Spatial Agent-Based Simulation Modeling in Public Health E-Book

S. M. Niaz Arifin

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Wiley Series in Modeling and Simulation
• Sprache: Englisch

Seitenzahl: 546

Veröffentlichungsjahr: 2016

Beschreibung

Presents an overview of the complex biological systems used within a global public health setting and features a focus on malaria analysis Bridging the gap between agent-based modeling and simulation (ABMS) and geographic information systems (GIS), Spatial Agent-Based Simulation Modeling in Public Health: Design, Implementation, and Applications for Malaria Epidemiology provides a useful introduction to the development of agent-based models (ABMs) by following a conceptual and biological core model of Anopheles gambiae for malaria epidemiology. Using spatial ABMs, the book includes mosquito (vector) control interventions and GIS as two example applications of ABMs, as well as a brief description of epidemiology modeling. In addition, the authors discuss how to most effectively integrate spatial ABMs with a GIS. The book concludes with a combination of knowledge from entomological, epidemiological, simulation-based, and geo-spatial domains in order to identify and analyze relationships between various transmission variables of the disease. Spatial Agent-Based Simulation Modeling in Public Health: Design, Implementation, and Applications for Malaria Epidemiology also features: * Location-specific mosquito abundance maps that play an important role in malaria control activities by guiding future resource allocation for malaria control and identifying hotspots for further investigation * Discussions on the best modeling practices in an effort to achieve improved efficacy, cost-effectiveness, ecological soundness, and sustainability of vector control for malaria * An overview of the various ABMs, GIS, and spatial statistical methods used in entomological and epidemiological studies, as well as the model malaria study * A companion website with computer source code and flowcharts of the spatial ABM and a landscape generator tool that can simulate landscapes with varying spatial heterogeneity of different types of resources including aquatic habitats and houses Spatial Agent-Based Simulation Modeling in Public Health: Design, Implementation, and Applications for Malaria Epidemiology is an excellent reference for professionals such as modeling and simulation experts, GIS experts, spatial analysts, mathematicians, statisticians, epidemiologists, health policy makers, as well as researchers and scientists who use, manage, or analyze infectious disease data and/or infectious disease-related projects. The book is also ideal for graduate-level courses in modeling and simulation, bioinformatics, biostatistics, public health and policy, and epidemiology.

Leseprobe

TABLE OF CONTENTS

COVER

WILEY SERIES IN MODELING AND SIMULATION

Title Page

COPYRIGHT

DEDICATION

LIST OF CONTRIBUTORS

LIST OF FIGURES

LIST OF TABLES

PREFACE

OUTLINE OF CHAPTERS

INTENDED AUDIENCE

ACKNOWLEDGMENTS

LIST OF ABBREVIATIONS

CHAPTER 1: INTRODUCTION

1.1 OVERVIEW

1.2 MALARIA

1.3 AGENT-BASED MODELING OF MALARIA

1.4 CONTRIBUTIONS

1.5 ORGANIZATION

CHAPTER 2: MALARIA: A BRIEF HISTORY

2.1 OVERVIEW

2.2 MALARIA IN HUMAN HISTORY

2.3 MALARIA EPIDEMIOLOGY: A GLOBAL VIEW

2.4 MALARIA CONTROL

CHAPTER 3: AGENT-BASED MODELING AND MALARIA

3.1 OVERVIEW

3.2 AGENT-BASED MODELS (ABMs)

3.3 HISTORY AND APPLICATIONS

3.4 ADVANTAGES OF ABMs

3.5 MALARIA MODELS: A REVIEW

3.6 SUMMARY

CHAPTER 4: THE BIOLOGICAL CORE MODEL

4.1 OVERVIEW

4.2 THE AQUATIC PHASE

4.3 THE ADULT PHASE

4.4 AQUATIC HABITATS AND OVIPOSITION

4.5 SENESCENCE AND MORTALITY RATES

4.6 MORTALITY IN THE CORE MODEL

4.7 DISCUSSION

4.8 SUMMARY

CHAPTER 5: THE AGENT-BASED MODEL (ABM)

