Patrick D. Finley

Modeling evacuation of a hospital without electric power

Hospital evacuations that occur during, or as a result of, infrastructure outages are complicated and demanding. Loss of infrastructure services can initiate a chain of events with corresponding management challenges. This report describes a modeling case study of the 2001 evacuation of the Memorial Hermann Hospital in Houston, Texas (USA). The study uses a model designed to track such cascading events following loss of infrastructure services and to identify the staff, resources, and operational adaptations required to sustain patient care and/or conduct an evacuation. The model is based on the assumption that a hospitals primary mission is to provide necessary medical care to all of its patients, even when critical infrastructure services to the hospital and surrounding areas are disrupted. Model logic evaluates the hospitals ability to provide an adequate level of care for all of its patients throughout a period of disruption. If hospital resources are insufficient to provide such care, the model recommends an evacuation. Model features also provide information to support evacuation and resource allocation decisions for optimizing care over the entire population of patients. This report documents the application of the model to a scenario designed to resemble the 2001 evacuation of the Memorial Hermann Hospital, demonstrating the models ability to recreate the timeline of an actual evacuation. The model is also applied to scenarios demonstrating how its output can inform evacuation planning activities and timing.

Modeling Hospitals' Adaptive Capacity during a Loss of Infrastructure Services

Resilience in hospitals - their ability to withstand, adapt to, and rapidly recover from disruptive events - is vital to their role as part of national critical infrastructure. This paper presents a model to provide planning guidance to decision makers about how to make hospitals more resilient against possible disruption scenarios. This model represents a hospitals adaptive capacities that are leveraged to care for patients during loss of infrastructure services (power, water, etc.). The model is an optimization that reallocates and substitutes resources to keep patients in a high care state or allocates resources to allow evacuation if necessary. An illustrative example demonstrates how the model might be used in practice.

Opinion dynamics in gendered social networks: an examination of female engagement teams in afghanistan

International forces in Afghanistan have experienced difficulties in developing constructive engagements with the Afghan population, an experience familiar to a wide range of international agencies working in underdeveloped and developing nations around the world. Recently, forces have begun deploying Female Engagement Teams, female military units who engage directly with women in occupied communities, resulting inmore positive relationships with those communities as a whole. In this paper, we explore the hypothesis that the structure of community-based social networks strongly contributes to the effectiveness of the Female Engagement Team strategy, specifically considering gender-based differences in network community structure. We find that the ability to address both female and male network components provides a superior ability to affect opinions in the network, and can provide an effective means of counteracting influences from opposition forces.

Using GIS Technology to Manage Information Following a Bioterrorism Attack

ABSTRACT The BROOM system was developed to collect, manage and analyze information from bioterrorist attacks on strategic buildings. GIS features help decision-makers and analysts rapidly assess the current status of contaminated facilities and develop optimized cleanup strategies. BROOM consists of networked server, desktop and PDA components. PDAs are deployed to collect samples of suspected bioagents, such as anthrax. Novel geostatistical methods are used to generate contaminant maps and define optimum locations for subsequent sampling. Efficiency and accuracy gains witnessed in field tests show that GIS technology can play a vital role in visualizing, managing and analyzing data from bioterrorism incidents

Creating interaction environments: defining a two-sided market model of the development and dominance of platforms

Interactions between individuals, both economic and social, are increasingly mediated by technological systems. Such platforms facilitate interactions by controlling and regularizing access, while extracting rent from users. The relatively recent idea of two-sided markets has given insights into the distinctive economic features of such arrangements, arising from network effects and the power of the platform operator. Simplifications required to obtain analytical results, while leading to basic understanding, prevent us from posing many important questions. For example we would like to understand how platforms can be secured when the costs and benefits of security differ greatly across users and operators, and when the vulnerabilities of particular designs may only be revealed after they are in wide use. We define an agent-based model that removes many constraints limiting existing analyses (such as uniformity of users, free and perfect information), allowing insights into a much larger class of real systems.

Joint Sandia/NIOSH exercise on aerosol contamination using the BROOM tool.

