Download - La informática en el ámbito de la salud una ayuda en la gestión del servicio de urgencias
Emilio Luque Computer Architecture & Operating Systems Department
University Autonoma of Barcelona (UAB)
Emergency Departments (ED)
are complex and
quite dynamic systems.
ED’s are overcrowded and work
with limited budget.
Patients must be
addressed with the best
quality.
Simulation:What if? Optimization:
The best solution for?
Supported by the MICINN Spain, under contract TIN2007-64974 and
the MINECO (MICINN) Spain, under contract TIN2011-24384
Emilio Luque
CAOS – HPC4EAS
Manel Taboada GIMBERNAT
Eduardo Cabrera
CAOS – HPC4EAS
Francisco Epelde
PARC TAULÍ
Ma. Luisa Iglesias
PARC TAULÍ
Optimization
Simulation
Variables Values Observability
Symptoms (patients) Healthy, Cardiac/respiratory arrest, severe/moderate
trauma, headache, vomiting, diarrhea E/I
Communication skills Low, Medium, High E
Level of experience (doctors)
Resident (1 to 5); Junior (5-10); Senior (10 - 15) and Consultant (over 15 years)
E/I
Level of experience
(triage nurses)
Low, Medium, High E/I
Level of experience (emergency nurses)
Low, Medium, High E/I
Level of experience (admissions)
Low, Medium, High E/I
Current state
/ Output Input
Next state /
Output
…. …. ….
Sx / Ox Ia (p1) Sy / Oy
Sx / Ox Ia (p2) Sz / Oz
Sx / Ox Ia (p3) Sx / Ox
…. …. ….
STATE Variables Values Observability
Name/identifier <id> Unique per agent I
Personal details
Gender, Medical history (cardiology, pulmonology,
neurological,…); Allergies (yes-no);
Treatments that received (classified into therapeutic groups:
bronchodilators, vasodilators, etc.);
Origin (national or immigrant)
I
Location Entrance, Admissions, Waiting Room, Triage, Treatment
Zone. E
Action
Idle, Requesting information from <id>, Giving information
to <id>, Searching, Moving to <location> , Waiting for
ambulance.
E
Physical condition Healthy; Hemodynamic-Constant; Barthel Index (degree of
dependence). E/I/N
1) Active Agents
Patients
Companions of patients
Admission personnel
Sanitarian technicians
Nurses (Triage, Emergency)
Doctors (Emergency,
Specialists)
2) Passive Agents
Information system
Loudspeaker system
Pneumatic pipes
Tests services
1 to 1(One-to-One) 1 to n (Multicast) 1 to Zone: individuals in Zone
(Area- Restricted Broadcast)
The Environment
The model should include the spatial topographical design from the ED
Arrival/dismissal
by own means
Arrival/dismissal
by ambulance
Agents interactions
ED functionality
Agents
A
N
D
Arrival/dismissal
by own means
Arrival/dismissalb
y ambulance
What if?
ED Simulator
Patients:
How many arrive to the service
How many leave the service
Times of staying in each area
Patients arrival:
Could arrive every 3 min. , but with different probabilities:
20% (4 pat/hr), 40% (9 pat/hr),
60% (13 pat/hr) , 80% (17 pat/hr)
Staff Number Junior Senior
Admission 1-2 2 min. 1 min. 15 sec.
Triage Nurse 1-3 8 min. 5 min.
Doctor 1-4 20 min. 15 min.
Configuration of the ED Staff
Input
Output
Given any objective (index) function f :
minimize Maximize
• Find the best/optimum solution from all the possible solutions.
AxA
Ax
xf
Af
oset;constraint
tosubject
min/max
:
xfxf o xfxf o
Axo
Is it always the "best solution" (the optimum) the most interesting for us?
Methodology
Parameter configuration:
A
N
D
Simulator: 2nd version
A, N, D = > 3D + P => 4D
~ 400 patients daily
Discrete
Methodology: Computational complexity
• Search space
– # Dimensions = Patients,
staff (D, N, A, …), T, B, …
– Each dimension=> range of possible values
– # Points = # simulations (indexes)(time)
COMBINATORIAL!
A
N
D
Multidimensional
A
N
D
P
A
N
D
P
T A
N
D
P
B
T
ABM
SIMULATOR PARAMETERS
I
N
D
E
X +
constraints
DSS
14 D, 9 N, 9 A = 1,134 cases
Patient
Arrival
20% (4 pat/hr)
40% (9 pat/hr)
60% (13 pat/hr)
80% (17 pat/hr) 25,000 ticks => 1 day
1,134 cases * 4 = 4,536 cases
Staff Time (ticks)
Senior Junior
Doctors 260 350
Nurses 90 130
Admission personnel 20 35
Quantity
1 - 4
1 - 3
1 - 3
Cost (€)
Senior Junior
1000 500
500 350
200 150
Quality Index:
Minimize patient “Length of Stay” (LoS) Constraint: Cost <= 3500 €
4,536 total cases => 2,408 cases under limit
Cost constraint <= 3500 € Average patient “LoS”
4 p/hr 9 p/hr
13 p/hr 17 p/hr
Optimum
Time
(ticks)
€ #
Staff
D N A
428 3,200 5 2 S 2 S 1 S
428 2,900 5 2 S 1 S 2 S
428 2,850 5 2 S 1 S 1 S, 1 J
Patient
Arrival
20% (4 pat/hr)
40% (9 pat/hr)
60% (13 pat/hr)
80% (17 pat/hr)
4 p/hr 9 p/hr
13 p/hr 17 p/hr
Optimum
Time
(ticks)
€ # Staff D N A
3,266 3,350 7 1 S, 3 J 2 J 1 J
Cost constraint <= 3500 € Average patient “LoS”
4 p/hr 9 p/hr
13 p/hr 17 p/hr
Cost constraint <= 3500 € Average patient “LoS”
4 p/hr 9 p/hr
13 p/hr 17 p/hr
Cost constraint <= 3500 € Average patient “LoS”
Optimal
vs
Suboptimal