Authors

Abstract

Individual electronic health records (EHRs) and clinical reports are often part of a largersequencefor example, a single patient may generate multiple reports over the trajectoryof a disease. In applications such as cancer pathology reports, it is necessary not only toextract information from individual reports, but also to capture aggregate information regardingthe entire cancer case based off case-level context from all reports in the sequence. Inthis paper, we introduce a simple modular add-on for capturing case-level context that isdesigned to be compatible with most existing deep learning architectures for text classificationon individual reports. We test our approach on a corpus of 431,433 cancer pathologyreports, and we show that incorporating case-level context significantly boosts classificationaccuracy across six classification taskssite, subsite, laterality, histology, behavior, andgrade. We expect that with minimal modifications, our add-on can be applied towards a widerange of other clinical text-based tasks.