Natural language processing (NLP) of clinical text offers great potential to expand secondary use of high-value electronic health record (EHR) data, but a barrier to adopting NLP is the high total cost of operation, driven mainly by the costs and limited availability of technical personnel in applied health research settings. To overcome this barrier we propose a cloud-based service systems model by which entire NLP systems deployed in the cloud are cloned and provided to the adopting institution for their exclusive and unlimited use. Useful algorithms that perform various information extraction and classification tasks are built in to the NLP system. A rationale and model for cloud-deployed NLP is presented and the inherent data security and patient privacy issues it raises addressed. Both technical and socio-institutional security issues are discussed in the context of the unique challenges associated with processing highly regulated clinical text in an unconventional computing environment. Results of a June 2010 survey of Institutional Review Board (IRB) managers in applied research settings are presented. Survey questions address IRB managers' readiness to approve research projects involving cloud-based NLP technologies. Useful next steps and information needs are presented in the conclusion.
展开▼
