Estimating the economic costs of skin cancer in New South Wales, Australia

Background Skin cancer is one of the most common cancers in the world. The increased incidence of skin cancer, combined with limited health care resources and tight budgetary conditions, has increased the importance of understanding the economic impact of skin cancer. This research estimates the economic cost of skin cancer in the Australian state of New South Wales. Method An incidence based approach is used to estimate lifetime costs of skin cancer. Both direct and indirect costs are considered - direct costs include resources associated with the management of skin cancer and indirect costs refer to productivity costs associated with morbidity and premature mortality. Diagnosis of skin cancer was determined according to ICD-10 codes using principal diagnosis. Linked administrative data and regression modelling are used to calculate costs; presented as Australian dollars for the year 2010. The human capital approach is used to value present and future productivity losses. Results The lifetime cost of the 150,000 incident cases of skin cancer diagnosed in NSW in 2010 is estimated at $536 million ($44,796 per melanoma and $2459 per non-melanoma). Direct costs accounted for 72 % of costs ($10,230 per melanoma and $2336 per non-melanoma) and indirect costs accounted for 28 % of costs ($34,567 per melanoma and $123 per non-melanoma). Direct costs are, on average, higher for females than males with indirect costs, on average, higher for males than females. Conclusion This research provides new evidence on the economic cost of skin cancer and provides policy makers with information of the potential monetary savings that may arise from efforts to reduce the incidence of skin cancer.


Background
Skin cancer is one of the most common cancers in the world [1][2][3][4]. The majority of skin cancers develop from exposure to ultraviolet radiation (UVR), particularly from sun exposure. The highest incidence rates of skin cancer worldwide are in Australia and New Zealand, where two out of every three people are likely to be diagnosed in their lifetime [5].
Basal cell carcinoma (BCC) is the most common form of skin cancer followed by squamous cell carcinoma (SCC) [1,2]. Together, BCC and SCC make up the majority of non-melanoma skin cancers (NMSC). Malignant melanoma accounts for less than five percent of skin cancer cases, yet it represents the vast majority of skin cancer deaths in Australia [6]. The incidence of skin cancer is increasing in Australia, and the incidence rate is greater than breast, prostate, lung and colon cancers combined. In terms of prevalence, more people have been diagnosed with skin cancer than all other cancers combined over the past three decades [7].
The management of skin cancer generally involves diagnosis, treatment and follow-up. Melanoma, SCC and BCC are typically detected opportunistically during specific skin examinations by skin cancer specialists or dermatologists, or during general health checks by a general practitioner. In Australia, the standard treatment for primary melanoma is wide local excision of the skin and subcutaneous tissues around the melanoma. The aim is complete surgical excision of all in situ and invasive melanoma components, confirmed by comprehensive histological examination [8]. Surgery is the prime treatment for NMSC: more than 70 % of the BCC lesions recorded in the 2002 National survey were surgically excised [9]. For BCCs not surgically excised, cryotherapy was more commonly used for upper and lower limb lesions than facial lesions, and 10 % of BCCs were treated with curettage and diathermy. Anecdotal evidence suggests that nonsurgical treatment has increased since 2002 for superficial BCC with imiquimod in particular. The majority of SCC lesions, regardless of body site, were treated by surgical excision [9,10]. Confirmation of complete removal of lesions is an essential part of management [10].
The post-treatment follow-up regimen is relatively intensive in Australia with clinical guidelines recommending post melanoma treatment follow-up visits every sixmonths for five years for patients with stage I disease, three-monthly or four-monthly for five years for patients with stage II or III disease, and yearly thereafter for all patients [8]. NMSC recommendations for follow-up have yet to be established for the detection of further primary tumours, however, some suggestions have been made that six-monthly follow-up for two years may assist in early detection of new primary tumours or of metastatic disease [10].
The incidence of skin cancer increases with age [11,12]. With most western countries experiencing a demographic transition towards an older cohort, including Australia, the incidence and prevalence of skin cancer is rising, along with the consequential economic impact, even though incidence in younger age groups (i.e., less than 55 years) is stabilising [12][13][14].
The economic impact of skin cancer can be considered as a combination of direct and indirect costs. Direct costs include the management of skin cancer from diagnosis, treatment to follow-up, and refer to the utilisation of health care resources such as hospital, medical and allied health care services. Indirect costs reflect the lost productivity resulting from an individual's inability to work (morbidity costs such as sick leave and early retirement) and premature mortality (defined as death before the age of 65 years, the upper limit of the working age in Australia).
The increased incidence and prevalence of skin cancers, combined with the current fiscal environment of limited health care resources and tight budgetary conditions, has increased the importance of understanding the economic impact of skin cancer. From a policy viewpoint, it is important to understand the resource requirements of the current skin cancer burden and the efficiency of competing strategies that are most likely to lower the incidence of skin cancer and, therefore, reduce its burden. Although several international studies have examined the direct cost of skin cancer treatment [15][16][17][18][19][20][21][22][23][24][25][26], and two have examined indirect costs [27,28] The authors did not report incidence rates or average cost per case, but noted that costs had risen by 14 % after adjusting for inflation since 1995 [32]. O'Dea estimated the cost of skin cancer in New Zealand in 2007 at $NZ123.10 million, equivalent to $NZ1,785 per case (AUD$1650 in 2010 prices) [30]. The general trend in these country specific studies is higher direct costs for NMSC (because of their relatively high prevalence) and higher indirect costs for melanoma (because of its relatively greater severity of illness in those of working age). Although the proportion of costs attributable to indirect costs in all three studies to date appears comparable at around 54 %, total costs appear to vary widely partly due to methodological differences.
New South Wales (NSW) is Australia's most populated state with an estimated population of 7.29 million (34.5 % of the population of Australia). One of the key objectives of the skin cancer prevention strategy for NSW (2012)(2013)(2014)(2015) is to increase and utilise evidence to inform future planning and development of skin cancer prevention strategies with a priority on: increasing the adoption of UVR protection behaviours; increasing shade provision; and improving polices to increase protection from UVR across a range of settings and life stages [33]. Although an understanding of the magnitude of the economic burden of skin cancer in NSW is important given its high incidence, there is currently a clear lack of such evidence. The objective of this study is to estimate the economic cost (both direct and indirect) of skin cancer in NSW. A lifetime approach is adopted, which estimates costs over the management of skin cancer through diagnosis, treatment and follow-up.

