Evaluation of an algorithm to identify women with carcinoma of the breast

Although claims data are increasingly being used to measure and manage the cost and quality of health care, few studies have evaluated algorithms developed for such analyses. Therefore, the present study was performed to evaluate prospectively a previously published algorithm used to identify women with the new diagnosis of carcinoma of the breast. This algorithm had been developed from the patterns of claims that suggested common clinical presentations of carcinoma of the breast. In the present study, this algorithm was used to identify 177 potential cases of women with newly diagnosed carcinoma of the breast from the claims database of a large health maintenance organization (HMO). The algorithm's positive predictive value for cases identified in the present study was 83% (147/177). To attempt to improve upon the positive predictive value, multiple modifications of the algorithm were performed. The previously defined best modification of the initial algorithm yielded a positive predictive value of 84% (147/174) in the present study with the loss of none of the true positive cases. These results demonstrate that logic-based algorithms can be used as a valid and efficient method of identifying large numbers of new breast cancer cases from claims data. This algorithm provides a powerful tool to perform health care analysis and research for women with newly diagnosed carcinoma of the breast.