• Document: How to extract transform and load observational data?
  • Size: 2.46 MB
  • Uploaded: 2018-10-12 00:42:22
  • Status: Successfully converted


Some snippets from your converted document:

How to extract transform and load observational data? Martijn Schuemie Janssen Research & Development Department of Pharmacology & Pharmacy, The University of Hong Kong Outline • Observational data & research networks • OMOP Common Data Model (CDM) • Transforming data to the CDM • Available tools for the CDM Observational data & research networks Observational health data Subjective: Objective: Assessment and Plan: • Complaint and • Observations • Diagnosis symptoms • Measurements – vital • Treatment • Medical history signs, laboratory tests, radiology/ pathology findings Observational health data • All captured in the medical record • Medical records are increasingly being captured in EHR systems • Data within the EHR are becoming increasingly structured The average primary care physician sees 20 different patients a day care data …that’s 100 visits in a week… care data …which can reflect a panel of over 2,000 patients in a year care data Databases for research • Diseases • Health care provided Evidence • Effects of treatments generation • Differences between patients • Electronic Health Records • Insurance claims • Registries • Personalized medicine Key point Existing observational health care data such as Electronic Health Records and insurance claims databases have great potential for research 现有的卫生保健观测数据,例如电子健康记 录、保险索赔数据库等具有巨大的潜力为研 究所用 Research networks • Data may be at different sites • Sites often cannot share data at the patient level • Data can be in very different formats Patient level, identifiable information Practice Hospital Claims Registry Aggregate summary Standardized analytics statistics Firewall Common Data Model Extract, Transform, Load (ETL) ETL ETL ETL ETL Patient level, identifiable information Practice Hospital Claims Registry Count people on drug A and B, and the number of outcomes X Firewall Common Data Model Extract, Transform, Load (ETL) ETL ETL ETL ETL Patient level, identifiable information Practice Hospital Claims Registry Count people on drug A and B, and the number of outcomes X Firewall Common Data Model Practice: 100 patients on A, 1 has X 200 patients on B, 4 have X Extract, Transform, Load (ETL) ETL ETL ETL ETL Hospital: 1000 patients on A, Patient 10 have level, X identifiable 2000 patients on B, 40 have X informationAggregated data Etc. Practice Hospital Claims Registry No privacy concerns Distributed research networks • Europe • United States • Asia AsPEN • Global Key point Observational data can often not be shared. Using a common data model allows analysis programs to ‘visit’ the data instead. 观测数据通常不是共享的。而使用公共数据 模型可以实现分析程序对不同来源数据的统 一“访问” OMOP Common Data Model (CDM) Common Data Model • A common structure Person - person_id - year_of_birth - month_of_birth - day_of_birth - gender_concept_id • A common vocabulary How do we store gender? - M, F, U - 0, 1, 2 - 8507 (male), 8532 (female), 8851 (unknown gender), 8570 (ambiguous gender) OMOP Common Data Model • Designed fo

Recently converted files (publicly available):