The impact of changing dental needs on cost savings from fluoridation

Authors



Dr Denise L Bailey
Melbourne Dental School
The University of Melbourne
720 Swanston Street
Melbourne VIC 3010
Email: dlbailey@unimelb.edu.au

Abstract

Background:  Although community water fluoridation has been one of the cornerstone strategies for the prevention and control of dental caries, questions are still raised regarding its cost-effectiveness. This study assessed the impact of changing dental needs on the cost savings from community water fluoridation in Australia.

Methods:  Net costs were estimated as Costs(programme) minus Costs(averted caries). Averted costs were estimated as the product of caries increment in non-fluoridated community, effectiveness of fluoridation and the cost of a carious surface. Modelling considered four age-cohorts: 6–20, 21–45, 46–65 and 66+ years and three time points 1970s, 1980s, and 1990s. Cost of a carious surface was estimated by conventional and complex methods. Real discount rates (4, 7 (base) and 10%) were utilized.

Results:  With base-case assumptions, the average annual cost savings/person, using Australian dollars at the 2005 level, ranged from $56.41 (1970s) to $17.75 (1990s) (conventional method) and from $249.45 (1970s) to $69.86 (1990s) (complex method). Under worst-case assumptions fluoridation remained cost-effective with cost savings ranging from $24.15 (1970s) to $3.87 (1990s) (conventional method) and $107.85 (1970s) and $24.53 (1990s) (complex method). For 66+ years cohort (1990s) fluoridation did not show a cost saving, but costs/person were marginal.

Conclusions:  Community water fluoridation remains a cost-effective preventive measure in Australia.

Abbreviations and acronyms:
CWF

community water fluoridation

NOHS

National Oral Health Survey

Introduction

In Australia, as in many other countries, community water fluoridation (CWF) has been one of the cornerstone strategies for the prevention and control of dental caries. CWF was first introduced in Australia in 1953 in Beaconsfield, Tasmania and into Melbourne, Victoria in 1977. CWF has been identified as an important area of public health policy in Australia.1 Internationally it has been recognized as one of the 10 great public health achievements of the last century.2 Despite this, questions are still raised regarding its cost-effectiveness.

Economic evaluation is commonly adopted by decision makers in the health sector to investigate the effectiveness of public health programmes and to help plan future initiatives. An economic evaluation is “the comparative analysis of alternative courses of action in terms of both their costs and consequences in order to assist policy decisions”.3 In the case of CWF, economic analysis to help guide decisions is inherently difficult, largely because it makes demands on epidemiological and demographic data that are hard to meet.

In addition, decisions regarding CWF have often been based on dated epidemiological evidence, international data and have also tended to concentrate on children and young adults, making relatively simple extrapolations to older cohorts.1 However, environments and health needs do not remain constant over time.4 There is a substantial body of literature on the economic value of CWF and the findings of more recent investigations5–7 continue to support its continued application. However, to date there have been few studies that focus particularly on how the temporal decline in dental caries in conjunction with greater rates of tooth retention have affected the economics of CWF in an ageing population. The aims of this study were to assess the impact that changing community oral health has on the cost savings that can be achieved from CWF and to specify an economic model relating oral health for a given effectiveness of CWF.

Methods

The form of economic evaluation used in this study was cost-benefit or averted costs analysis8 and was conducted from a societal perspective. The outcome of this analysis, expressed as net costs or savings, is derived from the difference between the programme costs and benefits rather than a ratio of costs to benefits as in a true cost-benefit analysis. The basic formula can be written as:

image

Programme costs

For this evaluation, which considers both the health gains and quantifiable deficits arising from CWF, programme costs were considered in terms of: (1) direct costs – the costs associated with fluoridating a reticulated water supply; and (2) indirect or consequent costs – in this preliminary model, these were restricted to increases in periodontal treatment costs (both prevention and therapy) arising from the decline in tooth loss (and hence an increase in the number of retained teeth) that can be directly attributed to CWF.

