Ó 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Community Dent Oral Epidemiol 2015; 43; 560–568 All rights reserved
Relative cost-effectiveness of home visits and telephone contacts in preventing early childhood caries Koh R, Pukallus M, Kularatna S, Gordon LG, Barnett AG, Walsh LJ, Seow WK. Relative cost-effectiveness of home visits and telephone contacts in preventing early childhood caries. Community Dent Oral Epidemiol 2015; 43: 560–568. © 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Abstract – Objectives: To evaluate the cost-effectiveness of a home-visit intervention conducted by oral health therapists relative to a telephone-based alternative and no intervention. Methods: A Markov model was built to combine data on dental caries incidence, dental treatments, quality of life and costs for a cohort of children from age 6 months to 6 years. The probabilities of developing caries and subsequent treatments were derived primarily from the key intervention study. The outcome measures were costs (US dollars), qualityadjusted life years (QALYs) and the number of carious teeth prevented. Oneway and probabilistic sensitivity analyses were used to test the stability of the model. Results: For every group of 100 children, the model predicted that having the home-visit intervention would save $167 032 and telephone contacts $144 709 over 5½ years relative to no intervention (usual care). The home visits and telephone intervention would prevent 113 and 100 carious teeth (per 100 children) relative to no intervention in a period of 5½ years. Sensitivity analysis showed that a lower rate of caries reduced the intervention’s cost-effectiveness primarily through reducing general anaesthesia costs. The home visits and telephone interventions resulted in 7 and 6 QALYs, respectively, gained over the usual care group for the 100 children over 5½ years. Both interventions were ‘dominant,’ as they saved costs and produced health benefits over usual care. Conclusions: Both the home visits and telephone-based community interventions conducted by oral health therapists were highly cost-effective than no intervention in preventing early childhood caries.
Early childhood caries (ECC) is a common problem worldwide, and high prevalence has been noted in the USA (47%) (1) Greece (17%) (2), China (64%) (3) India (75%) (4), England (38%) (5) and Australia (42%) (6). The risks of ECC development are early colonization with cariogenic bacteria, poor oral hygiene and frequent sugary food and drinks consumption (7). ECC is highly prevalent in low socioeconomic and indigenous populations and contributes significantly to toothaches, oral abscesses and hospital admissions (8–11). Children
Rongzhen Koh1, Margaret Pukallus2, Sanjeewa Kularatna3, Louisa G. Gordon3, Adrian G. Barnett4, Laurence J. Walsh1 and Wan Kim Seow1 1 Centre for Paediatric Dentistry, Oral Health Centre, The University of Queensland, Herston, Qld, Australia, 2Oral Health Program (Logan-Beaudesert Division), Metro South Health, Logan City, Qld, Australia, 3Centre for Applied Health Economics, Menzies Health Institute, Queensland, School of Medicine, Griffith University, Meadowbrook, Qld, Australia, 4 Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, Qld, Australia
Key words: cost-effectiveness; early childhood caries; home-visit intervention; Markov model; telephone intervention W. Kim Seow, Centre for Paediatric Dentistry, Oral Health Centre, The University of Queensland, Herston, Qld 4006, Australia Tel.: +61 7 33658061 Fax: +61 7 33658199 e-mail: [email protected]
Submitted 4 December 2014; accepted 3 June 2015
with ECC also have higher caries risk in later childhood, due to persistence of poor oral hygiene habits most likely related to lifestyles and social circumstances (12). Children with ECC often require general anaesthesia for dental treatment due to their young age, thereby significantly increasing the healthcare resources required. Prevention of ECC and the need for general anaesthesia for treatment would thus result in significant cost savings. To assess the best caries prevention options, it is pertinent to
Cost-effectiveness of home visits
assess the cost-effectiveness of ECC prevention programmes. Health economic evaluation can be used to assess health services to ensure there is cost-efficient resource allocation. A health intervention is considered cost-effective when, contrasted with the next best alternative, it produces acceptable costs and health benefits. To date, economic evaluations of dental care services are rare (13). Economic data are now recognized as important due to the fact that dental disease is very common and expensive for health care systems. Hence, there is a need to find cost-effective dental treatments and preventive programmes that are attractive for Governmental and private investment. Although we have previously examined the cost-effectiveness of telephone contacts compared with usual care in a previous report, we did not include the costs and outcomes of a home-visit intervention, utility analysis or out-of-pocket costs to families (14). Patient utilities add the important dimension of quality of life of patients experiencing poor health, whereas previously we looked at the economic costs and number of carious teeth. The aim of this study therefore was to undertake a cost-utility analysis of two ECC prevention methods by comparing their relative costs and outcomes.
