Assessing the Use of Agent-Based Models for Tobacco Regulation (2015)

Index

[Page numbers followed by b, f, or t refer to boxed text, figures, or tables, respectively.]

A

ABMs. See Agent-based models

Add Health. See National Longitudinal Study of Adolescent Health

Addiction

biology of, 50–51

individual differences in risk of, 37

modeling of, 46, 122, 131, 153

progression of smoking behaviors to, 48

Agent-based models

advantages, 162–164

applications, 2–3, 20, 25, 161, 164–166, 199–200, 217–218

best practices in development of, 9, 98, 180–187

case examples, 101–105, 168–172, 199–202, 229–237

Center for Tobacco Products objectives for, xvi, 3–4, 25–26

communicating results from, 9–10, 90f, 92f, 99–100, 105, 106, 111, 113, 186–187

complex adaptive systems and, 195, 196, 198–199, 205

data needs, 15, 149–153, 154

defined, 2, 162, 195

definition of agent in, 2, 195

design of experiments for, 66, 67, 179t, 185

as deterministic nonlinear models, 208–209

drawing conclusions from, 186

error-checking in implementation of, 182–183

evaluation of, 9, 98. See also Evaluation framework for ABMs

goals for the Institute of Medicine review of, 1–2, 2b, 20b, 21, 27–28

grammar for, 89–90t

inputs, 66–68, 143, 173

investigating surprising results of, 185

microsimulations and, 6–7, 77–79

misperceptions about, 188

model testing, 92f, 97, 111. See also Validation issues in modeling

for policy design and evaluation, 27, 87–88, 166–168, 169, 207–209

potential agent types in, 152–153

purpose of, xvi, 25, 161, 162–163, 205

rationale for use of, 4–5, 52–55

recommendations for development of, 6, 8, 10, 72, 80, 99–100

sensitivity analysis in, 6, 8, 71, 83, 92f, 179t, 184–185, 227–228, 241

stakeholder involvement in development of, 155–156

testing and calibration, 92f, 97, 111, 184

validation issues, 123–124, 130–133, 205–207. See also Validation issues in modeling

See also Computational models; Design of ABMs; Evaluation framework for ABMs; Individual-level models; Microsimulation; Structural models

Agents, defined, 2, 195

Aggregate models, 4, 21, 26b, 52–53, 72, 76–77. See also Macro-level models; Population models

American Time Use Survey, 151

Analytical models, 73–75

At-risk populations, 37. See also Lesbian, gay, bisexual, and transgender persons; Race; Socioeconomic status

B

BENESCO model, 26b

Biological systems as agents in ABMs, 153

Blowback, 64. See also Policy resistance; Unintended consequences

Bounded confidence opinion dynamics model, 125, 127–128. See also Deffuant–Weisbuch model

C

Calibration, 136, 177, 179t, 182, 184, 205, 206, 223, 224–227, 228–229, 239–240. See also Validation issues in modeling

Campaign Tobacco Free Kids, 35–36

Cancer Intervention and Surveillance Modeling Network, xv, 26b, 155–156

CANSAVE, 26b

Center for Tobacco Products

data collection for ABM, 15, 154–155

goals for ABM development, xiv, 3–4, 25–26

mission, 3, 38

objectives for using ABM, 1–2, 2b, 19–20, 27–28

recommendations for, 6, 8, 10, 15, 16, 72, 80, 99–100, 139, 154, 155

use of modeling by, xv–xvi, 1–2, 20, 24

Centers for Disease Control and Prevention, 35–36, 43

Cessation of tobacco use

data sources, 144

determinants of, 50–51

identifying social interaction effects on, 70

individual differences in, 37

modeling efforts to date, 26b, 52–53

modeling of influences on individual behavior, 46–47

patterns of, 50

physiologic processes in, 50–51

rationale for agent-based models of, 54

SnapDragon model, 12, 14, 130

social context of, 51

CISNET. See Cancer Intervention and Surveillance Modeling Network

Cognitive–behavioral modeling, 46

Committee on the Assessment of Agent-Based Models to Inform Tobacco Product Regulation, xvi, 1–2, 2b, 19–20, 20b, 21–22

