Automatic extracting of patient-related attributes: disease, age, gender and race

In the Evidence-based Medicine (EBM), PICO format is designed to easily and correctly search for the best available evidence. As the main element of PICO, the Patient/Problem (P) represents the attributes of patient in the clinical question and studies. In order to better understand the clinical problems, patient attribute identification is crucial and indispensable. Due to the richness of the human nature language, many issues like various term representations, grammar structures and abbreviations present challenges for automatically extracting the patient-related attributes from the unstructured data. In this paper, we employed the nature language processing (NLP) technologies to deeply analyze the linguistic characteristics of the attributes. Based on the NLP analysis results, we built the rule sets for different attributes and applied the rule-based approach to extract the patient-related attributes.