5.1 OVERVIEW

5.2 MODEL ARCHITECTURE

5.3 MOSQUITO POPULATION DYNAMICS

5.4 MODEL FEATURES

5.5 SUMMARY

CHAPTER 6: THE SPATIAL ABM

6.1 OVERVIEW

6.2 THE SPATIAL ABM

6.3 RESOURCE CLUSTERING

6.4 FLIGHT HEURISTICS FOR MOSQUITO AGENTS

6.5 SIMULATION RESULTS

6.6 SPATIAL HETEROGENEITY

6.7 SUMMARY

CHAPTER 7: VERIFICATION, VALIDATION, REPLICATION, AND REPRODUCIBILITY

7.1 OVERVIEW

7.2 VERIFICATION AND VALIDATION (V&V): A REVIEW

7.3 REPLICATION AND REPRODUCIBILITY (R&R): A REVIEW

7.4 SUMMARY

CHAPTER 8: VERIFICATION AND VALIDATION (V&V) OF ABMs

8.1 OVERVIEW

8.2 VERIFICATION AND VALIDATION (V&V) OF ABMs

8.3 PHASE-WISE DOCKING

8.4 COMPARTMENTAL DOCKING

8.5 SUMMARY

CHAPTER 9: REPLICATION AND REPRODUCIBILITY (R&R) OF ABMs

9.1 OVERVIEW

9.2 VECTOR CONTROL INTERVENTIONS

9.3 SIMULATION RESULTS

9.4 REPLICATION AND REPRODUCIBILITY (R&R) GUIDELINES

9.5 DISCUSSION

9.6 SUMMARY

CHAPTER 10: A LANDSCAPE EPIDEMIOLOGY MODELING FRAMEWORK

10.1 OVERVIEW

10.2 GIS IN PUBLIC HEALTH

10.3 THE STUDY AREA AND THE ABM

10.4 THE GEOGRAPHIC INFORMATION SYSTEM (GIS)

10.5 SIMULATIONS AND SPATIAL ANALYSES

10.6 RESULTS

10.7 DISCUSSION

10.8 CONCLUSIONS

CHAPTER 11: THE EMOD INDIVIDUAL-BASED MODEL

11.1 OVERVIEW

11.2 MODEL STRUCTURE

11.3 RESULTS

11.4 DISCUSSION

APPENDIX A: ENZYME KINETICS MODEL FOR VECTOR GROWTH AND DEVELOPMENT

A.1 OVERVIEW

A.2 STOCHASTIC THERMODYNAMIC MODELS

A.3 POIKILOTHERMIC DEVELOPMENT MODELS

A.4 THE SHARPE AND DEMICHELE MODEL

A.5 THE SCHOOLFIELD

ET AL.

MODEL

A.6 DEPINAY

ET AL.

MODEL

A.7 SUMMARY

APPENDIX B: FLOWCHART FOR THE ABM

B.1 FLOWCHART FOR THE AGENT-BASED MODEL (ABM)

APPENDIX C: ADDITIONAL FILES FOR CHAPTER 10

APPENDIX D: A POSTSIMULATION ANALYSIS MODULE FOR AGENT-BASED MODELS

D.1 OVERVIEW

D.2 SIMULATION OUTPUT ANALYSIS: A REVIEW

D.3 THE LINK MODEL

D.4 P-SAM ARCHITECTURE

D.5 POSTSIMULATION ANALYSIS AND VISUALIZATION

D.6 P-SAM PERFORMANCE

D.7 CONCLUSION

REFERENCES

INDEX

END USER LICENSE AGREEMENT

Pages

xv

xvii

xviii

xix

234

235

236

237

238

xx

xxi

xxii

xxiii

xxiv

xxv

xxvi

xxvii

xxix

xxxi

xxxii

1

2

3

4

5

7

8

9

10

11

12

13

14

15

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

57

58

59

60

61

62

63

64

65

66

67

68

69

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

95

96

97

98

99

100

101

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

269

268

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

Guide

Cover

Table of Contents

Preface

Begin Reading

List of Illustrations

CHAPTER 1: INTRODUCTION

Figure 1.1 Book components. Logical connections between the major components and the chapters are indicated. This book primarily concerns modeling & simulation (M&S), specifically agent-based modeling (ABM), in public health (malaria epidemiology). Chapters 2 and 3 present some general background of malaria and agent-based models (ABMs) and discuss the applicability of ABMs in malaria epidemiology research, which, in turn, broadly falls under the realm of computational biology. Chapters 4–6 describe the biological core model, the agent-based models (ABMs), and the spatial ABMs, and form the core of the book. Chapters 7–9 discuss the verification, validation, and replication issues of the ABMs. Chapter 10 presents a landscape epidemiology modeling framework, and Chapter 11 presents the

EMOD

individual-based model. Note that some chapters may overlap into multiple components. All components of the book share the global/public health implications.

CHAPTER 2: MALARIA: A BRIEF HISTORY

Figure 2.1 Life cycle of the malaria parasite.

CHAPTER 4: THE BIOLOGICAL CORE MODEL

Figure 4.1 Life cycle of mosquitoes in the models. The

An. gambiae

mosquito life cycle consists of

aquatic

and

adult

phases. The

aquatic

phase consists of three aquatic stages:

Egg (E)

,

Larva (L)

, and

Pupa (P)

. The

adult

phase consists of five adult stages:

Immature Adult (IA)

,

Mate Seeking (MS)

,

Blood Meal Seeking (BMS)

,

Blood Meal Digesting (BMD)

, and

Gravid (G)

. Each oval represents a stage in the model. Permissible time transition windows (from one stage to another) are shown next to the corresponding stage transition arrows as rounded rectangles. Note that adult males, once reaching the

Mate Seeking

stage, remain forever in that stage until they die; adult females cycle through obtaining blood meals (in

Blood Meal Seeking

stage), developing eggs (in

Blood Meal Digesting

stage), and ovipositing the eggs (in

Gravid

stage) until they die.

Figure 4.2 The egg hatching time distribution used in the core model. The

An. gambiae

egg hatching time distribution is strongly skewed to the right, with 89% of the eggs hatching during the second and third day after oviposition, 10% hatching during the next 4 days, and the remaining 1% hatching over the subsequent week [575]. The x-axis denotes hatching time (days), and the y-axis denotes the probability of hatching.