Communication

as ABM evaluation category, 9–10, 99, 111–112, 113

of ABM findings, 9–10, 90f, 92f, 99–100, 105, 106, 111, 113, 186–187

Compartmental models. See Aggregate models; Population models

Complex adaptive systems, 31–32, 213–214, 239

ABMs and, 195, 196, 198–199, 205

defined, 195, 213

epidemiologic example, 196–199

specification of agent behavior in, 196, 199

Computable general equilibrium, 219–220

Computational models, 1, 19, 25, 73–75.
See also Agent-based models

Conceptual frameworks, 46

for model evaluation, 92f, 112

for model formation, 95, 97, 105

in SnapDragon model, 124

for tobacco behavior, 46–47

Conclusions on

ABM and microsimulation, 7, 78–79

current data collection, 15, 154

human behavior inputs in ABMs, 6, 71–72

low- and high-dimensional modeling, 8, 79

quantifying uncertainty in models, 8, 83

SnapDragon model, 13–14, 137–139

Constraint interactions, 68

Contextual effects, 69, 81

Continuous Opinions and Discrete Actions, 128–129

Correlated effects, 69

CTP. See Center for Tobacco Products

Current Population Survey, 134, 144, 151

D

Data

existing sources and tools for tobacco, 143–149, 145–146b

to inform ABM, 7, 14–15, 53–54, 79, 80–82, 89, 94, 97, 101–104, 133–135, 143, 150–155, 170–171, 184, 200–201

recommendations for, 15, 154–155

shortcomings in, 148–149

SnapDragon use of, 123–124, 133–137

strategies for improving, 14–15, 149–153, 154

types of, for tobacco research and modeling, 143

Data needs for tobacco control ABM, 149–153

Deeming of tobacco products, 24, 35, 42

Deffuant–Weisbuch model, 127–128. See also Bounded confidence opinion dynamics model

Descriptive output validation, 205

Design document, 88, 95

Design of ABMs

agent perspective in, 180

assumptions in, 181–182, 188

best practices, 173, 179–187

case examples, 100–105

computer coding, 182–183

conceptual stage, 93, 110, 179–180

documentation, 184

environmental specifications, 177–178

evaluation of, 87–88. See also Evaluation framework for ABMs

factors to be considered in modeling individual behavior in, 45–48

key building blocks, 173–178

key steps in, 178–179t

level of detail, 7–8, 67–68, 78–79, 153, 183, 210–211, 214, 230, 237, 239–241

parameter estimation, 67–68, 94, 102, 132, 136, 138, 206–207, 208, 210–211

for policy development, 27, 54–55, 68, 87, 166–168, 187–188

recommendations for expertise in, 72–73, 155

repurposing, 188

role of model consumers in, 8, 54–55, 75, 83, 88, 95, 103, 155, 187–188

specifying agent characteristics in, 173–175, 195–196

specifying individual behavior in, 80–82, 175–176

time specifications, 176–177, 208

transparency in, 94, 186

understanding of human behavior for, 5–6, 71–72, 83

Discrete choice models, 80–82

Drug Policy Modelling Program, 105, 155–156

Dynamic Stochastic General Equilibrium modeling, 206–207

E

E-cigarettes, 23, 24, 35, 36, 42. See also Deeming of tobacco products; Smokeless tobacco products

