This paper describes the colorectal cancer burden in 2006 in two Italian areas covered by cancer registration starting from individual patterns of hospital care. The idea is to provide the distribution of health care expenditures according to the disease pathway from first diagnosis to possible recovery or death, for subgroups of cancer survivors homogeneous with respect to their health care needs.
The phase of care approach here used subdivides care into three clinically relevant intervals: the first year since diagnosis, the last year of life and the monitoring or continuing period. The main advantage of the phase of care approach is the possibility to describe the distribution by disease phase of prevalent cases and hospital care expenditures at a given date taking into account the individual patterns of care during the entire lifespan. The methodology requires the registration of new cancer cases and follow-up information, typically provided by the cancer registries, and the collection of data on treatments and corresponding costs, derived from other administrative sources. To our knowledge this represents the first experience in Italy in estimating prevalent cases and costs distribution according to a three- phase of care framework and linking, at individual level, CR information to the HDC database.
Data linkage is the preliminary step in order to reconstruct the patterns of care along the entire disease pathway. Clinical guidelines are insufficient to predict patterns of care and furthermore compliance with clinical guidelines might vary between geographical areas. Following the individual record linkage, DRG codes have been accurately selected in order to identify only colorectal cancer-related treatments, i.e. those treatments appropriately attributable to the colon and rectum primary. One of the advantages of this procedure is that it allows the direct estimation of the net cancer costs, without using a control cohort of non-cancer subjects matched to patients by sex, age, location area and phase of care, as elsewhere proposed in the literature . Another advantage is the direct identification of prevalent cases in the final phase by counting the number of deaths during the following year – information on life status follow-up is provided by the CR using the National Death Certificate Database– rather than estimating a survival curve, as elsewhere proposed .
Stage is a key variable in this study, because it determines the treatment approach and the corresponding patterns of care and costs. Stage classification is a complex issue, and different attitudes in the definition of stage at diagnosis might compromise the comparability of data between cancer registries. In our study, however, stage at diagnosis is used merely as a proxy of treatment approach and an accurate stage classification is not required.
In the disease phase framework the number of phases and length of each phase are a-priori defined and do not vary according to cancer site or patients characteristics; however, a twelve months initial phase might be too short compared to the time required to complete some first course treatments ; the continuing phase length may vary according to patient survival and corresponding care needs depend on additional prognosis factors, such as the presence of metastases and possible comorbidities, which are not taken into account in the analysis; a flexible estimation of the phase number and duration based on observed disease pathways of cancer patients cohorts might therefore be envisaged .
Short term survivors are attributed to the final phase only, regardless of their cause of death and the fact that they might have received first course procedures and treatments. This might lead to a possible overestimation of the number of terminal patients and of the total costs attributed to final phase. A correction factor based on the observed proportion of other causes of death in the incident cohorts could be implemented.
Costs of the initial phase refer to a cohort of patients diagnosed in 2000-2001 and do not include recent variations in the cost of drugs and new diagnostic procedures and treatments. Relevant changes in first course therapies of colorectal cancer patients have been occurred in Italy after 2005. As a further development other sources of data in addition to in-hospital records should be included, i.e. data on drugs consumption and outpatient treatments, in order to obtain a complete estimation of costs directly attributable to a specific cancer.
In Italy CR’s represent a reliable surveillance source which however covers a third of the national population and it is not representative of the whole country; furthermore, most of the areas covered by the CRs correspond only to portions of a region; finally, each region is an autonomous entity regarding health care administration. As a consequence of these features, extrapolating costs at national level is a very complex exercise. On the other hand extrapolation at regional level is in principle feasible, but requires some further methodological steps, such as projecting prevalent cases in areas not covered by CR’s, which goes beyond the scope of this pilot study, and will be the goal of a future development.
A validation of our cost estimates using comparable figures of health care expenditures documented in the regional budget plans of Veneto  and Tuscany  has been carried out: a total health expenditure of 44.5 million Euros and 33.8 million Euros, in VCR and TCR respectively, is estimated for 9640 and 8357 colorectal prevalent cases respectively in 2006, and the distribution of costs by phase of care is similar between the two cancer registries; the total costs per capita of colorectal cancer patients, regardless the phase of care, is about 4 300 Euros per year in the two cancer registries combined, and this value is consistent with a total health expenditures of about 5 000 reported in the regional budget 2008. The comparability of the results between the two CR areas confirms the consistency with the clinical guidelines and represents a first step for future analyses with the aim to extrapolate local results to the entire regions.
Finally, the methodology proposed and applied in this paper may be improved with a projection tool in order to evaluate the impact of specific public health interventions on patterns of care and costs of cancer patients. The analysis of the effects of cancer control strategies, such as screening programs or new treatments introduction, on the expected economic burden will be an important area for additional research in a public health perspective.