So:

image

Averted costs

Specification of the averted costs followed the method proposed by Griffin and more recently employed by O’Connell.5,7 This method allows modelling of caries incidence for a given effectiveness of CWF. In brief:

image

Substituting the above equations into the basic formula gives the formula for this analysis:

image

The net cost outcome is interpreted as the annual person-cost resulting from a one-year exposure to CWF. A negative cost implies cost savings (i.e., the averted costs are greater than the programme costs).

The economic model

To better understand how the net costs from CWF differ across age-cohorts and how this pattern has changed over time, the economic model considered three time points: 1970s, 1980s, 1990s; and four age-cohorts: children–teens (6–20 year-olds); young adult–middle age (21–45 year-olds); older middle age–young elderly (46–65 year-olds); and mid to old elderly (66+ years). Thus, a maximum of 12 (four cohorts by three decades) base estimates were obtained or constructed for the parameters required in the computation of net costs. These parameters, and their methods of estimation, are outlined below. All costs and savings referred to in this paper are calculated at the level of 2005 Australian dollars.

Estimation of parameters

The aims of the study required that changes in the parameters of the model be differentiated between cohorts and over time. This is generally expressed in the text as “x parameter in cohort c at time t”. In some instances the parameter is varied in the model in only one dimension, “c” or “t”, because of either the small variation, the complexity it would add to the model’s computations, or the difficulty obtaining data.

Programme cost parameters – direct costs: cost of CWF

Estimates of CWF costs in 2005, both fixed and ongoing, were obtained from Melbourne Water (http://www.melbournewater.com.au). Costing followed standard methodology in this type of evaluation.3 One-time fixed costs included equipment, installation, testing and safety equipment, and consultant engineering fees. Annual operating costs included consumables, labour and maintenance. The population of Melbourne receiving fluoridated water in 2005 was calculated from Australian Bureau of Statistics (ABS) annual regional population figures. From these calculations, a person fluoridation cost of A$0.27 was utilized in the model.

Programme cost parameters – consequent costs: cost of periodontal services

The consequent costs were considered as periodontal treatment costs arising in the additional number of dentate people that occurred as a direct consequence of CWF. Treatment of periodontal disease was included in this model as it is the fifth most prevalent health problem in Australia9 and, although other oral diseases may affect this additional number of dentate people it is beyond the scope of this research to model their effects.

The 1970s were taken as the base (zero consequent costs) and periodontal costs calculated from the product of excess number of dentate relative to the 1970s. For subsequent time points (1980s and 1990s), the excess number of dentate attributable to CWF was calculated according to the following formula:

image

where:

XDc,t = the excess no. of dentate attributable to CWF for cohort c at time t

D = number of dentate people

c = model cohorts

t = model time points 1980s and 1990s

T = model time point 1970s

XEc = differential in caries-related edentulous rates in fluoridated and non-fluoridated communities for cohort c.

The number of dentate people in cohort c at time t was estimated by subtracting the proportion of edentulous people from the total number of dentate people.10,11 The differential in caries-related edentulous rates between fluoridated and non-fluoridated communities for cohort c was obtained from the National Oral Health Survey (NOHS) 1987–8812 (this study was conceived and executed prior to the publication of the National Survey of Adult Oral Health 2004–06).13

The number of people in cohort c at time t requiring periodontal treatment N (Pc,t,N) was calculated as:

image

where:
XDc,t = the excess dentate attributable to CWF for cohort c at time t

Nc = proportion of people in cohort c requiring periodontal treatment N

The proportion of people in cohort c requiring periodontal treatment (either oral hygiene instruction, scaling and oral hygiene instruction or complex treatment) was derived using data from the NOHS 1987–88.12 A per person periodontal treatment cost (Cp) for cohort c at time t was calculated by summing the costs for the individual N strata within cohort c at time t, and dividing by the total number of dentate in cohort c at time t, as follows:

image

where:
Pc,t,N = number of people in cohort c at time t requiring periodontal treatment N

CN = cost of periodontal treatment N (i = OHI; ii = scaling and OHI; iii = complex periodontal treatment)

Dc,t = number of dentate people in cohort c at time t.