CHAPTER 5: THE AGENT-BASED MODEL (ABM)

Figure 5.1 A simplified class diagram of the model architecture in the ABM. Attributes and methods are omitted for simplicity.

Figure 5.2 A simplified class diagram for the agents in the ABM. Each

MosquitoAgent

class instance is inherited as a subclass of an

Agent

superclass instance. The

get

and

set

methods are used to access and modify the corresponding attributes, respectively. Attributes and methods that are added for the spatial extension are marked in gray.

Figure 5.3 A simplified class diagram for the environments in the ABM. A generic

Environment

class can be inherited as a

MosquitoEnvironment

class, which can be further specialized as either an

AquaticEnvironment

or a

HumanEnvironment

class. The

get

and

set

methods are used to access and modify the corresponding attributes, respectively. Attributes and methods that are added for the spatial extension are marked in gray.

Figure 5.4 A simplified class diagram for the agentlists in the ABM. Agentlists bind different sets of mosquito agents to their corresponding environments depending on their life-cycle phases. A generic

AgentList

class instance can be inherited as either an

AquaticAgentList

or an

AdultAgentList

class instance. An

AquaticAgentList

instance contains aquatic mosquito agents for a specific

AquaticEnvironment

instance and an

AdultAgentList

instance contains adult mosquito agents for a specific

MosquitoEnvironment

instance. The

get

and

set

methods are used to access and modify the corresponding attributes, respectively.

Figure 5.5 An event-action-list (EAL) diagram for the ABM. Each

squashed rectangle

represents an event-action pair, in which the

event

is denoted at the upper half, and the

action

is denoted at the lower half. Each

rectangle

represents the

list(s)

(data structures) of agents affected by the event-action pair.

Figure 5.6 The ordering of the key processing steps performed in a single time step of a simulation run. The

dependency relationships

between the processing steps are crucial to ensure the correctness of the model's output.

Figure 5.7 Dependency relationships in processing steps ordering can be viewed as a

directed acyclic graph (DAG)

. (a) The interdependent steps performed in a single time step of a simulation run are connected by directed dashed lines. (b) and (c) Two possible ordering that preserve the dependency relationships of all the major processing steps. By

topological sort

, (c) can be arranged so that all directed edges go from left to right, as in (b).

CHAPTER 6: THE SPATIAL ABM

Figure 6.1 Examples of three types of landscapes: (a) regular, (b) random, and (c) hybrid. Gray and white rectangles represent spatial resources and empty cells in the mosquito environment, respectively.

Figure 6.2 Screenshot of an early version of the landscape generator tool,

AnophGUI

.

Figure 6.3 Screenshot of the latest version of the landscape generator tool,

VectorLand

.

VectorLand

can generate landscapes with varying spatial heterogeneity of both types of resources: aquatic habitats and houses. Locations of resources can be controlled using the

Clustering

sliders across both axes. Additional statistics about the generated landscape and legends are also shown in separate panels.

Figure 6.4 Examples of landscapes with different clustering schemes of resources. Filled and open circles denote different types of resources. (a) Cluster around center. (b) Cluster around the first object. (c) Dynamic clustering. (d) Multiple clusters ().

Figure 6.5 Controlling the clusters along a specific axis. Twenty resources of both types are used. (a) Along

x

-axis. (b) Along both axes.

Figure 6.6 Foraging event for mosquito agents. (a) Agent movement in

Blood Meal Seeking

(

BMS

) stage. (b) Agent movement in

Gravid (G)

stage. (c) The

movement

event.

Figure 6.7 Flight heuristics for mosquito agents. Detailed activities by mosquito agents in host-seeking, oviposition, and foraging/search are shown.

Figure 6.8 Model verification results. Both the nonspatial and the spatial ABMs yield consistent results with identical parameter settings. Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

Figure 6.9 Results of using different landscape patterns. In the spatial ABM, the use of

regular

and

random

landscape patterns does not significantly alter the population levels. Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

Figure 6.10 Results for resource size variation with 1 and 16 aquatic habitats. For both the nonspatial and the spatial models, the numbers of aquatic habitats are increased as squares of the first 10 integers. This Figure shows

VA

with landscapes having 1 and 16 aquatic habitats.

Figure 6.11 Results for resource size variation with 49 and 100 aquatic habitats. For both the nonspatial and the spatial models, the numbers of aquatic habitats are increased as squares of the first 10 integers. This Figure shows with landscapes having 49 and 100 aquatic habitats.

Figure 6.12 Results for resource density variation. This Figure shows the effects of varying by increasing the number of

AH

s within the same hybrid landscape. Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

Figure 6.13 Results for system capacity variation. Using hybrid landscapes, the system capacity is gradually increased from to . Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

Figure 6.14 Sample landscapes. (a) A landscape with resource densities

below

the critical level. (b) A landscape with resource densities

above

the critical level.

Figure 6.15 Results for resource density

below

the critical level. in a landscape with and resource density

below

the critical level. Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

Figure 6.16 Results for resource density

above

the critical level. in a landscape with and resource density

above

the critical level. Each graph represents the average of 40 simulation runs and shows only relevant portions of the 1-year simulation results.