Methods Ethical clearance for the study was obtained from the human research ethics committees of The University of Queensland and Queensland Health. This economic study is predominantly based on data estimates from our earlier randomized controlled study on prevention programmes for ECC (14, 15). The methods of the prevention programmes are reported in detail elsewhere (15). All mothers presenting to public maternity health clinics of the Logan–Beaudesert area of Queensland State, Australia, were invited to participate in the study between January 2007 and June 2008 (15). Mothers who provided written informed consent to join the study were randomized to either the home visits or the telephone intervention groups. The interventions were provided when the children were aged 6, 12, 18, 30 and 42 months and clinical assessments were performed at 24, 36, 48 and 60 months. Children who received no intervention comprised the usual care group and resided in the same district. They were school children who were registered in the public school
dental programme to receive regular routine dental care at the community dental clinic.
Description of interventions The home-visit intervention consisted of five 6-monthly home visits by oral health therapists, where they provided dental examinations of the children and dental care instructions to the mothers, for durations of approximately 30 minutes (15). The telephone intervention consisted of five 6monthly telephone calls delivering dental care instructions by the oral health therapists. The telephone calls were between 15 and 20 minutes of duration, and instructions included tooth brushing and dietary advice. The usual care group received no prior dental contact, which is typical of dental services for the age group. The home visits and telephone intervention groups were examined clinically at ages 24, 36, 48 and 60 months by dental practitioners blinded to group allocation (15). The usual care groups were children aged 24 and 60 months who were the reference control children of the study and received dental examinations at the community dental clinic (15).
Markov model A Markov cohort model was developed to represent the three intervention arms and the subsequent events that could occur to children from each arm (Fig. 1) (14). The model was constructed in the software program TREEAGE PRO 2013 (TreeAge Software Inc., Williamstown, MA, USA). This enabled the input of the probabilities of developing caries lesions or not for each arm and allowed for changes over time so that children could develop caries, be successfully treated but then also have repeated episodes, either in the same tooth (different surface), same place (secondary caries) or a different tooth (16, 17). The model cohort starting age was 6 months (the age of first tooth eruption), and children were tracked until they were 6 years old (18). Transition probabilities for caries changed when the children reached 24, 36, 48 and 60 months of age. Children older than 6 years were excluded as we were interested in ECC only. The model had two health states: ‘Caries’ and ‘Healthy’. As newly erupted teeth are healthy, all children started in the healthy state. Children could move between these mutually exclusive health states once every 6 months, or remain in the same health state. According to the district public clinic waiting periods, waiting times for caries treatment can be more than 6 months.
Koh et al.
Fig. 1. The Markov Model outline for the Early Childhood Caries (ECC) intervention pathways.
Following treatment (a tooth extraction), all children returned to the ‘healthy’ state. However, as there is a possibility of a child developing new caries or secondary caries after restoration, the treated child could later return to the caries state. In the model, there was no distinction made between enamel caries, dentinal caries or pulpal involvement and these were combined into the one ‘caries’ health state. The model aggregated the total health costs, incidence of caries and quality-adjusted life years (QALYs) gained depending on the time children spent in each health state (18, 19).
Data and sources Costs. A societal perspective was taken and included costs to the health system and parents. The costs were based on those incurred by both the government and parents. Resources for the homevisit intervention included: vehicle costs (including lease and fuel), telephone appointments, administration costs, toothbrush and toothpaste packs and the staff costs for oral health therapists to provide at-home education. Costs for the telephone intervention included: telephone call costs; toothbrush and toothpaste pack; packing and posting the toothbrush and toothpaste pack; and administrative costs and staff time. Healthcare costs for all children (where applicable) included the following: restorations, restorations with crowns and extractions as well as indirect costs of the parents. Indirect costs consisted of travel, loss of income and privately purchased medications. The indirect costs were collected from a consecutive sample of 100 parents who presented to the community paediatric dental clinic with their children aged 5 years and younger with caries during the months of October and November 2013. The costs of dental care fees were based on the schedule of fees by the Australian Dental Association (20). Treatment costs were provided separately for each number of surface restorations (1–5
surfaces decayed) by the incisor and molar teeth categories, ensuring the complete range of treatment costs was evaluated. The restoration costs were collected separately for each number of surfaces by incisor and molar categories, and a weighted mean cost by surface number and type of tooth was used in the model. In the sensitivity analysis, the model tested the higher cost of stainless steel crowns as a restorative option. The cost of tooth extractions included the cost of general anaesthesia which is usually required for extractions in this young age group. We assumed that some children who waited 18 months or longer for treatment were prescribed antibiotics. The costs of antibiotics and analgesics were included according to the national price schedule for pharmaceuticals (21) for a proportion of the patients (48%) who had acute infections while on the waiting list for treatment. Health utilities and QALYs. Health utilities are similar to quality of life scores and are used in economic evaluations to calculate QALYs, the most widely accepted effectiveness measure, where survival (life years) is adjusted for utility scores. (18) Measuring utilities is more complex in early childhood because utility tools have not yet been developed for this age group. In this study, we used the Child Health Utility (CHU-9D) paediatric quality of life multi-attribute instrument which has been validated and evaluated in the UK for children aged 7–17 years (22). With permission from the developers, we adapted (with minimal rewording) the CHU-9D so that parents could respond as reasonable proxies for their children (23). The CHU9D describes nine dimensions each with five levels of severity; worried, sad, pain, tired, annoyed, school work, sleep, daily routines and activities. In this study, we substituted the question on school work which is not relevant for our cohort, with a question regarding coping with child care. The CHU-9D produces utility scores ranging from 0 to 1.0 with 1.0 being full health and 0 being death.
