The growth of digital clinical data has raised questions as to how best to leverage this data to aid the world of healthcare. Promising application areas include Information Retrieval and Question-Answering systems. Such systems require an in-depth understanding of the texts that are processed. One aspect of this understanding is knowing if a medical condition outlined in a patient record is recent, or if it occurred in the past. As well as this, patient records often discuss other individuals such as family members. This presents a second problem - determining if a medical condition is experienced by the patient described in the report or some other individual. In this paper, we investigate the suitability of a machine learning (ML) based system for resolving these tasks on a previously unexplored collection of Patient History and Physical Examination reports. Our results show that our novel Score-based feature approach outperforms the standard Linguistic and Contextual features described in the related literature. Specifically, near-perfect performance is achieved in resolving if a patient experienced a condition. While for the task of establishing when a patient experienced a condition, our ML system significantly outperforms the ConText system (87% versus 69% f-score, respectively).
展开▼
