I-Corps: A Machine Learning Tool for Medical Device History and Recalls

这个I-Corps项目的更广泛的影响/商业潜力是开发一个软件系统来预测医疗器械召回的可能性。2003年至2020年间，美国8.9%（4889件）的医疗器械因可能导致严重健康问题甚至死亡的问题而被召回。这些召回会影响患者的治疗结果，设备制造商和健康保险公司都会遭受巨大的经济损失。及时预测召回可能有利于医疗保健系统中的多个利益相关者。然而，制造商目前依靠自己的实验室研究、专有数据和客户反馈来评估设备安全性和预测召回。提出的技术使用先进的数据分析来研究医疗器械不良事件的历史，并为召回执行预测建模。这种基于数据分析的系统可以改善对医疗器械召回可能性的评估，通过避免故障和器械召回，对患者、医疗器械制造商、保险公司、监管机构和其他利益相关者都有好处。I-Corps项目的基础是开发一个数据分析平台，该平台使用来自多个数据源的历史数据来可视化和分析医疗设备召回的可能性。拟议的技术是一个在线决策支持系统，该系统使用数据分析从FDA的510(k)设备批准文件中提取见解，以研究和预测召回。该系统使用自然语言处理从设备文件中自动提取相关信息，并使用监督机器学习对医疗设备召回进行预测建模。该系统利用相关前代设备的特征和设备引用网络的特征，帮助制造商和分析人员探索不同医疗设备的召回概率。初步结果表明，该系统可以在及时预测设备召回方面提供显著的好处，这可能会提高患者的安全性，并减少医疗保健生态系统中几个利益相关者的经济损失。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of a software system to predict medical device recall likelihood. Between 2003 and 2020, 8.9% (4,889) of medical devices in the US were recalled due to issues that could cause serious health problems or even death. These recalls impact patient outcomes, and both device manufacturers and health insurers incur large financial losses. Timely prediction of recalls may benefit multiple stakeholders in the healthcare system. However, manufacturers presently rely on their own lab studies, proprietary data, and customer feedback to evaluate device safety and predict recalls. The proposed technology uses advanced data analytics to study the history of adverse events of medical devices and performs predictive modeling for recalls. This data analytics-based system may improve the evaluation of medical device recall likelihood with benefits to patients, medical device manufacturers, insurers, regulators, and other stakeholders by avoiding malfunctions and device recalls.This I-Corps project is based on the development of a data analytics platform that uses historical data from multiple data sources to visualize and analyze medical device recall likelihood. The proposed technology is an online decision support system that uses data analytics to extract insights from FDA’s 510(k) device approval files to study and predict recalls. The proposed system uses natural language processing to automatically extract relevant information from device files and performs predictive modeling for medical device recalls using supervised machine learning. The system leverages the characteristics of related predecessor devices and features from the device citation network to help manufacturers and analysts explore recall probabilities of different medical devices. Initial results suggest that this system may provide significant benefits in predicting device recalls in a timely manner, which may improve patient safety and reduce financial losses for several stakeholders in the healthcare ecosystem.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.