1. Data Quality in Healthcare
EMRs vs. Population Health

3. Vermont Blueprint for Health – Sprint Data Quality Program
Vermont Blueprint – Sprint Program
13 Health Service Areas
147 Practices
21 different EMRs
325k managed patients
Chronic Care Information Registry
All Payer Claims Database
Health Information Exchange
Support fot State Wide Pop Health Program – Community Health Teams – All Payer Model
Intensive Team Based Approach was used to attack the problem
Development of the Sprint Data Quality Program based on failures seen in the capabilities of data aggregation and reliable reporting on core measure sets

4. Overview of Data Quality Program
Population Health Management vs. Episodic Care
Sources & Types of Healthcare Information
Issues and Causes of Data Anomalies
Practice Management and ADT Systems Issues
EMR Clinical Data Issues
Data Transport Issues
Claims Data Issues
Remediation of Problem Data
Data Blocking Problem
Keeping Systems and Data Healthy
What we are going to talk about today

5. Population Health Management vs. Episodic Care
Episodic Care Teams Utilize a Single Patient’s Records
Regularly Permits Global Anomalies in Data to Occur
Population Health Teams Utilize the Entire Population’s Records
Inpatient and Outpatient EHR data
ADT Data
External Lab, Pharmacy and Prescription Feeds
Claims Data
Exposes Anomalies in Global Data
Core Measure Reporting in an Fee For Value Environment
Anomalies – Developed due to single record use – issues are not readily seen as each record is viewed individually.
these anomalies are exposed when EMR data is aggerated in a Pop health or data analytics system
Core measure reporting in an FFV program can be greatly affected by the hidden data issues

6. Sources & Types of Healthcare Data
Practice Management Systems (PM) & Admissions, Discharge & Transfer Data (ADT)
Patient Demographics, Identifiers & Insurance
Physician Demographics and Identifiers
Provider/Patient & Site Attribution
Patient Status Information
Hospital ADT – Diagnosis, Results, Procedures & Discharge Summary
Electronic Medical Records Systems
Structured Clinical Data
Coded vs. Uncoded
Free Text Data
Demographics are traditionally dirty – Records Matching is a problem due to missing Demographic components from various components
Duplicates cause issues
Different Identifier schemas create a one to many matching challenge
The same occurs in Provider demographics

7. PM & ADT Patient Demographics & Identifiers
Inaccurate/Misspelled Names
Outdated Contact Information
Multiple Instances of the Same Person
Patient Identifiers
Missing or Multiple Medical Record Number (MRN)
Erroneous MRNs (Third Party Systems)
Utilization of Modifiers
Data Anomalies from Merged Systems
Middle name initial missing causes issues with MPI matching
EMPI Enterprise M Patient Index
Enterprise Master Provider Index

8. PM & ADT Patient Panel & Attribution
Active and Inactive Patients
Deceased Patients
Special Patient Types
Dental Only
Special Clinic (HIV)
CFR 42 Part II Data
Insurance
Race/Ethnicity/Language

9. PM & ADT Patient Panel

10. PM & ADT – Provider Demographics & Identifiers
Provider Data
Outdated Provider List
Patient Attribution
Patients Attributed to Non-existing Providers
PCP Assignments – Other Provider Types
Site Attribution
Mismatched System Nomenclatures
Provider Identifiers
Missing NPI Numbers (Interface Issue)
Multiple NPI Numbers

11. Clinical Data Types & Issues
Structured Data
Despaired System Nomenclatures
CPT, ICD-9/10, SNOMED, CVX
Un-coded, Outdated Codes
Updated Code Sets
Free Text Data
Human Nomenclatures
Transport of Meaningful Data
Continuity of Care Documents
Machine Readable vs. Human Readable
Flat File Interfaces
HL7/FHIR
Labs - Pharmacy
Mapping and Translation
Creating Meaningful Population HM Data
use of different code sets - CPT vs CVX for Immunization; outdated loinc or CVX codes.
Another issue is the lab compendiums from the site to the hospital labs and back to the EMR must stay in sync

13. CCD Coding Issues

14. CCD Human Readable vs. Machine Readable
Human Readable Side – Observations – BMI and Height
Human Readable Side Has the Data – Good for viewing in a PDF as a document attached to a patient record

15. CCD Human Readable vs. Machine Readable
Machine Readable Side – BMI
Here is the machine readable side of the same CCD – The BMI code is not known – This data will fall on the floor and not be included in the pop health system and may or may not report as an error

16. CCD Human Readable vs. Machine Readable
Machine Readable Side – Height
Here we note that in the same message the height is included. – When we examined this entire CCD we found 15 such issues pertaining to important clinical data

17. Data Remediation Post Go-Live Remediation Significant Level of Effort
Intensive Resources
Multi-Level Problems
Pre Go-Live Remediation
Remarkably Easier to Accomplish
Establishment of Clean ADT Data
Cleaner Mapping of Clinical Data
Working with the EMR Vendor and HIE/IT Team
Catch it early – once data is committed to down stream systems – pop health or other EMRs and data analytics systems – it becomes much harder to deal with

18. Keeping Data Healthy Monthly Review of Provider List
Active/Inactive Patients
Deceased Patients
Notification to Receiving Parties
System Upgrades
Changes in Data Collected
Integration of PHRs & Portals
Maintaining Best Practices
Data Entry Policies and Support at the Practice Level
CCD Coding Review
Regular Review of Data Quality
It is important to conduct system reviews quarterly – as well as after version upgrades of the EMR

19. Sprint Program Outcomes
Data Quality Issues Reduced to <5% of All Data Collected
Corrected 4 Different EMR – CCD Global Issues
Developed a Reliable Work Around for Certain EMR Systems that Blocked Data
Developed a Direct Path for Practices that Could/Would Not Connect Thru the HIE

20. Data Blocking
Information blocking occurs when persons or entities knowingly and unreasonably interfere with the exchange or use of electronic health information.
Interference. Information blocking requires some act or course of conduct that interferes with the ability of authorized persons or entities to access, exchange, or use electronic health information.
This interference can take many forms, from express policies that prohibit sharing information to more subtle business, technical, or organizational practices that make doing so more costly or difficult.
Knowledge. The decision to engage in information blocking must be made knowingly.
No Reasonable Justification. Accusations of information blocking are serious and should be reserved for conduct that is objectively unreasonable in light of public policy.
Public policy must be balanced to advance important interests, including furthering the availability of electronic health information for authorized and important purposes.
Interesting Study by Julia Alder of the University of Michigan
A New Look at Economic Barriers to Interoperability
Julia Adler-Milstein, Phd
University of Michigan

21. Information Blocking Survey: Frequency
A New Look at Economic Barriers to Interoperability
Julia Adler-Milstein, Phd
University of Michigan

22. Frequency of Information Blocking Behaviors: EHR Vendors

23. Frequency of Information Blocking Behaviors: Hospitals/Health Systems