Poster sessions will be held during the conference and below is the list of posters that were accepted. Click on the title of a poster to jump to its abstract. Posters

Poster Listing

A Novel Solution to Decrease Open Defecation in an Urban Slum in Belén, Peru

Jason Kass (Toilets for People)

OBJECTIVE: Reduction of open defecation in poor communities is one of the goals of the WHO, to reduce the physical hazards and prevalence of contaminated of drinking water. To decrease open defecation in an urban slum setting in Belén, Peru, we examined the uptake and sustainability of a novel waste management system.
METHODS: We enrolled three households into a longitudinal study over a four-month period to examine the use of a composting toilet. Participating households were required to build a platform adjacent to their home to receive the bathroom. We installed a personal bathroom kit for each, including a composting toilet which relies on dry composition to decrease waste bulk and pathogen load. A mechanical counter recorded use for each toilet. Each household participated in both group- and individual-level education on the use and maintenance of the composting toilet. The outcomes included toilet counter readings, surveys and compost inspections for both quantitative and qualitative data regarding uptake and sustainability of the toilet.
RESULTS: All households reported consistent and exclusive use over the four-month span, and declined the buy-back option at the end of the study. 2 of 3 households continued to use and maintain the toilet according to original instruction over the four-month period. Quantitative data reflected 0.5 “ 2 uses per day per adult for >80% of the time, with approximately 15% of time showing inconsistent or sporadic use, likely due to counter failure during these periods.
CONCLUSION: Our study found that uptake of the composting toilet was 100% and satisfaction was high amongst participating households after four months. Education retention was 100% in 2/3 of households, and quantitative data suggested >80% uptake. We plan to expand our study to examine the uptake and sustainability of 20 toilets for a flood-prone jungle community of 14 households and 1 school.


A systems approach to coordination problem in Humanitarian Supply Chain Management

Lijo John (Indian Institute of Management Kozhikode), Anand Gurumurthy (Indian Institute of Management Kozhikode)

In the post-disaster response stages, often owing to multitude of reasons like lack of central authority, poor information exchange, large number of actors, scarce critical resources, etc. the coordination between the actors to effectively carry out the supply chain function is minimal. Poor coordination reduces the effectiveness and efficiency of the HSCM activities. In this study, we evaluate the coordination challenge in HSCM. We have modelled the HSCM as a combination to two main resources viz. inventory (or aid material) and manpower (volunteers). HSCM in this regard is assumed to function to achieve two main objectives, first, effective resource utilization (aid management) and second, helping beneficiaries through volunteers. Since HSCM is a not-for-profit supply chain, the cost of operations is man-hours required to achieve the objectives. The coordination problem is modelled using system dynamics methodology. The model has three major stock variables, viz. inventory, manpower and affected people. These stock variables measure the behaviour over time (BOT) of the system to shed light on the coordination problem in HSCM. The stock and flow model is simulated for a period of 100 days since the majority of relief activities takes place in the first 100 days post disaster and subsequently, the supply chain focus changes from relief to rehabilitation. The BOT of principle stock variables provide evidences for the coordination challenge and helps in understanding the systemic reasons contributing to the poor coordination in HSCM


Access-to-Medicines Vaccine Supply Chain Design: A Stakeholder Framework

Catherine Decouttere (KU Leuven), Nico Vandaele (KULeuven), Stef Lemmens (KULeuven), Mauro Bernuzzi (GSK Vaccines)

Supply chains supportive of Access-To-Medicines (ATM), like vaccine supply chains impose considerable additional challenges on the supply chain design process. We embed the modelling in a broader stakeholder based framework, which will substantially enhance the societal and human impact of the ATM supply chain service delivery. It boils down to a five-step approach: (1) stakeholder analysis and system delineation, (2) key performance indicator development, (3) scenario building and modelling, (4) scenario ranking and (5) scenario selection and implementation. We will briefly review this approach for a vaccine supply chain.


Adaptive Decision Making in Limited Data Environments

Kezban Yagci Sokat (Northwestern University), Irina Dolinskaya (Northwestern University), Karen Smilowitz (Northwestern University), Ryan Bank (Social Intelligence)

State of the art humanitarian logistics models have been developed over the past decades. Most of these models assume availability of data and use synthesis data. After a disaster, there is often limited information about infrastructure damage. Given limited information about road damage immediately after disaster, a field operations manager needs to deploy help. New data sources such as OpenStreetMap are emerging. Utilizing these new data sources as well the currently available data, we develop a framework to estimate incomplete information in limited data environments. The framework starts with data collection and processing which benefits from ARCGIS. We develop a model in ARCGIS to automate the data gathering and processing steps as much as possible. Once the data is processed and ready to be integrated in the models, we propose various imputation techniques: naïve methods (optimistic, pessimistic and neutral), unsupervised methods (clustering combined with simple averaging and adjacent arc averaging in the cluster) supervised learning based methods (classification tree). Then, the model is validated using different hold out methods (geographical or random). We present an application of this framework to a recent disaster where we estimate the status of a road, open, partially blocked or damaged. We measure the success by calculating percentage of the road that are identified accurately with each damage status. The study explores the impact of available level of information, dispersion of available data and imputation techniques used in approximating the incomplete information. Our results show that as the available information decreases success of imputation decrease in general. The dispersion of data did not change the success significantly for most cases. The success of a metric depends on the imputation method used. The study also investigates the impact of geographical detail. Our results suggest that higher granularity yields better estimates of the unknown information.