Fees for individual periodontal treatments at time t were estimated from the biennial Australian Dental Association (ADA) Survey of Dental Fees (available from http://www.ada.org.au). These one-off costs were multiplied by the average number of periodontal services per patient per year for cohort c.14

Averted cost or benefit parameters

Effectiveness of CWF

For this model, effectiveness of CWF was differentiated over time (for which there is strong evidence of a decline),15 but not over age-cohorts. While there is some suggestion that CWF delays the onset of dental caries rather than prevents it absolutely16-18– and, by implication, the effectiveness of CWF decreases with age, the evidence is still scant and yet to be widely accepted as a mechanism of action of CWF.

Estimates of the effectiveness of CWF were obtained from the literature.1,15,19–24 For each time point, an approximate mid-point of the reported range of effectiveness was accepted as the base estimate in the model, with the upper and lower levels used as the bounds of the range.

  •  1970s – reported range: 35 to 60%; 50% accepted
  •  1980s – reported range: 15 to 40%; 30% accepted
  •  1990s – reported range: 18 to 29%; 25% accepted

Annual caries increment

Obtaining estimates of annual caries increment in non-fluoridated communities for each age-cohort in the 1970s, 1980s and 1990s was approached through the construction of a two-dimensional (time-age) oral health status matrix. A search of the literature was undertaken to obtain published studies with DMFT/S data in any year between 1970 and 1999. The DMFT index is the sum of permanent teeth (T) that are decayed (D), missing (M), or filled (F) due to dental caries. The DMFS index is the tooth surface equivalent. A primary dataset was established from this and the NOHS.12 When multiple data sources were available to populate any one matrix cell, preference was given to studies reporting oral health status suitably broken down by age and fluoridation status. Where stratification by age within the published source did not coincide with the matrix, the data were entered to “best-fit” the required age-cohorts in the model.

This process populated the matrix for all but the 46–65 and 66+ cohorts in the 1970s. International sources25–27 were included to provide estimates for these cells. Table 1 provides details of the final oral health status matrix.25–47

Table 1.   Oral health matrix
YearAge range
06–1011–1516–2021–2526–3031–3536–4041–4546–5051–5556–6061–6566–7071–7576+
19703.58.8             
1972  12.6            
19732.36    21.121.1       
1976       2222  24.524.52626
19772.86.2             
19821.74.510.910.915.317.317.317.3       
19850.82.88816.716.7         
19870.52.25.3813.515.417.717.719.519.52020212121
1988  4.36.98.913.2     21.521.526.426.4
19900.131.2             
1991           22.822.823.924.3
19920.170.499916.516.516.516.521.521.521.521.5232323
19930.41              
19950.416.96.910.310.312.512.514.514.514.514.516.616.616.6
1996             24.924.9
19970.410.953.594.637.079.04         
1998           19.819.819.825.3
19990.411.17 3.66      18.318.319.519.5 

Annual increments from the matrix were achieved by taking the difference between the average of the DMF estimates across the decade at both the low and high end of an age-cohort and dividing this by the time over which the increment was calculated. Where the matrix showed no increment over the age-cohort (because the source data gave only an average DMF for that age range), the method of increment calculation was modified to allow the value from a neighbouring age-cohort to represent the low or high end estimate. The DMFT increments were subsequently converted to surface (DMFS) increments using conversion factors, for the final DMFS estimates. Conversion factors were sourced from the published literature. Where conversion factors were unavailable, a series of imputations was used. Published data provided the conversion factors for the 21–45 and 46–65 age-cohorts in the 1970s and 1980s,48 the 21–45 age-cohort in the 1990s44,49 and the 6–20 age-cohort in the 1990s.50 Around these figures, the remaining factors were imputed under a set of assumptions:

  •  The decline in the DMFS to DMFT ratio over time which was apparent from the 21–45 age-cohort would be observed in the 6–20 cohort.
  •  The plateauing of the conversion factor observed in the 46–65 age-cohort from the 1970s to 1980s would continue into the 1990s.
  •  The conversion factors in the 66+ years cohort would be the same as the 46–65 cohort across all time points.

