Unsupervised Extraction of Diagnosis Codes from EMRs Using Knowledge-Based and Extractive Text Summarization Techniques

Abstract

Diagnosis codes are extracted from medical records for billing and reimbursement and for secondary uses such as quality control and cohort identification. In the US, these codes come from the standard terminology ICD-9-CM derived from the international classification of diseases (ICD). ICD-9 codes are generally extracted by trained human coders by reading all artifacts available in a patient’s medical record following specific coding guidelines. To assist coders in this manual process, this paper proposes an unsupervised ensemble approach to automatically extract ICD-9 diagnosis codes from textual narratives included in electronic medical records (EMRs). Earlier attempts on automatic extraction focused on individual documents such as radiology reports and discharge summaries. Here we use a more realistic dataset and extract ICD-9 codes from EMRs of 1000 inpatient visits at the University of Kentucky Medical Center. Using named entity recognition (NER), graph-based concept-mapping of medical concepts, and extractive text summarization techniques, we achieve an example based average recall of 0.42 with average precision 0.47; compared with a baseline of using only NER, we notice a 12% improvement in recall with the graph-based approach and a 7% improvement in precision using the extractive text summarization approach. Although diagnosis codes are complex concepts often expressed in text with significant long range non-local dependencies, our present work shows the potential of unsupervised methods in extracting a portion of codes. As such, our findings are especially relevant for code extraction tasks where obtaining large amounts of training data is difficult.

Case-Based Interpretation of Best Medical Coding Practices—Application to Data Collection for Cancer Registries

“…While our approach focuses on assisting operators in their tasks, these approaches seek to replace the need for operators in their current capacity. They focus on analyzing and annotating medical reports [10,11].…”

BioNLP 2017

“…Following the past studies Aronson et al, 2007;Kavuluru et al, , 2013Koopman et al, 2015;Patrick et al, 2007;Ira et al, 2007), the CMC dataset is preprocessed with following steps:…”

“…A standard procedure in hospital is to assign the International Classification of Diseases (ICD) codes to diseases appearing in medical texts by professional coders. As a result, several recent studies have been devoted to automatically extracting ICD code from medical texts to help manual coders Aronson et al, 2007;Kavuluru et al, , 2013 2013; Koopman et al, 2015).…”

“…The reason to conduct our study on ICD-9-CM is to compare with the state-of-art methods, whose evaluations have mostly conducted on ICD-9-CM code (Aronson et al, 2007;Kavuluru et al, , 2013Patrick et al, 2007;Ira et al, 2007;Zhang, 2008). ICD-9-CM codes are organized hierarchically, and each code corresponds to a textual description, such as "786.2, cough".…”

Trends in Multi-document Summarization System Methods

“…Automatic summaries can be used to retrieve information from EHR and complete information can be stored there if summarization can be used to retrieve the information. [44] used knowledge based and extractive text summarization method to generate ICD codes from patient's EMR by computing the c-value of noun phrases extracted by TS. The noun phrases are run on name Entity Recognition system MetaMap for recognition of biomedical concepts in EMRs and mapped them to ICD codes using UMLS Meta thesaurus.…”