Electronic medical records for discovery research in rheumatoid arthritis

By Katherine P. Liao, Tianxi Cai, Vivian S. Gainer, Sergey Goryachev, Qing Zeng-Treitler, Soumya Raychaudhuri, Peter Szolovits, Susanne E. Churchill, Shawn N. Murphy, Isaac S. Kohane and 2 more. 2010

329 citations recorded by the source corpus.

Electronic medical records (EMRs) are a rich data source for discovery research but are underutilized due to the difficulty of extracting highly accurate clinical data. We assessed whether a classification algorithm incorporating narrative EMR data (typed physician notes) more accurately classifies subjects with rheumatoid arthritis (RA) compared with an algorithm using codified EMR data alone.

Subfields: Artificial Intelligence, Rheumatology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yuan Luo
2. Tristan Naumann
3. Ozlem Uzuner
4. Zexian Zeng
5. Peter Szolovits
6. Xiaoyu Li
7. Yu Deng
8. Dmitriy Dligach

In Rheumatology

1. Zexian Zeng

Citations in the corpus

• Natural Language Processing for EHR-Based Computational Phenotyping (2018, methodology): “One of themost important applications of computational phenotyping is diagnosis categorization, which enables the automated and efficient identification of patient cohorts for secondary analysis [15], [27], [28], [29], [30], [31].”
• Computational Phenotype Discovery Using Unsupervised Feature Learning over Noisy, Sparse, and Irregular Clinical Data (2013, methodology): “Some progress has been made using supervised machine learning to define disease-level cohorts [32,33], which avoids the manual phenotype specification but still includes all of the other effort associated with supervised methods.”
• Autoantibodies, autoimmune risk alleles and clinical associations in rheumatoid arthritis cases and non-RA controls in the electronic medical records (2013, unknown): “The RA case and non-RA control study populations were created with anonymized clinical data from the EMRs (14), with linkage to a biospecimen repository of discarded blood samples.”
• Methods to Develop an Electronic Medical Record Phenotype Algorithm to Compare the Risk of Coronary Artery Disease across 3 Chronic Disease Cohorts (2015, background): “Briefly, in RA, 63% were ACPA positive, 60% had 1 electronic prescription for methotrexate and 33% for a tumor necrosis factor antagonist [8].”
• Automatic identification of methotrexate-induced liver toxicity in patients with rheumatoid arthritis from the electronic medical record (2015, methodology): “This study used the RA EMR cohort, which included 5903 patients with RA followed at Partners HealthCare since 1992.(14)”
• Using Electronic Patient Records to Discover Disease Correlations and Stratify Patient Cohorts (2011, background): “…database knowledge discovery and association mining [16,17,18], identifying and classifying specific medical cases or conditions in an EPR [19,20,21,22], patient safety and automated surveillance of adverse events, contraindications and epidemics [23,24,25], comorbidity and disease…”
• Extracting research-quality phenotypes from electronic health records to support precision medicine (2015, background): “Many phenotype definitions of complex diseases, such as hypothyroidism [50], cardiovascular diseases [82-84], T2DM [57] and rheumatoid arthritis [56,85], have been deployed and validated across multiple institutions.”
• A systematic comparison of feature space effects on disease classifier performance for phenotype identification of five diseases (2015, background): “Commonly structured EHR-related information such as ICD-9 codes have been shown to be insufficient for producing state-of-the-art performance (Liao et al., 2010) which is why many phenotyping systems employ semantically-informed Natural Language Processing (NLP) methodologies to unlock the…”
• Using electronic health records to drive discovery in disease genomics (2011, background): “Rather, a multi-month iterative refinement of the NLP queries — involving a team of computer scientists, statisticians, and clinical and genomic experts — is required to attain positive predictive values of at least 95% for the intended phenotype”
• Development of phenotype algorithms using electronic medical records and incorporating natural language processing (2015, methodology): “12 This method was also successfully applied to EMR data at other institutions.”
• Clinical Natural Language Processing in languages other than English: opportunities and challenges (2018, background): “There have been a number of success stories in various biomedical NLP applications in English [8–19].”