Cost of a carious surface

The cost of a carious surface took into account both the direct dental restorative costs and the costs arising from lost productivity. Productivity losses result from the loss in work time due to attending a restorative visit.

To estimate the direct costs associated with restoring a carious surface, conventional and complex approaches were adopted. The conventional approach is a simple construct, utilizing the life and cost of a 1-surface restoration (i.e., amalgam). For the complex method proposed here, a service-mix approach was adopted. Elements required for the calculation were: (1) the cost of restoring a carious lesion in cohort c at time t; (2) both the life of an average restorative mix in cohort c at time t and the average life expectancy of the population at time t to estimate the number of replacements required over the remaining lifetime of the cohort; and (3) since restorations can only be placed in dentate, the proportion dentate in cohort c at time t.

Restorative services considered in the mix were restricted to 1, 2, and 3 or more surface-amalgams, composites and silicates/glass ionomers, as well as cast crowns. Data of this type, broken down by age-cohort, were available from the AIHW DSRU’s longitudinal study of dentist practice activity51 for 1983, 1988 and 1995 and used as representative of the 1970s, 1980s and 1990s, respectively. For each cohort, the total number of services at time t was calculated and used as the denominator in the conversion to a service percentage.

The cost of an average service for cohort c at time t was estimated by “weighting” the cost of the individual items of service by the service percentages and summing across the cohort. The cost of each restorative item was again sourced from the ADA Survey of Dental Fees. For the 1970s, costs were unavailable for silicates and 2- and 3-surface composites. Composite costs were imputed by increasing the 2- and 3-surface amalgam costs by the same factor (9%) observed across the 1-surface composite and amalgam costs. The cost of silicates was imputed from 1980s data where silicates were approximately 92% of the cost of composites.

The life of an average service was calculated from estimates from the published literature of restorative services in the chosen service-mix, taking the mid-point of the identified ranges and applying the weighting procedure as in the estimate of costs. For reasons of increasing complexity of the modelling, differentiation of the life of a service was restricted to the temporal dimension. The number of replacements of an average service was calculated by dividing the number of years remaining to a cohort by the life of an average service at time t. Average life expectancies were obtained from Australian Life Tables (Financial Demographics Pty. Ltd; http://www.findem.com.au) and assumed to be the same for all cohorts at time t.

The final step in the calculation of this parameter involved weighting for the number of dentate individuals and discounting. Discounting was performed using a base rate of 7% as recommended in the June 1997 New South Wales Treasury Policy and Guidelines Paper TPP97-2 (available from http://www.treasury.nsw.gov.au). It is recommended that 7% is adopted as the central real discount rate to allow comparison of analyses performed by different agencies, with sensitivity analyses performed at discount rates of 4% and 10%. The proportion of dentate people in cohort c, at time t, was estimated by subtracting the proportion of edentulous people from the total number of dentate people.10,11

The costs of lost productivity were estimated using data from wave 1 of the 2001–2002 Household, Income and Labour Dynamics in Australia (HILDA) survey conducted by the Melbourne Institute for Applied Economic and Social Research as reported by Rodgers.52 Average hourly income data took into account both part- and full-time employed rates and pensioner rates and was estimated to be $10, $19, $19 and $10 respectively, across the four cohorts, adjusting for 2005 Australian dollar values. Differentiation within the model was made across cohorts, but not over time. Productivity losses for the 6–20, 21–45, 46–65, and 66+ cohorts were calculated.