Ecological momentary sampling and assessment, 151, 170

Ecological perspective, 47–48, 47f

Economic implications of tobacco use, 23, 24, 34f, 45

Economic models of tobacco use behaviors, 47

Emergent behaviors, 130, 135, 196, 200, 206, 209

Energy Information Administration, 220, 224–225

Energy modeling

assessment, 224

calibration and uncertainty, 223, 224–228, 238–239

complexity and validity, 228–229

current, non-agent-based, 220–224

evolution of, 218, 243–244

examples of, 229–237

lessons from, for ABM design, 239–240, 242

parameter values, 223–224

terminology, 219–220

validation, 237–238

Energy Modeling Forum, 155–156

Equilibrium modeling, 73–76, 219–220

multiple equilibria, 74

in SnapDragon, 13–14, 132–133

Ethnographic research, 82, 103–104, 151, 200

European Smoking Prevention Framework, 147

Evaluation framework for ABMs, 28

activities, 93–95, 110–112, 124

case examples, 100–105

categories, 9–10, 98–99

conceptual basis, 91

descriptive grammar, 88–90, 89–90t, 106–108t

environmental factors, 96–97

evaluation questions, 91, 108–114

high-priority questions, 97–98, 108–114

implementation review, 110–111

logic model, 91, 92f

model suitability, 9–10, 94, 98–99

outcomes, 96, 113–114

outputs specification, 95, 112–113

policy outcomes, 10, 99, 112–113

purpose, 9, 87–88, 90, 97–98

recommendation for, 10, 99–100

resources, 91–93, 109–110

SnapDragon review, 124–137

structure, 91, 92f

technical best practices, 9–10, 98–99

Expectations interactions, 68–69

Expertise, 15–16, 72–73, 93, 109, 155. See also Interdisciplinary modeling team; Modeling, generally; Subject-matter experts

F

Family Smoking Prevention and Tobacco Control Act, 3, 24, 32–35, 38, 42, 44, 148

main provisions, 39b

Federal agencies in tobacco regulation, 43–44. See also U.S. Food and Drug Administration

Feedback mechanisms

aggregate modeling, 77

complexity of, in tobacco landscape, 32, 34f

individual–social behavior, 5–6, 68–73, 74, 77

policy effects, 64

in SnapDragon model, 12, 13, 130–131, 138

Findings on

SnapDragon model, 13–14, 137–139

value of ABMs for tobacco control policy, 5, 54

Food, Drug, and Cosmetic Act, 35, 42

Fundamental evaluation categories, 9, 98

Funding as ABM resource, 93

G

Gas models, 195, 196–197

Genetic factors in tobacco use, 47

Geographic context, 163

Grammar, ABM, 88–90, 89–90t, 106–108t

Graphic warning labels, 32–35. See also Warning labels

H

Health belief model, 46

Health departments as agents in ABMs, 152

Health risks, smoking-associated, 3, 23, 31, 50–51, 129

Heuristics, 82, 143, 176

High-dimensional modeling, 7–8, 79–80

Homophily, 70, 101, 107

Human resources, 15–16, 72–73, 93, 109, 155. See also Interdisciplinary modeling team; Subject-matter experts

I

Identification, 69–70, 207–209

Incentives for behavior, 5–6

Indian Country, 45

Indian Health Service, 43

Indirect policy models, 168, 172

Individual behavior

current conceptualizations of, in tobacco use, 45–51

data needs for modeling of tobacco use, 14–15, 149–153

importance of, in modeling tobacco use and control, 4–5, 52–54

representations in ABMs, 5–6, 8, 71–72, 83, 196

strategies for specifying, in ABMs, 80–82

See also Social context of tobacco use

Individual-level data, 133, 134, 143

Individual-level models, 7, 63, 77–78. See also Agent-based models; Micro-level models; Microsimulation

Infectious disease modeling, 169–170, 196–199, 201–202, 205, 209–210

Initiation of tobacco use

age at initiation, 132, 144

data sources, 144

determinants of, 48–50

e-cigarettes, 23

individual differences in, 51

modeling efforts to date, 26b, 52–53

motivation of individuals for, 37

patterns and trends, 3, 48, 132

progression to regular use, 48

rationale for agent-based models of, 52–54

Input validation, 205, 206

Interdisciplinary modeling team

information needs of, 93

knowledge broker, 95, 105

policy development and, 99, 106

resources for, 9, 98

role of, in model development, 17, 97, 157

SimAmph, 103–105

Ising spin models, 125

K

Knowledge broker, 95, 105

L

Legacy, 36

Legacy Tobacco Documents Library, 149

Lesbian, gay, bisexual, and transgender persons, 37, 41

Logic models, 91–92

Low-dimensional modeling, 7–8, 79–80

M

Macro-level models, 66. See also Aggregate models; Population models

Marketing. See Tobacco marketing

Marschak’s stability requirement, 204

Mental illness, smoking and, 37

Micro-level models, 63–64. See also Agent-based models; Individual-level models; Microsimulation

