Evaluation of a Method to Identify and Categorize Section Headers in Clinical Documents

摘要: Objective: Clinical notes, typically written in natural language, often contain substructure that divides them into sections, such as “History of Present Illness” or “Family Medical History.” The authors designed and evaluated an algorithm (“SecTag”) to identify both labeled unlabeled (implied) note section headers “history physical examination” documents (“H&P notes”). Design: SecTag uses a combination language processing techniques, word variant recognition with spelling correction, terminology-based rules, naive Bayesian scoring methods headers. Eleven physicians SecTag's performance on 319 randomly chosen H&P notes. Measurements: primary outcomes were the algorithm's recall precision identifying all document sections predefined list twenty-nine major sections. A secondary outcome was evaluate ability recognize correct start end boundaries identified sections. Results: 16,036 total 7,858 Physician evaluators classified 15,329 true positives 160 omitted by SecTag. 99.0 95.6% for 98.6 96.2% 96.6 86.8% determined starting ending text 94.8% 85.9% sections. Conclusions: accurately history documents. This type may assist applications, clinical decision support systems competency assessment medical trainees.