PaperlayerView exact revision on arXiv

Source-linked AI summary

Statistical physics of vaccination

Zhen Wang, Chris T. Bauch, Samit Bhattacharyya, Alberto d'Onofrio, Piero Manfredi, Matjaz Perc, Nicola Perra, Marcel Salathé, Dawei Zhao

arXiv:1608.09010v3physics.soc-phcond-mat.stat-mechcs.SIq-bio.PEstat.AP

TL;DR

The report addresses how mathematical epidemiology can explain and help control infectious diseases, particularly vaccination-related dynamics shaped by behavior and heterogeneous contacts. It reviews models from mean-field compartmental systems through networks, spatial structure, behavioral feedback, and digital epidemiology, using concepts from statistical physics. The review concludes that behavior-coupled and network-based models capture dynamics unavailable to simpler models, while digital data offer new high-resolution inputs but methods such as social-media forecasting retain important limitations.

  • Problem

    Infectious diseases remain difficult to eliminate because transmission is shaped not only by vaccination and disease processes but also by human behavior and heterogeneous social structure.

  • Method

    The report reviews mathematical models of disease, vaccination, behavior, networks, spatial structure, statistical-physics methods, and digital epidemiological data.

  • Results

    Coupled behavior-disease models generate emergent dynamics absent from models that ignore vaccination behavior, while network models reveal effects such as difficulty of eradication.

  • Takeaways & Limitations

    High-resolution digital sources can support network-based behavior-disease models by informing individual connections and population movement.

  • Takeaways & Limitations

    Social-media epidemiological methods have substantial limitations, including major forecasting errors and the discontinuation of Google Flu Trends.

Abstract

from arXiv · show

Historically, infectious diseases caused considerable damage to human societies, and they continue to do so today. To help reduce their impact, mathematical models of disease transmission have been studied to help understand disease dynamics and inform prevention strategies. Vaccination - one of the most important preventive measures of modern times - is of great interest both theoretically and empirically. And in contrast to traditional approaches, recent research increasingly explores the pivotal implications of individual behavior and heterogeneous contact patterns in populations. Our report reviews the developmental arc of theoretical epidemiology with emphasis on vaccination, as it led from classical models assuming homogeneously mixing (mean-field) populations and ignoring human behavior, to recent models that account for behavioral feedback and/or population spatial/social structure. Many of the methods used originated in statistical physics, such as lattice and network models, and their associated analytical frameworks. Similarly, the feedback loop between vaccinating behavior and disease propagation forms a coupled nonlinear system with analogs in physics. We also review the new paradigm of digital epidemiology, wherein sources of digital data such as online social media are mined for high-resolution information on epidemiologically relevant individual behavior. Armed with the tools and concepts of statistical physics, and further assisted by new sources of digital data, models that capture nonlinear interactions between behavior and disease dynamics offer a novel way of modeling real-world phenomena, and can help improve health outcomes. We conclude the review by discussing open problems in the field and promising directions for future research.

1. Introduction

Infectious diseases remain a major threat despite advances in prevention, while human behavior can impede elimination. This report reviews mathematical models that connect disease dynamics, behavior, physics-based methods, and emerging digital data.

  • Infectious diseases continue to burden societies, including through persistent vaccine-preventable diseases and newly emerging pathogens.
  • Mathematical and statistical models help determine control measures and understand or predict disease outbreak patterns.
  • Human behavior can prevent local elimination or global eradication when vaccine scares reduce confidence and enable disease resurgence.
  • Coupled behavior-disease models examine the two-way interplay between human behavior and infectious-disease dynamics.
  • The report surveys behavior-disease models, their connections to physics, and future open questions and research directions.
  • The review progresses from classical mean-field models to network, spatial, behavior-coupled, and digital-data approaches.

2. Basic concepts in infectious disease epidemiology

Infectious-disease epidemiology studies transmission, pathogen characteristics, infection time periods, population immunity, and interventions. Vaccination provides direct individual protection and indirect population-level protection, while behavior-linked measures can also alter transmission.