CHAPTER 8: VERIFICATION AND VALIDATION (V&V) OF ABMs

Figure 8.1 The phase-wise docking workflow. The programming language-specific implementations and versions of the ABMs developed in Java and C++ are denoted as

J1

,

J2

,

J3

, and

CPP

. Unidirectional solid arrows indicate immediate successorship between the implementations. Bidirectional solid arrow indicates

verification

relationships between

J1

&

CPP

. Dashed arrows indicate

validation

relationships with the core model.

Figure 8.2 Phase-wise docking results for Phase 1. Mosquito abundance graphs from

J1

&

CPP

outputs are plotted for different populations. The -axis denotes simulation time (in days) and the -axis denotes abundances.

Figure 8.3 Phase-wise docking results for Phase 3. Mosquito abundance graphs from

J3

&

CPP

outputs are plotted for different populations. The -axis denotes simulation time (in days) and the -axis denotes abundances.

Figure 8.4 The compartmental docking workflow. The programming language-specific implementations and versions of the ABMs developed in Java and C++ are denoted as

J1

,

J2

,

CPP1

, and

CPP2

. Bidirectional arrows indicate

verification

relationships between the four implementations. The dashed arrow, between the re-factored implementations

J2

and

CPP2

, indicates

internal verification

, since these two are docked with respect to each other. Unidirectional dashed arrows indicate

validation

relationships between the core model and the four implementations. source© 2010 IEEE. Reprinted, with permission, from Arifin

et al.

[22].

Figure 8.5 Simplified life cycle of mosquitoes for compartmental docking. Two slightly different versions were used as follows: (a) A model with a single aquatic habitat and (b) A model with multiple aquatic habitats. The

aquatic

phase is unchanged, consisting of three aquatic stages:

Egg

,

Larva

, and

Pupa

. The

adult

phase, however, consists of four adult stages:

Immature Adult

,

Blood Meal Seeking

,

Blood Meal Digesting

, and

Gravid

. Each oval represents a stage in the model. Male agents are omitted. Adult female agents cycle through obtaining blood meals, developing eggs, and ovipositing the eggs until they die.

Figure 8.6 Compartmental docking results for

Phase 1

. Each subFigure depicts the fourfold outputs from the four ABMs:

CPP1

,

J1

, and

CPP2/Java2

(since

CPP2

and

J2

were verified with respect to each other, their outputs were merged as a single graph). Different mosquito populations are shown: (a) adult agents, (b) all agents (adults and aquatic), (c) the 1-day-old equivalent larval population (see Eq. 4.9), and (d) biomass (see Eq. 4.8). The legend at the bottom shows the corresponding ABMs. Within each subfigure, each color-coded plot represents outputs from a specific ABM, with color keys presented in the legend. The -axis denotes simulation time (days) and the -axis denotes abundance. The first 100 days of the simulations are shown.

Figure 8.7 Compartmental docking results for

Phase 2

. Each subFigure depicts the fourfold outputs from the four ABMs:

CPP1

,

J1

, and

CPP2/Java2

. Different mosquito populations are shown: (a) adult agents, (b) all agents (adults and aquatic), (c) the 1-day-old equivalent larval population (see Eq. 4.9), and (d) biomass (see Eq. 4.8). The legend at the bottom shows the corresponding ABMs. Within each subfigure, each color-coded plot represents outputs from a specific ABM, with color keys presented in the legend. The -axis denotes simulation time (days) and the -axis denotes abundance. The first 100 days of the simulations are shown.

Figure 8.8 Compartmental docking results for

Phase 3

. Each subFigure depicts the fourfold outputs from the four ABMs:

CPP1

,

J1

, and

CPP2/Java2

. Different mosquito populations are shown: (a) all agents (adults and aquatic) and (b) biomass (see Eq. (4.8)). The legend at the bottom shows the corresponding ABMs. The -axis denotes simulation time (days) and the -axis denotes abundance. The first 100 days of the simulations are shown.

CHAPTER 9: REPLICATION AND REPRODUCIBILITY (R&R) OF ABMs

Figure 9.1 Coverage schemes for ITNs. (a) Household-level partial coverage with single chance. (b) Household-level partial coverage with multiple chances. denotes the number of persons in the house. (c) Household-level complete coverage.

Figure 9.2 Landscapes for Applying LSM in Isolation. The grid-based landscapes are digitized and reproduced from the GN-LSM study [222]. Each landscape contains 70 aquatic habitats (circles) and 20 houses (house icons). Within each landscape, the houses are arranged either diagonally, horizontally, or vertically. For each arrangement, seven scenarios of LSM are shown; from left to right: NOCTRL means no LSM; T1, T2, and T3 refer to targeted removal of aquatic habitats within 100, 200, and of surrounding houses, accounting for 4, 17, and 28 of 70 habitats, respectively. C1, C2, and C3 refer to nontargeted, random removal of the same numbers of aquatic habitats as the corresponding targeted interventions.

Figure 9.3 Landscape for Applying ITNs in Isolation. The landscape is digitized and reproduced from the GN-ITNs study [221]. It contains 90 aquatic habitats (circles) that are randomly distributed and 50 houses (house icons) that are arranged diagonally.

