MDEpiNet RAPID Meeting

Transcription

1 MDEpiNet RAPID Meeting BUILD, PCORnet & SENTINEL: Background, Data Model and Data Elements Jeffrey Brown, PhD Associate Professor May 25,

2 FDA Sentinel: Background 2007: FDA Amendments Act A mandate to create an active surveillance system Access data from 25 million individuals by July 2010 Access data from 100 million individuals by July : FDA launched the Sentinel Initiative 2009: Mini-Sentinel funded under Sentinel Initiative 2014: Funding awarded for Sentinel Operates under FDA s public health authority 2

3 Sentinel Partner Organizations Lead HPHC Institute Data and scientific partners Scientific partners Institute for Health 3

4 Sentinel Data Partners Starting

5 Sentinel Common Data Model and System Architecture 5

6 Key database model-related questions (2009 white paper) What does the system need to do? What data are needed to meet system needs? Where will the data be stored? How will the data be analyzed? Is a common data model needed, and if so, what will the model look like? 6

7 7

8 Uses of the Sentinel System Anticipated primary functions include Medical product safety surveillance Confirmatory safety studies (hypothesis evaluation) Safety surveillance data mining (hypothesis generation) Monitor adoption, diffusion, and use of new medical products Augment registry information (e.g., medical devices) Additional uses and needs identified by FDA Centers Quickly calculate background incidence rates for outcomes of interest Better information regarding the predictive value of diagnosis codes of interest in automated health care data Identify immediate adverse events (e.g., transfusion-related acute injury) Various additional uses were identified by non-federal stakeholders 8

9 Basic capabilities Timely access to data from large populations representative of individuals exposed to the products FDA regulates Ability to validate exposures and/our outcomes (e.g., through access to medical charts) Both routine and ad hoc surveillance Rapid response to public health emergencies 9

10 Mini-Sentinel Common Data Model v1.0 Described populations with administrative and claims data Had well-defined person-time for which medicallyattended events are known Data areas Enrollment Demographics Outpatient pharmacy dispensing Utilization (encounters, diagnoses, procedures) Mortality (death and cause of death) Based on medical encounters and patient demographics 10

11 Mini-Sentinel Common Data Model v1.0 Each Data Partner translated local source data to the common data model structure and format and documented the process in a detailed report Questions and issues were discussed on weekly teleconferences Transformed data were quality-checked and characterized using standard programs developed by the Mini-Sentinel Operations Center 11

12 Over time, we expanded the model 12

13 Over time, we on-boarded new Data Partners (some who contributed new data) 13

14 Critical questions for extracting data from EHR into analysis-ready form for secondary use Are the data elements needed captured in the data ecosystem? Are they recorded/captured in a systematic, consistent way Within facilities? Across facilities? By clinical staff within facilities? By clinical staff across facilities? Are they collected/input/stored in structured, semi-structured or unstructured/free-text form? Are there existing allowable values If no, can values be categorized without specific clinical knowledge/expertise into allowable values or categories? 14

15 Sentinel Common Data Model 15

16 Sentinel Common Data Model: One patient, one encounter DEMOGRAPHIC PATID BIRTH_DATE SEX HISPANIC RACE zip PatID1 2/2/1964F N ENCOUNTER PATID ENCOUNTERID ADATE DDATE ENCTYPE PatID1 EncID1 10/18/ /20/2005IP ENROLLMENT PATID ENR_START ENR_END MEDCOV DRUGCOV PatID1 7/1/ /31/2006Y Y PatID1 9/1/2007 6/30/2009Y Y DISPENSING PATID RXDATE NDC RXSUP RXAMT PatID1 10/14/ PatID1 10/14/ PatID1 10/17/ PatID1 10/17/ PatID1 10/21/ PatID1 10/21/ PatID1 10/21/ PatID1 10/22/ PatID1 10/23/ DIAGNOSIS PATID ENCOUNTERID ADATE PROVIDER ENCTYPE DX DX_CODETYPE PDX PatID1 EncID1 10/18/2005 Provider1 IP P PatID1 EncID1 10/18/2005 Provider1 IP S PatID1 EncID1 10/18/2005 Provider1 IP S PatID1 EncID1 10/18/2005 Provider1 IP P PatID1 EncID1 10/18/2005 Provider1 IP S PatID1 EncID1 10/18/2005 Provider1 IP S PatID1 EncID1 10/18/2005 Provider1 IP S PROCEDURE PATID ENCOUNTERID ADATE PROVIDER ENCTYPE PX PX_CODETYPE PatID1 EncID1 10/18/2005 Provider1 IP 84443C4 PatID1 EncID1 10/18/2005 Provider1 IP 99222C4 PatID1 EncID1 10/18/2005 Provider1 IP 99238C4 PatID1 EncID1 10/18/2005 Provider2 IP 27445C4 16

