Spatial Agent-Based Simulation Modeling in Public Health
Description
More details
Other editions
Additional editions
Persons
Content
List of Contributors xv
List of Figures xvii
List of Tables xxi
Preface xxiii
Acknowledgements xxix
List of Abbreviations xxxi
1 Introduction 1
1.1 Overview 1
1.2 Malaria 3
1.3 Agent-Based Modeling of Malaria 4
1.4 Contributions 4
1.5 Organization 5
2 Malaria: A Brief History 7
2.1 Overview 7
2.2 Malaria in Human History 7
2.2.1 The Malarial Path: Ancient Origins 8
2.2.2 Naming and Key Discoveries 9
2.2.3 Antimalarial Drugs 9
2.2.4 Prevention Measures 10
2.3 Malaria Epidemiology: A Global View 10
2.3.1 The Malaria Parasite 11
2.3.2 Geographic Distribution 12
2.3.3 Types of Transmission 12
2.3.4 Risk Mapping and Forecasting 13
2.4 Malaria Control 13
3 Agent-Based Modeling and Malaria 17
3.1 Overview 17
3.2 Agent-Based Models (ABMs) 17
3.2.1 Agents 18
3.2.2 Environment 19
3.2.3 Rules 20
3.2.4 Software for ABMs 20
3.3 History and Applications 21
3.3.1 M&S Organizations 21
3.4 Advantages of ABMs 23
3.4.1 Emergence, Aggregation, and Complexity 23
3.4.2 Heterogeneity 24
3.4.3 Learning and Adaptation 24
3.4.4 Flexibility in System Description 24
3.4.5 Inclusion of Multiple Spaces 25
3.4.6 Limitations of ABMs 25
3.4.7 ABMs vs Mathematical Models 27
3.4.8 Applicability of ABMs for Malaria Modeling 28
3.5 Malaria Models: A Review 29
3.5.1 Mathematical Models of Malaria 30
3.5.2 Agent-Based Models (ABMs) of Malaria 33
3.5.3 The Spatial Dimension of Malaria Models 35
3.6 Summary 36
4 The Biological Core Model 39
4.1 Overview 39
4.1.1 Relevant Terms of Interest 40
4.2 The Aquatic Phase 41
4.2.1 Egg (E) 42
4.2.2 Larva (L) 43
4.2.3 Pupa (P) 45
4.3 The Adult Phase 46
4.3.1 Immature Adult (IA) 46
4.3.2 Mate Seeking (MS) 47
4.3.3 Blood Meal Seeking (BMS) 47
4.3.4 Blood Meal Digesting (BMD) 47
4.3.5 Gravid (G) 47
4.4 Aquatic Habitats and Oviposition 48
4.4.1 Aquatic Habitats 48
4.4.2 Oviposition 48
4.5 Senescence and Mortality Rates 50
4.5.1 Senescence 50
4.5.2 Mortality Models: Basic Mathematical Formulation 51
4.6 Mortality in the Core Model 51
4.6.1 Aquatic (Immature) Mortality Rates 52
4.6.2 Adult Mortality Rates 53
4.7 Discussion 53
4.7.1 An Extendible Framework for Other Anopheline Species 53
4.7.2 Weather, Seasonality, and Other Factors 54
4.7.3 Mortality Rates 54
4.8 Summary 54
5 The Agent-Based Model (ABM) 57
5.1 Overview 57
5.2 Model Architecture 58
5.2.1 Object-Oriented Programming (OOP) Terminology 58
5.2.2 Agents 60
5.2.3 Environments 62
5.2.4 Event-Action-List Diagram 62
5.3 Mosquito Population Dynamics 64
5.4 Model Features 66
5.4.1 Processing Steps Ordering 66
5.4.2 Model Assumptions 67
5.4.3 Simulations 69
5.5 Summary 69
6 The Spatial ABM 71
6.1 Overview 71
6.2 The Spatial ABM 74
6.2.1 Definition of Terms 74
6.2.2 Landscapes 75
6.2.3 Landscape Generator Tools 76
6.3 Resource Clustering 79
6.4 Flight Heuristics for Mosquito Agents 81
6.5 Simulation Results 85
6.5.1 Model Verification 85
6.5.2 Landscape Patterns 86
6.5.3 Relative Sizes of Resources 87
6.5.4 Resource Density 88
6.5.5 Combined Habitat Capacity (CHC) 89
6.6 Spatial Heterogeneity 90
6.7 Summary 93
7 Verification, Validation, Replication, and Reproducibility 95
7.1 Overview 95
7.2 Verification and Validation (V&V): A Review 96
7.2.1 Acceptability Assessment and Quality Assurance (QA) 96
7.2.2 Verification and Validation (V&V) 98
7.3 Replication and Reproducibility (R&R): A Review 100
7.4 Summary 101
8 Verification and Validation (V&V) of ABMs 103
8.1 Overview 103
8.2 Verification and Validation (V&V) of ABMs 103
8.3 Phase-Wise Docking 105
8.3.1 Assumptions for the ABMs 105
8.3.2 Phase-Wise Docking Results 107
8.4 Compartmental Docking 110