  • Infectious-disease epidemiology studies contagious parasites and the biology and ecology underlying their transmission and control.
  • Transmission routes are classified as direct or indirect, depending on how pathogens move between infected and susceptible hosts.
  • The latent, incubation, and infectious periods distinguish infection, disease onset, and the ability to transmit infection.
  • Infectivity, virulence, and pathogenicity are pathogen characteristics that affect infectious-disease transmission and outcomes.
  • Herd immunity provides indirect protection because population immunity influences epidemic potential and susceptible individuals’ infection risk.
  • Vaccines produce protective immune responses, combining direct effects on vaccinated individuals with indirect effects that reduce population transmission.
  • Social distancing can reduce epidemic attack rates and may help explain multiple outbreaks or infection waves.

3. Basic concepts and methods in epidemiological modeling and vaccination: compartmental (mean-field) models

Compartmental mean-field models represent infection as transitions among epidemiological classes and use mass-action transmission to describe population-level dynamics. Their reproduction-number framework supports outbreak thresholds and vaccination coverage requirements.

  • Classical SIR models established influential deterministic foundations for modeling outbreaks and endemic infections.
  • McKendrick’s statistical-physics-inspired formulation modeled susceptible and infective individuals as particles whose encounters generate infection.
  • Compartmental models assign dynamic variables to epidemiological classes and describe transitions such as infection incidence from S to E.
  • The mass-action formulation embeds social contact behavior and biological contagion within the transmission-rate parameter β.
  • The basic reproduction number R0 determines whether infection invades a large homogeneous population and governs initial generational growth.
  • For measles with R0 about 15, the critical vaccine coverage is xc = 93.3%, requiring effective vaccination of nearly all newborns.

3.4. SIR and SEIR models for epidemic outbreaks

SIR and SEIR outbreak models relate epidemic thresholds, prevalence, and final attack rates to R0 and susceptible depletion. They explain initial growth, peak prevalence, incomplete contagion, and the limits of basic models for long-term endemicity.

  • R0 is the key threshold parameter: outbreaks require an initial effective reproduction number above 1.
  • Outbreak prevalence initially grows exponentially, then decelerates nonlinearly as susceptible depletion lowers the effective reproduction number.
  • For a wholly susceptible population, maximum prevalence increases with R0, reaching approximately 45% for Spanish flu and 80% for measles-like R0 ≈15.
  • The final attack rate increases concavely with R0, while infections that confer immunity do not infect the entire population.
  • Vaccination and transmission reduction can be incorporated to lower attack rates or slow outbreak progression, including through time-varying β(t).
  • In SEIR models, infection dies out when R0 < 1/S(0), whereas R0 > 1/S(0) produces an initial phase of exponential growth.

3.5. Emergence and features of endemicity: the SIR and SEIR models with vital dynamics.

Vital dynamics prevent infection extinction by rebuilding the susceptible population, producing endemic equilibria and recurrent epidemics. In SEIR models, periodic contact patterns can affect disease-free stability according to the full shape of transmission.

  • SIR model with vital dynamics: SIR and SEIR models without vital dynamics predict infection extinction because the susceptible population is depleted.Vital dynamics are therefore needed to sustain long-term endemicity and recurring epidemic patterns.
  • SIR model with vital dynamics: For R0 ≤ 1, the disease-free equilibrium is globally attractive; for R0 > 1, an endemic equilibrium appears and is globally attractive.Above threshold, infective prevalence increases strictly with R0.
  • SIR model with vital dynamics: For measles-like parameters R0 = 15 and life expectancy 75 years, the model predicts a natural period of approximately 2.01 years.This reproduces the observed biennial inter-epidemic period in UK and US cities before vaccination.
  • SIR model with vital dynamics: Births rebuild the susceptible pool after outbreaks, allowing infections to persist and generating progressively recurring epidemics.When γ >> µ, an initial outbreak rapidly depletes susceptibles, after which demographic replenishment supports later epidemics.
  • Periodic contact rates: With periodic contact rates, SIR disease-free stability depends on the average reproduction number, whereas SEIR stability depends on the whole shape of β(t).For SEIR systems, annual periodic forcing can also produce nonlinear resonances under childhood-infection parameter conditions.
  • Nonlinear contact rates and behavior: Mass-action models constrain behavior by treating transmission as constant or driven by exogenous periodic factors, although prevalence-dependent behavior can destabilize endemic dynamics.Delayed behavioral responses to prevalence can trigger sustained oscillations.

