Arthur Davidson, MD, MSPHDenver Public Health
[email protected] Committee on Recommended Social and Behavioral Domains and Measures for
Electronic Health RecordsNational Academies of Sciences Building
2101 Constitution Avenue NWWashington, DC 20418
April 8, 2014 1
How to Create Information Systems With Data that Flow Both Ways
Linking EHRs Between PH, Social Service Agencies, and Other Relevant Organizations
Agenda
• Conceptual Model• Linkage for Public Health Surveillance• Linkage for Public Health Intervention• Next Steps
Linkage Conceptual Model
ClinicalCare (EHR)
ClinicalCare (EHR)
Public Health Public
Health Social
ServiceSocial
Service
Schools/ChildcareSchools/Childcare
QuitlineQuitline
Linkage for Public Health Surveillance(Colorado)
• Transparent, distributed data network– Modeled on the Mini-Sentinel (FDA) project and local
experience with HMO Research Network (AHRQ) project
• Governance – Voluntary participation; unlike mandated reporting, data use
agreements established/required
• Privacy– Minimal data necessary to achieve stated goal (de-identified
to start)
• Technical – Infrastructure: 1) common data model, 2) emphasize data
quality assessment, and 3) federated query tool
Linkage for Public Health Surveillance(Colorado)
• Colorado Health Observation Regional Data Service (CHORDS)– Provide a "laboratory" to develop and evaluate scientific
methods to support public health surveillance – Afford Denver Metro and Colorado communities an
opportunity to use existing EHR data systems for public health surveillance
– Learn about barriers and challenges, both internal and external, to building a viable and accurate system of surveillance for public health events (e.g., conditions, behaviors and outcomes)
– Build an event agnostic infrastructure for public health surveillance, quality assessment, and research
e.g., Kaiser Permanente
CHORDS RegistriesColorado Health Observation Regional Data Service
Standard (Virtual) DataWarehouse
CHORDSQuery Service(PopMedNet )
PMN
Secure federated
query
Current registry efforts:•BMI•CVD risk•Tobacco use and SHS exposure•Mental health•Colorectal cancer•Adult obesity
PMNClient
e.g., Denver Health
Standard (Virtual) DataWarehouse PMN
Client
Secure federated
query
Authorize
Authenticate
CHORDS Project Source Years
Colorado Clinical Translational Science Institute focused on regional informatics infrastructure for research; existing Cancer Center informatics expertise
NIH 2008-2018
Grant allowed initiation of virtual data warehouse (VDW) at Denver Health complementing Kaiser Permanente local expertise AHRQ 2010-
2012
Evaluation resulted in selection of Mini-sentinel PopMedNet model used FDA post-marketing surveillance
2012-2013
BMI monitoring including social and environmental and data TCHFKP-CB
2011-2013
Cardiovascular risk reduction - Community Transformation Grant CDC 2011-2016
Tobacco use, second hand smoke exposure and cessation CDPHE 2012-2015
Mental health and substance use AHRQ 2013-2017
Capacity to Support Multiple Conditions
Link Clinical, Social and Environmental Data Across Multiple Delivery Systems
Pre-CHORDS (e.g., weight status surveillance):•BRFSS self-reported demographic, weight data •Survey ~12,000 Colorado/year = 700 Denver/year•Allows county-level estimates onlyCHORDS state:•Combine measured BMI data from multiple institutions•Include demographic data, residence location (geo-code)•Link geographically aggregated BMI data (e.g. census tract) with social and environmental data•Identify “place-based” interventions (e.g., social marketing, community resource development, and policy initiatives)•Pilot features of local data sharing network
Types and Sources of Geo-coded Social and Environmental Data for Mapping
Data Source Data type
Grocery stores Reference USA Points, aggregated into census tracts
Restaurants Reference USA Points, aggregated into census tracts
Food Deserts (USDA definition)
USDA Economic Research Council At census tract level
Walkability (based on number of street intersections per unit area)
Streetmap USA/ ESRI web distribution
Points, aggregated into census tracts
Green space/parks From wide variety of sources
Polygons (areas) , with points of park entrance
Poverty American Community Survey
Polygons (areas), aggregated into census tracts
Combined Measured BMI Data
Denver County (valid BMI) :– all ages: 184,644 (31%)– adults: 119,075 (26%)– children: 64,606 (51%)
Coverage varies widely:‒ > 50% for some communities‒ few with aberrant results
CHORDS BMI Registry DescriptionChildren <18 345,756Adults 364,889
Total Sample 710,645 Median Adult BMI by Year
2009 26.52010 26.52011 27.1
Median Child BMI Percentile2009 62.92010 64.42011 65.4
Proportion of Children with Obesity, Denver
Percent Overweight + Percent of Families in Poverty
Personal Prescription - example
Linkage Conceptual Model
ClinicalCare (EHR)
ClinicalCare (EHR)
Public Health Public
Health Social
ServiceSocial
Service
Schools/ChildcareSchools/Childcare
QuitlineQuitline
e-Referral between EHR and Quitline
Goal: Efficient EHR-mediated e-Referral (including patient preferences) to Quitline and timely acknowledgement/status messages returned to and posted within the EHR.
• North American Quitline Consortium• Consensus process with ~15 Quitline vendors/service
providers• Message requirements:
– Content: define common data elements– Structure: HL7 2.x and c-CDA formats– Transport: sFTP, Direct, web-service (WSDL-SOAP)
What next?
CHORDS•Build out standard data model (i.e., add tables, required content/variables [IZ], extend time range)•Conduct comprehensive data quality assessment•Compare member-vs. visit-based denominator estimates•Expand stakeholders (PH and clinical)•Address duplicatesQuitline•Set e-Referral standards, assure meets PH needs•Vet with EHRA and standards development organization•Consider as model for PH related HIE and e-referral for other community-based services
Sustainability strategy
Enhance “event agnostic” distributed surveillance/research network (e.g., breadth of use cases and stakeholders, depth of content)
Study utility of system to multiple stakeholders (i.e., communities, elected officials, and individuals)
Facilitate incorporation of new social/environmental data (e.g., barriers and assets)
Standardize approach to geocoding and de-duplication Target outreach and community-based interventions (i.e., policy,
systems and environmental changes) to those who need them Assess impact on health disparities reduction Develop cadre of applied researchers and methods
Discussion/Questions?