Microsimulation, 6–7, 26b, 66, 77–79

MIDAS. See Models of Infectious Disease Agent Study

Misperceptions about ABMs, 188

Model suitability, 9–10, 98–99

in SnapDragon evaluation, 129–133

Modeling, generally

applications of, xv, 1, 27

comparative features, 63–64

current and past approaches to in tobacco control, xv–xvi, 26b, 52–53

expertise for, 15–16, 72–73, 155

individual versus aggregate level specification, 76–77

limitations of, 8, 82

quantifying uncertainty in, 8, 83

tobacco control, xv, 25, 26b

Models of Infectious Disease Agent Study, 155–156, 165, 169–170, 187

Morbidity, tobacco use, 3, 23

Mortality, tobacco use, 3, 23

Multiple equilibria, 74–75

N

National Energy Modeling System, 220–221, 222f

National Longitudinal Study of Adolescent Health (Add Health), 81, 144, 147, 149

National Youth Tobacco Survey, 144, 145, 148

New tobacco products, regulation of, 40, 42. See also E-cigarettes

Nicotine content, 35, 42

Nicotine pharmacology, 50–51

O

Obesity, models of, 3, 25, 166, 172

Online media, 151–152

Opinion dynamics, 11, 12, 14, 121–122, 124–129, 130–133, 138. See also Social Network Analysis for Policy on Directed Graph Networks (SnapDragon)

P

Packaging, the U.S. Food and Drug Administration authority for, 40–41. See also Warning labels

Parameters in models, 1, 67–68, 74, 94, 178, 179t

Parental influences, 49, 147

PARTE framework, 94, 173–178, 174f

PATH. See Population Assessment of Tobacco and Health

Peer influences, 49, 69–70, 81–82 data sources, 147, 149

Peer review of ABM, 10, 93, 100, 112

Physical space in ABM, 88, 89t

Pipe tobacco, 42

Plug-in electric hybrid vehicles, 101–103, 106–108t

Policy design and implementation

ABM design for, 54–55, 79–80, 87, 166–168, 187–188

blowback, 64. See also Policy resistance case example of ABM development for, 101–103, 168–172