Annotating & Conducting Syndromic Surveillance in Social Media Communication

Scott Appling (Georgia Institute of Technology), Erica Briscoe (Georgia Institute of Technology), Edward Clarkson (Georgia Institute of Technology)

Social media data is rich with communications discussing wellbeing about individuals, people they observe, including symptoms and signs. These communications are highly useful for conducting syndromic surveillance which is tasked with understanding the likely state of existing situations of interest and emerging events related to the health of local and global populations. We present work in progress on developing a robust social media contextually focused annotation coding scheme.


Commodity Distribution Point Program

Joseph Whitney (New York City Emergency Management), Tony Maldonado (New York CIty Emergency Management)

This poster outlines New York City Emergency Management’s Commodity Distribution Point Program.


Converting adversity into opportunity Rural Housing 2005 Earthquake

Abrar Ismael (Earthquake Reconstruction Rehabilitation Authority), Qasim Jamal (Earthquake Reconstruction Rehabilitation Authority)

The earthquake of 2005, destroyed over 600,000 houses in the affected area, spread over 30,000 sq km and killed over 73000 people, in the Northern Pakistan. The initial damage assessment conducted by the World Bank and Asian Development Bank indicated that a very large number of houses had been destroyed/ damaged as the result of the 2005 earthquake, predominantly in Rural Areas. This report was confirmed by Earthquake Reconstruction & Rehabilitation Authority (ERRA) “Damaged and Eligibility Assessment Survey” (DEAS). The programme was quite difficult due to the geographic spread of the affected areas, and in many cases, the remote location and very difficult terrain. Northern Pakistan is prone to seismic activities and the traces of earthquakes and construction of seismic resistant buildings can be traced back to Indus valley civilization (2500 BC). However, the communities living in these areas had forgotten these practices with the passage of time.

A joint effort between ERRA, UN-Habitat and other stakeholders culminated in the development of the ERRA (Rural Housing Reconstruction Programme - RHRP). To implement this programme, an owner driven strategy was formulated and indigenous technologies and practices of construction were incorporated/ reintroduced. Women of the area were provided a leading role in this program to monitor and supervise the stages of construction. The program also assisted the Local Governments to register the residents of far flung remote areas, as due to geographical location these areas were inaccessible to National Registration Database. The project got completed in a record period of two years and earned UN-Sasakawa award in 2011.


Empowering women in Uganda through Industrial Engineering

Kaitlin Rizk (Days for Girls)

Days for Girls (DfG) is a non-profit organization that is dedicated to improving knowledge about and access to menstrual hygiene materials around the world. They have designed an innovative re-usable pad that is affordable for women and girls to use during their menstrual period, which allows them continue with their routine such as school and work. DfG has an enterprise model where it teaches groups of women how to make re-usable pads & soap and grow the knowledge into a micro-enterprise. DfG centers (in Uganda, Nepal, South Africa and Ghana) then supply materials for making these products which empowers these women and enables them make an income by selling the reusable pads & soap.

My work focuses on the creation of a set ordering and inventory system for DfG-Uganda which supplies 140 micro-enterprises and sells kits to customers all-over Africa. With consultation, I decided to use the Kanban system in the design because of its ease for users to know when to order. This system works by using pushes and pulls to signal ordering and re-stocking while keeping a constant stock in place. Two bins of materials are set next to each other. Once one is empty this is a signal to reorder material and use the second bin until the material is replenished. Each bin has a Kanban card on them to explain how much to re-order. The system is based off of set order and re-order quantities. The Kanban card is placed at a “to be ordered” station until the material is brought and then is moved to an “ordered” station. This system avoids counting materials and for quick replacement of stock. The Days for Girls staff simply have to check for empty bins of materials.


Equitable models for prepositioning of supplies in preparation for a disaster

German Andres Velasquez Diaz (NC State University), Maria Mayorga (NC State University), Eduardo Ayres Cruz (Universidade Federal do Parana)

In numerous disaster situations it is possible to plan in advance in order to improve the efficiency of the post-disaster response. Inherent high levels of uncertainty in such scenarios advocate for the use of stochastic approaches. In this paper we formulate three mathematical programs and heuristics approaches for equitably prepositioning relief supplies during the preparedness phase of predictable disasters such as floods and hurricanes. First, we provide a deterministic mathematical model and two robust counterparts. Regarding the robust counterpart optimization techniques, two different types of uncertainty sets are studied: interval and ellipsoidal sets. Then, heuristic approaches are provided for each of the mathematical models proposed. The proposed models consider multiple relief item types, budgetary constraints and equity constraints while integrating supplier selection, inventory and facility selection decisions. Proposed mathematical models and heuristics are evaluated in two different experiments: in the first experiment, we test the performance of the heuristics by comparing the heuristic solutions to the optimal solutions found with their respective mathematical models. In the second experiment, we assess the robustness of the solutions obtained with the robust counterparts and their respective heuristics by comparing them to the solution generated with the deterministic approach in a number of realizations under different test problems. Finally, we study the performance of our mathematical models and heuristics when applied to a realistic scenario by developing a case study based on Wake County, NC.


