Mp50-14 Mining Kidney Stone Composition From Millions of Clinical Notes Within the Electronic Health Record
Year: 2018
Type: article
Abstract: You have accessJournal of UrologyStone Disease: Epidemiology & Evaluation II1 Apr 2018MP50-14 MINING KIDNEY STONE COMPOSITION FROM MILLIONS OF CLINICAL NOTES WITHIN THE ELECTRONIC HEALTH RECORD Ryan Hsi, Daniel Lee, and Cosmin Bejan Ryan HsiRyan Hsi More articles by this author , Daniel LeeDaniel Lee More articles by this author , and Cosmin BejanCosmin Bejan More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2018.02.1625AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES The electronic health record (EHR) is an underutilized data source that is usable for research. ″Big Data″ approaches through natural language processing (NLP) can capture novel phenotyping information to improve care, create new knowledge, and personalize care and decision-making. To date, EHR-based tools are nonexistent for phenotyping kidney stone formers. We developed an NLP system for the task of stone composition identification from millions of clinical notes in the Vanderbilt EHR. METHODS Annotation of all the text mentions of chemical stone composition from 400 de-identified clinical notes with mentions of ″stone analysis″ or ″kidney stone″ were performed by two urologists. We randomly selected 70% of the resulted annotations for training and the remaining 30% for validation. Our NLP algorithm consisted of analyzing the annotations from the training set and developing regular expressions for finding predefined patterns of stone compositions in text. We evaluated the algorithm on the validation set and determined positive predictive value (PPV), sensitivity, and F1 score, which is the harmonic mean of PPV and sensitivity. To maximize PPV and minimize false positives, only annotations containing % composition were included. The algorithm was applied to the entire de-identified medical record. Descriptive statistics with distributions by age, race, and sex were extracted by stone composition category for calcium oxalate monohydrate (COM), calcium oxalate dehydrate (COD), mixed calcium oxalate (mixed CaOx), hydroxyapatite (HA), brushite, and mixed calcium phosphate (mixed CaP) stone formers (mixed CaP). RESULTS The performance of the algorithm on the training set for the 6 stone types ranged 94-100% for PPV, 94-100% for sensitivity, and 96-98% for F1 score. The respective performance on the test set was 91-100%, 80-90%, and 89-92%, except for mixed CaOx with 100% PPV, 40% sensitivity, and 57% F1 score. The inter-annotator agreement between the two annotators was high (Cohen′s kappa=0.845). As of October 2017, the Vanderbilt EHR contains ˜2.8 million patients with >120 million notes. When applied to the entire medical record, the algorithm identified 2658 total patients with 1944, 729, 409, 229, 99, and 1285 respective individuals with % COM, COD, mixed CaOx, HA, brushite, and mixed CaP with significant differences in mean age, sex, and race (each p<0.001). CONCLUSIONS Our novel NLP algorithm to phenotype kidney stone formers by stone composition has high PPV, sensitivity, and F1 score. Mining stone composition from large-scale EHR is feasible with high precision from clinical notes. © 2018FiguresReferencesRelatedDetails Volume 199Issue 4SApril 2018Page: e680-e681 Advertisement Copyright & Permissions© 2018MetricsAuthor Information Ryan Hsi More articles by this author Daniel Lee More articles by this author Cosmin Bejan More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...
Source: The Journal of Urology
Authors: Ryan Hsi, Daniel Lee, Cosmin Bejan
Language: English
Subfield: Pulmonary and Respiratory Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2795881294