challenge of anticipating effects of, 64–65

challenges using ABMs for, 207–209

choice of model for, 211–212

computational modeling for, 1, 19, 25, 53, 73–75

data needs, 14–15

data on policy interaction effects, 148–149

evaluating effects of, 65–66

evaluation of ABM effectiveness in, 1–2, 3–4, 10, 19, 21, 99, 112–113

historical use of tobacco modeling for, 26b

identifying social interaction effects for, 68–71

indirect policy models, 168, 172

modeling rationale, 27, 66–68

prospective models, 167–168, 169–171, 181

recommendations for model development to support, 16, 56

retrospective models, 168, 171, 181

structural modeling for, 73–79, 204–205

use of ABMs for, 1, 2–3, 4, 16–17, 20, 53–54, 68, 77–78

use of multiple models and methods in, 187

Policy realism in ABM, 90t

Policy resistance, 64, 164

Policy testing, 92f, 94–95, 111

Population Assessment of Tobacco and Health, 147–148

Population models, xvi, 4, 19, 25, 26f, 52.
See also Aggregate models

Predictive output validation, 205–206

Preference interactions, 69

Prevalence of tobacco use

among specific populations, 37, 49

estimates by young people of, 49

modeling efforts to date, 26b

price sensitivity, 76

Prospective policy model, 25, 167–168, 169–171, 181

Public education, the U.S. Food and Drug Administration authority for, 41

Public health, ABM use in, 165–166, 169–171, 172

Public health standard for review of tobacco products, 24

Q

Qualitative data, 82, 133–134, 143, 151

Quantitative aggregated data, 80–82, 133, 134–135, 143, 152

R

Race, 37, 51, 134, 148, 152, 162

Randomized controlled trials, 65–66

“Real Cost, The,” 41

Recommendations on

ABM development, 6, 72

Center for Tobacco Products model development, 16, 155

data collection, 15, 154

high- and low-dimensional modeling, 8, 80

model evaluation, 10, 99–100

SnapDragon model, 14, 139

use and development of models, 16, 156

use of findings from models, 8, 83

Reinforcement of smoking behaviors, 51

Relapse of tobacco use, 5, 50–51, 53, 132

Resources

for development and use of models, 15–16, 72–73, 155

as evaluation criteria for ABMs, 9, 91–93, 98, 109–110

Retrospective policy models, 168, 171, 181

Return on investment, 87, 113

S

Sales, tobacco. See Tobacco distribution and sales

SAMMEC model, 26b

Sandia National Laboratories, xvi, 3–4, 25–27, 119, 120–121

Schelling segregation model, 72, 200

Secondhand smoke, 23, 144

Sensitivity analysis, 6, 8, 71, 82–83, 92f, 179t, 184–185, 227–228, 241

SIENA, 147

SimAmph, 103–105, 106–108t

SimSmoke, xv, 26b

SIS process, 196–199, 203, 209–210

Smallpox epidemic, modeling of, 169–170, 201–202, 205

Smokeless tobacco products, 3, 23, 24, 38, 40–41, 44–45. See also Deeming of tobacco products; E-cigarettes

SnapDragon. See Social Network Analysis for Policy on Directed Graph Networks

Social cognitive theory, 46

Social contagion, 70, 131

Social context of tobacco use

cessation and, 51

data sources, 144–147, 149–150

ecological perspective, 47–48, 47f

endogenous effects, 69–70, 81

rationale for ABMs to study, 4–5, 53–54, 164

significance of, among youth, 49–50

tobacco use landscape, 32–38

Social dynamics, 81, 89t, 126, 162

Social interactions, modeling of, 5, 52–54, 68–70, 81–82

Social media, 151–152

Social multipliers, 64

Social Network Analysis for Policy on Directed Graph Networks (SnapDragon)

