Sifting Through Lines, Events, and Trees: Stories of EHR Visual - PDF document

5/28/2014 Sifting Through Lines, Events, and Trees: Stories of EHR Visual Exploratory Analysis Christopher Herrick, Vivian Gainer, Alyssa Goodson, Shawn Murphy Victor Castro, Kay Mahamaneerat, Yanbing Wang, and Taowei David Wang tdwang@partners.org May 29, 2014 • Research Information Services & Computing (Shawn Murphy) • Curate clinical data for research • IRB-approved, HIPAA-compliant, auditable • i2b2, RPDR, mi2b2 1

5/28/2014 • Research Information Services & Computing • Curate clinical data for research • IRB-approved, HIPAA-compliant, auditable • i2b2, RPDR, mi2b2 Dynamic Query Lifelines2 EventFlow Treemap + Piccolo • Research Information Services & Computing • Curate clinical data for research • IRB-approved, HIPAA-compliant, auditable • i2b2, RPDR, mi2b2 - Mediated access to clinical PACS - Cloud storage - Download to local machines All data on this slide is de-identified 2

5/28/2014 Dynamic Query • Research Information Services & Computing • Curate clinical data for research • IRB-approved, HIPAA-compliant, auditable • i2b2, RPDR, mi2b2 All data on this slide is de-identified Dynamic Query • Research Information Services & Computing • Curate clinical data for research • IRB-approved, HIPAA-compliant, auditable • i2b2, RPDR, mi2b2 Filter All data on this slide is de-identified 3

5/28/2014 Lifelines2 Data Extract, Transform, Load (ETL) • Searching for the right patients for a comparison is difficult • SQL is complex, prone to mistakes • Mistakes are hard to spot Lifelines2 Data ETL: Glargine & Pancreatic Cancer • Searching for the right patients for a comparison is difficult • SQL is complex, prone to mistakes • Mistakes are hard to spot Order of Cancer and Glargine is not enforced. cancer glargine c Nor is the time taken from Glargine to Cancer taken into account. All data on this slide is simulated (i.e. made up) 4

5/28/2014 Lifelines2 Data ETL: Glargine & Pancreatic Cancer • Searching for the right patients for a comparison is difficult • SQL is complex, prone to mistakes • Mistakes are hard to spot Order of Cancer and Glargine is not enforced. cancer glargine spike c Nor is the time taken from Glargine to Cancer taken into account. All data on this slide is simulated (i.e. made up) Diabetes Treatment Progression: SGLT-2 • Type 2 diabetes mellitus (DM2) has complex causes and treatments • New class of diabetes drug: SGLT-2 • Interested in its usage, in particular: • Characteristics of patients on different drug classes • Demographics, measurements before/after taking a drug class, etc. • What drugs were used prior to prescribing SGLT-2 • EventFlow • Phenotypes of patients on SGLT-2 vs other diabetes patients • Treemap 5

5/28/2014 EventFlow Years Before 8 6 4 2 SGLT-2 All data on this slide is simulated (i.e. made up) EventFlow Years Before 8 6 4 2 SGLT-2 All data on this slide is simulated (i.e. made up) 6

5/28/2014 Comparing Large Trees: Diabetes Drugs • Two populations of patients (e.g. CGLT-2 vs Other) • Hierarchically organized diagnosis code (e.g. ICD9) (SGLT-2) (Other) All data on this slide is simulated (i.e. made up) Treemap+ Piccolo All data on this slide is simulated (i.e. made up) 7

5/28/2014 Treemap+ Lifelies2 Piccolo All data on this slide is simulated (i.e. made up) Thank you tdwang@partners.org 8