3.6. Mass vaccination and herd immunity for vaccine preventable infections: the SIR model with vaccination at birth

The vaccinated SIR model represents perfect vaccination at birth and replaces the basic reproduction threshold with a vaccination-adjusted threshold. Coverage at or above xc = 1 − 1/R0 eliminates infection, while insufficient coverage lowers prevalence and lengthens inter-epidemic periods.

  • Model assumptions: The model vaccinates a fixed proportion x of newborn susceptible individuals using a perfect vaccine with lifelong protection.Under this assumption, vaccine coverage equals the successfully vaccinated fraction.
  • Threshold dynamics: Vaccination changes the threshold from the basic reproduction number to a vaccination reproduction number, with a disease-free equilibrium at (1 − x, 0).The disease-free equilibrium is globally attractive when the vaccination reproduction number is at most one.
  • Threshold dynamics: The critical coverage is xc = 1 − 1/R0; coverage x ≥ xc eliminates infection by leaving only a globally asymptotically stable disease-free equilibrium.Below this threshold, the endemic equilibrium persists.
  • Post-vaccination dynamics: Insufficient vaccination produces lower prevalence and a longer inter-epidemic period than the unvaccinated regime.The longer period follows from reduced replenishment of the susceptible class through vaccination.
  • Post-vaccination dynamics: For vaccination begun at t = 20 years, coverage x = 0.95 eliminates the infection, whereas x = 0.80 and x = 0.90 leave it endemic with much lower prevalence.The remaining endemic cases also show a marked increase in the inter-epidemic period.

3.7. Return to susceptibility: SIS and SIRS models

SIS and SIRS models sustain endemicity through loss of infection-acquired immunity, while stochastic formulations reveal extinction risks that deterministic mean-field models can miss. The review therefore connects immunity loss, imperfect vaccination, and population-size effects to epidemic persistence.

  • Return to susceptibility: Loss of infection-acquired immunity rebuilds susceptibility, allowing endemicity even without vital dynamics.In SIS models, recovered individuals return directly to susceptibility; SIRS models insert a recovered-state sojourn.
  • SIS and SIRS dynamics: Under periodic transmission, extinction versus endemicity depends on whether the average reproduction number is below or above one.With constant transmission, R0 > 1 permits a globally attractive endemic state in both SIS and SIRS settings.
  • Vaccination with return to susceptibility: Imperfect vaccines can leave recipients fully susceptible, provide temporary immunity, or reduce susceptibility, making elimination more difficult.For some SIS vaccination models, bistability means elimination requires coverage above a higher threshold x*.
  • Stochastic epidemic models: Mean-field models neglect correlations by approximating averages of products with products of averages, so deterministic equations are approximations to stochastic epidemic processes.The deterministic model emerges in the limit N → +∞.
  • Stochastic epidemic models: Small populations require birth-death stochastic models, while system-size approximations connect master equations to Fokker–Planck and stochastic differential equations.These approximations replace integer-valued state vectors with real-valued approximations and noise terms.
  • Stochastic extinction: During inter-epidemic periods, stochastic extinction can occur at low prevalence even when deterministic periodically forced models remain positive.Vaccination increases this relevance by further reducing prevalence.
  • Stochastic extinction: The critical community size marks a heuristic population threshold above which infection is likely to persist and below which extinction is likely.Prevaccination measles data place this threshold near 250,000–400,000 people.

3.9. Space and beyond: mean-field metapopulations

Metapopulation models extend epidemic dynamics beyond homogeneous mixing by representing populations as connected patches, while revealing how mobility, vaccination, and contact topology shape spatial spread.