Ethics
Ethics clearance was obtained from the New South Wales Ministry of Health (2012/09/417).

Economic approach
Economic costs may be estimated using either the prevalence or incidence-based approach to costing. A prevalence-based approach provides estimates of costs for the total population for one year, or costs accumulated over a longer time horizon [34]. An incidencebased approach follows a disease cohort for the duration of the disease and estimates discounted costs [34], it is the most commonly used method as it allows policy makers to understand the potential impact of reducing the incidence of a disease by adopting cost-effective strategies. This analysis uses the incidence based approach to estimate lifetime economic costs. This lifetime perspective reflects the recommended clinical management of skin cancer in Australia [8,10], managing melanoma and NMSC for five and two years respectively, post diagnosis. Total economic costs in 2010 are derived by multiplying the number of incident cases in 2010 by estimates of the average lifetime direct and indirect costs.

Epidemiological data
In Australia, melanoma is notifiable to cancer registries. To estimate the incidence of melanoma by age and gender, data were obtained from the NSW Central Cancer Registry (CCR) which at the time of writing was available up to 2008 [35]. These data were combined with Australian Bureau of Statistics (ABS) estimates of the resident NSW population in 2010 [36].
Unlike melanoma and other invasive cancers, NMSC is not notifiable by law to cancer registries in Australia, despite being the most commonly diagnosed cancer. Consequently, rates of NMSC have been estimated from population surveys [5,9,37,38]. The most recent survey was conducted by the National Cancer Control Initiative (NCCI) in 2002 [9]. To estimate the incidence of NMSC by age and gender, data from the NCCI survey were combined with ABS estimates of the resident NSW population in 2010 [36].