8.4.1 Implementations of the ABMs 111
8.4.2 Assumptions for the ABMs 112
8.4.3 Compartmental Docking Results 114
8.5 Summary 116
9 Replication and Reproducibility (R&R) of ABMs 121
9.1 Overview 121
9.1.1 Simulation Stochasticity 122
9.1.2 Boundary Types 123
9.2 Vector Control Interventions 124
9.2.1 Larval Source Management (LSM) 125
9.2.2 Insecticide-Treated Nets (ITNs) 126
9.2.3 Population Profiles for ITNs 127
9.2.4 Coverage Schemes for ITNs 127
9.2.5 Applying LSM in Isolation 130
9.2.6 Applying ITNs in Isolation 132
9.2.7 Applying LSM and ITNs in Combination 132
9.3 Simulation Results 134
9.3.1 Simulation Stochasticity 134
9.3.2 LSM in Isolation 134
9.3.3 Impact of Boundary Types 137
9.3.4 ITNs in Isolation 138
9.3.5 LSM and ITNs in Combination 143
9.4 Replication and Reproducibility (R&R) Guidelines 147
9.5 Discussion 150
9.6 Summary 152
10 A Landscape Epidemiology Modeling Framework 155
10.1 Overview 155
10.2 GIS in Public Health 159
10.3 The Study Area and the ABM 160
10.3.1 Features of the Spatial ABM 161
10.3.2 Vector Control Intervention Scenarios 162
10.4 The Geographic Information System (GIS) 163
10.4.1 The GIS-ABM Workflow 163
10.4.2 GIS Processing of Data Layers 164
10.4.3 Feature Counts 165
10.5 Simulations and Spatial Analyses 165
10.5.1 Output Indices 166
10.5.2 Hot Spot Analysis 167
10.5.3 Kriging Analysis 167
10.6 Results 168
10.6.1 Mosquito Abundance 168
10.6.2 Oviposition Count per Aquatic Habitat 171
10.6.3 Blood Meal Count per House 174
10.7 Discussion 177
10.7.1 Stochasticity and Initial Conditions 177
10.7.2 Model Calibration and Parameterization 178
10.7.3 Emergence 178
10.7.4 Complexity 179
10.7.5 Data Resolution (Granularity) 179
10.7.6 Spatial Analyses 180
10.7.7 Habitat-based Interventions 181
10.7.8 Miscellaneous Issues 181
10.8 Conclusions 182
11 The EMOD Individual-Based Model 185
Philip A. Eckhoff and Edward A. Wenger
11.1 Overview 185
11.1.1 Motivation: Modeling of Malaria Eradication 186
11.1.2 Questions that Arise in the Context of Malaria Eradication 187
11.1.3 Spatial Heterogeneity and Metapopulation Effects 188
11.1.4 Implications for Model Structure 190
11.2 Model Structure 193
11.2.1 Human Demographics and Synthetic Population 193
11.2.2 Vector Ecology 194
11.2.3 Vector Transmission 195
11.2.4 Within-Host Disease Dynamics 197
11.2.5 Human Migration and Spatial Effects 198
11.2.6 Stochastic Ensembles 200
11.3 Results 201
11.3.1 Single-Village Simulations 201
11.3.2 Spatial Simulations: Garki District 202
11.3.3 Madagascar 203
11.4 Discussion 206
Appendix A Enzyme Kinetics Model for Vector Growth and Development 209
A.1 Overview 209
A.2 Stochastic Thermodynamic Models 210
A.3 Poikilothermic Development Models 210
A.4 The Sharpe and DeMichele Model 214
A.5 The Schoolfield et al. Model 217
A.6 Depinay et al. Model 219
A.7 Summary 221
Appendix B Flowchart for the ABM 223
B.1 Flowchart for the Agent-Based Model (ABM) 223
Appendix C Additional Files for Chapter 10 233
Appendix D A Postsimulation Analysis Module for Agent-Based Models 239
D.1 Overview 239
D.2 Simulation Output Analysis: A Review 240
D.3 The LiNK Model 243
D.4 P-SAM Architecture 245
D.5 Postsimulation Analysis and Visualization 247
D.6 P-SAM Performance 250
D.7 Conclusion 254
References 255
Index 279
PREFACE
In today's scientific world, computational science is considered the third pillar of scientific inquiry, along with the two traditional pillars of theory and experimentation. Although science is still carried out as an ongoing interplay between theory and experimentation, the increased scale and complexity of both have compelled computational science to be an integral aspect of almost every type of scientific research.
