Abstract
There is a great and growing need to ascertain what exactly is the state of a patient, in terms of disease progression, actual care practices, pathology, adverse events, and much more, beyond the paucity of data available in structured medical record data. Ascertaining these harder-to-reach data elements is now critical for the accurate phenotyping of complex traits, detection of adverse outcomes, efficacy of off-label drug use, and longitudinal patient surveillance. Clinical notes often contain the most detailed and relevant digital information about individual patients, the nuances of their diseases, the treatment strategies selected by physicians, and the resulting outcomes. However, notes remain largely unused for research because they contain Protected Health Information (PHI), which is synonymous with individually identifying data. Previous clinical note de-identification approaches have been rigid and still too inaccurate to see any substantial real-world use, primarily because they have been trained with too small medical text corpora. To build a new de-identification tool, we created the largest manually annotated clinical note corpus for PHI and develop a customizable open-source de-identification software called Philter (“Protected Health Information filter”). Here we describe the design and evaluation of Philter, and show how it offers substantial real-world improvements over prior methods.
Highlights
Structured electronic health records (EHRs) fields, primarily comprises elements such as high-level demographics and billing codes (ICD), are currently the most utilized in determining the state of a patient, in terms of clinical care details or disease state
The Inter-Rater Reliability, for Protected Health Information (PHI) vs. Safe tokens, between first and second pass annotators in the UCSF corpus was greater than 99.99%, with the second annotator identifying an average of 39 additional PHI tokens and converting an average of 21 tokens from PHI to Safe per 500 notes
Primary and Secondary result metrics on both corpora are displayed in Table 1, with precision listed as a reference
Summary
Structured electronic health records (EHRs) fields, primarily comprises elements such as high-level demographics and billing codes (ICD), are currently the most utilized in determining the state of a patient, in terms of clinical care details or disease state. Many of these fields are often used in clinical research, and are starting to be used to determine human phenotypes[1] for genome-wide association studies, and can be used to facilitate automated improvements of healthcare decision-making[2]. There have been many studies that have successfully used data from clinical notes for discoveries, including detection of drug adverse outcomes[5], identification of off-label drug use[6], surveillance of disease states[7], and identification of clinical concept relatedness[8]
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call