This document discusses using discrete event simulation (DES) to support decision making in emergency departments. DES allows modeling of dynamic patient flow and testing of "what if" scenarios. The document outlines best practices for setting up successful DES projects including defining objectives, gathering quality data, validating models, and including frontline staff. Case studies demonstrate how DES has been used at hospitals to evaluate options for capacity changes, process improvements, and reducing wait times.
4. Eric Hamrock, MBA, PMP- Sr. Project Administrator,
Johns Hopkins Health System
Kerrie Paige, PhD- President, NovaSim, LLC
DES for Decision Making in
the Emergency Department
4/11/2013
5. Roadmap
⢠What is discrete event simulation and why
do we use it?
⢠Setting up a project for success
⢠Gathering the right information
⢠Model validation
⢠Case Studies: Using simulation to support
decisions at JHHS
6. WHAT IS DES AND WHY USE
IT FOR DECISION MAKING IN
THE EMERGENCY
DEPARTMENT?
7. Discrete Event Simulation isâŚ
⢠A powerful predictive analytics tool
⢠Able to capture the dynamic interactions and
variation inherent in any Emergency Department
⢠Useful for predicting the impact of a wide range of
what-if scenarios
⢠Highly visual
⢠Great for getting everyone on the same page
⢠Robust, flexible, powerful
⢠A way to create an environment of objectivity
⢠A method to view outcomes of complex systems in
a simpler way
9. What if WeâŚ
⢠Add capacity (i.e. space, staff)?
⢠Change a practice pattern?
⢠Receive more or fewer patients?
⢠Get a different mix of patients?
⢠Reduce boarding time?
⢠Adopt a new model of care (fast track, short
stay, etc.)?
⢠Want to evaluate proposed lean initiatives?
⢠Reconfigure or add additional care areas?
11. Keys to a Successful DES
ProjectâŚ
⢠Ensure a strong project champion is identified
⢠Define a clear scope and desired outcomes
for the simulation
⢠Assess what sources and quality of data are
available
⢠Is the process appropriate for DES or would
another tool work better?
⢠Include front line staff and those affected by
the process when available
⢠Educate the team on DES
14. DES Pitfalls
⢠Scope creep
⢠Did not define the question clearly up
front?
⢠Lack of project champion buy-in
⢠Lack of front-line staff buy-in
⢠Quality/accuracy of data
⢠DES not the right tool
15. Phases of a Simulation Study
Process Analysis
⢠Historical data
⢠Flow charts
⢠Interviews
⢠Time studies
Modeling
⢠Configure model
⢠Validate
Scenarios
⢠Create
⢠Compare & evaluate
Reporting
⢠Communicate
⢠Knowledge transfer
19. Start by Gathering the Right Information
and Checking it Carefully
⢠Carefully mine any transactional data available for
arrivals, LOS, pathways, D/C
⢠Supplement with manual time studies as needed
⢠Check patient flow and critical process elements
through clinic visits and interviews with critical staff
⢠Verify the model by following simulated patients
through the system
⢠Validate the model - did the baseline model predict
the same performance we observed in the real
system?
20. How Do We Know When to Trust
the Model?
⢠When it looks and âfeelsâ like the real system
â Do we have the right number of arrivals of each patient type
into each part of the system?
â Does the hourly census pattern in each area of the ED
match reality?
â Are patients waiting where they should be and for the right
period of time?
â Is overall time in system about right?
â Bottom line: Does the clinical staff believe it?
⢠It is very common that the initial model does NOT validate
very well
â Tracking down the problem is often enlightening in itself
â It is also proof that validation is a critical part of the process
23. All ED models help us
evaluate
⢠Overall ED Performance â now and in the future
â LWBS
â Bed utilization
â Expected patient wait times
â ED census levels
⢠Potential trouble spots
â Are there any areas with unsustainable bed occupancy
levels? Are we on the edge of the cliff?
â Which parts of the process are particularly sensitive to
volume, patient mix or dwell time changes?