In February of 2005, a joint exercise involving Sandia National Laboratories (SNL) and the National Institute for Occupational Safety and Health (NIOSH) was conducted in Albuquerque, NM. The SNL participants included the team developing the Building Restoration Operations and Optimization Model (BROOM), a software product developed to expedite sampling and data management activities applicable to facility restoration following a biological contamination event. Integrated data-collection, data-management, and visualization software improve the efficiency of cleanup, minimize facility downtime, and provide a transparent basis for reopening. The exercise was held at an SNL facility, the Coronado Club, a now-closed social club for Sandia employees located on Kirtland Air Force Base. Both NIOSH and SNL had specific objectives for the exercise, and all objectives were met.

A Spatial Biosurveillance Synthetic Data Generator in R

Objective To develop a spatially accurate biosurveillance synthetic data generator for the testing, evaluation, and comparison of new outbreak detection techniques. Introduction Development of new methods for the rapid detection of emerging disease outbreaks is a research priority in the field of biosurveillance. Because real-world data are often proprietary in nature, scientists must utilize synthetic data generation methods to evaluate new detection methodologies. Colizza et. al. have shown that epidemic spread is dependent on the airline transportation network [1], yet current data generators do not operate over network structures. Here we present a new spatial data generator that models the spread of contagion across a network of cities connected by airline routes. The generator is developed in the R programming language and produces data compatible with the popular `surveillance’ software package. Methods Colizza et. al. demonstrate the power-law relationships between city population, air traffic, and degree distribution [1]. We generate a transportation network as a Chung-Lu random graph [2] that preserves these scale-free relationships (Figure 1). First, given a power-law exponent and a desired number of cities, a probability mass function (PMF) is generated that mirrors the expected degree distribution for the given power-law relationship. Values are then sampled from this PMF to generate an expected degree (number of connected cities) for each city in the network. Edges (airline connections) are added to the network probabilistically as described in [2]. Unconnected graph components are each joined to the largest component using linear preferential attachment. Finally, city sizes are calculated based on an observed three-quarter power- law scaling relationship with the sampled degree distribution. Each city is represented as a customizable stochastic compartmental SIR model. Transportation between cities is modeled similar to [2]. An infection is initialized in a single random city and infection counts are recorded in each city for a fixed period of time. A consistent fraction of the modeled infection cases are recorded as daily clinic visits. These counts are then added onto statically generated baseline data for each city to produce a full synthetic data set. Alternatively, data sets can be generated using real-world networks, such as the one maintained by the International Air Transport Association. Results Dynamics such as the number of cities, degree distribution power- law exponent, traffic flow, and disease kinetics can be customized. In the presented example (Figure 2) the outbreak spreads over a 20 city transportation network. Infection spreads rapidly once the more populated hub cities are infected. Cities that are multiple flights away from the initially infected city are infected late in the process. The generator is capable of creating data sets of arbitrary size, length, and connectivity to better mirror a diverse set of observed network types. Conclusions New computational methods for outbreak detection and surveillance must be compared to established approaches. Outbreak mitigation strategies require a realistic model of human transportation behavior to best evaluate impact. These actions require test data that accurately reflect the complexity of the real-world data they would be applied to. The outbreak data generated here represents the complexity of modern transportation networks and are made to be easily integrated with established software packages to allow for rapid testing and deployment. Randomly generated scale-free transportation network with a power-law degree exponent of λ =1.8. City and link sizes are scaled to reflect their weight. An example of observed daily outbreak-related clinic visits across a randomly generated network of 20 cities. Each city is colored by the number of flights required to reach the city from the initial infection location. These generated counts are then added onto baseline data to create a synthetic data set for experimentation. Keywords Simulation; Network; Spatial; Synthetic; Data

The Role of Community Structure in Opinion Cluster Formation

Opinion clustering arises from the collective behavior of a social network. We apply an Opinion Dynamics model to investigate opinion cluster formation in the presence of community structure. Opinion clustering is influenced by the properties of individuals (nodes) and network topology. We determine the sensitivity of opinion cluster formation to changes in node tolerance levels through parameter sweeps. We investigate the effect of network community structure through rewiring the network to lower the community structure. Tolerance variation modifies the effects of community structure on opinion clustering: higher values of tolerance lead to less distinct opinion clustering. Community structure is found to inhibit network wide clusters from forming. We claim that advancing understanding of the role of community structure in social networks can help lead to more informed and effective public health policy.

