We have constructed a comprehensive natural history and cervical screening model, using observed data on actual screening behaviour, and extensively calibrated and cross-validated outputs against observed data from the NCSP in Australia. We estimate that the program cost to government, excluding administrative overheads, was approximately $194.8 million in 2010. In total, about $174.2 million (89%) of the total expenditure was estimated to be associated with the prevention of invasive cervical cancer (primary screening of asymptomatic women, follow-up of women with cervical abnormalities, and diagnosis and treatment for women with precancerous lesions). Within this, a total of 1.7 million primary screening smears costing $96.7 million were estimated to have been conducted, a further 188,900 smears costing $10.9 million were conducted to follow-up low grade abnormalities, 70,900 colposcopy and 34,100 histological evaluations costing $21.2 million were performed, and about 18,900 treatments for precancerous lesions were performed, costing $45.5 million after accounting for both the treatment and post-treatment follow-up. We also estimated that $20.5 million (11%) of the total expenditure was associated with work-up and treatment for the approximately 761 women who were diagnosed with invasive cervical cancer, and care for the 213 women who died from cervical cancer.
To our knowledge this is the first study that provides such detailed estimates of resource utilisation in relation to any organised cervical screening program in a developed country. The findings of this evaluation provide a comprehensive source of information on the NCSP which has not previously been available. Screening histories for women who participate in the cervical screening program are collected by State and Territory level Pap test registers [1, 2] and the collected data are collated by the Australia Institute of Health and Welfare (AIHW) to produce annual public reports summarising participation rates in the program, early re-screening rates, abnormalities detected by cytology and histology, cervical cancer incidence and cervical cancer mortality in Australia. The most recent report for the period 2009–2010 contains expanded information on performance indicators and also includes overall numbers of cytology and histology tests . However, information on the detailed breakdown of the indications for cytology testing (primary screening, follow-up, or test-of-cure), and the resource utilisation associated with colposcopy, biopsy, treatment, and test-of-cure is either not yet uniformly collected (as for in the case of colposcopy) or has not yet been able to be routinely reported. Where comparison was possible, we found good agreement between our findings and the data published by the AIHW. We estimate that a total of about 2.1 million smears were performed in 2010, 2 million of these were performed in women aged 20–69 years; which is consistent with the number of tests performed in 2010 recently published by AIHW (2,109,131 in women of all ages and 2,025,860 in women aged 20–69 years) . We estimate that there were 84,600 and 27,700 cytology tests which were found to have low-grade and high-grade cervical abnormalities respectively in women aged 20–69 years, and that 33,200 women with abnormal cytology had a follow-up histological evaluation. These findings are broadly consistent with the number of cytology tests with an abnormal result in 2010 (78,510 tests associated with low-grade abnormalities and 28,491 tests associated with high-grade abnormalities), and the number of cytology tests performed which were followed by a histological evaluation test in 2009 (38,859 tests) published by AIHW .
The previously available information on the cost of cervical cancer screening in Australia is limited. A 1993 report of the annual cost of the National Cervical Screening Program to government estimated these costs as $138 million; [31, 32] applying the health services consumer price index [33, 34] this is equivalent to $302 million in 2010, but it is not clear whether downstream diagnostic and treatment costs (or overheads costs) were taken into account in the estimate. Annual reports on Australia’s national public health expenditure published by the AIHW have included some information about expenditure on the cervical screening program in 2007, the AIHW reported that cervical screening-related expenditure by federal and state governments was $113.2 million, with an average expenditure of $5.33 per person (including both males and females) . Of the $113.2 million, 70% ($79.3 million) was spent by the federal government on screening-associated costs including incentive payments, Medicare benefits for general practitioner consultations, pathology testing and other benefits related to collecting cytology samples, and departmental expenses [35, 36]. The remaining $33.9 million was spent by State and Territory governments on program implementation and promoting community awareness of the screening program . Although the overall estimate includes some of the direct costs included in our calculations, and additionally includes overheads and program promotion costs not included in our estimates, it does not appear to include costs associated with follow-up and diagnostic management or treatment for precancerous lesions. Therefore the findings are not directly comparable with our results. In 2005–06, the AIHW provided an estimate of government expenditure for women presenting with symptoms indicative of cancer of $20.1 million,  which appears broadly comparable with our estimated costs for cancer work-up and treatment of $21.5 million.
The total expenditure on health goods and services in Australia in FY2009-10 has been estimated at $121.4 billion by the AIHW,  and therefore our estimate suggests that secondary prevention with cervical screening comprises approximately 0.16% of the health-related expenditure in Australia. However, our findings represent an underestimate of the total costs related to cervical cancer prevention and treatment, for several reasons. Firstly, the total costs also include costs related to primary prevention via the implementation of the National HPV Vaccination Program, which was beyond the scope of the current evaluation. Secondly, we did not consider administrative overheads for the screening program, which may include substantial costs related to implementation at the State and Territory level as well as practice incentive payments for screening in women aged 20–69 years who have not had a cervical smear in the last four years, and other incentives for practices that engage with the program and screening registers . In this study we did not incorporate overhead costs related to running the screening registries and sending screening reminder letters. These costs vary from state to state and are not available in the public domain. They are not readily amenable to the type of modelled analysis used in our paper since the overhead costs are not related to the number of women screened, but are likely to relate to state-specific infrastructure and funding arrangements. The NCSP Renewal process includes a process to explore an option for a real or virtual national register system  and also an option to change from the current reminder-based system to an invitational system of call and recall. In consequence the administrative overheads for the screening program may potentially change. Additionally, because our evaluation was conducted from the health-services perspective, we considered only the costs to government. We did not consider additional societal costs to women or their families that may be related to cervical screening. For example, in the state of NSW, up to 30% of women choose to have an adjunctive liquid-based cytology smear, for which they pay out of their own pocket [14, 39], and therefore these costs are incurred outside the screening program. If a societal perspective was taken, this cost as well as women’s out-of-pocket expenses for transportation, lost of productivity, capital cost and depreciation, and other program operational costs including regular program overheads, screening participation incentives, and the costs associated with awareness and other health promotion campaigns would need to be included in the evaluation.