Typically, computational science uses computer simulations (to construct computational models) and quantitative analysis techniques in order to analyze and solve scientific problems. In particular, modeling & simulation (M&S) techniques are being increasingly used to model complex systems, which in general exhibit complex properties such as heterogeneity, dynamic interactions, emergence, learning, and adaptation. With the ever-widening availability of computing resources, the increasing pool of human computational experts and due to its unconstrained applicability across academic discipline boundaries, the importance of M&S continues to grow at a remarkable rate.
Agent-based modeling and simulation (ABMS) is a class of M&S techniques for simulating the actions and interactions of autonomous agents with a view to assessing their effects on the simulated system as a whole. Having its roots from the investigation of complex systems, complex adaptive systems, artificial intelligence, and computer science, ABMS combines elements of game theory, complex systems, emergence, computational sociology, multiagent systems, and evolutionary programming. The suite of models developed using ABMS, known as agent-based models (ABMs), have applications in diverse real-world problems and have become increasingly popular as a modeling approach in almost all branches of science and engineering.
In public health research, epidemics and infectious disease dynamics modeling can be termed as a signature success of ABMS. Uses of M&S in public health include synthesizing knowledge from disparate disciplines, filling the gaps in existing knowledge, conducting cost-benefit trade-off studies, and generating hypotheses. As such, an increasing number of U.S. universities are incorporating systems science and M&S into their curricula and research programs through the schools of public health and other health-related academic departments.
A major objective of this book is to present a practical and useful introduction to the important facets of a sufficiently complex M&S project that largely involved the evolution of a complex ABM. The ABM was developed by experts from multiple academic disciplines. Thus, major portions of the contents of this book materialized as a result of interdisciplinary, collaborative research efforts concerning ABMS (from Computer Science and Engineering) and malaria epidemiology (from Biological Sciences) at the University of Notre Dame [547].
Malaria is one of the oldest and deadliest infectious diseases in humans, and the control of malaria represents one of the greatest public health challenges of the twenty-first century. According to the latest estimates (released in December 2014), the World Health Organization (WHO) reported about 198 million cases of malaria in 2013 and an estimated 584,000 deaths, with half of the world's population (about 3.3 billion) being at risk [567]. Human malaria is transmitted only by female mosquitoes of the genus Anopheles, which are regarded as the primary vectors for transmission.
The ABMs presented in this book were developed by following a conceptual, biological core model of Anopheles gambiae (An. gambiae for short) for malaria epidemiology. The notion of this core model plays a central role in the long development process of multiple versions of the ABMs, as well as in conducting such crucial steps as model verification, validation, and replication. Evolution of the core model has been guided by relevant biological features concerning An. gambiae, which were iteratively refined and incrementally added to the existing pool of model features. Subsequently, the ABMs were updated to reflect the changes.