Estimating Disease Burden from a Potential A(H7N9) Pandemic Influenza in the United States

Walter Silva (University of South Florida), Tapas Das (University of South Florida)

Periodic emergence of A(H7N9) influenza virus in China since winter of 2013 resulted in 770 laboratory-confirmed cases of human infections causing 306 deaths (39.75% fatality rate) till May, 2016. Researchers have developed early estimates of some of the epidemiological parameters to characterize A(H7N9) virus in China. Though most of the infection cases have been attributed to animal to human transmission of the virus, human to human transmission has been suspected in a handful of families with multiple infected members. Our goal in this paper is to assess disease burden on the U.S. in the event of an A(H7N9) pandemic outbreak.

We incorporated estimated epidemiological parameters and U.S. demographic data in an agent-based simulation model. The model mimics the A(H7N9) virus transmission process that is expected to occur during a pandemic. The model results are used to estimate disease burden. We considered non-pharmaceutical interventions as the only containment measure (without vaccines and antivirals). To reduce the computational needs of the simulation model, we first divided the 50 states of the U.S. in three clusters based on urban population density. Then, representative states from each cluster were separately simulated as outbreak areas. Infection attack rates (IARs) and the number of infected were considered as the measures of disease burden. These measures were stratified over three age-groups (≤19 yrs, 20-64 yrs and 65+ yrs).


GPU Supported Large Scale Simulation Models for Influenza Pandemic Outbreaks

Zhila Nouri (University of South Florida), Hanisha Tatapudi (University of South Florida), Walter Silva (University of South Florida), Yicheng Tu (University of South Florida), Tapas K. Das (University of South Florida)

Influenza epidemics are a serious global concern that threatens to turn into a pandemic. Health care professional and disease control experts and other medical practitioners are constantly seeking better understanding of the pandemic impacts while exploring effective pharmaceutical and non-pharmaceutical interventions to contain pandemic spread. In this pursuit, researchers have adopted different models based on differential equations, Markov chain Monte Carlo simulation, agent-based (AB) simulation, etc. Among these, AB simulation models appear most versatile capable of considering large population sizes, demographic variations, individual schedules and travels, contact and infection processes, and disease natural history.

However, the AB simulation models incorporating large populations have computational limitations. Addressing this limitation requires fast and highly parallel computing facility with large and high-throughput memory, which is often not supported by current CPUs.

Over the past few years, with massive computing power and high-speed memory, graphical processing units (GPUs) have become a part of many high-performance computing systems. Originally designed for graphics processing, the use of general-purpose computing on GPUs (GPGPU) has been boosted in recent years with the development of software frameworks such as compute unified device architecture (CUDA) and open computing language.
We develop a GPU supported AB simulation model capable of simulating a country wide influenza pandemic outbreak in the U.S. We consider census data to locate 318 million people as well as the schools, workplaces, and community/social gathering places across the country. We also incorporate detailed travel patterns of the people to facilitate movement and tracking of virus spread. A preliminary version of the model is tested for a A(H7N9) pandemic outbreak.


Human in the Loop Optimization for Recovery from Extreme Events

Aybike Ulusan (Northeastern University), Ozlem Ergun (Northeastern University), Casper Harteveld (Northeastern University)

We consider the problem faced by contractors for collecting debris from a road network to the disposal facilities in the aftermath of a disaster. The problem has a multi-objective nature which embodies implementable division of a service region among subcontractors such that the assigned workload among different subcontractors is balanced and the time to complete all debris collection operations is minimized. More specifically, the subcontractors’ operating time and the profit gained from all the operations should be similar while functioning on geographically contiguous and distinct regions. Such region assignment prevents conflicts between different subcontractors, it is easier in terms of execution, and provides clear responsibilities. The complex nature of the problem gives rise to the need for novel solution approaches. In this project, we are investigating a unique interdisciplinary approach that combines data driven mathematical modeling with human-in-the-loop game based experiments. Our goal is to find evidence that the complex multi-objective problems modeled to improve resiliency are best optimized by the collaboration of humans and algorithms, and build a decision support tool in order to aid decision making after an external disruptive event. We leveraged game-based simulations in lab experiments to investigate the contribution of humans. To test our conjectures, initially we deployed an Excel-based prototype and collected player decisions and observations in a within-subjects experimental pilot study. Building on all the experiments, our objective is to create a digital tool that utilizes both human input and optimization algorithms that would enable industry and public service decision makers to be more effective in the design and management of complex infrastructure networks under extreme events.


