Transcript Clinical Informatics & Applications

Clinical Informatics &
Applications
Javed Mostafa
Biomedical Research & Imaging Center
School of Information & Library Science
Translational & Clinical Sciences Institute
May 15, 2009
EPID 896 Clinical Research Curriculum seminar
Outline
• Informatics
– Roots and evolution
– Emergence of clinical informatics
• Data management and data mining
• Carolina data warehouse for health
Informatics
• A discipline which is concerned with effective
and efficient use of computing to promote
discovery, creativity, decision-making, and
productivity
– A wide variety of sub-disciplines exists
An analogy
Electrical
Engineering
Engineering
& Applied Math
Mechanical
Engineering
Civil
Engineering
Few Informatics Examples
Informatics in Relation to Medicine &
Health
• Many associated domains exist, sometimes
leading to confusion .. .
– Bionformatics
– Health informatics
– Biomedical informatics
– Medical informatics
– Clinical informatics
• Additionally … nursing informatics, public health
informatics …
Huang, R.Q. (2007). Competencies for graduate curricula in health, medical and biomedical informatics: a
framework, Health Informatics Journal, Vol 13(2): 89–103.
Clinical Informatics
• American Medical Informatics Association
(AMIA) recently approved the Core Content of
of Clinical Informatics
– Clinical Informaticians transform health care by
analyzing, designing, implementing, and
evaluating information and communication
systems
… that enhance individual, population health outcomes,
improve patient care, and strengthen the clinicianpatient relationship
Gardner et al. (2009). Core content for the subspecialty of clinical informatics. JAMIA, 16(2), 153-157.
Critical Areas of Clinical Informatics
Clinical Care
The Health
System
Clinical Informatics
Information &
Communications
Technology
• Care – provision of service to an individual
• Health system – organization, policies, quality,
data management
Critical Areas in CI: Information
Systems
• System development & integration
• Networks
• Security
• Data representation, manipulation, and
sharing
A Key Challenge in CI: Data
Management
• Volume of data growth is rapid
• Type of data is heterogeneous
• Need systematic way to aggregate
– For retrieval and analysis
• To support decision making, quality control, and longterm projects such as research
Hersh, W. (2009). Information Retrieval: A Health and Biomedical
Perspective, NYC:Springer.
Evolution of Data Management
1960s
1970s
1980s
Hierarchical
Traditional
files
1990s
Objectrelational
Relational
Network
Objectoriented
2000+
Web-integrated
multimedia DBs
?
Relational Model
Relation is a term that comes from mathematics
and represents a simple two-dimensional
table. Representation based on logical
associations only! No pointers …
Name
Relation = Table
Job
Branch
Relational Model
• 1980-1990+
• E.F. Codd proposed the Relational Model
• Simple and elegant and scales with ease
• Combined with Structured Query Languages
(SQL) offers a powerful mechanism for data
organization and access
DW Multidimensional Model
• Example of Two- Dimensional vs. MultiDimensional
Two Dimensional Model
T hree d imensio nal d at a cub e
REGION
REG1
P
R
O
D
U
C
T
REG2
REG3
P123
P124
P125
P126
:
:
P
r
o
d
u
c
t
P1 2 3
r
r t e tr 4
a
u
Q
l Q tr 3
a
c
Q
F i s tr 2
Q
1
r
Qt R e g 1
Reg 2 Reg 3
P1 2 4
P1 2 5
P1 2 6
:
:
Region
Multidimensional Star Schema
• Star schema:
– Consists of a fact table with a single table for each
dimension.
DW OLAP
• OLAP – OnLine Analytical Processing
– Fast analysis of shared multidimensional
information (FASMI)
• Data mining is a critical aspect of OLAP
DW Data Mining
• Prediction:
– Determine how certain attributes will behave in the future.
• Identification:
– Identify the existence of an item, event, or activity.
• Classification:
– Partition data into classes or categories.
• Optimization:
– Optimize the use of limited resources.
• Referred to as PICO …
Carolina Data Warehouse for Health Evolution
•
UNC health care system started developing electronic medical records almost 20
years ago
•
Inpatient and outpatient care in UNC hospitals, clinics and affiliated satellite
practices throughout central North Carolina
•
Paperless with full nursing notes, physician order entry, progress notes, laboratory,
procedure notes, discharge summaries, medication lists, and the ability to write
prescriptions available on-line
•
24/7 used by over 1900 physicians, 3000 nurses, with hundreds of thousands of
patients each year
•
Two years ago UNC Health Care System (UNCHCS) initiated development of an
enterprise-wide data warehouse, the Carolina Data Warehouse for Health (CDWH), to meet the dual challenges of enhancement of quality of care and clinical
research with our patient populations (invested > $7 million so far)
CDW – H Strategic Vision
Source
Databases
Applications, Analysis
Tools, Search, Query,
Mining, ..
GE IDX
External
Collaborators
caBIG
Pay 4 Perform
Cohort Analysis
Public Health
Quality Reporting
Patient Safety
Siemens
DSS
Portal Layer
Outcomes
Tumor
Registry
Selective
Text
Extraction
Clinical Registries
Rim?
Systems
SOA
SOA
SOA
Applications
Applications
Applications
Collaboration Layer
WebCIS
CDR
Security Layer
Other
Operational
Systems
Research
Genomics,
Proteomics, etc.
Administrative
Pillar
Information Federation Layer
Staging
Images
Extract
Transform
Load
Cleansing
Linking
Conforming
Data
Warehouse
PubMed
dbSNP
Other…
Federated Data Sources
External & Internal
Biological, Images, Literature, etc.
Secure Exchange
of information
with outside
entities.