Diagnosis of skin cancer
Diagnosis of melanoma (both in-situ and invasive) was determined by the principal diagnoses codes in the Australian modified International Statistical Classification of Diseases and related health problems, 10th revision (ICD-10-AM), specifically codes C43 (malignant melanoma of skin) and D03 (melanoma in situ) [44]. Diagnosis of NMSC was identified using codes C44 (malignant neoplasm of skin) and D04 (carcinoma in situ of skin). Diagnosis was also derived from MBS and PBS utilisation and self-report data from the 45&Up study.

Analysis criteria
Costs were calculated for, and compared across, three diagnostic groups: melanoma; NMSC; and, neither melanoma or NMSC. Participants self-reported having either melanoma or NMSC in the 45&Up survey by their responses to the following two questions: "Has a doctor EVER told you that you have melanoma?" "Has a doctor EVER told you that you have skin cancer (not melanoma)?" Self-reported responses were validated by linked records in the CCR, APDC and MBS. Participants without CCR, APDC or MBS records were classified as diagnosis uncertain if their evidence of diagnosis -for either skin cancer -was only self-identification in the 45&Up survey or only relevant MBS records. This group was omitted from analysis due to the ambiguity of their status. The group 'neither melanoma or NMSC' had no data relevant to melanoma or NMSC diagnosis and is used as the control group in the regression analysis.

Direct costs
Cost groupings were derived from the following sources (data periods appear in brackets): Variations in data coverage reflect the nature of linking multiple patient-level data sources. To ensure consistency, data from these files were analysed only for their overlap period, 01/06/2004 to 31/12/2011. MBS, PBS and APDC cost estimates were converted to 2010 prices using appropriate price inflators [45].
Estimation of lifetime direct costs used longitudinal methods to estimate average costs (MBS, PBS and APDC), for each diagnosis group, for each calendar year since diagnosis -for 5 years post diagnosis for melanoma and 2 years post diagnosis for NMSC. A cost per annum from year of diagnosis reflects that any cost difference between those with and without a diagnosis is likely to be most pronounced in the first years of diagnosis. Regression analysis is used to estimate the effect on the cost associated with a diagnosis of either melanoma or NMSC, compared to the control group, i.e., those without a diagnosis. The dependent variable is the average annual cost for the year as identified as the year from diagnosis. The covariates included in this model are listed in Table 2 and include the years from diagnosis as an explanatory variable.
The analysis can be represented by the following equation: From each data source (MBS, PBS, APDC) cost data was obtained for each calendar year of the collection period (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011). The data sets were merged on person number and year. The dataset had one record for each participant for each calendar year of the collection period. Records contained fields for calendar year MBS, PBS, APDC costs and, for participants with skin cancer, years since year of diagnosis. Costs are inclusive of patient and government contributions. Participants who died during the collection period only had records for the calendar years in which they were alive. Mortality information was sourced from ABS mortality [6] and the NSW Registry of Births, Deaths and Marriages (RBDM) [43]. Regression modelling for average lifetime direct costs used General Estimating Equations (GEE) analysis. Given a priori evidence of skewed cost data (i.e., large number of zero or small cost observations with a small number of observations with very large costs), GEEs were run with a gamma family and log link. Standard diagnostic tests were conducted (e.g. correlations between independent variables and comparisons of residuals and predicted values). Robust variance estimators were included to ensure more robust estimates of standard errors. Margins (estimates) were derived for average treatment costs for each year from diagnosis by each diagnosis status (melanoma, NMSC) and compared to average annual direct costs for people with no skin cancers. These incremental results were then multiplied by skin cancer incidence figures to derive lifetime direct costs for NSW in 2010.
All regressions were conducted with STATA 12 software. Regression results are available from authors upon request.

Calculating indirect costs of skin cancer
Indirect costs quantified in this analysis include morbidity and premature mortality for those of working age.