Improving Access to Health Commodities by Strengthening the Supply Chain Management Workforce: The People that Deliver Initiative and the Country Partnership Program

Dominique Zwinkels (UNICEF/People that Deliver Initiative)

From 2011-2015 the People that Deliver (PtD) Initiative focused on creating global awareness of and engagement around the importance of the health supply chain workforce. Starting in 2016, PtD aims to focus more of its advocacy efforts and partner resources directly at the country level. This effort intends to support Ministries of Health, NGOs and implementing partners to plan, implement, and sustain reforms which strengthen human resources (HR) for health supply chain management (SCM).

PtD has successfully built awareness at a global level for the public health supply chain workforce and has applied, through PtD partner organizations, some of its tools and approaches in its select focus countries (Burkina Faso, Dominican Republic, Ethiopia, Indonesia, Liberia, Mozambique, Namibia). For example in Namibia PtD leveraged the expertise of member organizations to provide technical assistance in planning, deployment, training and retention of the SCM workforce; documented the process and lessons learned; and drafted a case study on the process for PtD to share globally. Namibia and PtD collaborated on four key interventions: development of a SCM competency framework, identification of number and types of SC personnel required, targeted capacity building, and identification of context-specific incentives to encourage staff retention.

In order to move PtD’s efforts from global advocacy to effect real sustainable country-based change, more engagement at the country level must occur. Many country supply chain programs have expressed desire and need for assistance in the HR for SCM realm. Therefore, PtD has created the “Country Partnership Program” where it acts as a “broker of services”, matching country partner needs with PtD member organizations that can supply the service. Some of the services included: tailored HR for SCM assessment, development of a comprehensive HR for SCM strategy, development of competency frameworks, development of a costed and resourced annual training plan, etc.


Integrating Supply Chains for Emergencies and Ongoing Operations in UNHCR

Tina Rezvanian (Northeastern University), Ozlem Ergun (Northeastern University)

This poster presents a warehouse location model for the joint prepositioning of emergency response and ongoing operations. In this study, we designed a network topology that offers valuable cost cutting and responsiveness properties. Response time is optimized along with the total costs associated to deterministic and uncertain demands. To formulate such a mathematical program, two models with conflicting objective functions are combined. Developed models input data provided by the United Nations High Commissioner for Refugees (UNHCR)and the results of the integrated model are visualized through trajectories presenting the impacts of using such mathematical programs. Another focus of this study is to incorporate political and security factors along with other crucial elements to the practice of humanitarian operations management which are overlooked in the existing studies of the field. Some of these measures are accessibility, co‐location, security, and human resources which turned out to be influential factors for reducing cost and response time. Valuable Numerical and visual insights are driven upon the inclusion of political and security factors.


INTREPID: Enhancing Field Based Data Collection

Rebecca Hutchins (Booz Allen Hamilton), Taylor Hemby (Booz Allen Hamilton)

Due to our increased mobility, the emergence of new infections and resurgence of old ones, disease outbreaks are increasingly likely and the associated risks continue to increase. To address this increasing risk, there needs to be renewed focus and innovation with respect to the detection and management of outbreaks. Traditional epidemiological approaches such as contact tracing have proven effective in managing smaller scale outbreaks, but a number of challenges have kept these approaches from being effective as the size of the outbreak increases. With the INTREPID platform, we leverage the latest advancements in technology to enhance traditional surveillance approaches and improve disease detection and management.

INTREPID is a customizable, modular, platform comprising mobile devices with INTREPID data collection & analytic software. The INTREPID mobile software provides positive identification of patients, electronic data capture to improve data accuracy, and an analytics/visualization platform to evaluate data and extract key insights in near real-time.

The goal of INTREPID is to improve public health response and specifically to reduce overall morbidity and mortality related to preventable diseases through the application of modern technology. It provides a novel data collection approach that uses patient tracking via biometrics along with near real-time data analysis, allowing for maximized focus of resources to under-served areas. By simplifying data collection and analysis, INTREPID enables insight into the current situation to help improve overall public health response efforts.


