Computer-assisted de-identification of free-text nursing notes

Abstract

Medical researchers are legally required to protect patients' privacy by removing personally identifiable information from medical records before sharing the data with other researchers. Different computer-assisted methods are evaluated for removing and replacing protected health information (PHI) from free-text nursing notes collected in the hospital intensive care unit. A semi-automated method was developed to allow clinicians to highlight PHI on the screen of a tablet PC and to compare and combine the selections of different experts reading the same notes. Expert adjudication demonstrated that inter-human variability was high, with few false positives and many false negatives. A preliminary automated de-identification algorithm generated few false negatives but many false positives. A second automated algorithm was developed using the successful portions of the first algorithm and incorporating other heuristic methods to improve overall performance. A large de-identified collection of nursing notes was re-identified with realistic surrogate (but unprotected) dates, serial numbers, names, and phrases to form a "gold standard" reference database of over 2600 notes (approximately 340,000 words) with over 1800 labeled instances of PHI. This gold standard database of nursing notes and the Java source code used to evaluate algorithm performance will be made freely available on the Physionet web site in order to facilitate the development and validation of future de-identification algorithms.