  • Spatial dynamics: Spatial epidemic models must address human movement that differs from Fickian diffusion, including rapid commuting between highly clustered cities.Adding spatial diffusion to SIR or SEIR models yields reaction-diffusion analogies but can require complex integro-differential formulations.
  • Metapopulation framework: Metapopulation models divide populations into patches with local SIR-type dynamics linked by population fluxes such as commuting.This framework models geographical spread while retaining deterministic epidemic dynamics within each patch.
  • Metapopulation framework: Patch-coupled systems encode compartment-specific mobility through a diagonal mobility matrix and inter-patch transmission through a connection matrix.The state variables describe epidemiological compartments in each patch, while matrix structure captures heterogeneous mobility and cross-patch transmission.
  • Spatial dynamics: Linearization around spatially homogeneous solutions decomposes the dynamics into independent systems indexed by the connection matrix eigenvalues, paralleling Turing-bifurcation analysis.The same methodology can also be applied when population fluxes are nonlinear.
  • Vaccination and spatial structure: Vaccination can alter spatio-temporal dynamics in multi-patch systems, with metapopulation models explaining observed time-series decorrelation among major UK cities after measles vaccination.The cited example connects vaccination introduction with changes in correlations across cities.
  • Beyond geographical patches: Increasing movement and outbreak data enable larger-scale modeling of inter-patch contact topology, while multi-group models generalize patches to social or age groups.Age-structured models are used to study nonlinear vaccination effects, including changes in average age at infection; continuous infection or vaccination age can produce integro-differential models.

4. Basic concepts and methods in (non-behavioral) epidemiological modeling and vaccination: network models

Network models extend epidemic analysis beyond homogeneous mixing by representing heterogeneous contacts, enabling analytical study of spreading thresholds and vaccination strategies. The review covers standard network models, percolation methods, multilayer structures, and the limitations of network-based simulations.

  • Network foundations: Random-graph, small-world, and evolving scale-free models provide idealized frameworks for representing distinct network structures and dynamics.The Erdős–Rényi model uses probabilistic links, the Watts–Strogatz model combines short paths with clustering, and the Barabási–Albert model combines growth with preferential attachment.
  • Network foundations: Network theory captures heterogeneous contact structures that classical SIS and SIR models omit under the homogeneous-mixing assumption.The network topology can inform transmission routes and control strategies such as contact tracing.
  • Percolation and spreading: Percolation theory uses generating functions to identify the critical transmissibility Tc at which outbreaks expand from finite clusters to a giant component.For T > Tc, the network contains an epidemic-scale giant component.
  • Percolation and spreading: For degree distributions P(k) ∼ k^-ζ with ζ < 3, the percolation threshold tends to zero as network size approaches infinity.Under these conditions, epidemics can spread for arbitrarily small infection probabilities and rapid recovery.
  • Simulation limitations: Network-based simulations are sensitive to whether the sampled structure represents relevant community variation and rare epidemiologically important contacts.Idealized networks and analytical tools can help identify which structural elements determine epidemic dynamics.
  • Vaccination over networks: In scale-free networks with diverging second degree moment, complete vaccination is required for eradication.The reported condition is xc = 1 when <k^2> → ∞.
  • Vaccination over networks: Vaccination requirements depend on spreading rate and epidemic threshold, while targeted vaccination is especially effective in highly heterogeneous scale-free networks.A small fraction of highly connected nodes can substantially affect infection spread, but targeted strategies are less efficient in networks with limited heterogeneity.

5. Non-behavioral epidemiological vaccination on networks

Network vaccination models show that epidemic control depends strongly on contact-network heterogeneity and on how vaccination targets nodes. Strategies using degree, centrality, acquaintance, random-walk, or community information can outperform uniform vaccination, especially in heterogeneous networks.

  • Network vaccination immunizes selected nodes with a perfect vaccine so they can no longer transmit disease to neighbors.The effective spreading rate is βI/γ.
  • In homogeneous networks, vaccination above the critical threshold xc prevents large epidemic outbreaks.
  • For scale-free networks with degree exponent ζ < 3, xc approaches 1 under both SIS and SIR dynamics, making uniform vaccination largely inefficient.The divergence of ⟨k2⟩ indicates that complete network vaccination is required in the model.
  • Vaccinating high-degree nodes raises the epidemic threshold, and degree-based targeting can yield an exponentially small critical vaccination threshold.
  • Simulations find that targeted vaccination sharply reduces SIS prevalence near its threshold, whereas random vaccination declines slowly and vanishes only as x approaches 1.
  • Centrality-based strategies exploit network position: eigenvector, betweenness, random-walk, and closeness measures identify nodes whose immunization can suppress or delay spread.Random-walk centrality can produce the fewest infected cases at low coverage, while closeness targeting postpones epidemic spreading.
  • Acquaintance vaccination is more efficient than random vaccination across both high and low degree-exponent values.
  • When only local network information is available, D-steps vaccination offers a strong efficiency–flexibility compromise, while community bridge finding can outperform hub-targeting strategies.

