The economic and operational
value of using drones to
transport vaccines
Leila Haidari, MPH
Patrick Wedlock, MSPH
July 26, 2016
For more information, please visit: hermes.psc.edu
Simulation Modeling for Technology Assessment
•
•
•
•
•
•
Technology assessment involves forecasting how a technology will perform in the real world and affect
or be affected by the system the new technology will inhabit.
Investing in, developing, or implementing a technology without understanding its potential in the real
world could waste substantial time, effort, and resources. It could even lead to unanticipated negative
effects.
Trials can provide data but can require time, effort, and resources to design and run. They alone also
provide data on one set of specific locations and circumstances.
Basic mathematical models (e.g., spreadsheet and process optimization models) can provide a general
sense of what may occur and help analyze specific aspects of the system but may not capture all of the
complex interactions of the entire system.
Detailed simulation models are virtual representations of the entire system in which the new technology
will reside and therefore serve as a “virtual real world” to test and assess new technologies. (Think “Sim
City”)
Simulation modeling is used extensively in certain industries to test and evaluate new products,
strategies, and people.
–
An example is flight and aerospace simulators to evaluate and train pilots, test new flight and aerospace technologies,
equipment, vehicles, planes, spaceships, etc. These simulators are designed to duplicate the experience of flying through the air
or space and every step, procedure, phenomena, and equipment involved.
For more information, please visit: hermes.psc.edu
HERMES: A “Flight Simulator” for Supply Chains
Data on Supply
Chain Structure,
Storage Locations,
Transport,
Capacities,
Personnel, etc.
HERMES is a software platform that
can rapidly generate a detailed
discrete-event simulation model of
any health product supply chain
Standard
input deck
Economic
Metrics
Total
Costs
Discrete event
simulation
model of supply
chain
Supply
chain
performance
metrics
Unit
Costs
Cost
Drivers
HERMES models can include virtual
representations of:
• Every storage location, storage
device, transport route, transport
vehicle, transportation device,
personnel, and product in the supply
chain
• Each health center, outreach location,
all healthcare workers at these
locations, and people arriving to get
vaccinated
• Millions of vaccine vials being
shipped, unpacked, stored, repacked, shipped, and eventually
administered or wasted at virtual
locations
For more information, please visit: hermes.psc.edu
Example Topics HERMES Can Address & Sample Publications
•
Introducing new vaccines and technology
– Lee BY, Assi T, Rajgopal J, Norman BA, Chen S, Brown ST, Bailey RR, Kone S, Kenea H, Welling J, Connor DL, Wateska AR, Jana A, Wiringa AE, Van Panhuis WG,
Burke DS. (2012) Impact of introducing the pneumococcal and rotavirus vaccines into the routine immunization program in Niger. Am J Public Health, 102(2):26976.
– Norman BA, Nourollahi S, Chen S, Brown ST, Claypool EG, Connor DL, Schmitz MM, Rajgopal J, Wateska AR, Lee BY. (2013) A passive cold storage device economic
model to evaluate selected immunization location scenarios. Vaccine, 31(45):5232-8.
+ 3 more
•
Altering characteristics of vaccines and other technologies
– Lee BY, Cakouros BE, Assi TM, Connor DL, Welling J, Kone S, Djibo A, Wateska AR, Pierre L, Brown ST. (2012) The impact of making vaccines thermostable in Niger’s
vaccine supply chain. Vaccine, 30(38):5637-43.
– Lee BY, Assi T, Rookkapan K, Connor DL, Rajgopal J, Sornsrivichai V, Brown ST, Welling J, Norman BA, Chen S, Bailey RR, Wiringa AE, Wateska AR, Jana A, Van
Panhuis WG, Burke DS. (2011) Replacing the measles ten-dose vaccine presentation with the single-dose presentation in Thailand. Vaccine, 29(21):3811-7.
+ 2 more
•
Changing configuration and operations of the supply chain
– Assi TM, Brown, ST, Kone S, Norman BA, Djibo A, Connor DL, Wateska AR, Rajgopal J, Slayton RB, Lee BY. (2013) Removing the regional level from the Niger
vaccine supply chain. Vaccine, 31(26):2828-34.
