Purpose: Aim of the project was the identification of 2007 cancer incidence for Austria based on a longitudinal billing dataset for the years 2006 and 2007. Grading according to age and sex as well as comparison with the Austrian cancer registry and an analysis of the usage potential are included.
Method: Data extraction and filtering for cancer incidence is realized in several steps based on a relational database including drug prescription, stationary visits and extramural health service for anonymized Austrian inhabitants. First inpatient datasets are filtered by relevant ICD-10 diagnoses for 2007. Then the number of identified persons is reduced by the patients having a similar hospital stay in 2006 (only new diseases should be detected). Next the procedure includes only those patients with a readmission to hospital. Patients with only one hospital stay are excluded because it is very likely that they just had one control visit. Afterwards the drug prescription database is searched for relevant prescriptions in cancer therapy given before the first hospital stay occurs, which leads to the elimination of the identified persons from the list of the new emerging cases.
Result: Using the declared methods for determination of liver and breast cancer disorders in the year 2007 results in 906 respective 4882 new cases. In comparison with data collected by the cancer registry conducted by Statistik Austria which reports 892 respective 4833 new cases high accordance can be seen. The difference is in both cases less than 2% and can be explained by exclusion of private financed hospitals in the billing data sets. Using data sources from intramural as well as extramural health care institutions on the single patient level with fine time resolution facilitates highly reliable results.
Conclusion: Using billing data for identification of new tumor cases in comparison with overall cases reported by data from Statistik Austria leads to the insight that this method is highly reliable. The main benefits are first getting the whole patient way through the health care system after the time point of cancer detection. Second, the analysis of the medical treatment and drug data prior to tumor detection leads to additional insights about high risk factors and high risk groups. This knowledge can subsequently be used for identification of groups for screening.
See more of: The 33rd Annual Meeting of the Society for Medical Decision Making