Final averted costs were quantified as the product of these three parameters: effectiveness of CWF, annual caries increment, and the cost of a carious surface.

Sensitivity analyses

A one-way sensitivity analysis was undertaken to test the robustness of the results to estimated parameter values.3 This provided a plausible range of net costs and savings that could be achieved from CWF. For this analysis the following assumptions were varied: (1) previously published upper and lower levels of estimated effectiveness of CWF for each time point; (2) social discounting at 4% (worst case) and 10% (best case); (3) caries increment in a non-fluoridated community varied at +/− 10%; (4) restorative costs varied at +/− 10% of estimated costs.

Results

The results are presented using conventional methods for base, best and worst-case scenarios. This is followed by a comparison of conventional and complex methods. Summary data are presented (Fig 1 and Table 2); additional data are available from the authors on request.

Figure 1.

 Net costs by time and age cohort (scale of y-axis differs between age cohorts).

Table 2.   Comparison of conventional and complex methods, by year and age cohort; base and worst case
  Net costs ($)
Base caseWorst case
6–2021–4546–6566+Total6–2021–4546–6566+Total
1970sconv−26.80−33.60−6.09−1.61−56.41−12.30−15.84−3.10−0.90−24.15
comp−81.77−162.66−39.11−10.51−249.45−31.65−69.64−18.40−5.42−107.85
1980sconv−8.20−22.21−2.75−0.08−26.45−12.57−6.17−0.430.44−5.45
comp−24.24−105.68−17.63−4.46−121.34−6.65−31.22−5.33−1.15−31.22
1990sconv−6.44−12.41−2.080.52−17.75−3.01−4.28−0.061.01−3.87
comp−14.80−48.58−12.802.33−69.86−6.01−20.06−5.26−0.45−24.53

Figure 1 and Table 2 summarize the temporal trends in net costs from CWF for each age-cohort under base, worst and best-case assumptions. For 6–20-year-olds, CWF demonstrated cost savings at each time point with savings highest in the 1970s and declining over time. Under base-case assumptions, savings ranged from $26.80 in the 1970s to $6.44 in the 1990s. Under worst-case assumptions, savings were still demonstrable –$12.30 in the 1970s, and plateauing in 1980s and 1990s to approximately $3. Under best-case assumptions, cost savings ranged as high as $55.52 in the 1970s, $19.09 in the 1980s and $12.98 in the 1990s. A similar pattern is observed for both the 21–45 and 46–65 age-cohorts. For the 21–45 cohort, base-case assumptions savings ranged from $33.60, $22.21 and $12.41 in 1970, 1980 and 1990, respectively. The 46–65 age-cohort also demonstrated cost savings at each time point, albeit decreasing over time ($6.09, $2.75 and $2.08 in 1970, 1980 and 1990, respectively).

For the 66+ years cohort, cost savings were demonstrated at all time points under best-case assumptions. Under base-case assumptions cost savings were demonstrated in both the 1970s ($1.61) and 1980s ($0.08), but by the 1990s there was a small but marginal cost of $0.52. Similarly, under worst-case assumptions small but marginal costs were demonstrable in both the 1980s ($0.44) and 1990s ($1.01).

Figure 2 brings this information together to compare the net costs across age-cohorts under base-case assumptions. Totals, and how they are broken down across the age-cohorts, are charted. Cost savings ranged from $56.41 in the 1970s to $17.75 in the 1990s. At all time points, the greatest savings were observed in the 21–45 age-cohort, followed by the 6–20 age-cohort. The decrease in savings over time for the 21–45 age-cohort was reasonably even. However, for the 6–20 age-cohort a much greater decline was observed between 1970s and 1980s, followed by a plateauing in the 1990s. By contrast, smaller cost savings and smaller variations over time are observed in the two older cohorts: 46–65 years and 66+ years.

Figure 2.

 Net costs, by age cohort and time (base case).

