1. Modeling Patient Survivability for Emergency Medical Service Systems
Laura A. McLay
Virginia Commonwealth University
Statistical Sciences & Operations Research
4 October 2007
In conjunction with the Hanover Fire and EMS Department in Hanover County, Virginia

2. Emergency Medical Service Systems
Motivation
Goal: design next-generation emergency medical service (EMS) systems that
deliver advanced medical care quickly
save lives
Optimization models for EMS systems straightforward if goal is to deliver medical care quickly
Optimization models for saving lives not so clear
Agenda: introduce a new approach to modeling patient survivability
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

3. Emergency Medical Service (EMS) Systems
EMS systems measured according to how they respond to cardiac arrest (CA calls)
CA victims have 1-8% chance of survival
CAs cause 400,000 – 460,000 deaths per year
Ambulances employ either
Paramedics (ALS)
Emergency medical technicians (BLS)
Most EMS systems in the US have moved from BLS to ALS since the 1960s
CAs motivated change
Development of CPR and defibrillators
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

4. Emergency Medical Service Systems
EMS Models
Why not redesign EMS systems to optimize patient survivability?
Focus on CA 911 calls
What does the medical community know?
What helps CA patients?
Early bystander intervention
Early CPR
Defibrillators at scene in 4 minutes
What doesn’t?
Paramedics at scene in 8 minutes (as opposed to basic medical care)
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

5. Why don’t paramedics save lives?
System designed to cover 80% of calls for service in 9 minutes
Rule of thumb for survivability:
90% survival rate if defibrillation within one minute
Survival reduces about 10% every minute thereafter
CDF of calls for service covered
Time in minutes 9
80%
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

6. Maximizing Patient Survivability
Traditional models for EMS systems maximize a proxy for patient survivability
cover the most area possible in a given amount of time
cover the largest population in a given amount of time
cover the most calls for service in a given amount of time
Objective: Directly tie ambulance service to patient outcomes
Why hasn’t this been done before???
EMS systems designed prior to the information age
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

7. Maximizing Survivability
Objective: Directly tie ambulance service to patient outcomes
Traditional operations research optimization models for EMS systems maximize a proxy for patient survivability
Examples:
cover the most area possible in a given amount of time
cover the largest population in a given amount of time
cover the most calls for service in a given amount of time
Does this “just in time” modeling approach save lives?
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

8. Anatomy of a 911 call
Defibrillation should occur within six minutes from CA
Response time measures time from ambulance dispatch, not time from CA
Response time (patient)
Cardiac arrest
911 call
Ambulance dispatched
Ambulance arrives at scene
Response time (EMS)
Defibrillation or care provided
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

9. Maximize CA patient survivability
Not all ways of reaching 80% coverage are equal
System that responds robustly to CA calls will respond well to all calls
CDF of calls for service covered
Time in minutes 9
80%
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

10. Existing OR models for EMS
Goal of operations research models to determine
what type of resources (ambulances) to purchase
where to place ambulances
how to staff ambulances
how to dispatch ambulances
how to accurately measure time traveling and other parameters
Operations research methods
Simulation, optimization, queuing
Issues considered
Busy vehicles, back-up coverage, vehicle types, dynamic issues
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

11. Existing OR models for EMS, cont’d
Much research in 1970s and 1980s
No CAD systems, dispatch centers pencil and paper
Data difficult to obtain so reasonable assumptions made
Information age in 1990s and beyond
CAD systems in dispatch centers collect lots of data
Patient billing data links EMS to patient outcomes
The proxies for patient survivability don’t do what we want them to do
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

12. Case study: Hanover County, Virginia
Anecdotes from an ambassador to the EMS community
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

13. Emergency Medical Service Systems
Hanover County map
The basics:
Population = 100,000
Area = 474 mi2
70% rural with small pockets of suburbs
EMS a branch of the Fire Department
EMS all volunteer-run (BLS) until recently
Staff (ALS) work on weekdays
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

14. Emergency Medical Service Systems
Hanover County Goals
Their goals:
Cover 80% of calls within 9 minutes
(currently covering ~50% of calls)
Understand if not meeting goal due to geography or insufficient resources
Decide which resources to purchase
My goals:
Is covering 80% of calls within 9 minutes really the goal?
Is 80% coverage realistic in a semi-rural county?
Are they measuring what they really need to measure to reach their goals?
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

15. Response Time: Stopping the Clock
Priority 1 (life threatening) calls require ALS response (60% of calls)
24% of calls could potentially be CAs
11% of calls are “Chest Pain/Heart Problems”
13% of calls are “Breathing Difficulty”
Double coverage by BLS ambulance or fire truck if ALS not immediately available (12% of calls)
Response time defined when ALS arrives
Issue: models depends on response time
Can we stop the clock when the first responder arrives?
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

16. Emergency Medical Service Systems
The Problem is Complex
Vehicles that can respond to calls
ALS ambulance [2 people]
BLS ambulance [2 people]
ALS QRV (non-transport unit) [1 person]
Fire truck (BLS) [3 people]
Police car (AED) [1 person]
Two types of vehicles = hard problem
Five types of vehicle = great problem
Impact out-of-service times, response times, service times, turnaround times.
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay

17. Emergency Medical Service Systems
Final messages
The dispatch center and CAD system are backbone of entire EMS system
Software constrains how systems works
Constant customer interaction and feedback
Need input from (real) doctors
It’s truly a systems problem
Police cars have defibrillators
Other counties rely on Hanover County EMS
Be a good first responder—learn CPR
SAMSI
4 October 2007
Emergency Medical Service Systems
Laura A. McLay