De-identification of Unstructured Clinical Data for Patient Privacy Protection

The adoption of Electronic Health Record (EHR) systems is growing at a fast pace in the United States and in Europe, and this growth results in very large quantities of patient clinical information becoming available in electronic format, with tremendous potential, but also equally growing concern for patient confidentiality breaches. Secondary use of clinical information is essential to fulfil the promises for high quality healthcare, improved healthcare management, and effective clinical research. De-identification of patient information has been proposed as a solution to both facilitate secondary use of clinical information, and protect patient information confidentiality. Most clinical information found in the EHR is unstructured and represented as narrative text, and de-identification of clinical text is a tedious and costly manual endeavor. Automated approaches based on Natural Language Processing have been implemented and evaluated, allowing for much faster de-identification than manual approaches. This chapter introduces clinical text-de-identification in general, and then focuses on recent efforts and studies at the U.S. Veterans Health Administration. It includes the origins and definition of text de-identification in the United States and Europe and a discussion about text anonymization. It also presents methods applied for text de-identification, examples of clinical text de-identification applications, and U.S. Veterans Health Administration clinical text de-identification efforts.