Table 2 compares the net costs resulting from the conventional method with those of the complex method under base and worst-case assumptions and at each time period. At all time points and age-cohorts cost savings resulting from the complex approach were substantially higher than the conventional approach. In contrast to the conventional method which produced marginal costs for the 66+ years cohort in the 1980s (worst-case) and in the 1990s (both base and worst-case) at no point using the complex approach and under either base or worse-case assumptions did CWF not show a cost saving.

Discussion

This is the first study to consider how changing dental health status and needs affect the economics of CWF in an ageing population. The study’s ability to take into consideration the effects of an ageing population, lower rates of edentulism, and consequent higher rates of periodontal treatment need was one of its most innovative features.

The results of this economic model of water fluoridation suggest that, even in an era of declining levels of dental decay and supplementary sources of fluoride, CWF should continue to be supported. However, at a planning and policy level, consideration will need to be given to preventive strategies that contain costs of treatment arising out of greater tooth retention in an ageing population. Results showed annual per person savings ranging from just over A$56.41 in the 1970s to just under A$18 in the 1990s, using the conventional method; under the comprehensive method these values ranged from A$249.45 in the 1970s to A$69.86 in the 1990s.

However, at each of the time points analysed the cost savings from CWF declined with age, largely as a result of a plateau in the amount of new tooth decay experienced, lower averted costs of decayed surfaces experienced later in life and estimates of higher periodontal treatment needs compared with those of younger age groups.

Furthermore, sensitivity analyses found the net costs of CWF to be sensitive to the following factors: discount rate, effectiveness of CWF, caries increment, and the cost of a restoration/average service. When a worst case analysis was conducted, CWF remained cost-effective. Only for the 66 years and older cohort in the 1990s did CWF not show a cost savings and the costs per person in this cohort were marginal.

Although appropriate data sources were available for the majority of assumptions required by the analysis, an analysis of this nature is inherently difficult, largely because of the demands it places on epidemiological data. In the current study, estimates of dental health status in a non-fluoridated population were required for four age-cohorts at three time points. As such, several methodological issues warrant discussion. For example, in several instances, the data available failed to meet this requirement and the matrix was populated with a mixed fluoridation history DMF estimate. Inclusion of DMF estimates would lower the estimated increment and therefore underestimate any cost saving demonstrable from CWF. Future studies are required to avoid these estimations. Also, even though a new method for the estimation of the cost of a carious surface has been proposed here, future studies are indicated using more refined methods for the estimation of the cost of carious surfaces.

Additionally, it was central to this investigation to construct a model that allowed for variability in the parameters so that age-cohorts and time periods could be appropriately differentiated. While this was achievable for the parameters in the main, the cost per person of water fluoridation was included as a constant in the model over time because of the difficulty in obtaining estimates for the earlier time periods. Since the direction of any possible difference was unknown – that is whether fluoridating domestic water supplies is less or more expensive now than when it was first introduced – it was decided not to make any assumptions regarding this parameter.

Despite these limitations, the data presented here provide additional information to assist in making evidence-based decisions on the continued application or extension of CWF programmes. Results indicate that, for situations equivalent to those prevailing in Australia, even in an era of declining levels of dental decay and supplementary sources of fluoride, when averaged over the whole community population, CWF continues to be a cost-effective preventive measure. However, as the trend towards greater tooth retention in an ageing population continues, the prospect that the cost-saving benefits of CWF might be offset to some degree by the potentially high costs of periodontal treatment needs later in life emerges. This highlights the need for public dental health professionals and policy-makers to begin to give consideration to community-based strategies which will be effective in containing these costs.

Acknowledgements

Adele Campain, the lead researcher for this project, passed away in May 2008. The authors wish to acknowledge the help and support of Adele’s family in the preparation of this manuscript.

The authors would also like to acknowledge the assistance of staff from Dental Health Services Victoria. This research was supported by funding from NHMRC Strategic Research Development Grant #219207.

Ancillary