Figure 9.4 Sample landscapes for applying LSM and ITNs in combination. Each landscape contains 200 aquatic habitats and different densities of houses (). (a) with human population of 100. (b) with human population of 350. (c) with human population of 1000. Aquatic habitats and houses are shown as circles and house-shaped icons, respectively. For 240 distinct parameter combinations (see Table 9.3), similar landscapes are generated and 50 replicated simulations are run for each.

Figure 9.5 Sufficient number of replicated simulation runs can smooth out the simulation stochasticity effects. The importance of performing multiple simulation runs (instead of a single run) can be seen by comparing abundances for

maximum

,

minimum

, and

average

cases. Four LSM scenarios are shown (see Figure 9.2): (a) and (b) refer to scenarios C1 and C2, respectively, and use absorbing boundaries with nontargeted, random removal of aquatic habitats. (c) and (d) refer to the same scenarios and use nonabsorbing boundaries. Within each scenario, the three time-series plots represent the maximum, the minimum, and the average mosquito abundances, respectively, obtained across all 50 replicated runs in each time step.

Figure 9.6 The Figure depicts the full 1-year results of applying LSM in isolation with absorbing boundaries as we replicate the results of GN-LSM [222]. Each subFigure represents a specific LSM scenario. The -axis denotes simulation time (days) and the -axis denotes mosquito abundance. The three targeted interventions T1, T2, and T3 refer to the removal of aquatic habitats within 100, 200, and of surrounding houses, accounting for 4, 17, and 28 of 70 habitats, respectively. C1, C2, and C3 refer to nontargeted, random removal of the same numbers of aquatic habitats as the corresponding targeted interventions. Within each subfigure, the

Diagonal

,

Horizontal

, and

Vertical

plots represent abundances (for the specified LSM scenario) for three different arrangements of houses in the landscapes (see Figure 9.2 for the landscapes). With an

absorbing boundary

, mosquitoes are killed when they hit an edge of the landscape's boundary. Each simulation is run 50 times and the average results are reported. This Figure represents averages of a total of 900 () simulations.

Figure 9.7 The Figure depicts the full 1-year results of applying LSM in isolation with nonabsorbing boundaries as we replicate the results of GN-LSM [222]. Each subFigure represents a specific LSM scenario. The -axis denotes simulation time (days) and the -axis denotes mosquito abundance. The three targeted interventions T1, T2, and T3 refer to the removal of aquatic habitats within , , and of surrounding houses, accounting for 4, 17, and 28 of 70 habitats, respectively. C1, C2, and C3 refer to nontargeted, random removal of the same numbers of aquatic habitats as the corresponding targeted interventions. Within each subfigure, the

Diagonal

,

Horizontal

, and

Vertical

plots represent abundances (for the specified LSM scenario) for three different arrangements of houses in the landscapes (see Figure 9.2 for the landscapes). With a

nonabsorbing boundary

, when mosquitoes hit an edge of the landscape's boundary, they enter the landscape from the edge directly opposite of the exiting edge and thus are not killed due to hitting the edge. Each simulation is run 50 times and the average results are reported. This Figure represents averages of a total of 900 () simulations.

Figure 9.8 The Figure depicts the full 1-year results of applying ITNs in isolation with household-level partial coverage and single chance for host-seeking as we replicate the results of GN-ITNs [221]. Each subFigure represents a specific combination for the three ITNs parameters of coverage, repellence, and mortality. The -axis denotes simulation time (days) and the -axis denotes mosquito abundance. Each row represents a specific coverage () for ITNs (e.g., ). Each column represents a specific repellence () for ITNs (e.g., ). Within each subfigure, each color-coded plot represents a specific mortality () value for ITNs (e.g., ), with mortality () color keys at the bottom of the Figure The Figure represents averages of a total of 3000 () simulations. Nonabsorbing boundaries are used. For the partial coverage schemes, see Section 9.2.4.

Figure 9.9 The Figure depicts the full 1-year results of applying ITNs in isolation with household-level complete coverage as we replicate the results of GN-ITNs [221]. Each subFigure represents a specific combination for the three ITNs parameters of coverage, repellence, and mortality. The -axis denotes simulation time (days) and the -axis denotes mosquito abundance. Each row represents a specific coverage () for ITNs (e.g., ). Each column represents a specific repellence () for ITNs (e.g., ). Within each subfigure, each color-coded plot represents a specific mortality () value for ITNs (e.g., ), with mortality () color keys at the bottom of the Figure The Figure represents averages of a total of 3000 () simulations. Nonabsorbing boundaries are used. For the complete coverage scheme, see Section 9.2.4.

Figure 9.10 The Figure depicts the full 1-year results of applying ITNs in isolation with household-level partial coverage and multiple chances for host-seeking as we replicate the results of GN-ITNs [221]. Each subFigure represents a specific combination for the three ITNs parameters of coverage, repellence, and mortality. The -axis denotes simulation time (days) and the -axis denotes mosquito abundance. Each row represents a specific coverage () for ITNs (e.g., ). Each column represents a specific repellence () for ITNs (e.g., ). Within each subfigure, each color-coded plot represents a specific mortality () value for ITNs (e.g., ), with mortality () color keys at the bottom of the Figure The Figure represents averages of a total of 3000 () simulations. Nonabsorbing boundaries are used. For the partial coverage schemes, see Section 9.2.4.