Morbidity estimates
The Australian Burden of Disease study provide information on total years lived with a disease and the loss of health (referred to as a disability weight -DW) associated with that disease [46]. Disability weights are based on a scale ranging from 0 to 1 where 0 represents perfect health and 1 represents death [46]. For skin cancer, a range of DWs are used to reflect health states in relation to sequelae. Across the entire disease spectrum, the respectively. An estimate of the average health years of life lost due to skin cancer are used as a proxy for morbidity costs in this analysis and are derived by dividing total years lived with skin cancer with the relevant DW. The human capital approach is used in this analysis to value the loss of productive life. The approach equates the value of a human life to the discounted market value of the output produced by an individual over an expected lifetime. In other words it uses forgone income to estimate forgone productivity [34,47]. The value of a healthy year of life is equivalent to the average annual earnings in NSW for 2010 -$61,105 for males and $42,238 for females [48]. A further adjustment is made to this value to reflect the likelihood of being employed -82 % in males and 68 % in females [49].

Premature mortality
Premature mortality costs are derived by valuing potential years of life lost (PYLL) due to skin cancer before the age of 65. The most comprehensive source of skin cancer mortality data in NSW is provided by the ABS [6]. Dividing ABS data on PYLL with number of deaths provides an estimate of average years of life lost per death.
ABS data does not, however, report age of death so an alternate means was required to estimate average years of life lost due to skin cancer before the age of 65. The CCR provides individual level data on age at melanoma diagnosis and age at death [35]. For melanoma, CCR data suggest that 38 and 35 % of total years of life were lost in those dying before the age of 65 years, for males and females respectively. In the absence of similar CCR data for NMSC, this proportion is applied to ABS data on PYLL and deaths to estimate the average years of productive life lost per incident case of skin cancer.
As above for morbidity, average annual earnings, adjusted for employment, is used as a proxy for the value of a productive year. All future costs are converted to present value using a 3 % discount rate.

Results
Epidemiology Table 3 provides an overview of skin cancer epidemiology in NSW. In 2010, there were an estimated 3797 new cases of melanoma (2295 male and 1502 female) and 148,610 new cases of NMSC (86,812 male and 61,798 female). Equivalent age-standardised incidence rates are, for melanoma, 65 and 42 per 100,000 males and females, respectively; and, for NMSC, 2449 and 1716 per 100,000 males and females, respectively. For melanoma, the average years of healthy life lost are 3.66 for males (equivalent to 254 days) and 1.77 years for females (equivalent to 123 days). For NMSC, the average years of healthy life lost are 0.05 years for males (equivalent to 16 days) and 0.02 years for females (equivalent to 6 days).
In 2010, 359 men and 155 women lost their lives to melanoma corresponding to an age-standardised death rate of 9.8 and 3.6 per 100,000 for males and females, respectively. NMSC claimed the lives of 104 men and 41 females, corresponding to an age-standardised death rate of 2.9 and 0.8 per 100,000 for males and females, respectively. For melanoma, PYLL per death are 9.34 years for males and 10.14 years for females. For NMSC, PYLL per death are 5.44 years for males and 2.53 for females. For those dying of melanoma before the age of 65 years

Discussion
In conducting this study a range of data sources and methods were used. As such, a number of potential limitations and strengths of the analysis need to be considered.

Limitations
First, while the CCR data informed the incidence of melanoma in NSW, partial data sources were relied on to develop an understanding of the incidence of NMSC. These data sources are dated and may not reflect recent changes in incidence of NMSC. Second, our analysis did not consider the cost of skin cancer by stage of diagnosis, type of treatment, treatment provider or socio-economic status. Evidence suggests that there are variations in costs across these categories [15][16][17][18]. In our study, although CCR provided information on staging of disease at time of diagnosis, no other data set had comparable data. Third, the base year for the analysis is 2010. This year is appropriate given data availability but it is acknowledged that skin cancer management may have changed over recent years. Fourth, a limitation of productivity estimates is a lack of complete Industry data. A recent study conducted by Safe Work Australia on exposure to direct sunlight and the provision of sun exposure controls in Australian workplaces, provides evidence that certain workers have a higher likelihood of being exposed to direct sunlight [50]. Further, the report suggests that the provision of sun protection (i.e. sunscreen, protective clothing, hats, sunglasses and being able to reorganise work outside peak UVR hours) was affected by worker employment and demographic characteristics [50]. The lack of data precludes a more robust assessment of Industry-related costs due to skin cancer in this analysis. Fifth, the analysis did not value the contribution made by carers. A report by Access Economics for Carer's Australia examined and valued the amount of informal care being provided in Australia [51]. The report suggests that in 2010, over 1 in 8 Australians (2.87 million people) were estimated to be providing informal care with each carer providing up to 9 h per week in informal care. In the absence of any data related to the number of carers' for skin cancer patients, carer costs were excluded.