17 Sentinel Common Data Model: Device table? ENCOUNTER PATID ENCOUNTERID ADATE DDATE ENCTYPE PatID1 EncID1 10/18/ /20/2005IP PROCEDURE PATID ENCOUNTERID ADATE PROVIDER ENCTYPE PX PX_CODETYPE PatID1 EncID1 10/18/2005 Provider1 IP C4 PatID1 EncID1 10/18/2005 Provider1 IP C4 PatID1 EncID1 10/18/2005 Provider1 IP C4 PatID1 EncID1 10/18/2005 Provider2 IP C4 DEVICES PATID ENCOUNTERID ADATE PROVIDER ENCTYPE PX PX_CODETYPE UDI PatID1 EncID1 10/18/2005 Provider2 IP C PatID1 EncID1 10/18/2005 Provider2 IP C

18 Ways Sentinel might assist device surveillance Data for largest national insurers can be linked to UDI registries Sentinel data model can be used to capture UDIs Sophisticated analytic tools to analyze data in Sentinel data model format 18

19 PCORnet: the National Patient-Centered Clinical Research Network PCORnet is a large, highly representative, national patient-centered clinical research network. Our vision is to support a learning U.S. healthcare system and to enable large-scale clinical research conducted with enhanced quality and efficiency. Our mission is to enable people to make informed healthcare decisions by efficiently conducting clinical research relevant to their needs. 19

20 PCORnet embodies a community of research by uniting people, clinicians & systems 20 Patient-Powered Research Networks (PPRNs) 13 + = Clinical Data Research Networks (CDRNs) PCORnet A national infrastructure for people-centered clinical research 20

21 Resulting in a national evidence system with unparalleled research readiness Sex Race Age Pool of patients Female Male PCORnet represents: ~110 million patients White Non-White Missing 0 4 who have had a medical encounter in the past 5 years For clinical trials 42,545,341 For observational studies *some individuals may have visited more than one Network Partner and would be counted more than once 83,131,450 Missing 21

22 Underpinned by a Common Data Model Same data are represented differently at different institutions (e.g., Race)

23 Underpinned by a Common Data Model Same data are represented differently at different institutions (e.g., Type of Encounter) SITE 1 Social Work Visit Allied Health Office Visit Nurse Visit Procedure Visit Employee Health Vascular Lab Sleep Study Visit Social Work Visit SITE 2 Office Visit Specimen Postpartum Visit Clinical Support Initial Prenatal SITE 3 Home Care Visit Office Visit Therapy Visit Orders Only Cardiology Testing Hospital Encounter Common Data Model Ambulatory Visit (AV) Emergency Department (ED) ED Admit to Inpatient (EI) Inpatient Hospital (IP) Non-Acute Inst. Stay (IS) Other Ambulatory (OA) Other (OT) Unknown (UN) No Information (NI) In order to be able to trust results of an analysis, we need to have consistent representations

24 Underpinned by a Common Data Model Same data are represented differently at different institutions (e.g., Race) SITE 1 Caucasian African American Asian Multiple Race Blank SITE SITE 3 African American American Indian Asian American White Other Unknown Common Data Model Value Set 01 = American Indian or Alaska Native 02 = Asian 03 = Black or African American 04 = Native Hawaiian or Other Pacific Islander 05 = White 06 = Multiple Race 07 = Refuse to Answer NI = No Information UT = Unknown OT = Other In order to be able to trust results of an analysis, we need to have consistent representations

25 Data model is necessary but not sufficient condition for a viable distributed network Other requirements Guiding principles for data model and analytics Conformance with the data model Consistency in implementation across sites Data quality and characterization Distributed analytic tools Secure communications Data partners buy-in 25

26 Thank you! 26