OUTLINE OF CHAPTERS
Chapter 1 of this book introduces the reader to its major components, presents a brief introduction to malaria and ABMs, and lists our specific contributions. Chapters 2 and 3 present general introductions to malaria and ABMs. Their purpose is to collectively serve as a concise background for readers who are less familiar with the disease and its epidemiological aspects, and why ABMs are particularly useful in modeling diseases like malaria.
Chapter 4 thoroughly describes the biological core model of An. gambiae. After defining some relevant terms of interest, it addresses several important features of the mosquito life cycle, including development in different life-cycle stages, aquatic habitats, oviposition, vector senescence, and density- and age-dependent mortality rates. It also discusses some of the key features, characteristics, and limitations of the core model.
Chapter 5 discusses the design and implementation of a simplified, fixed version of the ABM. Since the ABM is developed in the Java object-oriented programming (OOP) language, we present some relevant OOP terminology. We then describe thearchitecture of the ABM and present class diagrams to elaborate the agents and their environments. In order to capture the major daily events of a typical simulation in a standard fashion, a new type of descriptive diagram, called the Event-Action-List (EAL) diagram, is presented. The chapter also describes the mosquito population dynamics and some of the other characteristics and features of the ABM, including processing steps ordering, initialization, and simulation assumptions.
Chapter 6 presents a spatial extension of the ABM. In general, an ABM can be applied to a domain with or without an explicit representation of space. However, analysis of spatial relationships is fundamental to epidemiology research, as demonstrated by several recent studies. In some cases, an explicit spatial representation may be desired for certain aspects of the ABM to be modeled more realistically. For example, in a malaria ABM, some frequent events performed by the mosquito agents such as obtaining a successful blood meal (host-seeking) or finding an aquatic habitat to lay eggs (oviposition) can be spatially modeled in the landscape in which the agents move. These aspects are also affected by the underlying spatial heterogeneity, which defines the spatial distribution of resources and directly affects the mosquito population in the ABM. In Chapter 6, we describe the modeling aspects of the spatial ABM, the mosquito agents and their spatial movement, the landscapes, and the resource-seeking events. We also describe a custom-built landscape generator tool that is used to generate landscapes with desired characteristics for the spatial ABM and present results concerning the effects of varying landscape patterns, the relative size and density of the aquatic habitats, the overall capacity of the system, and the effects of spatial heterogeneity of the landscapes.
Chapters 7-9 describe the techniques and results of verification, validation, and replication of the ABMs, which in general deal with the measurement and assessment of accuracy of M&S research. They also present the results of examining the impact of two malaria control interventions, namely, larval source management (LSM) and insecticide-treated nets (ITNs). We investigate the effects of LSM and ITNs, applied both in isolation and in combination, on the mosquito agent populations. We compare our results to those reported by previously published malaria models and recommend guidelines for future ABM modelers, summarizing the insights and experiences gained from our work of replicating earlier studies.
Chapter 10 presents a landscape epidemiology modeling framework that integrates a Geographic information system (GIS) with the spatial ABM. The idea of integrating GIS with ABMs is not new, and several studies in multiple domains (e.g., urban land-use change, military mobile communications) have shown such integration. GIS and spatial statistical methods have also been extensively used in entomological and epidemiological studies. In particular, for malaria as a disease, GIS applications have been used for measuring the distribution of mosquito species, their habitats, the control and management of the disease, and so on. However, with the exception of the individual-based model named EMOD (which is presented in Chapter 11), no ABM-based malaria study has yet shown how to effectively integrate an ABM withGIS and other geospatial features and thereby harness the full power of GIS. There is also a vacuum of knowledge in building robust integration frameworks that can guide the use of geospatial features (related to malaria transmission) as model inputs, as opposed to simply use these features as cartographic outputs from the models (as done by most previous studies). In Chapter 10, we show how to effectively integrate simulation outputs from our spatial ABM with a GIS. For a study area in Kenya, we construct different landscape scenarios and perform spatial analyses on the simulation results. Results indicate that the integration of epidemiological simulation-based outputs with spatial analyses techniques within a single modeling framework can be a valuable tool for conducting a variety of disease control activities such as exploring new biological...
