Subject Data Characterization Report Meeting with University of Nebraska (C4 GPC) Date and Time Wednesday, November 11, 2015 at 10am EST Location Web Join link: Join WebEx meeting Call-in toll-free number: 1-888-687-4838 Meeting number: 642 757 567 Attendees University of Nebraska/GPC Network: Bob Greenlee, Laurel Verhagen, Jim Campbell, Jim McClay, Russ Waitman, Jerry Parker, Abu Mosa HPHCI: Jessica Sturtevant, Casie Horgan, Lindsey Petro Duke: James Topping Agenda: Item Topic 1 Discuss High-Level Metrics 2 Discuss Data Report 3 Other Questions 4 CDRN Liaison Update: James Topping Introduction 5 Next Steps Action Items: Task Team & Person Follow up with DRN OC Team (Shelley Rusincovitch) regarding possible CDM Forum discussion on encounter type mapping HPHCI Further investigation into three remaining data questions on DC report GPC (Jim Campbell) Provide U Nebraska/GPC team with location of DC SQL code HPHCI Follow up with DRN OC Team regarding version control tracking for SAS programs HPHCI Follow up with DRN OC Team regarding presentation of Data Characterization queries and process in the future HPHCI Due Date Status Current DataMart Status: Data Summary Status DataMart ready to respond to Prep-to-Research Query Awaiting follow-up on remaining data discussion topics DataMart Refresh Approved Awaiting follow-up on remaining data discussion topics Minutes: Discuss High-Level Metrics - No comments Discuss Data Report DEMOGRAPHIC - From Data Characterization Report: UNKNOWN from Clarity mapped to UN and NULL from Clarity mapped to NI ENROLLMENT - Geographic and encounter-based enrollment described in DC Report - Not all patients in the DEMOGRAPHIC table are in ENROLLMENT due to the enrollment criteria o This is helpful to know going forward Restriction of patients included in DEMOGRAPHIC not necessary at site/CDRN level The Coordinating Center will restrict (if necessary) based on study-specific criteria The purpose of DC is to characterize and better understand how data elements like enrollment are defined ENCOUNTER - Some patients in DEMOGRAPHIC table do not have associated encounter o From DC Report: This is due to conversion of historical data o Helpful to know going forward - Linear drop in encounters starting in January 2014 o Awaiting further investigation by Jim Campbell (post-Thanksgiving) o Could be due to date obfuscation method (dates shifted backward 0-365 days) o Could be due to encounters table being limited to a required # of encounters within a period of time similar to those who fit the encounter-based enrollment algorithm - ENCOUNTER table should not be limited by ENROLLMENT o Do not want to limit source data, keep source data and allow analytics to limit as required for each analytic purpose - Encounters mapped to “other” include lab draws, testing, and administrative events (appointments, canceled appointments, etc) – due to Epic system o We should be mapping encounters for health care utilization - Face-to-face encounters with a health care interaction were mapped to specific categories - DRN OC will circle back with team (Shelley Rusincovitch) regarding possible discussion on how to handle administrative encounters in a CDM Implementation Forum - ED encounter to IP admission considered two encounters - Telephone/email interaction (advice given) = Other Ambulatory as stated in PCORnet CDM o Consider removing for CDM V3 implementation - Transition to Epic in April 2012 would explain why encounters before that date were strictly “other” o Encounter metadata from historical data not sufficient quality to be mapped - DRG missingness so high because only hospitalization encounters are coded with DRGs and a small proportion of all encounters were hospitalizations - 771 values outside of CDM specifications in admit_source o Mismapped and will be investigated by Jim Campbell DIAGNOSIS - Small number of diagnoses per encounter due to large number of strictly administrative events that do not have an associated procedure/diagnosis - In DX_source, may be useful if time/resources allow to map inpatient events still in progress to IN (interim) according to PCORnet CDM - From DC Report: In PDX: ED, AV, and OA encounters do not prioritize diagnosis and thus were classified as “X” according to CDM specifications. Inpatient diagnosis not yet processed by billing were classified as “other” o Helpful information going forward PROCEDURES - Note: distribution of this data similar to data in diagnosis and encounter o o VITAL - Will be investigated further May be due to backwards date shifting as discussed above Height frequency looks as expected for an adult population Weight frequency: o Majority between 101-275 (expected with adult population) o 0 records for 276-350 with a number of records for 350+ -- possible extract error that will be investigated by Jim Campbell ETL ADD - Billing information encounters distinguished from EHR data by charge codes in PROCEDURE table (PX_source) - Unshifting dates a possibility o Consult IRB Other Questions - Location of SQL code for this query sent to all meeting attendees for reference-also available via the query tool o Past modifications to account for running on Unix machine - Software/version tracking – input and suggestions welcome - SAS decisions made by PCORI - For Phase 2, analytic and DC code will be in SAS - Version control for SAS DC: must circle back with Duke team (owners of DC workstream) - Timelines for analytics and code would be useful, as well as information on budget, priorities, etc James Topping Introduction - DCRI - Will be working as DRN OC liaison - james.topping@duke.edu