Cost-effectiveness of home visits
The mean utility value for the ‘caries’ health state was obtained from data provided by the parents of 100 patients with ECC (the same parents who provided their personal cost information).
Transition probabilities The matrix of transition probabilities that determined the movement between health states was based on the proportion of children developing caries and being treated. The probability of caries development was adopted from data obtained from the cohort in the ECC prevention programme (14, 15). At the community dental clinic, all teeth present were examined for dental caries. The decayed, missing and filled teeth (dmft) score was used to assess caries experience. A tooth was classified as decayed (d) if the carious lesion progressed beyond the dentino-enamel junction and involved dentine. Precarious white spot lesions were not included in the dmft scores. At the 5-year visit, the clinical examination included bitewing radiographs. The clinical examinations were completed by 10 calibrated operators who had undergone pre-trial training to ensure standardized examination. The intra- and interexaminer kappa scores were 0.91 and 0.93, respectively, showing excellent agreement. The bitewing radiographs were scored by a single author (RK) who had been previously calibrated for assessment of bitewings. The mean kappa statistic for intra-examiner consistency was 0.91. This examiner was calibrated against another author (WKS), and the interexaminer kappa statistic was 0.89. The treatment probabilities were available from the clinical database of our cohort studies (Table 1) and from an additional random selection of 100 children younger than 6 years of age who presented with caries and were treated from 1 January 2009. These children were included on the basis of having attended the clinics during the same period when the research was conducted. In the model, the rates were converted into probabilities using a rate to probability formula (1 – e rate 9 time) (24). The data were available for the intervention groups after following them until the age of 6 years (24, 36, 48 and 60 months) and the usual care group at 24 and 60 months (Table 1). The model duration was 5½ years when the children reached the mixed dentition age.
Analysis The analysis adhered to consensus modelling guidelines (25–27). Costs and effects were discounted at
5% per year to adjust to present values in line with Australian practice. To ensure that the maximum waiting time for treatment in the model reflected that observed currently in the dental service (18 months), the model was calibrated and the probability of treatment was altered so that all patients received treatment within 18 months. The probability of a new case being treated within 6 months was 0.79 (95% CI 0.76 to 0.82). Incremental cost-effectiveness ratios (ICER) were generated by calculating the incremental costs divided by the incremental effects (number of carious lesions prevented and QALYs gained, separately) for the intervention group (home visits and telephone intervention separately) over the usual care group. We also compared the home visits and telephone interventions. To aid interpretation, the costs and caries outcomes were presented per 100 children. Results are presented in 2014 US dollars (USD) following conversion of 2013 Australian dollars (AUD) at the rate of 1.55 (28). In one-way sensitivity analyses, model inputs were varied across a range of plausible high and low values to assess volatility of the base results. For costs, high and low values were determined from the Australian Dental Association item ranges or tested between 15% (Table 1). For indirect costs and utility of caries, the 10th and 90th percentiles were considered as the low and high values. For all probabilities, the 95% confidence interval limits were used as high and low values. A probabilistic sensitivity analysis was also performed by resampling 1000 times at random from the probability distributions assigned to each parameter. The cost estimates were assigned gamma distributions which accounted for their right skewness. Beta distributions were used for probabilities and utilities as they are bounded between zero and one. Using 1000 Monte Carlo simulations where each simulation randomly selected a value from the assigned distributions, the ICERs were recalculated. These Monte Carlo simulations provided a range of plausible cost and effect pairings and the likelihood of the interventions being cost-effective for different incremental cost per QALY thresholds.
Results Table 1 shows the model inputs and intervention costs. The proportion of parents that indicated an income loss was 17/75 (23%). The highest category of cost was loss of income for parents
Koh et al. Table 1. Model inputs, input description and values Sensitivity values Description
Model starting age for children 6 months Model duration 5.5 years Model cycle length 6 months Discount rate 5% Probabilities Incidence of caries in HV 24 months 0.02 36 months 0.07 48 months 0.15 60 months 0.45 Incidence of caries in TI 24 months 0.04 36 months 0.11 48 months 0.23 60 months 0.42 Incidence of caries in UC 24 months 0.20 60 months 0.60 New patient treated