Figure 9.11 Percent reductions in mosquito abundance by ITNs, applied in isolation, comparing household-level partial coverage (with multiple chances for host-seeking) and complete coverage. Each subFigure represents a specific combination of coverage scheme (partial or complete) and repellence () for ITNs: (a) and (b) show the partial scheme with and , respectively. (c) and (d) show the complete scheme with and , respectively. The -axis denotes ITNs coverage and the -axis denotes ITNs mortality. The upper row represents household-level partial coverage and the lower row represents household-level complete coverage, as marked on the left. Each column represents a specific repellence () value, as marked on the top. ITNs are applied at day 100 in the grid-based landscape with 50 houses having a total human population of 185. The percent reduction (PR) values are represented as filled contour plots in each subFigure The color bar on the right quantifies the PR isolines. Results for the entire parameter space are depicted in Figure 9.12.

Figure 9.12 Percent reductions in mosquito abundance by ITNs, applied in isolation, comparing household-level partial coverage (with multiple chances for host-seeking) and complete coverage. This Figure shows results for the entire parameter space. For other details, see Figure 9.11.

Figure 9.13 Percent reductions in mosquito abundance as a function of LSM coverage and ITNs coverage when LSM and ITNs are applied in combination. The -axis denotes ITNs coverage and the -axis denotes LSM coverage. Each subFigure represents a filled contour plot where the isolines are labeled with specific percent reduction (PR) values, with specific combination of density of houses () and mortality () for ITNs: subFigure (a)–(c) represent with of

Low

,

Medium

, and

High

, respectively; subFigure (d)–(f) represent with of

Low

,

Medium

, and

High

, respectively. ITNs repellence () is fixed at 0.5. Each simulation is run for 1 year; both LSM and ITNs are applied at day 100 and continued up to the end of the simulation. The color bar on the right quantifies the PR isolines. The Figure represents average percent reduction values of a total of 6000 () simulations. For ITNs, household-level complete coverage scheme is used (see Figure 9.1c). Nonabsorbing boundaries are used. Sample landscapes with the three levels are shown in Figure 9.4. The Figure depicts selected results that involve a subset of the parameters from Table 9.3.

Figure 9.14 The Figure depicts percent reductions results that involve the entire parameter space from Table 9.3 as a function of LSM coverage and ITNs coverage when LSM and ITNs are applied in combination. This Figure shows results for the entire parameter space. For other details, see Figure 9.13.

CHAPTER 10: A LANDSCAPE EPIDEMIOLOGY MODELING FRAMEWORK

Figure 10.1 The study area. (a) Kenya boundary and administrative units (provinces). (b) Study area with selected data layers; the outlined polygon represents a subset of villages selected for the simulation runs in this study. (c) Village cluster in Asembo. (d) Legends.

Figure 10.2 Selected sets of GIS features for Kenya. (a) Villages. (b) Water sources.

Figure 10.3 Maps for the

mosquito abundances

index. (a) Abundance map for baseline. (b) Legends: for clarity, houses and pit latrines are not shown. (c) Abundance map for . (d) Abundance map for . (e) Abundance map for . (f) Abundance map for .

Figure 10.4 Kriged maps for the

mosquito abundances

index. (a) Kriged abundance map for baseline. (b) Legends. (c)–(f) The four intervention scenarios.

Figure 10.5 Maps for the

oviposition count per aquatic habitat

index. Oviposition counts are categorized using the same quantitative scale and are shown using graduated symbols. For clarity, houses and pit latrines are not shown. Hot spots and cold spots are spatially clustered. (a) Baseline. (b) Legends. (c)–(f) The four intervention scenarios.

Figure 10.6 Kriged maps for the

oviposition count per aquatic habitat

index. (a) Baseline. (b) Legends. (c)–(f) The four intervention scenarios.

Figure 10.7 Maps for the

blood meal count per house

index. Blood meal counts are categorized using the same quantitative scale. For clarity, houses and pit latrines are not shown. Hot spots and cold spots are spatially clustered using two confidence intervals (CIs) of 95 and . (a) Baseline. (b) Legends. (c)–(f) The four intervention scenarios.

Figure 10.8 Kriged maps for all scenarios for the

blood meal count per house

index. (a) Baseline. (b) Legends. (c)–(f) The four intervention scenarios.

CHAPTER 11: THE EMOD INDIVIDUAL-BASED MODEL

Figure 11.1 The EMOD model of the mosquito feeding cycle with outcomes.

Figure 11.2 Simulation of Namawala, Tanzania, as described in [160], shows changes in EIR due to seasonal variations at baseline (a), with IRS (b), and with IRS and a transmission-blocking vaccine (c). Adding interventions adds variability to the simulation, with the transmission-blocking vaccine providing an additional reduction in transmission of over a factor of 2, which would be difficult for further interventions targeting indoor feeding to achieve on top of IRS.

Figure 11.3 Outcomes of different intervention combinations in the Garki District [163]. The light gray is baseline, the dark gray adds in IRS, the dark line is with infrequent drug campaigns, and the dark dashed line with frequent drug campaigns. The dots are field data from the Project.

Figure 11.4 Modeled distribution of detected prevalence in Madagascar. This can be compared to estimates for prevalence from the Malaria Atlas Project [232].

