Computerized decision support systems aim to transform clinical data into useful information to help clinicians. Such systems have been developed at Vanderbilt University Medical Center's emergency department to detect pneumonia and asthma, initiate appropriate treatment guidelines, and forecast ED crowding levels. Key challenges include integrating systems with clinical workflows and gaining support for changes through understanding stakeholders and change management processes. Lessons learned emphasize addressing important problems, engaging relevant stakeholders, supporting long-term evaluation of systems, and considering both technical and behavioral aspects of medical informatics.

1. Computerized Decision Support:
From Data to Information
Dominik Aronsky, MD, PhD
Dept. of Biomedical Informatics &
Emergency Medicine
Vanderbilt University Medical Center
Nashville Tennessee
and
ii4sm, Basel, Switzerland

2. 2
?
Clinicians:
MD,
RN,
admin
….
L
hunter & gatherer
Computerized Decision Support
Clinical
information
system
J
information manager

3. Sir William Osler
3
“Medicine is
a science of uncertainty
and
an art of probability”
Fundamental impact
on how we deal
with data in medicine

4. 4
Computerized Decision Support
collect report
store
3.18 elevated
IF…THEN
…
ELSE
….
ENDIF
Decision Support Systems

5. 5
Practicing Medicine in the ED
 Multitasking
 Communication challenges
 Interruptions
 Workflow disruptions
 Hand-offs
 Team work
Challenges: Information management
Workflow optimization
1
2

6. 6
Computerized Decision Support
ED Information System Infrastructure:
ED whiteboard: “patient tracking”
Applications / Research:
Pneumonia detection system
Asthma decision support system
Forecasting ED overcrowding
1
2

7. 7
Tracking in Other Industries

8. 8
Tracking Patients in Healthcare –
It is a Simple World
2001

9. 9
Electronic Whiteboard
Electronic Tracking Board
(Version 0.1a)
2001

10. 10

11. 11
ADT System
Registration
information
Disposition
information
Hospital
Bed Board
Application
Computerized
Patient Record
Computerized
Provider Order
Entry System
Radiology
System
Enterprise
Data
Warehouse
Whiteboard
Information
Radiology
Exam
Status
Bed
Request
Status of
Bed
Request &
Diversion
Status
Patient
information
Patient
location
Orders
Hospital Information System
ED Triage
ED Order
Tracker
Triage
Information
Order
Status
Whiteboard
Screenshot
Viewer
Whiteboard
Screenshots
ED Information System
Subject
Recruitment
Waiting
Room
ED Bed
Board
Registration
log
Treatment
Area
Staff
Roster
Recent
Discharges
ED Patient Tracking Board

12. 12

13. 13
ED Discharge
Application

14. 14
Discharge

15. 15
ED Whiteboard “Movie”
Original intent:
• Bridging downtime periods
Unintentional (positive) consequences
• Review: appropriateness of ED diversion episodes
• Malpractice claims
• State investigations

16. 16
Whiteboard:
Return on Investment
Direct benefit:
Additional revenues:
> $ 1.4 million / year
……
Indirect benefit:
more accurate documentation
> $ 1.5 increased MD billing
JCAHO visit 2009
……

17. 17
Computerized Decision Support
ED Information System Infrastructure:
ED whiteboard
Applications / Research:
Pneumonia detection system
Asthma decision support system
Forecasting ED overcrowding
1
2

18. 18
Pneumonia
Pneumonia ?

19. 19

21. 21
Pneumonia Care Process

22. 22
ResultsMinutes
Month

23. 23
Computerized Decision Support
ED Information System Infrastructure:
ED whiteboard
Applications / Research:
Pneumonia detection system
Asthma decision support system
Forecasting ED overcrowding
1
2

24. 24
Asthma Detection: Objectives
Screening:
• Identify eligible patients early
• Screen all ED patients automatically
• Screen all ED patients in real-time
Workflow Integration:
• No additional data entry
• Inform clinicians before initial evaluation
Generalizability:
• Use only electronically recorded data
• Use only common data elements
Goal  Alert clinicians about asthma guideline eligible patients
 Overcome behavioral barrier of initiating guideline

25. 25
Asthma Detection System
Computerized
Nurse Triage
• Coded chief complaint
• Coded asthma history
• Vital signs
• Demographics
Billing Record
Database
• Prior visit codes
– In- or outpatient
– ICD-9 = 493.*
Electronic
Medical Record
• Problem list (text)
– History of asthma
• Medication List (text)
– Asthma medications

26. 26
Asthma Detection System
Method: Developed & implemented a Bayesian Network

27. 27
Prospective Evaluation
• Study period: 4 weeks (Jan 27 - Feb 24, 2006)
• 2,006 encounters; 153 asthma patients (7.6%)
Sensitivity (fixed) Specificity Positive PV Negative PV
90% 89.9% 42.5% 99.1%
AUC = 97.1%
(CI: 95.5% - 98.1%)

28. 28
Initiating Asthma Guideline

29. 29
Computerized Decision Support
ED Information System Infrastructure:
ED whiteboard
Applications / Research:
Pneumonia detection system
Pneumococcal vaccination system
Forecasting ED overcrowding
1
2

30. 30
death was “a result of gross deviations from the standard of
care that a reasonable person would have exercised in this situation.”

31. 31

32. 32
Forecasting ED Crowding
Problem
 No tools available to measure objectively
and manage proactively
Research opportunity
Using ED whiteboard data:
Develop a real-time prediction instruments to alert about impending ED
diversion

33. 33
Forecasting ED Crowding
http://mac01xd.mc.vanderbilt.edu:8080/crowd-war/netica
Bayesian network:
Data collection from ED, OR, hospital, access center, etc., over 2 years
Identified 11 variables predictive of ED diversion: prospective evaluation

34. Forecasting ED Crowding
34
http://mac01xd.mc.vanderbilt.edu:8080/crowd-war/netica

35. 35

36. The “Divide”
IT in Medicine Medical Informatics
36
Computer scientist
IT manager
nurses
physicians
technicians
(Bio-) Medical
Informatics

37. 37
ED: Computerized Decision Support
fast
intuitive
rich in content
optimized for workflow

38. 38
Psycho-social, organizational, political aspects…
“Change Management”
Decision Support Systems

39. 39
Creating a Culture of Informatics
billing
informatics
physicians
hospital
registration
……
nursing
Ambulance
services

40. Lessons learnt
40
- “Is it an important problem?” (Don Lindberg)
- Who cares?
- A very long way from design, implementation,
to evaluation.
- “Get (institutional) support”
- “If it can happen - it will” (Murphy)
- People – Process – Technology: understand the
data, workflow and processes
- “So what?” (Reed Gardner)
- “Change management” (Nancy Lorenzi)
- “Medical Informatics is a behavioral science.”
(Homer Warner)
… if ONE of them does not apply: Have Fun J

41. 41
Acknowledgment
National Library of Medicine:
• R21 LM009747
• R21 LM009002
• T15 007450 (Biomedical
Informatics Training Program)

42. 42
Questions