Locating Rehabilitative Shelters for Survivors of Domestic Human Trafficking

Renata Konrad (Worcester Polytechnic Institute), Kayse Maass (University of Michigan), Andrew Trapp (Worcester Polytechnic Institute)

Human trafficking is the trade of persons for the gain of others. Forced labour and sexual exploitation are estimated to generate $150 billion (U.S.) globally in illegal profits each year, and involve over 20 million victims worldwide. Although only a small percentage of trafficking survivors utilize group residential rehabilitative facilities, or “long-term shelters,” these facilities are an important component of protection and prosecution efforts of anti-trafficking organizations. In addition to safe housing, for successful rehabilitation, shelters need to be linked to a network of services such as education, psycho-social care, legal advocacy, medical treatment, and life-skills training. Shelters entail a large investment, typically from the non-profit sector, in a small population of survivors. Factors that influence the geographic location of shelters include access to necessary survivor services security issues, and expected future demand. These factors need to be balanced against establishment and maintenance costs as well as shelter capacity. As such, the decisions of whether to open and where to locate a shelter are complex. The discipline of operations research offers tools that can efficiently evaluate many alternatives to find better ways to allocate scarce resources. While there has been a great deal of work in operations research and analytics in the general area of facility location, shelter location for trafficking survivors has received scant attention. We propose a preliminary multi-objective hub-location optimization model which evaluates geographic locations for rehabilitative shelters. The model considers the demand for shelters balanced against access to services, costs and site security as well as proximity to the survivor’s home for possible reintegration with their family.


Logistics Shelter Support Program

James Papagni (New York City Emergency Management), Diandra Hayban (New York City Emergency Management)

New York City Emergency Management (NYCEM) and its partner agencies created the Logistics Shelter Support Program(LSSP) in 2007 in response to Hurricane Katrina. The LSSP was engineered using data collected from logistics studies and with recommendations from mass care providers. The program was ultimately created to provide shelter support for adults, people with special medical needs,people with disabilities and access and functional needs, children, infants, and companion animals in the event of an emergency. It was designed to meet the needs of approximately 70,000 individuals in designated emergency shelters across the five boroughs of NYC for up to seven days. One of the main components of the program is an Emergency Supply Stockpile which consists of 6,100 pallets of mass care supplies.

Past Activations: The program was successfully deployed in preparation for Hurricane Irene and Super Storm Sandy. Most recently it was mobilized in preparation for Hurricane Joaquin.

Lessons Learned and Improvements: Through past activations as well as regular training and exercises, there have been many lessons learned and improvements implemented to strengthen the program. Some of the most recent and notable improvements include:
• Addition of a robust field labor program, to assist shelter staff with the stockpile supplies and in turn make deployment and demobilization faster and more efficient.
• Increased usage of the program as whole through planning and exercise to bolster our ability to respond to a “no-notice” emergency.
• Integration of new commodities to better support infants and those with disabilities and access and functional needs.


Managing Volunteer Convergence at Disaster Relief Centers

Hussain Abualkhair (North Carolina A&T State University), Lauren Davis (North Carolina A&T State University)

Disaster relief centers are the retailers of relief supply chain that provide to disaster victims. They deal with uncertain supply quantities and equipment, uncertain time of the donation and an uncertain number of volunteers. Volunteer convergence, which is the physical movement of volunteers, is one of the biggest challenges for relief center managers. Several emergency management professionals describe volunteer convergence as the “Secondary disaster within the disaster”. To better understand the volunteer convergence phenomenon and find effective managing strategies, it should be studied from the quantitative perspective. In this paper, we use data collected from a case study of volunteer convergence following the 2011 Tuscaloosa tornado in Alabama. The problem is modeled as a queuing system that consists of two multiple server queues representing donor and victim arrival, respectively, and random arrivals and departures of volunteers (servers). The queuing system is modeled as an agent-based simulation model to find a good volunteer assignment policy that reduces customers (donors and victims) waiting time and time in system. Additionally, we are looking to improve the performance of volunteers by reducing the idle times.


Modeling Pick-up Enterprises for Disaster Response

Doug Bodner (Georgia Institute of Technology)

Response to large-scale disasters typically takes the form of a “pick-up” enterprise consisting of numerous government agencies, private firms, and non-governmental organizations. Enterprises are characterized by complexity, no central locus of control, and adaptive behavior among participating organizations. Due to these factors, there may be unintended consequences and secondary effects that impede response effectiveness. For instance, local populations may react to information or authority decrees in unanticipated ways such as looting or dissension. Processes of multiple organizations may interact in unexpected ways resulting in non-cooperative behavior. This research presents an approach to enterprise modeling and simulation that can be used by decision-makers to identify and mitigate such unintended consequences under different disaster scenarios, as well as craft strategies for effective multi-organization disaster response. The model incorporates agent-based and system dynamics simulation and includes responding organizations, supply networks, response assets, infrastructure, and local populations. Since each disaster is unique, the disaster is parameterized as to type, scope and effects. The modeling methodology is detailed, and results from modeling the 1991 Bangladesh cyclone and Operation Sea Angel are demonstrated.


Multi-criteria framework for a reliable aid distribution

Fabiola Regis (Tecnologico de Monterrey), Jaime Mora-Vargas (Tecnologico de Monterrey), Angel Ruiz (Université Laval)

Humanitarian logistics’ decisions need to be made in a short periods of time and in a challenging context .The scarce resources, uncertain demand and the state of the infrastructure requires humanitarian logistics most of the times global solutions and actions. The need of several actors having different goals, interests, capacities and expertise, makes coordination a key aspect to support and improve the humanitarian supply chain. Coordination has therefore become a key aspect and has attracted interest from researchers in the last decade, leading to the development of interesting modeling approaches to support and improve the humanitarian supply chain decisions. Although some of these contributions propose multi-objective models considering, for example, minimization of costs, unmet demand, response time and maximization of equity, very little is said on how to balance the relative importance of these objectives according to the perspective of the different stakeholders and in how to translate their qualitative goals into quantitative ones to be considered in optimization models.