Although our findings appear to be in good agreement with the available data, the predicted levels of resource utilisation depend on a number of factors. We assessed screening behaviour in women of various age groups and with different screening histories over a 10 year period using data from Victoria (1995–2007), and then projected forward to estimate behaviour and the consequent resource utilisation in the year 2010. Newer processes for management of low grade abnormalities, and post-treatment management and test-of-cure only began to be rolled out from mid-2006,  and therefore information on compliance with these recommendations is much more limited than is the case for attendance for routine screening. While we were able to derive compliance estimates based on the data from the VCCR after the introduction of the updated guidelines, practices and behaviour may have changed over time. Any differences between the assumed and actual behaviour will affect the accuracy of our estimates of resource utilisation in the program. Additionally, based on data showing that cervical cancer incidence and mortality rates have stabilised since about 2003–2004,  we assumed that rates in 2010, which were not yet reported at the time of writing, would be similar to those observed over the period 2004–2007; again, if actual rates are eventually found to be substantially different from these levels this would have an impact on the interpretation of our findings. For the current analysis we also assumed that the population-based HPV vaccination program, rolled out in 2007, did not have a discernable impact on the rate of detected precancerous abnormalities in adult women (aged 18 years or older) within the screening program by 2010. This assumption is supported by the findings of a recent study in Victoria, which identified a reduction in the incidence of histologically-confirmed high-grade precancerous lesions in females younger than 18 years by the end of 2009 (which may be an early marker of the effects of the vaccination program), but identified no significant effect on the incidence of high grade abnormalities in women above the age of 18 years at this stage .
As a strength of our analysis, we used extensive screening data from VCCR to explicitly model for the heterogeneity of re-screening rates based on measured covariates for age, time elapsed since last screen and recommended follow-up after last screen. We did not model for residual individual-level heterogeneity in the re-screening rates due to unmeasured covariates, due to a lack of information on such covariates to the required level of detail. Given the explicit modelling of covariates for re-screening rates, we would expect that residual individual-level heterogeneity to have limited bias on the measures of cost-effectiveness.
The annual expenditure on cervical screening and treatment for CIN (excluding cancer treatment-related expenditure) in Italy, France and UK has previously been estimated to be €181.5 million (~A$232 million) with an estimated 6.4 million women screened in 2005 , €196.5 million (~$251 million) with 6.1 million women screened in 2004  and £157 million (~$249 million) with 3.4 million women screened in 2006–07 , respectively. Our estimates, based on these data, of the associated estimated annual expenditure on cervical screening and treatment for CIN per women screened in Italy (~$36), France (~$41) and UK (~$73) is lower than the estimated ~$91 per woman screened in Australia (this estimate excludes a calculated $11 per women in Australia for cancer treatment, for comparability with the other estimates). The differences might be due to a number of factors, and it is not clear in how much detail the costs in the other evaluations have been modelled. Some of the factors impacting the comparison, might include, for example, the differences in screening recommendations (e.g. screening interval and management of abnormal cytology), screening participation, local screening and diagnostic test performances, the underlying disease prevalence in the population and health system costs.
We have estimated that the average annual cost of screening per adult woman in the population is $23, and that a substantial proportion of this is related to delivery of primary screening tests and follow-up of low grade cytological abnormalities. In the future, due to the effects of the National HPV Vaccination Program, it is expected that the number of cervical abnormalities and cases of invasive cervical cancer detected by the organised cervical screening program will decrease as more vaccinated women enter and age within the screening program over time. An ongoing body of work will use the model platform described here to predict the magnitude and timing of the impact of the National HPV Vaccination Program on resource utilisation in young women within the screening program, and will also assess the impact of changing screening recommendations on resource utilisation, costs, distribution of cancer stage at diagnosis and the effectiveness of screening. For example, we have previously used the platform to evaluate the impact of moving to a three-yearly screening interval under different systems of screening organisation (reminder-based versus call-and-recall organisation) , finding that three-yearly screening would be cost saving and would not have a substantial impact on the number of cervical cancer cases and cancer deaths. Ongoing work will also consider the impact of raising the age of starting screening, and the effect of further increases in screening interval which could be considered in conjunction with the implementation of primary HPV screening . These evaluations, conducted using the calibrated and validated model platform described in the current study, are intended to assist policy-making in designing optimal strategies for cervical cancer prevention in future.
Some changes to screening programs may affect cancer incidence, but not, to the same degree, mortality because actions to increase compliance to recommendations may impact on stage distribution of detected cancers and thus mortality. Therefore, not only changes in cancer incidence but also changes in stage distribution in relation to screening and vaccination at different ages should be evaluated. The current model is calibrated to observed data on stage at diagnosis, and incorporates heterogeneity in screening uptake and compliance with screening and follow-up recommendations. These aspects of the model structure will allow some insight in future evaluations into how changes would affect not only numbers of cancer cases, but also age distribution, stage at diagnosis, and therefore survival, in women who are unscreened or underscreened, as well as those who attend regularly or are overscreened.