Biosurveillance/SNOMED-CT Natural Language Processing

生物恐怖主义仍然是对我们公共卫生的重大威胁。早期发现与接触生物恐怖制剂一致的患者症状发生率增加，是遏制生物恐怖袭击和对接触者进行更迅速治疗的一个重要方法。通常，在确认发生了与生物恐怖主义有关的接触之前，就已经有大量人出现了与接触生物恐怖主义制剂相一致的情况。这提供了从患者记录中捕获和分析信号的机会。我们建议从临床记录（逐节）中提取自动化临床信息，并使用SNOMED-CT进行编码，可作为监测数据的基础。我们认为，可以从临床记录中提取这些数据（按临床记录章节列出的代码集），这些数据包含有关患者表现、发现、药物、过敏和合并症的重要和突出的医学事实（代码）。在这项研究中，我们将分析SNOMED-CT的能力，以提供足够的内容覆盖星座的症状与接触剂， 生物恐怖主义（即炭疽、天花、蓖麻毒素和辐射暴露）。我们的方法建立在我们实验室已有的大量研究基础上。自1987年以来，我们一直在研究使用受控医学词汇编纂医学内容的方法。在这项研究中，我们将采用马约词汇服务器（MVS）在马约生物医学信息学实验室开发，并已在马约，约翰霍普金斯大学和退伍军人管理局医学中心都取得了巨大的成功。我们的MVS工具包的性能2已经过诊断验证，灵敏度为99.7%，特异性为97.9%。 这个建议涉及的实际问题，导致了数据的互操作性。这项研究的成果将有助于我们的国家倡议，为安全有效的生物安全生物监测解决方案铺平道路。

Bioterrorism remains a significant threat to our public health. Early identification of an increased rate of occurance of patient presentations consistent with an exposure to an agent of bioterrorism is one important method to contain bioterror attacks and effect more rapid treatment of exposed individuals. Often presentations consistent with an exposure to an agent of bioterrorism occur in significant numbers prior to the recognition that a bioterrorism related exposure has occurred. This presents an opportunity to capture and analyze signals from patient records. We propose to provide an abstraction of automated clinical information from the clinical record (section by section) that will be coded using SNOMED-CT which can serve as the substrate for surveillance data. We believe that this data (sets of codes by section of the clinical record) which holds the important and salient medical facts (codes) regarding the patients' presentation, findings, medications, allergies and co-morbidities could be abstracted from clinical records. In this study, we will analyze SNOMED-CT's ability to provide adequate content coverage for constellations of symptoms associated with exposures to agents of bioterrorism (i.e. Anthrax, Small Pox, Ricin, and Radiation exposure). Our method builds on the considerable body of research already available within our laboratory. We have been researching methods for codifying medical content using controlled medical vocabularies since 1987. For this study, we will employ the Mayo Vocabulary Server (MVS) developed in the Mayo Laboratory of Biomedical Informatics and has been used at Mayo, Johns Hopkins University and the VA medical centers all with great success. Our performanc2 of the MVS toolkit has been validated for diagnoses where we showed a sensitivity of 99.7% and a specificity of 97.9%. This proposal deals with practical issues that lead the way toward interoperable data. The fruits of this research will assist our national initiatives to pave the way toward a safe and effective BioSecure biosurveillance solution.