Strengths
A strength of the analysis is the use of linked epidemiological data. Data linkage transforms routinely collected administrative data into a powerful resource for research. For the current study, the linking of administrative data provided a rich source of complementary information on: diagnosis of melanoma (CCR); costs associated with skin cancer (MBS, PBS and the APDC); and skin cancer mortality (CCR, APDC). Second, maximising the utility of the linked data set required a flexible data analysis that was logical and robust. The linkage process followed a sequential strategy that aimed to minimise the number of false positive records. The diagnosis criteria for melanoma and NMSC were informed using relevant international classification of disease coding. The statistical approach enabled a comprehensive and rigorous assessment of the lifetime costs of skin cancer in NSW. The costing approach was broadly consistent with other studies analysing Medicare-linked population-based databases in the United States [15][16][17][18]52]. Third, consistent with other costing studies, our analysis has attempted  [30] to place a monetary value on indirect costs [27][28][29]53]. Our analysis only considers the economic value of productive years of life lost is a conservative estimate of mortality. Other studies quantify the economic value of all years of life lost, not just that before the age of 65 years [28].

Comparability with other studies
In spite of methodological differences, the estimated cost of skin cancer in NSW is generally consistent with previous Australian and International studies ( , and the New Zealand study AUD$802 [30]. Previous estimates for the cost of skin cancer in Australia only valued direct costs. Compared with our estimate of the average direct cost per incident case of melanoma (AUD$10,230), other Australian estimates ranged from AUD$14,805 in a 1993-94 study [55] to AUD$4561 in a 2000-01 study [54]. Compared with our estimate of the average direct cost per incident case NMSC (AUD$2336), other Australian estimates ranged from AUD$1496 for the 1993-94 study [55] and AUD$979 for the 2000-01 study [54].

Conclusion
This study provides new evidence on the economic costs associated with skin cancer in NSW, Australia. Although this analysis is based on the latest epidemiological and economic evidence, there are still large knowledge gaps in understanding the wider impact of skin cancer on society. This lack of data means that the study most likely under-estimates the true cost of skin cancer in NSW.
A key factor underpinning the strategic plan of the Cancer Institute NSW is the knowledge that both melanoma and NMSC are highly preventable. The most effective means of reducing risk of developing skin cancer is to avoid direct exposure to UVR during the time of day when solar UVR levels are moderate to extreme. As a consequence of this analysis, we are in a better position to quantify the savings to society of reducing the incidence of skin cancer through preventive strategies such as sunscreen or mass media campaigns efforts. These savings are likely to be significant given the average lifetime cost of skin cancer in NSW is AUD$44,796 per melanoma case and AUD$2459 per NMSC case.

Competing interests
Cancer Institute NSW provided funding to access linked data and conduct the study.
Authors' contributions CD was responsible for all facets of this study and preparing the manuscript. RL contributed to the analysis and manuscript preparation. JB contributed to methods, analysis, interpretation and manuscript preparation. MC contributed to all facets of this research including writing ethic application, assisting in data access and manuscript preparation. AW contributed to methods, analysis and manuscript preparation. DP contributed to all facets of this research including methods, data access, analysis, interpretation and manuscript preparation. AS contributed to conceptual design, interpretation and manuscript preparation. All authors read and approved the final manuscript.