edw 09112014 01a - medicine.utah.edu · • 2010-‐date epic for ... cognos business intelligence...
TRANSCRIPT
9/11/14
1
1
UUHC Enterprise Data Warehouse Overview & 2014 Update
Vikrant G. Deshmukh, Ph.D. Cheri Y. Hunter
Michael B. Strong, M.D.
2
Acknowledgements • Dr. Jim Turnbull • Jim Livingston • Travis Gregory • Nancy Brazelton • Ming-‐Chieh Tu • Reed Barney • Cary MarVn • Micky Daurelle • Charlton Park • Darryl Barfuss
• Dr. Michael White • Dr. Reid Holbrook • Dr. Jeffrey Lin • Kip Williams • Joshua Spuhl • Mingyuan Zhang • Grant Lasson • Dr. Bruce Bray • Dr. Wendy Chapman
3
Acknowledgements
• Database Services • Clinical InformaVon Systems (EMR, Other Ancillary ApplicaVons)
• Enterprise IntegraVon (HL7)
• Enterprise Server Management
• Storage Management Services
• Decision Support • Quality and PaVent Safety
• Strategic Planning • University of Utah Medical Group
• Utah PopulaVon Database
• Biomedical InformaVcs
4
Outline
• IntroducVon • Infrastructure • Metadata • Quality • IntegraVon • Access
5
INTRODUCTION IntroducVon
6
What is a Data Warehouse?
“A data warehouse is a subject-‐oriented, integrated, 6me-‐variant and non-‐vola6le collec6on of data used in an organiza6on”
What is a Data Warehouse? William H. Inmon, Prism, Volume 1, Number 1, 1995.
9/11/14
2
7
CharacterisVcs Characteris2c Descrip2on Example Subject-‐oriented
Data available on an enVre subject area
Visits, Orders, Respiratory, etc.
Integrated Data from mulVple sources integrated in a single place
Orders data integrated across Cerner, Epic
Time-‐variant Data from a specific Vme-‐period
Nursing documentaVon data in the only available from 2007 onwards
Non-‐volaVle New data constantly added, but old data is not deleted
Financial records from two reVred systems (Allegra & IDX) and current (Epic)
What is a Data Warehouse? William H. Inmon, Prism, Volume 1, Number 1, 1995. 8
Enterprise Data Warehouse
An Enterprise Data Warehouse (EDW) contains all the relevant current and historical data for the enVre organizaVon.
9
Data-‐marts
A data mart is a logical subunit of a data warehouse which contains data related to a single subject area (e.g. Orders), department/service line (e.g. Cardiovascular), business unit, etc.
10
High Level Overview
11
Brief History of Systems
Transac2onal Systems • 1993-‐1999 OACIS (inpaVent) • 1995-‐date Epic Care Ambulatory • 1995-‐2010 Allegra • 1995-‐2010 IDX • 1999-‐2003 e-‐Chart (inpaVent) • 2003-‐2014 Cerner (inpaVent)
– 2007 electronic nursing documentaVon
– 2009 computerized provider order entry
• 2010-‐date Epic for Business • 2014-‐date Epic OneChart
InpaVent
Analy2c Systems • 1990 HL7 integraVon engine • 1993 DisVnct data store
populated using HL7 • 1999 Cognos business intelligence
(BI) suite • 2000 Integrated financial data
(batch & HL7) • 2002 Subject-‐oriented data-‐marts • 2005 Corporate Radar web-‐based
tools for reporVng and dashboards
• 2011 SAP Business Objects Enterprise Business Intelligence
12
How is an EDW different? Transac2onal Database (EMR) Analy2cal Database (EDW)
OpVmized for simple, single transacVon queries.
OpVmized for complex queries on paVent populaVons.
Data from a single system. Data from mulVple source systems.
Highly normalized data structures.
De-‐normalized or dimensional data structures.
Raw data on transacVons. Details + aggregate data. Used by clinical, financial and other applicaVon end-‐users.
Used by analyVcal users, researchers, etc.
9/11/14
3
13
Types of Schemas Normalized
Dimensional
14
Data Warehousing Approaches Inmon Kimball
Top-‐down: first common relaVonal model, then data-‐marts
Bomom-‐up: first several data-‐marts, then integrate
RelaVonal model – 3rd Normal Form (3NF).
Dimensional model – Star schema and Snowflake.
Time to fully build an EDW and then develop data-‐marts
Faster development cycle – data-‐marts built first
Higher consistency across the EDW PotenVal for silos in separate data-‐marts
Most granular data for reporVng as the need arises.
May require addiVonal development if a dimension was not planned iniVally
15
The Data Integration Challenge (1)
16
The Data Integration Challenge (2)
• Three textual descriptions
• Three numeric properties
• Three date/time properties
• Three common attributes of these objects
• Three attributes that are unique to each object
• Three uses of each object
17
The Data Integration Challenge (3) • Malus
domestica • Fruit • Sweet • Things you eat • Made in
California • On sale at
Smiths (expires 09/13/2014)
• Volkswagen Beetle
• Year • Trim • License plate# • VIN# ? • Things you
drive • Made in
Germany
• Citrus sinensis • Fruit • Citrus fruit • OJ! (no, not
Simpson!) • Sweet • Tangy • Things you eat • Made in Florida
18
Data Integration Considerations (1)
• Like Vs. Unlike attributes • Old Vs. New sources/attributes • Necessary Vs. Unnecessary attributes • Feasible Vs. Unfeasible approaches to
integration
9/11/14
4
19
Data Integration Considerations (2)
• What is the intended use? • How soon does it need to be deployed? • How will the users access the data? • How reusable are these data for users
across the organization?
