Luther Stephen L, McCart James A, Berndt Donald J, Hahm Bridget, Finch Dezon, Jarman Jay, Foulis Philip R, Lapcevic William A, Campbell Robert R, Shorr Ronald I, Valencia Keryl Motta, Powell-Cope Gail
Stephen L. Luther, James A. McCart, Bridget Hahm, Dezon Finch, Philip R. Foulis, William A. Lapcevic, Robert R. Campbell, and Gail Powell-Cope are with the HSR&D Center of Innovation on Disability and Rehabilitation Research, James A. Haley Veterans Hospital, Tampa, FL. Donald J. Berndt is with the University of South Florida College of Business Administration, Tampa. Jay Jarman is with the East Tennessee State University Department of Computing, Johnson City. Ronald I. Shorr is with the North Florida/South Georgia Veterans Health System, Gainesville, FL. Keryl Motta Valencia is with the VA Caribbean Healthcare System, San Juan, PR.
Am J Public Health. 2015 Jun;105(6):1168-73. doi: 10.2105/AJPH.2014.302440. Epub 2015 Apr 16.
We determined whether statistical text mining (STM) can identify fall-related injuries in electronic health record (EHR) documents and the impact on STM models of training on documents from a single or multiple facilities.
We obtained fiscal year 2007 records for Veterans Health Administration (VHA) ambulatory care clinics in the southeastern United States and Puerto Rico, resulting in a total of 26 010 documents for 1652 veterans treated for fall-related injury and 1341 matched controls. We used the results of an STM model to predict fall-related injuries at the visit and patient levels and compared them with a reference standard based on chart review.
STM models based on training data from a single facility resulted in accuracy of 87.5% and 87.1%, F-measure of 87.0% and 90.9%, sensitivity of 92.1% and 94.1%, and specificity of 83.6% and 77.8% at the visit and patient levels, respectively. Results from training data from multiple facilities were almost identical.
STM has the potential to improve identification of fall-related injuries in the VHA, providing a model for wider application in the evolving national EHR system.
我们确定了统计文本挖掘（STM）能否在电子健康记录（EHR）文档中识别与跌倒相关的损伤，以及来自单一或多个机构的文档训练对STM模型的影响。
我们获取了美国东南部和波多黎各退伍军人健康管理局（VHA）门诊诊所2007财年的记录，共得到26010份文档，涉及1652名因跌倒相关损伤接受治疗的退伍军人以及1341名匹配的对照。我们使用STM模型的结果在就诊和患者层面预测与跌倒相关的损伤，并将其与基于病历审查的参考标准进行比较。
基于单一机构训练数据的STM模型在就诊和患者层面的准确率分别为87.5%和87.1%，F值分别为87.0%和90.9%，灵敏度分别为92.1%和94.1%，特异性分别为83.6%和77.8%。来自多个机构训练数据的结果几乎相同。
STM有潜力改善VHA中与跌倒相关损伤的识别，为在不断发展的国家EHR系统中更广泛的应用提供了一个模型。