calibration and verification, 136–137

conclusions and findings from review of, 12–14, 137–139

current state of, xvi, 10–11, 120–121

data use in, 133–137

description of, 27, 121, 124

descriptive grammar applied to, 106–108t

development, 120–121

health risk modeling in, 129–130

imitation, smoking adoption, 12–13, 131

lack of feedback mechanism from behavior to opinion in, 12, 13, 130–131, 138

media influence in, 120, 123, 128

opinion dynamics, 121–122, 124–129, 130–133, 138

parameters, 13, 136–138

product switching in, 122–123, 136

purpose of, 4, 27, 120, 121

recommendations on use of, 14, 139

structure and dynamics, 11–12, 121–124

suitability evaluation, 129–133

time path to equilibrium in, 132–133

use of social networks in, 11, 121, 123, 127–128

user classification and status in, 121, 129

Social psychology, 125

Social space in ABM, 89t

Socioeconomic status, 37, 51, 134, 152, 162

Sociophysics, 125

State and local governments

as agents in ABMs, 152

regulatory authority of, 42, 44–45

revenue from tobacco for, 36–37

Structural models, 5–6, 66–68, 72, 203–205, 207–208

defined, 5, 66–67, 73

types of, 73–78

Stylized facts, 133–134, 135, 143, 151

Subject-matter experts, xvii, 6, 17, 72, 92f, 93, 155, 186

recommendations for, 6, 72, 155

Substance abuse, 103–105, 200–201, 209

System dynamics models, 4, 5, 21, 52, 66, 87

T

Taxes, tobacco, 36–37, 44, 45, 64, 71, 78

Time-use data, 151

Tobacco Centers of Regulatory Science, 15, 155

Tobacco control

agents and relationships in, 33f, 34f, 35–37, 43–44

challenges for the U.S. Food and Drug Administration, 34–36

challenges to anticipating policy effects, 64–66

complexity of, 3, 31–38

federal agency authority of, 38–44

modeling efforts to date, 1, 25, 26b, 52–53

rationale for ABMs in study of, 1–3, 4, 52–55

SnapDragon modeling of, 122–124

state and local authority, 42, 44–45

Tribal authority, 42, 45

the U.S. Food and Drug Administration authority for, 3, 19, 23–24, 38–42

Tobacco Control Act. See Family Smoking Prevention and Tobacco Control Act

Tobacco distribution and sales

data sources, 148

minimum purchase age requirements, xv, 41–42, 44

regulatory authority, 3, 41, 44

retailer density, 35, 44, 170–171

See also Tobacco industry

Tobacco industry

as agents in ABMs, 53, 152

challenges to tobacco regulation from, 35, 36

data collection by, 149

marketing, 3, 40–41

modeling policy effect on behavior of, 71

See also Tobacco distribution and sales;

Tobacco manufacture

Tobacco manufacture

nicotine content, 35

regulatory authority, 3, 39–40

See also Tobacco industry

Tobacco marketing, 3, 40–41

Tobacco Prevalence and Health Effects Model, 26b

Tobacco Products Scientific Advisory Committee, 35, 38

Tobacco-related disease and death, 3, 23, 31. See also Health risks, smoking-associated

Tobacco use

characteristics of user population, 37

data sources, 14, 144–149

data types for research and modeling of, 133, 143

economic burden, 23

factors to be considered for modeling individual behavior, 45–51

health risks, 3, 23, 31, 50–51, 129

patterns and trends, 3, 22–23, 31, 37

progression of, 48–49

rationale for ABMs in study of, 4–5, 52–55

strategies for improving data collection on, 14–15, 149–154

as wicked problem, 32–37

See also Cessation of tobacco use; Initiation of tobacco use; Prevalence of tobacco use; Social context of tobacco use

Tobacco Use Supplement to Current Population Survey, 144

Tolerance, nicotine, 50

TPSAC. See Tobacco Products Scientific Advisory Committee

Transparency, 186

U

Uncertainty

communication of, 8, 27, 82–83, 186–187

fundamental model, 238–239

managing and communicating, 186–187

quantification of, 8, 82–83, 94, 228, 238

recommendation for consideration of, 8, 83

Unintended consequences, 37, 64, 167, 171. See also Blowback; Policy resistance

U.S. Department of Health and Human Services, 43

U.S. Environmental Protection Agency, 43–44, 221

U.S. Food and Drug Administration challenges for, in tobacco regulation, 34–36

limits to authority of, 42

recommendations for, 8, 16, 83, 156

tobacco regulation authority, 3, 19, 23–24, 38–42

use of models by, xv–xvi, 1, 14–16, 19–20, 154–156

V

Validation issues in modeling, 123–124, 130–133, 205–207, 219, 228–229

in model evaluation, 92f, 94, 97

in SnapDragon model, 126, 135, 136

Verification, 94, 238

in case studies, 104

in SnapDragon Model, 136

W

Warning labels, 32–35, 40–41

Wicked problems, 32, 37

Withdrawal, nicotine, 50–51

Y

Youth

determinants of tobacco initiation in, 48–50

e-cigarette use, 23

parental influences, 49, 147

progression of smoking behaviors, 48–49

projected mortality among tobacco users, 3, 23

restrictions on advertising to, 40

social context of tobacco use among, 49–50

tobacco initiation rates, 3, 22–23

tobacco prevention campaigns, 41

tobacco use data sources, 144–147

This page intentionally left blank.