Transcript Electronic Health Data for Research

How do data and
creativity relate?
2007
“gets the right care to
people when they
need it and then
captures the results for
improvement”
Requesting Data
http://www.popmednet.org/
• 12 networks
• Opt in and out of query requests
Tracking Definitions
https://www.phekb.org/
Inventory
Automated
queries
Custom
queries
Statistical
Tools
WWAMI region Practice &
Research Network
• ~50 Primary care WWAMI
clinics
• ~15 data connected clinics
• CHCs and RHCs
• Underserved populations
• Many serving rural populations
• Collaboration with national
network of practice based
research networks
Native American Clinical
Research Network
• ~20 Tribal communities
• 1 data connected clinic
• Serving American Indian
populations in urban and
reservation community settings
Examples
• Behavioral health
• Benchmarking – 1.5 million patient lives
• Chronic kidney disease
• Colorectal cancer screening
• Northwest Pharmacogenomic Research Network
• Acute Pain
• Lung cancer screening
• HealthShare Montana HIE Research Portal
• Chronic opioid therapy best practices for noncancer chronic pain
• Joined Alcohol and Drug Abuse Institute’s
Clinical Trial Network
Seeking
New Grants
Individual
trials
• Behavioral
Health
• Teratogens
• Diabetes
Networks
• Alcohol and
Drug Abuse
Institute –
Clinical Trials
Network
• PCORNet
• CMMI Practice
Transformation
Network
Hurdle free inventory
Discover the breadth and depth of data
across the network
Federated Information Dictionary Tool
https://dataquest.iths.org/
• Creating views of metadata and simple
interactive summaries to convey content
and usability
• Improve comprehension of the dataset to
allow for creative thinking about research
questions and goodness of fit for research
projects
• Share a standard data dictionary
Pitfalls
• Denominator, do you have the
population you think you have
• Extraction errors
• Inconsistent data over time
• Variability in code uses across sites
(ICD-9/10, CPT)
• Lack of consistent harmonized codes
across (medications, labs)
• Medication data are complex, no dose
quantities unless it’s fill data
• Lack coded key info (med hx, fam hx,
smoking status, race/ethnicity)
• Difficulty defining events or
relationships (i.e., a “visit” or “stay,”
who the primary care provider is)
Recommendations
• Get good consultation
• Create data validation procedures
• Use test datasets early in your studies to
work out the kinks
• Expect messiness
• Create the right exclusion and inclusion
criteria
• Understand the data provenance