CDW-H: As It Is Now …
• A retrospective, persistent record of cleansed,
transformed, and stored data originating from
operational systems
• The “one source of truth” for reporting, analytic,
and data mining
• Data organized logically into subject areas for the
user’s benefit without regard to its source
system
• Reports, analytics, and decision making will be
consistent across the entire organization
CDW-H: As It Is Now …
• Data is refreshed periodically (24-48 hrs) and is
not real time data
• CDW is not designed to replace or augment daily
operational activities, but to support those
activities through analytical retrospective
processes
• Designed to address overall organizational
priorities under the governance of the CDW
Oversight and Operations Committees
CDW-H: As It Is Now …
• Major Subject Areas in CDW include:
Account
Allergy
Ambulatory Claim
Charge
Contact Information
Core Measures
Diagnosis
Drug
Drug Order
Health Maintenance
Immunizations
Lab Results
Medications
Observation
Order
Organization
Patient
Patient Infection
Notes and Reports include:
•Ancillary Reports
•Cardiology Reports
•Clinical Notes
•ECG Reports
•GI Reports
Patient Readmission
Patient Visit Provider
Payer
Payment
Problem
Procedure
Provider
Vital Signs
Data Set Size
•
•
•
•
•
•
•
•
Number of Tables in Staging area:
Number of Columns in Staging area:
Number of Tables in ADS:
Number of Columns in ADS:
Number of Tables in Inpatient Datamart:
Number of Columns in Inpatient Datamart:
Number of Tables in Diabetes Datamart:
Number of Columns in Diabetes Datamart:
• Total number of unique Patients:
• Total number of unique Accounts:
219
3,849
202
2,840
81
1,581
21
504
1.8 Million
4.5 Million
Data Marts
•
Focused subset of atomic store data to support specific analytical requirements
……
•
The data is organized by Dimension and Facts
•
Fact Tables contain the desired detailed information
– Diabetes Facts: Last A1c, Last LDL, BP, Bilateral Amputee, Onset Date, Insulin Use, Micro
Albumin, etc.
•
Dimensions are distinct threads of information that allow the facts to be
summarized in specific ways
– Diabetes Dimensions: Patient, Clinic, Provider, Date, Visit, etc.
•
Dimensions are expanded fully to provide the aggregation required
– For example, the date dimension would specify the calendar date, the day of the week,
weekday / weekend, month, quarter, and year.
Topics Covered in the Diabetes Data Mart
Subject Areas:
•Allergies
•Discharge Medications
•Drug Orders
•Health Maintenance
•Lab Results
•Patient Diagnosis
•Patient Medications
•Patient Problems
•Patient Procedures
•Patient Providers
•Visits
•Vitals
Dimensions:
•Allergen
•Clinic
•Date
•Diagnosis
•Division
•Drug
•Drug Order Master
•Health Maintenance
Category
•Health Maintenance
Standard QA
•Hospital Service
•Lab Tests
•Order Master
•Patient
•Procedure
•Provider
•System User
•Visit
•Vital Master
Facts:
•Diabetes
•Diabetes Clinical Measures
•…
Diabetes: Dimensions and Facts
Research Portal: Gateway for
Researchers and Students
• An application to expose the various key
features of the CDW-H in a user friendly way
– Metadata and business terms
• A portal to find useful related resources and
services related to the CDW-H
• Currently, offers a Cohort Discovery Service as
a pre-research step
Medical Record Access: Challenges
De-identified data
Limited data set
Conduct Study - with
a particular cohort
Clinical data set
Access & Approval
De-identified view
IRB Approval/Data Use Agreement
De-identified view
Extracted Data for a Study
Summary of Access Rules
• The following table summarizes the basic
documentation requirements
Level of Access
Rule
Scope of Data
De-identified
No authorization
needed
Must not contain any
HIPAA defined data
elements that may
potentially reveal
identity
Limited data set
Requires signed Data
Use Agreement
Largely de-identified
PHI but may include
some identifiers
Complete set
Authorization /
Wavier of
Authorization
PHI that includes
identifiers beyond
the limited fields
Cohort Selection Demo
• Project Summary Descriptions:
– Need to determine which woman with digital
mammograms performed at UNC between May 2007 and
June 2008 who also have a documented history or new
diagnosis of cardiovascular disease
•
•
•
•
•
Logon to portal
Construct cohort query
Review the results
Refine cohort query
Review the results
Logon to portal
Display of main CDW home page which will be the main source of information
about CDW
Click on ‘Research Tools’ tab, to start a new search, click on ‘create A New Cohort
search’ button
A default cohort selection query panel with available class and objects on the
left side
Class &
Objects
Filter Area
Remove filter criteria by highlighting and clicking on remove button, or pressing
delete button, or dragging and dropping on the class list
Remove Button
Drag & Drop
Create a new filter with drag and drop Gender Code in filter area; choose filter condition
as ‘Equal to’ from the list; and enter ‘F’ for female
Similarly, drag and drop Radiology procedure name in the filter area; click on
drop down to pick values from list; search for word ‘digital’
Highlight Radiology procedure name displayed for digital search and click on
include button
Drag and drop Radiology Report Date column in the filter area; choose between
filter condition; and choose date from calendar
Click on Run Query button to execute the cohort query selection
Report display by current age, gender, and race
Edit the query to refine the selection by clicking on Edit Query button; add
diagnosis code to filter area; choose between condition; enter heart disease
codes; and click on Run Query to execute the refined query
Query result for patient who had digital mammograms between May 2007 and
Jun 2008 and also had heart disease
TraCS Service Center
• Please visit: http://tracs.unc.edu
• Check the Research Resources area …
• A set of consultants
– Clinical Research Analysts
– System/Business Analysts
– DB Programmer
Questions?
• Javed
• [email protected]