This work addresses five performance criteria, which consider the economic, egalitarian and utilitarian humanitarian logistics’ aspects. In order to develop a multi-criteria framework for a reliable aid distribution.


Opportunities and Pitfalls in current disease forecasting models and methods

Keith Paarporn (Georgia Inst. of Technology), Ceyhun Eksin (Georgia Inst. of Technology), Jeff Shamma (KAUST), Bradford Taylor (Georgia Inst. of Technology), Joshua Weitz (Georgia Inst. of Technology)

Accurate forecasting of infectious diseases through modeling can enable effective interventions impacting resource management and disease control. Valid models vary between diseases by accounting for different biological and social factors that affect disease spread. Forecasting involves fitting these tailored dynamic models to data and then extrapolating them forward in time. Most forecasting methodologies treat dynamics as deterministic, and assume that individuals do not change their behavior during an epidemic. In this poster, we present studies that address these issues. Typically, an estimate of the severity of an emerging epidemic results from identifying a model whose dynamics best fit case data. However, the best fit model may not be accurate due to ‘process noise'. Instead, we propose a method that solves the inverse problem: what set of models feature dynamics compatible with the case data? We show that ignoring process noise leads to overconfidence of estimates by applying our method to the 2014-2015 Ebola outbreak in West Africa. Second, we focus on two disease spread models that account for individual response to disease prevalence. The first model allows healthy individuals to take protective measures based on their awareness of disease prevalence in their community. We show that behavior response to disease prevalence reduces the total number of infections relative to the no behavior response model. The second model incorporates the concern sick individuals may have with spreading the disease to their healthy contacts by allowing them to take preemptive measures. Likewise, healthy individuals use protective measures in response to concern for disease contraction. We propose a game-theoretic model that captures the self-interests of individuals during disease spread. We show that there is a critical level of concern, i.e., empathy, by the sick individuals above which disease is eradicated quickly. Overall, we highlight the opportunities and pitfalls in disease forecasting methods.


Optimizing WHO-EPI Vaccine Distribution Networks

Jayant Rajgopal (University of Pittsburgh), Jung Lim (University of Pittsburgh), Maryam Mofrad (University of Pittsburgh), Lisa Maillart (University of Pittsburgh), Bryan Norman (University of Pittsburgh)

Despite vast improvements in coverage thanks to organizations such as the WHO and GAVI, infectious diseases remain a global concern and millions of children in low and middle income countries around the world remain at risk for diseases preventable via immunization. In addition to the lack of adequate resources in these countries, this can also be attributed to inefficiencies associated with how the WHO-EPI vaccine distribution chains are designed and operated. The basic structure of the distribution network and its operation in virtually every country is identical and almost no thought has been given to rethinking the basic design of the distribution networks along the lines of other modern supply chain networks. A country’s typical EPI network is designed as a four- (or occasionally, three- or five-) level hierarchical structure with strictly arborescent vaccine flows emanating from a single source. The network is supplemented by flows from a subset of nodes in the network, to represent ad hoc outreach sessions to remote locations.

In this research we develop mathematical models to address two main issues. First, we wish to optimally design distribution networks that permit more general structures common in other distribution networks (e.g., lateral shipments, skipping of levels in the chain, delivery loops, etc.), while also considering any country-specific constraints/policies and options for cold storage and transportation resources to be used. Second, we study outreach activities that supplement static immunization facilities and look for ways to systematize and optimize these so that overall coverage rates are maximized for a fixed budget, or alternatively, costs are minimized for providing a minimum desired coverage level.


Post Ebola - Free Health Care Delivery

Bram Dingemans (CAIPA Ltd), Rae Peters (CAIPA Ltd), John McGhie (CAIPA Ltd)

In the aftermath of the Ebola crisis the Sierra Leonean government has been scaling down emergency treatment services, including Ebola Treatment Centres throughout the country. The focus has shifted from emergency management on to building a Free health Care System which can be sustained. CAIPA, a partnership between Crown Agents (CA) and International Procurement Agency (IPA) has been heavily involved in the Department for International Development’s response to Ebola in Sierra Leone since October 2014 (DFID). CAIPA has continued to be active in shaping Sierra Leone’s health supply chain and the reinvigoration of public health initiatives.

