以機器學習方法藉由急診病人主訴資料預測醫生之檢驗項目

Abstract

急診壅塞的問題長期存在於台灣的醫療體系當中，本研究主要的目的在於藉由急診病患的主訴資料，建構基於機器學習的檢驗項目預測模型，針對病患該次就診最主要痛苦訴求，從非結構化的文本資料當中，找出可以預先執行的醫學檢驗事項，在醫生會診之前，完成應該有的檢驗，讓醫生可以根據檢驗的結果，快速判定病人應有的後續處置，用以改善並加速現有急診流程，提供更好的急診醫療品質及更快速的確診及服務。

The problem of Emergency Department Overcrowding has long existed in Taiwan's medical system. The main purpose of this study is to construct a diagnostic item predictive model based on patient’s chief complaint by using machine learning. From the unstructured text materials, find out the diagnostic items that can be performed and complete the due test before the doctor's consultation, so that the doctor can quickly determine the follow-up treatment that the patient should have based on the results of the test. Improve and accelerate the existing emergency procedures, provide better emergency medical quality and faster diagnosis and services.