Presentation is loading. Please wait.

Presentation is loading. Please wait.

Methods We employ the UMLS Metathesaurus to annotate ICD-9 codes to MedDRA preferred terms (PTs) using the three-step process below. The mapping was applied.

Similar presentations


Presentation on theme: "Methods We employ the UMLS Metathesaurus to annotate ICD-9 codes to MedDRA preferred terms (PTs) using the three-step process below. The mapping was applied."— Presentation transcript:

1 Methods We employ the UMLS Metathesaurus to annotate ICD-9 codes to MedDRA preferred terms (PTs) using the three-step process below. The mapping was applied to two databases: an electronic health record (EHR) and an administrative claims database. Normalized analyses are then conducted across sources to explore drug-condition associations. Results 21% of 17,505 MedDRA PTs have ≥1 ICD-9 annotation. Of the 14,246 distinct ICD-9 codes in the EHR, 86% mapped to MedDRA and 13% were E/V codes; 65 codes were unclassified, representing <0.01% of 32m condition eras in the data. In claims, 99.97% of 1.2b condition eras were successfully classified. Conclusions Establishing MedDRA as a common language to use in drug safety analyses allows for meaningful comparisons across disparate data sources. Mapping ICD-9 references to MedDRA categories enables systematic analyses of observational data that can be integrated into the overall pharmacovigilance process. Effectiveness of normalization can vary by source database, as utilization of ICD-9 codes differ based on intended purpose. In both databases, unclassified codes were rarely used and commonly appeared to be invalid ICD-9 codes or didn’t refer to true medical conditions. The construction and annotation of the SafetyWorks condition ontology demonstrates the value and power of MedDRA and the UMLS Metathesaurus in enabling analytic and knowledge exploration methodologies. Next Steps l Improvements to the ICD-9/MedDRA mapping by making use of the MTHICD9 UMLS vocabulary and more subtle relations available in the MRREL source. l Investigation of a subset of SNOMED CT as the primary medical conditions ontology, and in using it in conjunction with MedDRA for knowledge exploration and display purposes. References l U.S. National Library of Medicine, Unified Medical Language System at http://www.nlm.nih.gov/research/umls/. l Merrill GH, Ryan PB, Painter JL. Using SNOMED to Normalize and Aggregate Drug References in the SafetyWorks Observational Pharmacovigilance Project. KR-MED. May 2008. l Ryan PB, Powell GE. Exploring Candidate Differences Between Drug Cohorts Prior to Exposure: A Systematic Approach Using Multiple Observational Databases. International Society for Pharmacoeconomics and Outcomes Research, May 2008. l Ryan PB, Mera R, Merrill GH. Opportunities and Challenges in Leveraging Observational Data for Pharmacovigilance. AMIA Pharmacovigilance and Informatics Summit, June 2007. Background Observational databases, such as electronic health records and administrative claims databases, can provide a valuable supplement in assessing effectiveness and long-term safety outcomes of medicines. The SafetyWorks project at GlaxoSmithKline has developed an integrated set of methods to support the use of large observational data sources for monitoring and assessing drug safety. A key challenge in using these data sources to supplement pharmacovigilance activities is to establish a common vocabulary for making disparate analyses meaningfully comparable. A critical step in this is the use of the SafetyWorks condition ontology in normalizing condition references across disparate data sources. The Medical Dictionary for Regulatory Activities (MedDRA) is a unified standard terminology for recording and reporting adverse drug event data (Figure 1). However, many observational databases use the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9) to assign codes to diagnoses and procedures associated with hospital utilization. SafetyWorks uses MedDRA as a reference ontology and annotates source ontologies, such as ICD-9, to enable raw data from disparate sources to be normalized into a common set of condition concepts to be used for analysis. Figure 2 illustrates the extraction of raw data from the GlaxoSmithKline Healthcare Information Factory (a repository of large disparate observational databases), the normalization and aggregation of the raw data by means of medical condition and drug ontologies, and the use of this normalized and aggregated data in observational screening and risk estimation. Defining medical conditions by mapping ICD-9 to MedDRA: A systematic approach to integrating disparate observational data sources for enabling enhanced pharmacovigilance analyses Patrick B Ryan, Jeffery L Painter, Gary H Merrill GlaxoSmithKline Research & Development©, Research Triangle Park, NC, USA Method 1: Direct CUI mapping Table 1: Coverage of ICD9 codes and instances Method 2: CUI equivalence from LLT to PT Method 3: ICD-9 Boosting Normalize & Aggregate Data Extract Data Construct Ontologies Annotate Ontologies HIF UMLS Observational Screening Risk Estimation Safety Scientist Review Figure 2: SafetyWorks analysis process Figure 1: MedDRA Hierarchy


Download ppt "Methods We employ the UMLS Metathesaurus to annotate ICD-9 codes to MedDRA preferred terms (PTs) using the three-step process below. The mapping was applied."

Similar presentations


Ads by Google