Figure 11.5 Modeled vector population density distributions in Madagascar. (a) Adult vector distribution averaged daily over a simulated 3-year period. (b) Sporozoite rates for all local

Anopheles

averaged over the same period. Average rates are highest in tropical wet and dry climate zones due to the high rates at the end of the wet season, when the average age of the vector population increases without a large influx of emerging mosquitoes.

Figure 11.6 (a) Separatrix plots of the modeled probability of successful Eradication in Madagascar, showing the division of the high probability of success region separated from the low probability of success region. (b) The estimate variance. The two parameters shown here are ITN coverage and a linear scaling of migration rates. In these simulations, all vectors are indoor feeders and bed nets have perfect efficacy and do not decay. For realistic vector populations including outdoor biting, bed nets alone are insufficient for elimination and must be combined with other interventions. Increasing migration rates given broadly uniform high coverage spreads out the transition from low probability of success to high probability of success.

APPENDIX A: ENZYME KINETICS MODEL FOR VECTOR GROWTH AND DEVELOPMENT

Figure A.1 Energy states of the kinetic model (redrawn from [488]). The directed arrows indicate possible transitions.

Energy State 1

predominates at low temperatures.

Energy State 2

represents the active enzyme configuration and predominates over the mid-temperature range.

Energy State 3

predominates at high temperatures.

APPENDIX C: ADDITIONAL FILES FOR CHAPTER 10

Figure C.1 Clipped eater sources for Kenya. The Figure shows different water source features, including rivers, wetlands, and other types of water points, clipped within Kenya.

Figure C.2 Clipped village projections for Kenya.

Figure C.3 Polygon creation process. The polygon is created using the ArcGIS software release 10 [16]. A new shapefile is created for the desired polygon. The boundary for the polygon is created in the editing mode and the area of the polygon is calculated using the

Calculate Areas

tool. Finally, the polygon area is highlighted in the editing mode.

Figure C.4 Clipped habitats within the selected polygon. A new shapefile is created for the desired polygon. The boundary for the polygon is created in the editing mode and the area of the polygon is calculated using the

Calculate Areas

tool using the ArcGIS software release 10 [16]. Finally, the polygon area is highlighted in the editing mode.

Figure C.5 Data conversion to raster format, Part 1. The figure shows the values and counts of house features in the attribute table after the raster conversion. From these attributes, the counts of features per grid cell are generated.

Figure C.6 Data conversion to raster format, Part 2. The Figure shows all features converted into the raster format within the polygon area for the ABM. The interpretation of each (as supplied by the ABM) for each feature and count of features per grid cell are also linked.

APPENDIX D: A POSTSIMULATION ANALYSIS MODULE FOR AGENT-BASED MODELS

Figure D.1 The P-SAM architecture.

Figure D.2 P-SAM Infection Statistics tab.

Figure D.3 P-SAM Roaming Infection Statistics tab.

Figure D.4 P-SAM birth and death statistics tab.

Figure D.5 Example of a

pathogen transmission graph

. Nodes represent macaques and edges represent infection events. Loops represent autoinfection.

Figure D.6 P-SAM Summary Statistics tab.

Figure D.7 P-SAM performance. For each phase, the seven data points represent average values of output files, obtained by varying parameter settings of the LiNK model. The same seven sets of parameters are used for all phases.

Phase 1

bypasses the GUI.

Phase 2

eliminates unnecessary existence checks for

infection

hash used in the

Writer

, thus reducing search penalties for

macaque

nodes in the hash. Finally,

Phase 3

optimizes some hash operations using sets.

List of Tables

CHAPTER 3: AGENT-BASED MODELING AND MALARIA

Table 3.1 Applications of Agent-Based Models (ABMs)

Table 3.2 Malaria Models: A Comparison of Features

CHAPTER 4: THE BIOLOGICAL CORE MODEL

Table 4.1 Summary of Updated Features in the Core Model

Table 4.2 Symbols and parameters used in the core model and the ABMs. Parameters are listed in order of appearance in the text