20
Types of Schemas Normalized
Dimensional
21
Example of a Snowflake Model
22
INFRASTRUCTURE
23
EDW & UUHC Infrastructure
• Tier-‐1 system • MulVple copies – redundancy
– Downtown Data Center (primary) – Hot Site Data Center (failover) – Disaster Recovery Site Data Center (scheduled)
• MulVple nodes – Oracle RAC – 2 Rack-‐mount Servers (768 GB RAM, 16 proc. EA) – 4 Blade Servers (256 GB RAM, 12 proc. EA)
24
Resource VirtualizaVon
9/11/14
5
25
Resource UVlizaVon – All
Courtesy: Brad Millem 26
EDW Resource UVlizaVon – Groups
Courtesy: Brad Millem
27
METADATA
28
What is Metadata?
“Metadata is all the informa6on in the data warehouse environment that is not the actual data itself.”
Kimball, Ralph; Ross, Margy (2008-‐04-‐21). The Data Warehouse Toolkit: The Complete Guide to Dimensional Modeling (Kindle LocaVons 944-‐945). Wiley. Kindle EdiVon.
29
Metadata Contributors
• EMR Developers • Workflow Engineers • EMR Implementers • Clinical Users • EDW Architects • Analysts
30
Metadata Example (1)
Courtesy: Dr. Reid Holbrook
9/11/14
6
31
Metadata Example (2)
32
QUALITY
33
Defining Data Quality
• “Fitness for use” – OperaVons – Decision-‐making – Planning, etc.
Juran's Quality Handbook: The Complete Guide to Performance Excellence; Joseph M. Juran and Joseph A. De Feo; 6th EdiVon (May 2010); McGraw-‐Hill.
34
Assessing Quality Criteria Defini2on Example
Completeness Does the data provide a complete picture of events in a given domain?
Visits data-‐mart -‐ complete picture of all encounter/visit-‐level events.
Accuracy How well do the data reflect enVVes and relaVonships in the real world?
CPOE process in the EMR -‐ modeled in the Orders data-‐mart in terms of medicaVon and lab orders, other orders, lab results, medicaVon administraVons, etc.
Timeliness Are the data available at the Vme needed, without delay?
HL7 feed instead of a nightly extract for medicaVon orders
SyntacVc correctness
Are the data captured and stored in a well-‐conceived data-‐model?
Determining the most suitable data-‐model (See 3NF, dimensional, etc.)
SemanVc correctness
Are the data encoded using proper healthcare terms?
Determining if NDCs were used in documenVng medicaVon dispense, administraVon and billing
Adapted from: Clinical and Business Intelligence: Data Management – A Founda6on for Analy6cs – Data Governance, Health InformaVon Management Systems Society, April 2013
35
Monitoring Completeness (1)
Courtesy: Joshua Spuhl, Reed Barney 36
Monitoring Completeness (2)
Courtesy: Mingyuan Zhang
9/11/14
7
37
Monitoring Performance
Courtesy: Brad Millem 38
Quality and PaVent Safety
Courtesy: Darryl Barfuss
39
INTEGRATION
40
Epic EMR IntegraVon
Courtesy: Ryan Derrick
41
EMR DownVme SoluVon (1)
Courtesy: Michael R. Donnelly 42
EMR DownVme SoluVon (2)
Courtesy: Michael R. Donnelly
9/11/14
8
43
Novel ApplicaVons & IntegraVon
Courtesy: Michael Sherwood
FAZST
AC Lite – Epic BOE
44
Cerner – Allegra & Cerner – Epic
Courtesy: Michael R. Donnelly
45
EHR & ancillary
applicaVons
EDW
Provide data to other systems
Enterprise Data Warehouse
Support Research Infrastructures e.g. FURTHeR
Provide data to Research Groups.
e.g. PORC, SORG, UPDB
Quality & Regulatory Agencies
e.g. UHC, CDC, NSQIP
Enterprise ReporVng
Infrastructure
Provide data for reporVng
Quality & PaVent Safety, OperaVonal Excellence
46
ACCESS
47
Methods
• Enterprise Business Intelligence – Business Objects
• Warthog Targeted Chart Review Tool • Direct access to the database – SQL Navigator, Toad, ODBC, etc.
• Legacy tools – Corporate Radar • Other custom tools and applicaVons
48
The “Clinical Universe”
Courtesy: Reed Barney
9/11/14
9
49
Drag Objects to Include
Courtesy: Reed Barney 50
Add Filter Objects
Courtesy: Reed Barney
51
Example Report
Courtesy: Reed Barney 52
A Report With Drill-‐down
Courtesy: Reed Barney
53
Drill down to individual study
Courtesy: Reed Barney 54
Warthog: IniVal Cohort
Courtesy: Reed Barney
9/11/14
10
55
Warthog: Counts by Code
Courtesy: Reed Barney 56
Warthog: Text Search
Courtesy: Reed Barney
57
Warthog: Text Search Results
Courtesy: Reed Barney 58
Warthog: Text Results Review
Courtesy: Reed Barney
59
Warthog: Create QuesVons
Courtesy: Reed Barney 60
Warthog: Answer Grid
Courtesy: Reed Barney