Performance of Automatic De-identification Across Different Note Types

Free-text clinical notes detail all aspects of patient care and have great potential to facilitate quality improvement and assurance initiatives as well as advance clinical research. However, concerns about patient privacy and confidentiality limit the use of clinical notes for research. As a result, the information documented in these notes remains unavailable for most researchers. De-identification (de-id), i.e., locating and removing personally identifying protected health information (PHI), is one way of improving access to clinical narratives. However, there are limited off-the-shelf de-identification systems able to consistently detect PHI across different data sources and medical specialties. In this abstract, we present the performance of a state-of-the art de-id system called NeuroNER1 on a diverse set of notes from University of Washington (UW) when the models are trained on data from an external institution (Partners Healthcare) vs. from the same institution (UW). We present results at the level of PHI and note types.

Transferability of Neural Network-based De-identification Systems

Methods and Materials: We investigated transferability of neural network-based de-identification sys-tems with and without domain generalization. We used two domain generalization approaches: a novel approach Joint-Domain Learning (JDL) as developed in this paper, and a state-of-the-art domain general-ization approach Common-Specific Decomposition (CSD) from the literature. First, we measured trans-ferability from a single external source. Second, we used two external sources and evaluated whether domain generalization can improve transferability of de-identification models across domains which rep-resent different note types from the same institution. Third, using two external sources with in-domain training data, we studied whether external source data are useful even in cases where sufficient in-domain training data are available. Finally, we investigated transferability of the de-identification mod-els across institutions. Results and Conclusions: We found transferability from a single external source gave inconsistent re-sults. Using additional external sources consistently yielded an F1-score of approximately 80%, but domain generalization was not always helpful to improve transferability. We also found that external sources were useful even in cases where in-domain training data were available by reducing the amount of needed in-domain training data or by improving performance. Transferability across institutions was differed by note type and annotation label. External sources from a different institution were also useful to further improve performance.

A Context-Enhanced De-identification System

Many modern entity recognition systems, including the current state-of-the-art de-identification systems, are based on bidirectional long short-term memory (biLSTM) units augmented by a conditional random field (CRF) sequence optimizer. These systems process the input sentence by sentence. This approach prevents the systems from capturing dependencies over sentence boundaries and makes accurate sentence boundary detection a prerequisite. Since sentence boundary detection can be problematic especially in clinical reports, where dependencies and co-references across sentence boundaries are abundant, these systems have clear limitations. In this study, we built a new system on the framework of one of the current state-of-the-art de-identification systems, NeuroNER, to overcome these limitations. This new system incorporates context embeddings through forward and backward n-grams without using sentence boundaries. Our context-enhanced de-identification (CEDI) system captures dependencies over sentence boundaries and bypasses the sentence boundary detection problem altogether. We enhanced this system with deep affix features and an attention mechanism to capture the pertinent parts of the input. The CEDI system outperforms NeuroNER on the 2006 i2b2 de-identification challenge dataset, the 2014 i2b2 shared task de-identification dataset, and the 2016 CEGS N-GRID de-identification dataset (p<0.01). All datasets comprise narrative clinical reports in English but contain different note types varying from discharge summaries to psychiatric notes. Enhancing CEDI with deep affix features and the attention mechanism further increased performance.