental history and sociodemographic. All scales showed strong validity and reliability: factor loadings >0.50 and item total correlations >0.65. Two scales (impact of travel and impact on family) demonstrated known-groups validity by distinguishing location groups (Cronbach alpha = 0.94). Test-retest (agreement ICC) was >0.85 for all scales. The two lowest ICCs were the dental history and demographic scales (0.86 and 0.88, respectively). The perception of the impact of indirect costs, as measured by the questionnaire, was significantly higher (p<0.001) in the remote group than both the major city and regional groups, which were not significantly different to each other. DISCUSSION Dental caries is Australias most prevalent health problem, with recent estimates suggesting 11 million people suffering new decay each year.17 A previous survey by the Australian Research Centre for Population Oral Health (ARCPOH) has indicated that rural and remote residents experience a similar standard of oral health to urban residents but are disadvantaged in some aspects of access to dental care.18 More recently, the National Survey of Adult Oral Health 200406 revealed that tooth loss, mean decayed and decayed, missing, filled (DMF) teeth were all higher outside of capital city locations, and that
274

dental attendance pattern was worse (in terms of frequency, continuity and reason for visit), but there was no difference by location in financial barriers to dental care.8 The results of this study would appear to dispute these findings in that we concluded that people living in remote locations within Australia are likely to suffer both a perceived and real financial barrier to receiving dental care due to the indirect costs associated with receiving routine dental treatment. However, there are some important differences in methodology to consider. The ARCPOH data collection was administered via a telephone survey with a much larger sample size than our study, and thereby was probably able to capture a more generalizable population than would be found within a clinical trial. Our sampling provides a very limited view of access, since all subjects were patients. There is no representation of those who have not accessed care, which would presumably include those persons with the most severe access problems. As such, no conventional outcome measure related to access (e.g., use of services) is presented. It could reasonably be argued that our results may in fact underestimate the perceived and real indirect financial barriers to access in remote locations. Whereas we assessed dental health using a clinical measure (DMFS), the ARCPOH study utilized selfreported dental health status in which respondents were asked to rate their dental health using a six-point scale ranging from excellent to very poor. While a patient-reported approach is more likely to take account of the functional and social impact of oral disease and its treatment, a possible complication of incorporating internal (patient reported) views lies in the fact that those views are likely to be influenced by social experience.19 For example, patients living in rural or remote areas may not be aware of treatment modalities available more widely in major cities, and would possibly view their dental health differently if placed in a different social situation. A more likely explanation is that patient expectation of what treatment is realistically available outside major urban centres, or a more pragmatic view of appropriate dental care is being recorded. Despite this, the inclusion of such instruments within clinical trials is seen to be important as some treatment costs or effects are known only to the patient. In addition, there may be a desire to know the patient perspective about the cost or effectiveness of a treatment, and a systematic assessment of the patients perspective may provide valuable information that can be lost when that perspective is filtered through a clinicians evaluation of the patients response. The main weakness of our study was an inability to accurately score DMFS at the baseline visit due to the unreliable charting methods of the dental practitioners involved in the study. Only 30 per cent of the practices involved in the study routinely performed a full mouth charting at each recall visit. Twenty per cent of patient
Australian Dental Journal 2007;52:4.

files indicated that a full charting had never been undertaken. Practices that utilized a central computerbased record system were more likely to have undertaken at least one full charting procedure, and twice as likely to have completed a full charting at each recall visit. Where possible we extrapolated data from previous visits and radiographs; this was found to be a less than satisfactory method. CONCLUSIONS It is clear from this research that patients do consider indirect costs to be an important factor in the overall cost of dental treatment, and that these costs are more of a burden in remote areas, primarily due to transportation requirements and impact on family life. The instrument developed to measure the patient perception of these costs demonstrated acceptable reliability and validity, although further development is planned within a more generalizable population. While this research supports the finding that patients living in remote locations showed higher rates of DMFS due to dental decay, we would question the claim that there is no difference by geographic location in terms of financial barriers to dental care, both perceived and real. It will be important to include reliable and valid measures of outcomes related to access in future studies to corroborate these results. ACKNOWLEDGEMENTS This research has received funding from the Oral Health Foundation, the National Health and Medical Research Council (project grant 402466), the Dental Board of NSW and the Australian Dental Research Foundation. The support of the Australian Dental Association (NSW), Colgate, Australian Healthcare Management and GC (Australia) is gratefully acknowledged. REFERENCES
1. Rosenberg D, Kaplan S, Senie R, Badner V. Relationships among dental functional status, clinical dental measures, and generic health measures. J Dent Educ 1988;52:653-657. 2. Cushing AM, Sheiham A, Maizels J. Developing socio-dental indicators the social impact of disease. Community Dent Health 1986;3:3-17.

3. Locker D. Measuring oral health: a conceptual framework. Community Dent Health 1988;5:3-18. 4. Slade GD, Spencer JA. Development and evaluation of the Oral Health Impact Profile. Community Dent Health 1994;11:3-11. 5. Slade GD. Derivation and validation of a short-form oral health impact profile. Community Dent Oral Epidemiol 1997;25:284290. 6. Australian Institute of Health and Welfare. Rural, regional and remote health: a guide to remoteness classifications. AIHW Cat No. PHE 53. Canberra: AIHW, 2004. 7. Australian Research Centre for Population Oral Health. Geographic distribution of the dentist labour force. Aust Dent J 2005;50:119-122. 8. Australian Institute of Health and Welfare. Australias dental generations. The National Survey of Adult Oral Health 200406. AIHW Cat No. DEN 165. Canberra: AIHW, 2007. 9. Curtis BH, Evans W, Sbaraini A, Schwarz E. Recruitment and standardization of a group of Australian dentists for a multipractice study on dental caries prevention. Aust Dent J 2007;52:106-111. 10. Zwick WR, Velicer WF. Comparison of five rules for determining the number of factors to retain. Psychological Bulletin 1986;99:432-442. 11. Field AP. In: Discovering statistics using SPSS. 2nd edn. London: Sage, 2005:619-680. 12. Kaiser AB, Hennekens CH, Saslaw MS, Hayes PS, Bennett JV. Seroepidemiology and chemoprophylaxis disease due to sulfonamide-resistant Neisseria menigitidis in a civilian population. J Infect Dis 1974;130:217-224. 13. Feldt LS, Brennan RL. Reliability. In: Linn RL, ed. Educational measurement. 3rd edn. New York: Macmillan, 1989:105-146. 14. Nunnally J. Psychometric theory. New York: McGraw-Hill, 1994. 15. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychological Bulletin 1979;86:420-428. 16. Feise RJ. Do multiple outcome measures require p-value adjustment? BMC Med Res Methodol 2002;2:8. 17. Brennan DS, Spencer AJ. Oral health trends among adult public patients. AIHW Cat No. DEN 127. Canberra: AIHW, 2004. 18. Australian Research Centre for Population Oral Health. Oral health and access to dental care in Australia Comparisons by cardholder status and geographic location. Aust Dent J 2005;50:282-285. 19. Sen A. Health: Perception versus observation. BMJ 2002;324:860-861.

Address for correspondence/reprints: Dr Bradley Curtis 13511 Brentwood Lane Carmel, Indiana 46033 United States of America Email: bradcurtis13@msn.com

Australian Dental Journal 2007;52:4.

275