Table 4.3 Larval development parameters for

An. gambiae

CHAPTER 7: VERIFICATION, VALIDATION, REPLICATION, AND REPRODUCIBILITY

Table 7.1 Methodologies and Techniques Commonly Used for V&V in M&S Research

CHAPTER 8: VERIFICATION AND VALIDATION (V&V) OF ABMs

Table 8.1 V&V Techniques Used for the ABMs

Table 8.2 Simplified Stage Transition Times for Phase-Wise Docking

Table 8.3 Compartmental Docking Issues in Phase 1

Table 8.4 Compartmental Docking Issues in Phases 2–3

CHAPTER 9: REPLICATION AND REPRODUCIBILITY (R&R) OF ABMs

Table 9.1 Population Profiles for Varying Levels of ITNs Coverage

Table 9.2 Parameter Space for ITNs

Table 9.3 Parameter Space for LSM and ITNs

Table 9.4 Percent Reductions in Abundance with LSM

CHAPTER 10: A LANDSCAPE EPIDEMIOLOGY MODELING FRAMEWORK

Table 10.1 Vector Control Intervention Scenarios

Table 10.2 GIS Feature Types and Counts for the ABM

CHAPTER 11: THE EMOD INDIVIDUAL-BASED MODEL

Table 11.1 Summary of Issues for Eradication Modeling

Table 11.2 Summary of Features Desired for Eradication Modeling

APPENDIX A: ENZYME KINETICS MODEL FOR VECTOR GROWTH AND DEVELOPMENT

Table A.1 Entropy and Enthalpy of Activation

Table A.2 Enzyme, Reaction, and Rate Constant

APPENDIX D: A POSTSIMULATION ANALYSIS MODULE FOR AGENT-BASED MODELS

Table D.1 Perl Extension Modules Used

Wiley Series in Modeling and Simulation

Mission Statement

The Wiley Series in Modeling and Simulation provides an interdisciplinary and global approach to the numerous real-world applications of modeling and simulation (M&S) that are vital to business professionals, researchers, policymakers, program managers, and academics alike. Written by recognized international experts in the field, the books present the best practices in the applications of M&S as well as bridge the gap between innovative and scientifically sound approaches to solving real-world problems and the underlying technical language of M&S research. The series successfully expands the way readers view and approach problem solving in addition to the design, implementation, and evaluation of interventions to change behavior. Featuring broad coverage of theory, concepts, and approaches along with clear, intuitive, and insightful illustrations of the applications, the Series contains books within five main topical areas: Public and Population Health; Training and Education; Operations Research, Logistics, Supply Chains, and Transportation; Homeland Security, Emergency Management, and Risk Analysis; and Interoperability, Composability, and Formalism.

Advisory Editors • Public and Population Health

Peter S. Hovmand, Washington University in St. Louis

Bruce Y. Lee, University of Pittsburgh

Founding Series Editors

Joshua G. Behr, Old Dominion University

Rafael Diaz, Old Dominion University

Homeland Security, Emergency Management, and Risk Analysis

Forthcoming Titles

Zedda • Risk and Stability of Banking Systems

Interoperability, Composability, and Formalism

Operations Research, Logistics, Supply Chains, and Transportation

Public and Population Health

Arifin, Madey, and Collins • Spatial Agent-Based Simulation Modeling in Public Health:Design, Implementation, and Applications for Malaria Epidemiology

Forthcoming Titles

Hovmand • Modeling Social Determinants of Health

Kim and Hammon • Modeling and Simulation for Social Epidemiology and Public Health

Training and Education

Combs, Sokolowski, and Banks • The Digital Patient: Advancing Healthcare, Research, and Education

Forthcoming Titles

Tolk and Ören • The Profession of Modeling and Simulation

SPATIAL AGENT-BASED SIMULATION MODELING IN PUBLIC HEALTH

Design, Implementation, and Applications for Malaria Epidemiology

S. M. NIAZ ARIFIN

Department of Computer Science and EngineeringUniversity of Notre DameIN, USA

GREGORY R. MADEY

Department of Computer Science and EngineeringUniversity of Notre DameIN, USA

FRANK H. COLLINS

Department of Biological SciencesUniversity of Notre DameIN, USA

Copyright © 2016 by John Wiley & Sons, Inc. All rights reserved

Published by John Wiley & Sons, Inc., Hoboken, New Jersey

Published simultaneously in Canada

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, [phone], fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, [phone], fax [phone], or online at http://www.wiley.com/go/permission.

The contents of this work are intended to further general scientific research, understanding, and discussion only and are not intended and should not be relied upon as recommending or promoting a specific method, diagnosis, or treatment by health science practitioners for any particular patient. The publisher and the author make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of fitness for a particular purpose. In view of ongoing research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of medicines, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each medicine, equipment, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. Readers should consult with a specialist where appropriate. The fact that an organization or Website is referred to in this work as a citation and/or a potential source of further information does not mean that the author or the publisher endorses the information the organization or Website may provide or recommendations it may make. Further, readers should be aware that Internet Websites listed in this work may have changed or disappeared between when this work was written and when it is read. No warranty may be created or extended by any promotional statements for this work. Neither the publisher nor the author shall be liable for any damages arising herefrom.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at [phone], outside the United States at [phone] or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Arifin, S.M. Niaz, author.

Spatial agent-based simulation modeling in public health : design, implementation, and applications for malaria epidemiology / S.M. Niaz Arifin, Gregory R. Madey, Frank H. Collins.

p. ; cm.

Includes bibliographical references and index.

ISBN 978-1-118-96435-4 (hardback)

I. Madey, Gregory Richard, author. II. Collins, Frank H., author. III. Title.

[DNLM: 1. Malaria–epidemiology. 2. Computer Simulation. 3. Geographic Information Systems. 4. Models, Theoretical. 5. Spatial Analysis. WC 755.1]

RA644.M2

614.5′32090285–dc23

2015033121

To my parents:

Engineer S. M. Golam Mostofa

B.Sc. Engg. (Civil), FIE (B), PGD (CS)

My Father and Guide

Professor Parvin Akhter Jahan

M.A. (Economics), B.A. (Honors)

My Mother and Best Friend

and my wife:

Rumana Reaz Arifin

B.S., M.S.

My Soulmate

and my sister:

Mafruhatul Jannat

Ph.D., M.S., B.S.

We Grew up Together

- S. M. Niaz Arifin