CAIPA has used their existing supply chains, on behalf of DFID, to successfully coordinate the donation of pharmaceuticals and medical supplies to 78 public health facilities (hospitals and community health centres) across the country. This has fed into the Sierra Leonean government’s free healthcare (FHC) initiative, which was designed to address high maternal and child mortality rates. This quick-impact solution is designed to integrate into the existing infrastructure. One of CAIPA’s key objectives is to engage in local stakeholder engagement with the Ministry of Health, Social Services and districts to ensure long term sustainability of the Free Health Care Supply Chain. By using the existing resources and supply chain infrastructure, CAIPA has made sure that the right organisations have received essential pharmaceutical and medical supplies at the right time and in optimal conditions.

The post-Ebola period has presented unique challenges for CAIPA in their role in health systems strengthening in Sierra Leone. Key lessons can be learnt from these operations and the adaption of emergency response supply chain into sustainable health structures.


Predicting Association between Prediabetes and Gum Disease

Zeynab Bahrami Bidoni (Georgia Institute of Technology), Paul Griffin (Georgia Institute of Technology)

Diabetes is a disease that has been associated with an increased risk for a number of serious, sometimes life-threatening complications. Some of those risks include, poor dental health. Studies have shown that people with diabetes are more likely to have periodontal disease than people without diabetes, probably because diabetics are more susceptible to contracting infections. Researchers have also discovered that uncontrolled gum disease may even increase the progression of diabetes for pre-diabetic adults. In this study, we proposed a novel computational approach for modeling the typical bionetwork organization based on relationships between demographic attributes, health behaviors, prediabetes and periodontal disease which serves as a biomarker for presymptomatic detection of prediabetes and Gum diseases in the future. We applied our method to forecast the two-way correlation between pre-diabetes and periodontal diseases on kids with age range of 1 to 17 years old.


Refugee Resettlement in Georgia

Unaiza Ahsan (Georgia Institute of Technology), Wes Stayton (Georgia Institute of Technology), Oleksandra Sopova (Kansas State University), Bistra Dilkina (Georgia Institute of Technology)

According to the United States High Commission for Refugees (UNHCR), there are 65.3 million forcibly displaced people in the world today, 21.5 million of them being refugees. This has led to an unprecedented refugee crisis where people from conflict zones are mass migrating to safer lands and the need to resettle them in these countries is critical. Diverse agencies are helping refugees coming to United States to resettle and start their new life in the country. One of the first and most challenging steps of this process is to find affordable housing that also meets a suite of additional constraints and priorities. These include being within a mile of public transportation and near schools, hospitals, faith centers and international grocery stores. We detail an interactive data-driven web-based tool, which incorporates in one consolidated platform most of the needed information. The tool searches, filters and demonstrates a list of possible housing locations, and allows for the dynamic prioritization based on user-specified importance weights on the diverse criteria. The platform was created in a partnership with New American Pathways, a nonprofit that supports refugee resettlement in the metro Atlanta, but exemplifies a methodology that can help many other organizations with similar goals.


Risk pooling in a two-location perishable inventory system

Can Zhang (Georgia Tech), Turgay Ayer (Georgia Tech), Chelsea White (Georgia Tech)

We study a joint ordering and transshipment decision problem in a two-location perishable inventory system. This problem is motivated by a platelet inventory management problem in a local two-hospital system where there are couriers running between the two hospitals on an hourly basis. We formulate the problem as a Markov decision process. We then derive several structural properties of the optimal policies, which lead to a simple heuristic policy that we show can significantly reduce waste while maintaining the same service level compared with the current practice and several other heuristic policies.

Further, our research leads to several interesting insights that are significantly different from results in both the single-location perishable and the two-location non-perishable cases: i) unlike the single-location case where the optimal ordering quantity in the perishable inventory setting is always lower than in the non-perishable setting, we show for the two-location case that permits transshipment, we can be less conservative and order strictly more; ii) we show that the optimal transshipment quantity in the perishable inventory setting is always at least as high as in the non-perishable setting; and iii) we show that the value of risk pooling in the perishable inventory setting is in general higher than in the non-perishable setting, and unlike the non-perishable case, the value of risk pooling can be strictly positive even when the demand at one location is deterministic.


The expansion of Project Last Mile: Balancing global scale and local fit to bring private sector expertise to public health agencies.

Katherine LaMonaca (Yale University), Leslie Curry (Yale University), Adrian Ristow (The Coca-Cola Company), Trip Allport (3 Degrees Ventures), Alexandra Scott (Global Environment and Technology Foundation), Erika Linnander (Yale University)

Background: Although the global health community increasingly establishes partnerships across industry boundaries to address complex challenges, examples of scalable, multi-country partnerships are limited. In particular, the requirements of early-stage alignment for successful expansion are not well understood.

Methods: Project Last Mile (PLM) aims to transfer Coca-Cola’s logistic, supply chain, and marketing expertise to improve public health systems across Africa. Since PLM’s initial work in Tanzania and Ghana, a Global Development Agreement has formed between Coca-Cola, USAID, the Global Fund, and the Bill & Melinda Gates Foundation to partner with 10 African countries by 2020. We report PLM’s expansion to three countries in sub-Saharan Africa, drawing on interactions with PLM partners to describe country-specific approaches and identify factors that promote early-stage alignment with country priorities.

