Liu, Hongfang

AbstractPrecision medicine is at the forefront of biomedical research. Cancer registries provide rich perspectives and electronic health records (EHRs) are commonly utilized to gather additional clinical data elements needed for translational research. However, manual annotation is resource‐intense and not readily scalable. Informatics‐based phenotyping presents an ideal solution, but perspectives obtained can be impacted by both data source and algorithm selection. We derived breast cancer (BC) receptor status phenotypes from structured and unstructured EHR data using rule‐based algorithms, including natural language processing (NLP). Overall, the use of NLP increased BC receptor status coverage by 39.2% from 69.1% with structured medication information alone. Using all available EHR data, estrogen receptor‐positive BC cases were ascertained with high precision (P = 0.976) and recall (R = 0.987) compared with gold standard chart‐reviewed patients. However, status negation (R = 0.591) decreased 40.2% when relying on structured medications alone. Using multiple EHR data types (and thorough understanding of the perspectives offered) are necessary to derive robust EHR‐based precision medicine phenotypes.

Abstract Objective To assess clinical documentation variations across health care institutions using different electronic medical record systems and investigate how they affect natural language processing (NLP) system portability. Materials and Methods Birth cohorts from Mayo Clinic and Sanford Children’s Hospital (SCH) were used in this study (n = 298 for each). Documentation variations regarding asthma between the 2 cohorts were examined in various aspects: (1) overall corpus at the word level (ie, lexical variation), (2) topics and asthma-related concepts (ie, semantic variation), and (3) clinical note types (ie, process variation). We compared those statistics and explored NLP system portability for asthma ascertainment in 2 stages: prototype and refinement. Results There exist notable lexical variations (word-level similarity = 0.669) and process variations (differences in major note types containing asthma-related concepts). However, semantic-level corpora were relatively homogeneous (topic similarity = 0.944, asthma-related concept similarity = 0.971). The NLP system for asthma ascertainment had anF-score of 0.937 at Mayo, and produced 0.813 (prototype) and 0.908 (refinement) when applied at SCH. Discussion The criteria for asthma ascertainment are largely dependent on asthma-related concepts. Therefore, we believe that semantic similarity is important to estimate NLP system portability. As the Mayo Clinic and SCH corpora were relatively homogeneous at a semantic level, the NLP system, developed at Mayo Clinic, was imported to SCH successfully with proper adjustments to deal with the intrinsic corpus heterogeneity.

The burden and predictors of hospitalization over time in community-based patients with peripheral artery disease (PAD) have not been established. This study evaluates the frequency, reasons and predictors of hospitalization over time in community-based patients with PAD. We assembled an inception cohort of 1798 PAD cases from Olmsted County, MN, USA (mean age 71.2 years, 44% female) from 1 January 1998 through 31 December 2011 who were followed until 2014. Two age- and sex-matched controls ( n = 3596) were identified for each case. ICD-9 codes were used to ascertain the primary reasons for hospitalization. Patients were censored at death or last follow-up. The most frequent reasons for hospitalization were non-cardiovascular: 68% of 8706 hospitalizations in cases and 78% of 8005 hospitalizations in controls. A total of 1533 (85%) cases and 2286 (64%) controls ( p < 0.001) were hospitalized at least once; 1262 (70%) cases and 1588 (44%) controls ( p < 0.001) ≥ two times. In adjusted models, age, prior hospitalization and comorbid conditions were independently associated with increased risk of recurrent hospitalizations in both groups. In cases, severe PAD (ankle–brachial index < 0.5) (HR: 1.25; 95% CI: 1.15, 1.36) and poorly compressible arteries (HR: 1.26; 95% CI: 1.16, 1.38) were each associated with increased risk for recurrent hospitalization. We demonstrate an increased rate of hospitalization in community-based patients with PAD and identify predictors of recurrent hospitalizations. These observations may inform strategies to reduce the burden of hospitalization of PAD patients.

Abstract Existing general clinical natural language processing (NLP) systems such as MetaMap and Clinical Text Analysis and Knowledge Extraction System have been successfully applied to information extraction from clinical text. However, end users often have to customize existing systems for their individual tasks, which can require substantial NLP skills. Here we present CLAMP (Clinical Language Annotation, Modeling, and Processing), a newly developed clinical NLP toolkit that provides not only state-of-the-art NLP components, but also a user-friendly graphic user interface that can help users quickly build customized NLP pipelines for their individual applications. Our evaluation shows that the CLAMP default pipeline achieved good performance on named entity recognition and concept encoding. We also demonstrate the efficiency of the CLAMP graphic user interface in building customized, high-performance NLP pipelines with 2 use cases, extracting smoking status and lab test values. CLAMP is publicly available for research use, and we believe it is a unique asset for the clinical NLP community.

Objective The primary aim was to compare independent and joint performance of retrieving smoking status through different sources, including narrative text processed by natural language processing (NLP), patient-provided information (PPI), and diagnosis codes (ie, International Classification of Diseases, Ninth Revision [ICD-9]). We also compared the performance of retrieving smoking strength information (ie, heavy/light smoker) from narrative text and PPL Materials and Methods Our study leveraged an existing lung cancer cohort for smoking status, amount, and strength information, which was manually chart-reviewed. On the NLP side, smoking-related electronic medical record (EMR) data were retrieved first. A pattern-based smoking information extraction module was then implemented to extract smoking-related information. After that, heuristic rules were used to obtain smoking status-related information. Smoking information was also obtained from structured data sources based on diagnosis codes and PPI. Sensitivity, specificity, and accuracy were measured using patients with coverage (ie, the proportion of patients whose smoking status/strength can be effectively determined). Results NLP alone has the best overall performance for smoking status extraction (patient coverage: 0.88; sensitivity: 0.97; specificity: 0.70; accuracy: 0.88); combining PPI with NLP further improved patient coverage to 0.96. ICD-9 does not provide additional improvement to NLP and its combination with PPI. For smoking strength, combining NLP with PPI has slight improvement over NLP alone. Conclusion These findings suggest that narrative text could serve as a more reliable and comprehensive source for obtaining smoking-related information than structured data sources. PPI, the readily available structured data, could be used as a complementary source for more comprehensive patient coverage.

Research in clinical information retrieval has long been stymied by the lack of open resources. However, both clinical information retrieval research innovation and legitimate privacy concerns can be served by the creation of intrainstitutional, fully protected resources. In this article, we provide some principles and tools for information retrieval resource‐building in the unique problem setting of patient‐level information retrieval, following the tradition of the Cranfield paradigm. We further include an analysis of parallel information retrieval resources at Oregon Health & Science University and Mayo Clinic that were built on these principles.