A General Model of Resource Production and Exchange in Systems of Interdependent Specialists

Infrastructures are networks of dynamically interacting systems designed for the flow of information, energy, and materials. Under certain circumstances, disturbances from a targeted attack or natural disasters can cause cascading failures within and between infrastructures that result in significant service losses and long recovery times. Reliable interdependency models that can capture such multi-network cascading do not exist. The research reported here has extended Sandias infrastructure modeling capabilities by: (1) addressing interdependencies among networks, (2) incorporating adaptive behavioral models into the network models, and (3) providing mechanisms for evaluating vulnerability to targeted attack and unforeseen disruptions. We have applied these capabilities to evaluate the robustness of various systems, and to identify factors that control the scale and duration of disruption. This capability lays the foundation for developing advanced system security solutions that encompass both external shocks and internal dynamics.

Exploring the Value of Learned Representations for Automated Syndromic Definitions

Objective To better define and automate biosurveillance syndrome categorization using modern unsupervised vector embedding techniques. Introduction Comprehensive medical syndrome definitions are critical for outbreak investigation, disease trend monitoring, and public health surveillance. However, because current definitions are based on keyword string-matching, they may miss important distributional information in free text and medical codes that could be used to build a more general classifier. Here, we explore the idea that individual ICD codes can be categorized by examining their contextual relationships across all other ICD codes. We extend previous work in representation learning with medical data [1] by generating dense vector embeddings of these ICD codes found in emergency department (ED) visit records. The resulting representations capture information about disease co-occurrence that would typically require SME involvement and support the development of more robust syndrome definitions. Methods We evaluate our method on anonymized ED visit records obtained from the New York City Department of Health and Mental Hygiene. The data set consists of approximately 3 million records spanning January 2016 to December 2016, each containing from one to ten ICD-9 or ICD-10 codes. We use these data to embed each ICD code into a high-dimensional vector space following techniques described in Mikolov, et al. [2], colloquially known as word2vec. We define an individual code’s context window as the entirety of its current health record. Final vector embeddings are generated using the gensim machine learning library in Python. We generate 300-dimensional embeddings using a skip-gram network for qualitative evaluation. We use the TensorFlow Embedding Projector to visualize the resulting embedding space. We generate a three-dimensional t-SNE visualization with a perplexity of 32 and a learning rate of 10, run for 1,000 iterations (Figure 1). Finally, we use cosine distance to measure the nearest neighbors of common ICD-10 codes to evaluate the consistency of the generated vector embeddings (Table 1). Results T-SNE visualization of the generated vector embeddings confirms our hypothesis that ICD codes can be contextually grouped into distinct syndrome clusters (Figure 1). Manual examination of the resulting embeddings confirms consistency across codes from the same top-level category but also reveals cross-category relationships that would be missed from a strictly hierarchical analysis (Table 1). For example, not only does the method appropriately discover the close relationship between influenza codes J10.1 and A49.2, it also reveals a link between asthma code J45.20 and obesity code E66.09. We believe these learned relationships will be useful both for refining existing syndrome categories and developing new ones. Conclusions The embedding structure supports the hypothesis of distinct syndrome clusters, and nearest-neighbor results expose relationships between categorically unrelated codes (appropriate upon examination). The method works automatically without the need for SME analysis and it provides an objective, data-driven baseline for the development of syndrome definitions and their refinement. References [1] Choi Y, Chiu CY-I, Sontag D. Learning Low-Dimensional Representations of Medical Concepts. AMIA Summits on Translational Science Proceedings. 2016;2016:41-50. [2] Mikolov T, Sutskever I, Chen K, Corrado GS, Dean J. Distributed representations of words and phrases and their compositionality. InAdvances in neural information processing systems 2013 (pp. 3111-3119).