+ 2 more
•
Investing or allocating resources
– Haidari LA, Connor DL, Wateska AR, Brown ST, Mueller LE, Norman BA, Schmitz MM, Paul P, Rajgopal J, Welling JS, Leonard J, Chen S, Lee BY. (2013) Augmenting
transport versus increasing cold storage to improve vaccine supply chains. Plos One, 8(5):e64303.
+ 1 more
•
Optimizing vaccine delivery
– Brown ST, Schreiber B, Cakouros BE, Wateska AR, Dicko HM, Connor DL, Jaillard P, Mvundura M, Norman BA, Levin C, Rajgopal J, Avella M, Lebrun C, Claypool E,
Paul P, Lee, BY. (2014) The benefits of redesigning Benin's vaccine supply chain. Vaccine, 32(32):4097-103.
For more information, please visit: hermes.psc.edu
HERMES Global Work
Formation of HERMES
Logistics Team
Vaccine Supply
Chain Redesign
Vietnam
Kenya
Niger
Decade of Vaccines
Senegal
Benin
Mozambique
Thailand
Passive Vaccine
Storage Devices
India
2016
2015
2014
2013
2012
2011
2010
2009
Unmanned
Aerial Vehicles
For more information, please visit: hermes.psc.edu
HERMES Graphical User Interface (GUI)
For more information, please visit: hermes.psc.edu
HERMES Modeling Assessment of UAVs for Vaccine Transport
Vaccine article: http://dx.doi.org/10.1016/j.vaccine.2016.06.022
Forbes post: http://www.forbes.com/sites/brucelee/2016/06/30/the-next-newfrontier-for-drones-saving-lives/
For more information, please visit: hermes.psc.edu
HERMES Simulation Modeling Used to Compare Two Transport
Systems: Traditional Multi-Tiered Land Transport Versus UAV Transport
Traditional Multi-Tiered Land Transport System
(TMLTS) in Gaza Province, Mozambique
Unmanned Aerial Vehicle (UAV) Transport
‒ Trucks deliver vaccines from provincial depot to
district stores on a fixed monthly schedule
‒ Trucks are used to bring the vaccines to a depot for
storage, but UAVs transport all vaccines from the
depot to health facilities
‒ A mix of truck/motorbike deliveries and pick ups
via public transportation to move vaccines from
district stores to health centers monthly
‒ UAVs distribute up to 1.5L of packaged vaccines to
health centers up to a 75 km radius from a hub which
is co-located next to a supply depot
‒ Personnel costs for EPI logistics include 18% of a
health worker’s time at each health center
‒ Personnel costs for EPI logistics include 10% of a
health worker’s time at each health center reached by
UAVs
For more information, please visit: hermes.psc.edu
Full Province Scenario (Baseline Comparison)
TMLTS for All Locations
Results:
• 94% vaccine
availability
• $0.41 logistics
cost per dose
administered
UAVs Supplying Selected Locations (and
TMLTS for Remaining Locations)
Results:
• 96% vaccine
availability
• $0.33 logistics
Hub 1
cost per dose
administered
Hub 2
Provincial store
Hub 3
Provincial store
District stores (12)
District stores without hubs (3)
Health Centers (123)
Health Centers (123)
District stores with hubs (3)
For more information, please visit: hermes.psc.edu
≤75 km Subset Scenario: Focusing Just on Service Delivery
Locations that are <75 km from a Single Provincial Store Hub
TMLTS for All Locations
Results:
• 97% vaccine
availability
• $0.31 logistics
cost per dose
administered
Provincial store
UAVs for All Locations
Results:
• 100% vaccine
availability
• $0.22 logistics
cost per dose
administered
Provincial store
District stores (7 total)
Hub at Provincial store (1)
Health Centers (69)
Health Centers (69)
For more information, please visit: hermes.psc.edu
≤75km Subset: UAV Cost Savings Remain Robust in a Set of
Sensitivity Analyses
Logistics Cost Savings Per Dose Administered (USD)
$0.00
Geography: road speed
Baseline mean: 59 km/hr
$0.20
100 km/hr
(no effect)
Throughput (population)
mean: 720
newborns
Geography: road distance
mean: 39 km
Baseline mean: 360 newborns annually
$0.10
$0.30
5 km/hr
mean: 180
newborns
mean: 154 km
High
Baseline mean: 77 km
Population distribution
Baseline: Current Gaza population distribution
Evenly distributed
(no effect)
Percentage of health centers (HCs) unreachable for part of the year