Results: PLM works with ministries of health and local Coca-Cola bottling companies to plan interventions that align with public sector priorities and private sector expertise, resulting in a variety of unique programs tailored to local context. Current partnerships exist with three sub-Saharan African governments to expand medicine pick-up points for patients with chronic illnesses, optimize distribution networks, develop logistics staff capacity, and improve cold chain preventative maintenance. These interventions emerged through the (1) identification of opportunities by established in-country development partners (2) targeted assessment of opportunities by experienced individuals in the Coca-Cola system, (3) alignment with longer-term investments in supply chain development, and (4) strategic use of diverse funding sources from project partners.

Conclusions: PLM’s country-specific approach demonstrates the need for significant investment in early-stage alignment and shows the benefit of flexible funding mechanisms like the Global Development Agreement to promote adaptation of partnership models. These findings can inform the development of other partnerships seeking to scale across country settings.


The International Association of Public Health Logisticians (IAPHL) - Equipping supply chain managers for systems change

Andrew Brown (International Association of Public Health Logisticians (IAPHL)), Motomoke Eomba (John Snow Incorperatied), Walter Proper (John Snow Incorperated), Lea (John Snow Incorperated)

With 1/3 of the world’s population without access to essential medicines and 2016 ushering in the era of the Sustainable Development Goals, there has never been a more important time to renew our focus on developing the health supply chains in countries with the greatest health needs. Health supply chains are complex systems and are very weak in many developing countries, particularly sub Saharan Africa, with a lack of supply chain professionals a critical factor.

IAPHL is a professional association for logisticians and health supply chain managers (co-sponsored by USAID and JSI) with approximately 4000 members from 134 countries. IAPHL supports the professionalization of supply chain managers and others working in the field of public health logistics, with particular focus on developing countries, equipping individuals to strengthen the health systems in which they work. This year, with support from the inSupply project (funded by Bill and Melinda Gates Foundation), IAPHL is undergoing a strategic review. We have been communicating with members and the wider international community to explore how IAPHL can be of even more support to members.

Our members:
• Work at all levels of the supply chain, with the majority at central and regional levels
• Work across a variety of commodity areas
• Note increasing knowledge & networking as their most important objectives of membership
• Would like to see more training and professional development opportunities

For the future our members are looking for services that:
• Encourage networking with other professionals and South to South exchange
• Provide supply chain knowledge and competency development
• Provide updates and opportunities
• Advocate for health supply chain professionals.

IAPHL is working for the common good by giving public health supply chain professionals the support they need to make health system changes in their particular contexts.


Truth-inducing mechanism for medical surplus products allocation

Can Zhang (Georgia Tech), Atalay Atasu (Georgia Tech), Turgay Ayer (Georgia Tech), Beril Toktay ()

We study a product allocation problem faced by a Medical Surplus Recovery Organization (MSRO) that recovers and manages medical surplus products to fulfill the needs of under-served regions where recipients’ needs are unknown to the MSRO. Using a game theoretic analysis, we first identify loss of effectiveness caused by competition among multiple recipients in a recipient-driven model implemented by the MSRO. We then prove that the optimal truth-inducing mechanism has a very simple structure, and we show that our proposed provider-driven model significantly improves the total value provision compared with the recipient-driven models.


Using Mobile Pharmacies To Address Equity In Pharmaceutical Supply Chains In Low-resource Settings

Biplab Bhattacharya (University at Buffalo), Rajan Batta (University at Buffalo), Li Lin (University at Buffalo)

Low-resource communities are often faced with disease burden that they do not have the infrastructure or capital to tackle. Drug stock-outs have been a major problem that such communities face. These governments rely heavily on aid from donor organizations. Supply chain research have over time focused on medical drug distribution from an equality perspective or from an objective of maximizing profits or minimizing costs. However, the needs of medical drug supplies are more complex then what can be addressed by an equal distribution approach.

Factors like the demographic make-up of the community, medical drug availability over time, disease burden, socio-economic indicators and access to medical drugs all play a part in effectively and accurately creating a health equity model. Equity, different from equality, appropriately gives a measure to identify medical drug demand. This measure is dynamic as all the above-identified factors are dynamic in nature as well. This makes the supply chain shift from a business-centric to a community-centric focus. The goal of this research is to create a framework to assess inequity and to use that framework as an input to models that will help tackle inequity.

Low-resource settings are challenged with sparse drug availability in outlets at rural settings. Medicine outlets, being cash constrained, are strapped to keep up with demand. This results in people from the served community having to commute long distances from one provider to another to avail the required drugs. A mobile-outlet location-tour model can address this issue using inequity index as a criterion to choose communities to serve. The model uses two parts to make stocking and routing decisions based on the demand at each community for each drug over a time horizon and taking into consideration the capacity of the mobile-outlet and the physical drug specifications.