Clinical Data and Public Health Surveillance: Improving Accuracy Using a Statewide Master Patient Index Rachel Zucker, MPH September 13, 2016
What is CHORDS? • The Colorado Health Observation Regional Data Service • Facilitates electronic health record (EHR) data sharing and aggregation from 11 participating healthcare sites representing over 3 million registered patients in the Denver Metro region. • Early projects have focused on public health – Tobacco use and exposure – Obesity – Cardiovascular risk
Our Technology • Data Sharing • Common Data Model: Application: – Virtual Data Warehouse – PopMedNet ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
CHORDS Federated Query Overview Data Partners UC Anschutz Data Customer PMN Client Data Mart with Question Query Administrator Administrator Query VVVVV DO I =1 TO X; DO I =1 TO X; AJK ASKALF AJK ASKALF HJHHJKKKGFJK HJHHJKKKGFJK KFKLAFL;LKAKA KFKLAFL;LKAKA JGKGLKDGSHKJGL 0000 0000 JGKGLKDGSHKJGL LJGSKFLKG LJGSKFLKG JALFLKLKFALK JALFLKLKFALK LKLKAF LKLKAF JFLS JFLS LFALFKFLAKDF LFALFKFLAKDF JFLAKFLKFLADFLFALK JFLAKFLKFLADFLFALK KFF KFF Data ETL from May include PMN Data Firewall PopMedNet researcher, Mart Electronic Health Mart Client public health (PMN) Query (VDW) department/ Record agency, etc. Portal VDW = Virtual Data Warehouse ETL = extract transform and load ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
The problem? • Duplicate records for patients across institutions – Leads to artificially inflated counts for record # and prevalence estimates Patient 1 is an overweight Patient 1 is later seen at man whose height and the ER for difficulty weight are recorded in his breathing. His height and regular provider’s EHR. weight are recorded. ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
Query: How many patients does CHORDS have data for? CHORDS will overestimate the number of patients ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
Query: Return count of patients with BMI>25 in Denver Metro region CHORDS counts Patient 1 as two overweight people ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
The problem? • Scattered data for case identification – Leads to artificially deflated counts Patient 1 is an obese man Patient 1 is later seen at who sees his regular the ER for an asthma primary care physician attack ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
Query: Return patients with BMI>30 who have had an asthma attack The relevant criteria are in two different records; Patient 1 is not counted ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
The problem? • Both lead to inaccurate estimates for public health monitoring and inaccurate results for research • We need to link patient records to mitigate these problems ACCORDS – A DULT AND C HILD C ONSORTIUM FOR H EALTH O UTCOMES R ESEARCH AND D ELIVERY S CIENCE University of Colorado Denver | Anschutz Medical Campus
Potential solutions? Manual Linkage PPRL MPI Heavy computational “Delegating” Most resource intensive demand demand Most data exposure Manual steps Less resource intensive Additional governance required
Defining Requirements • No centralization of records within CHORDS • Minimal modification to CHORDS VDW • Limited resources (staff time, funding, technical) available to support record linkage – especially at some sites • Timeliness of data
Solution: MPI • Share existing statewide HIE MPI (CORHIO) • Best fit for existing CHORDS infrastructure and resources • Early efforts: – Obtaining required data elements – 2 site match – Use cases – Preliminary data
Getting the MPI – Ad Hoc
Getting the MPI – Periodic
Using the MPI
Matching Across Institutions 800,000 700,000 600,000 TCH 500,000 UCH 400,000 DH KP 300,000 No auto match 200,000 100,000 0 KP DH UCH TCH 20
Considerations • Ensuring Data Security • Defining Needed Fields to link MPI #s • Defining Data Transfer Processes • Will all partners work with HIE? • HIE’s purpose vs. CHORDS’ purpose
Questions? Thank you to the CHORDS team: Dr. Art Davidson and Emily McCormick at Denver Health and Jessica Bondy, Bryant Doyle and Dr. Lisa Schilling at Anschutz Medical Campus. Additional thanks to Dr. Michael Kahn at Children’s Hospital CO, and Dr. John Steiner and David Tabano at Kaiser Permanente CO. We would like to thank the Colorado Health Institute for supporting this presentation.
Recommend
More recommend
