operations research for public health preparedness -...
TRANSCRIPT
Operations Research for Public Health Preparedness
Jeffrey W. HerrmannAssociate ProfessorDept. of Mechanical Engineering& Institute for Systems ResearchA. James Clark School of EngineeringUniversity of Maryland
Rutgers UniversityJanuary 30, 2012
Focused Issue on Operations Engineering and Analysis
Homeland Security Department
IIE Transactions
2012 Industrial and Systems Engineering Research Conference
Hilton Bonnet Creek, Orlando, FloridaMay 19-23, 2012
The process of responding to a bioterrorism attack starts with detection and investigation.
Outbreak
EnvironmentalMonitoring
SyndromicSurveillance
ClinicalDiagnosis
Outbreak Investigation
Decision to respond
Distribution to PODsDispensing at PODs
Prepositioning
Jeffrey W. Herrmann 4
Points of Dispensing (PODs) provide mass vaccination or dispensing of medication.
Jeffrey W. Herrmann 5
The medication supply chain must move quickly to distribute medications.
StrategicNationalStockpile
LocalDepot(RSS)
POD
POD
POD
POD
residents
residents
residents
residents
POD residents
Planning Problems• What is the best way to preposition
medication?• How should medication be delivered to
points of dispensing (PODs)?• What is the best POD layout?• How many staff do we need?• How long will people wait in line?
Prepositioning medical countermeasures affects the mortality of an anthrax attack.
Exposed Prodromal Fulminant
Not exposed Potentialexposure
Prophylaxed Recovered
Dead
Potential exposure = someone will seek prophylaxis but cannot become ill.
1 2 3
4 5 6
7 8 109 1211
13 14
17
1615
22
18
19
20
21
23
24
25 26 27
28
28 compartments track the number of exposed, ill, recovered, and dead.
The compartment changes first after changing disease status and then after changing treatment status.
Scenario timeline.Time (hours) Event0 Attack occurs.48 Attack detected. PODs start opening.53 Local supplies become available.64 (or 76) Push pack supplies becomes available.84 Vendor-managed inventory becomes available.96 All PODs at maximum capacity.
Population size = 5,000,000.Number exposed = 50,000, 500,000, and 1,250,000.Percentage of non-exposed persons who will seek prophylaxis (potential exposures) = 1%, 10%, and 50%.Adherence rate = 65% and 90%.
Model output can estimate exposed population by condition over time.
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
0 48 96 144 192 240 288 336 384 432 480 528 576 624 672 720
Number
Time (hours)
Exposed Population by Condition
Incubation Prodromal Fulminant Adhering Recovered Dead
Deaths decrease as more medication is predispensed.
N = the number exposed.b = the fraction of non-exposed persons who will seek prophylaxis (potential exposures).alpha = prophylaxis adherence ratet = delay until push pack is available
Deaths as Medkits Distribution Increases (alpha = 90%, t = 24 hours)
0%
10%
20%
30%
40%
50%
60%
0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 3,500,000 4,000,000 4,500,000 5,000,000
Number of Medkits Distributed
Percen
tage
of tho
se exposed
who
die N=1,250,000, b=0.5
N=1,250,000, b=0.1N=1,250,000, b=0.01N=500,000, b=0.5N=500,000, b=0.1N=500,000, b=0.01
Jeffrey W. Herrmann 13
The last leg of medication distributionis from a local RSS to the PODs.
Jeffrey W. Herrmann 14
PODs will receive batch deliveries, which must arrive before supplies are exhausted.
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
0:00 6:00 12:00 18:00 24:00 30:00 36:00
Time from start (hh:mm)
Qua
ntity
(reg
imen
s)
DeliveredSlackDemand
Jeffrey W. Herrmann 15
Inventory slack routing problem (ISRP) formulation maximizes minimum slack.
