Objective: the study evaluates the accuracy of an algorithm based on hospital discharge data (HDD) in order to estimate breast cancer incidence in three italian regions (Emilia-Romagna, Toscana and Veneto) covered by cancer registries (CR). The evolution of computer-based information systems in health organization suggests automatic processing of HDD as a possible alternative to the time-consuming methods of CR. The study intends to verify whether HDD quickly provides reliable cancer incidence estimates for diagnosis and therapy evaluations.
Design and setting: an algorithm based on discharge diagnosis and surgical therapy of hospitalized breast cancer patients was developed in order to provide breast cancer incidence. Results were compared with the corresponding incidence data of cancer registries. The accuracy of the automatic method was also verified by a direct record-linkage between HDD output and registries' files. The overall survival of cases lost to "HDD method" was analyzed.
Results: in the period covered by the study (3,125,425 person/year) CR enrolled 6,079 incident cases, compared to 6,000 cases recorded through the HDD flow. Incidence rates of the two methods (CR 194.5; HDD 192.0 x 100.000) showed no statistical differences. However, matched cases by the two methods were only 5,038. The sensitivity of the HDD algorithm was 82.9% and its predictive positive value (PPV) was 84.0%. False positive cases were 9.9%. On the other hand, 12.3% CR incident cases were not identified by the algorithm: these were mainly made up of older women, not eligible for surgical therapy. Their three-years survival was 62.0% vs 88.8% of the whole incidence group.
Conclusion: HDD flow performance was similar to observations reported in the literature. The agreement between HDD and CR incidence rates is a result of a cross effect of both sensitivity and specificity limitations of the HDD algorithm. This can seriously impair the reliability of the latter method with regard to the evaluation of diagnostic and therapeutic strategies in cohort studies (i.e. the most effective approach to health setting in oncology).s.