Baseline mean: 10 years for land transport, 375,000 km for UAVs
Vaccine introductions
Baseline: mid-2015 Mozambique EPI schedule
Low
80% HCs unreachable,
4 months annually
Seasonality (impassable roads)
Vehicle lifetime
70% of population placed
at 3 urban centers
75% increase
(no effect)
Rota, IPV, MSD, &
HPV introductions
75% reduction
Across all sensitivity analyses
UAVs are cost saving versus the
TMLTS
For more information, please visit: hermes.psc.edu
≤75km Subset: Potential for UAVs to Produce Cost Savings Also
Depends on Flight Conditions and Equipment Costs
$10
$70,000
$9
$60,000
$8
$7
$50,000
$6
$40,000
$5
$30,000
$4
$3
No cost savings
achieved with
4-week delays
$2
$1
$0
$20,000
$10,000
Hub cost per year (USD)
UAV cost per round trip (USD)
Maximum UAV and Hub Costs* to Produce Cost Savings Over TMLTS
(EPI with Rota, IPV, MSD, and HPV Introductions),
Where Each Flight Has a 50% Probability of Delay
UAV
Hub
$0
No delay
1 week
2 weeks
3 weeks
4 weeks
Duration of delay
*Costs include energy, amortization, and maintenance
For more information, please visit: hermes.psc.edu
≤75km Subset: UAV Carrying Capacity Thresholds Exist for UAVs to
Achieve Cost Savings Over TMLTS
Minimum individual UAV carrying capacity (by volume) necessary to
achieve cost savings over TMLTS
EPI
EPI + Rota + IPV
+ MSD + HPV
Assuming current useful UAV/hub costs and useful lifetimes
>0.20L
>0.40L
Assuming estimated at-scale* useful UAV/hub costs and useful lifetimes
>0.15L
>0.20L
*At-scale estimates are based on unverified costs provided by a UAV manufacturer
Current UAV capacity of
1.5L achieves cost savings
For more information, please visit: hermes.psc.edu
Limitations
• Sensitivity analyses were performed on a subset of the service delivery locations (≤75 km
radius) and examined a set of key factors but did not explore all possible factors that may
affect the relative value of the UAVs versus the traditional multi-tiered land-based transport
system (TMLTS)
• There are alternative land-based transport scenarios other than the TMLTS which may
perform better
• Currently did not explore changing shipping/ordering policies, integrating other
commodities, and behavioral factors such as non-compliance with procedures, which future
simulation experiments can explore
• The experiments made assumptions (e.g., acceptance and compliance with stated policies
and procedures); field testing could help test the strength of these assumptions and
iteratively guide refinements in subsequent simulation experiments
• Field testing will likely identify additional regulatory hurdles or human factors not explored in
these models that may be subsequently incorporated into the model
For more information, please visit: hermes.psc.edu
Summary
• Simulation modeling can help with technology development and assessment by
providing a "virtual laboratory" to evaluate and test technology
• The UAVs provide cost savings while maintaining equivalent or slightly better
vaccine demand fulfillment compared to the TMLTS in a variety of scenarios
• Major drivers of costs savings: road speed of traditional land vehicles, number of
people needing to be vaccinated, and road distance
• UAVs can still provide cost savings if weather conditions have a 50% probability
of delaying flights by up to a 1 week
For more information, please visit: hermes.psc.edu
Acknowledgments
• We wish to acknowledge our collaborators
and co-authors at VillageReach
• This study was supported by the Bill &
Melinda Gates Foundation.
For more information, please visit: hermes.psc.edu
HERMES Logistics Modeling Team
Shawn T. Brown, PhD
[email protected]
Bruce Y. Lee, MD, MBA
[email protected]
Diana L. Connor, MPH
Jay DePasse, B.S.
Marie Ferguson, MSPH
Leila A. Haidari, MPH
Daniel Hertenstein, BS
James Leonard
Marie Spiker, MSPH, RD
Joel Welling, PhD
Eli Zenkov, BS
Patrick Wedlock, MSPH
For more information, please visit: hermes.psc.edu