{ }1, , ; 1, ,
max minv
vjv V j rS s
= ==
… …
( ){ }1min /vj
vj vjk k vj vjkks T Q L t w
σ∈= + − +
( ), vjk
vjk abka b E
Q q∈
= ∑
( )( )
, : 1, , ; 1, ,
ab vj ab
abk vj va b t t k
q I t v V j rσ≤ ∈
≤ = =∑ ∑ … …
, 1 , 1 1, , ; 2, ,vj v j v j vt t y v V j r− −≥ + = =… …
vj
vjk vk
q Cσ∈
≤∑
( )2 11 1
1, ,vrV
vjk kv j
q T T L k n= =
= − =∑∑ …0vjt ≥
Patients wait in line for flu shots in Silver Spring, Maryland.
Waiting at a POD
Jeffrey W. Herrmann 17
From Patient Flow …Arrival Triage Registration
Education
ScreeningVaccinationExit
Holding Room
Symptoms Room
Consultation
Jeffrey W. Herrmann 18
…to Simulation Model• Simulation model created using Arena®.
Queueing Network Analysis
Queue Server
Workstation
Patient
Jeffrey W. Herrmann 20
Performance Measures
i i i iCT w t W= + +
11
1 I
i ii
TCT rCTr =
= ∑
1
1, ,min i
i Ii i
m rRt r=
⎧ ⎫= ⎨ ⎬
⎩ ⎭…Clinic Capacity
Station utilization
Station cycle time
Total cycle time
i ii
i i
r tum k⋅
=⋅
Jeffrey W. Herrmann 21
Clinic Planning Model Generator
Clinic Generator
spreadsheet Planning model
Clinic Template
User input
The Clinic Generator spreadsheet takes inputs from the user and modifies the Clinic Template file
to create a custom clinic planning model that the user can modify as desired.
Jeffrey W. Herrmann 22
Comparing the approximations to simulation yielded mixed results.
30374451586572798693
100107114121128135142149156163170177184191198205212219226233240247254261268275282289296303310317
324331338345352359366373380
8.00
%
13.0
0%
18.0
0%
23.0
0%
28.0
0%
33.0
0%
38.0
0%
43.0
0%
48.0
0%
53.0
0%
58.0
0%
63.0
0%
68.0
0%
73.0
0%
78.0
0%
83.0
0%
88.0
0%
93.0
0%
98.0
0%
Resident arrival rate to clinic (percentage of clinic capacity)
Entit
y cy
cle
time
in c
linic
(min
)
Average of cycle time from simulation with lower and upperbound
Cycle time from formula
Cycle time from simulation (min) with lower and upper bound (confidence interval for 95%)
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
15.0
0%
20.0
0%
25.0
0%
30.0
0%
35.0
0%
40.0
0%
45.0
0%
50.0
0%
55.0
0%
60.0
0%
65.0
0%
70.0
0%
75.0
0%
80.0
0%
85.0
0%
90.0
0%
95.0
0%
100.
00%
Resident arrival rate to clinic (percentage of clinic capacity)
Entit
y cy
cle
time
in c
linic
(min
)
Average of cycle time from simulationwith lower and upper bound
Cycle time from formula
Jeffrey W. Herrmann 24
Public health emergency preparedness planners around the country are using CPMG.
Download: www.isr.umd.edu/Labs/CIM/projects/clinic
States whose health planners are using our software.
Jeffrey W. Herrmann 25
Final Thoughts• Modeling should create a conversation,
not just answer a question.
Jeffrey W. Herrmann 26
Acknowledgements• Centers for Disease Control and Prevention• National Association of County and City
Health Officers• Montgomery County Department of Health
and Human Services, Public Health Service– Kay Aaby– Rachel Abbey– Carol Jordan– Kathy Wood
… and the students:
Jeffrey W. Herrmann 28
For more information• Visit the project web site at
http://www.isr.umd.edu/Labs/CIM/projects/clinic/• Download the Clinic Planning Model Generator at
http://www.isr.umd.edu/Labs/CIM/projects/clinic/cpmg.html
• Get publications about the work at http://www.isr.umd.edu/Labs/CIM/projects/clinic/publications.html
• Contact Jeffrey Herrmann at [email protected]