Developing case-finding algorithms for second events of oropharyngeal cancer using administrative data: A population-based validation study.

Abstract

Second event (recurrence or second primary cancer)-free survival is an important indicator for assessing treatment efficacy. However, second events are not explicitly documented in administrative data such as cancer registries. Thus, validated algorithms using administrative data are needed to identify second events of oropharyngeal cancers. The algorithms were developed using classification and regression tree models. Data from chart review served as the reference standard. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated. The high-sensitivity algorithm achieved 87.9% (95% confidence interval: 82.2%-93.6%) sensitivity, 84.5% (81.1%-87.8%) specificity, 61.2% (54.1%-68.4%) PPV, 96.2% (94.2%-98.1%) NPV, and 85.2% (82.3%-88.1%) accuracy. The high-PPV algorithm obtained 52.4% (43.6%-61.2%) sensitivity, 99.1% (98.2%-100.0%) specificity, 94.2% (88.7%-99.7%) PPV, 88.2% (85.3%-91.0%) NPV, and 88.9% (86.3%-91.5%) accuracy. The validity of the algorithms for identifying second events following primary treatment of oropharyngeal cancers was acceptable.