⢠Business cases, providing evidentiary support for
proposed change
25. From a Basic Model, Many
Issues were Analyzed
⢠Addition of a psych pod
⢠Impact of reduced wait times for diagnostic
imaging and labs
⢠Change in capacity/schedule in fast track
⢠Reduced wait for an inpatient bed
⢠Change in nursing bed assignment patterns
⢠Revised allocation of ED beds
⢠Reduced housekeeping (bed turnover) times?
⢠Increased patient volumes
26. Case Study: Johns Hopkins Bayview
Medical Center
⢠380 bed area-wide trauma center
⢠Several throughput improvement
projects over the years
⢠Most recent: detailed model to support
capacity calculations for new ED
⢠Ground breaking for new ED
department, Spring 2013
27. Facility Size
Demand Analysis
9/4/2013
Emergency Department Operations
Review
28
Volume 59,275 63,000 75,000
Current Needed
Minor Care 8 8 8 9
Emergency 21 36 38 45
Specialty 4 7 7 7
Total 33 51 53 61
Ultimate recommendation:
55 beds for the near term, 60-62 for the longer term
28. Facility Size
Simulation: 65,000 Patients
9/4/2013
Emergency Department Operations
Review
29
Emergent
Beds
43 40 37 36 35
Bed
Utilization
66.8% 71.0% 75.4% 78.5% 80.0%
90th %ile
wait time
Level 3
0 mins 4.6 mins 23.5 mins 32.7 mins 48.4 mins
LWBS <1.0% <1.0% <1.0% <1.0% <1.0%
Total
Beds
58 55 52 51 50
Total includes specialty beds and 8 minor care
Add chairs for Psychiatric Evaluation Patients (6)
29. Case Study:
The Johns Hopkins Hospital
⢠How can we best manage throughput?
â Adding or reallocating capacity?
â Changing practice patterns?
â Reducing dwell times?
â Reducing boarding time?
â Adjusting unit operating hours?
31. Many Interventions
Considered
⢠Capacity reallocation/addition
⢠Dwell time reductions (via process
improvements)
⢠Practice changes
⢠Shifting demand elsewhere
⢠Interdepartmental process changes
Ultimate goal: work toward a âzero-waitâ ED
32. Time to First Beds
(Additional Beds)
0.0 min
20.0 min
40.0 min
60.0 min
80.0 min
100.0 min
120.0 min
140.0 min
Base 1 Bed 2 Beds 3 Beds 4 Beds 5 Beds 6 Beds 7 Beds 8 Beds 9 Beds 10 Beds
Additional Beds
Main ED EACU ICU RAP SuperTrack
33. Time to First Bed (Dwell Time
Reduction)
0.0 min
20.0 min
40.0 min
60.0 min
80.0 min
100.0 min
120.0 min
140.0 min
Base -5% -10% -15% -20% -25% -30%
Dwell Time Reduction
Main ED EACU ICU RAP SuperTrack
34. Dwell Time vs Total Beds
80
85
90
95
100
105
110
115
60% 65% 70% 75% 80% 85% 90% 95%
Total
Beds
Dwell Time Factor
Combination of factors necessary to get 95% of patients to first bed within 30 minutes
35. Bed Utilization vs. Service Level
(% of patients who wait less than 5 Min)
100% 100% 100% 100% 99%
98%
96%
92%
86%
76%
61%
41%
16%
0%0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
39% 44% 49% 54% 59% 63% 68% 73% 78% 83% 88% 93% 98% 100%
Bed Utilization
37. Lessons Learned
⢠Start simple, build from there
⢠If you have multiple EDs in your system or you are going to be running
models often over the years, take some time to build a robust, easy-to-
use interface
⢠Model building process can provide value just by getting the team to
think through all aspects of the patient flow process
⢠Extremely useful for presenting ideas/ selling concepts to senior
management
⢠Itâs ok to start with less than perfect data, but validate end results
closely before using for decisions
⢠Sometimes results are non-intuitive, that doesnât make them wrong
⢠Look beyond the immediate symptoms â issues may be originating in
another department
⢠Models often point out issues that had been previously unrecognized
⢠Invaluable for testing ideas before implementation
⢠Great for ED sizing/bed allocation questions