6. Rationale for studying behavior-disease dynamics

Although vaccine access has improved and elimination has become more feasible, human behavior increasingly constrains disease control. The review therefore motivates coupled behavior–disease models that use social context and increasingly detailed behavioral data to understand and predict these dynamics.

  • Coupled behavior–disease models are studied to represent the two-way interplay between behavior and infection dynamics and potentially improve explanation and prediction.The report surveys this literature and connects it to physics-inspired concepts and methods.
  • Vaccines and large-scale immunization campaigns have substantially reduced infectious-disease mortality, making global measles eradication increasingly conceivable.Measles deaths were estimated to fall from 766,000 annually in 2000 to 164,000 in 2008.
  • As access barriers recede, vaccine refusal and hesitancy can prevent local elimination or global eradication and enable disease resurgence.The review cites measles–mumps–rubella and oral polio vaccine scares as examples.
  • Observed increases in vaccine coverage during and after outbreaks demonstrate that vaccinating choices respond to disease prevalence.This coupling is especially evident near the elimination threshold.
  • Social context influences vaccination choices through peer groups, medical professionals, and social contact networks.The Disneyland, California measles outbreak illustrates the role of social factors in decision-making.
  • Herd immunity can foster complacency, so vaccine hesitancy and refusal may become more common as elimination thresholds are approached.
  • Models incorporating behavior can explain prevalence time series more effectively than models neglecting behavior, with little or no parsimony penalty, and have shown retrospective predictive power.

7. Basic concepts in behavioral modeling

Behavioral models range from phenomenological descriptions to mechanistic game-theoretical and psychological accounts. Their trade-off is between simple representations that capture observed effects and richer models that specify decision mechanisms and social processes.

  • Behavioral models are grouped into phenomenological, game-theoretical, and psychological approaches.
  • Phenomenological models: Phenomenological models describe observed behavioral effects on transmission without specifying the psychological mechanisms producing them.
  • Phenomenological models: A phenomenological transmission function may saturate at high prevalence when individuals reduce contact through measures such as hand-washing.
  • Phenomenological models: The contagion metaphor models the spread of ideas or behaviors through social networks but can overlook subtleties such as descriptive and injunctive social norms.
  • Phenomenological models: Phenomenological models can test whether behavior influences disease spread and can clarify how simple information transmission produces complex population-level outcomes.Mechanistic models are preferable when specific behavioral mechanisms are needed.
  • Game theoretical models: The Prisoner’s Dilemma frames vaccination as a conflict between individual optimization and group welfare, with non-vaccinators potentially benefiting from herd immunity.
  • Game theoretical models: Game-theoretical vaccination models define vaccinator and non-vaccinator strategies, assign payoffs, and analyze Nash equilibria or evolutionarily stable states.
  • Psychological models: Psychological models formulate vaccination decisions using mechanisms such as social learning, subjective expected utility, and prospect theory.The evolutionary-game model includes parameters for social learning and perceived disease risk relative to vaccine risk.

8. Behavior-vaccination dynamics in well-mixed (mean-field) populations

Voluntary vaccination couples disease dynamics to information, memory, imitation, and social norms, producing complex behavior-disease feedback. These models show that delayed or prevalence-dependent responses can generate oscillations, bistability, and persistent endemic infection despite high temporary uptake.

  • Voluntary vaccination decisions depend on perceived disease risks, disease spread, and vaccine side-effect risks, making their dynamics difficult to represent quantitatively.
  • Prevalence-dependent vaccination can make disease elimination impossible, unlike constant vaccination governed by the threshold RV = (1 −x)R0.
  • Delayed information produces stable oscillations through Hopf bifurcations, whereas the corresponding unlagged system has a globally asymptotically stable endemic state with only damped oscillations.
  • Sustained oscillations require a sufficiently steep behavioral response and an intermediate memory-delay window, with the result persisting for several realistic delay kernels.
  • 96% peak vaccine uptake during high perceived risk remained insufficient for elimination because long-term average coverage stayed below 80%, versus a critical coverage of xc ≈0.93.
  • Imitation-game models represent social learning in vaccination, while payoff-dependent strategy differences determine whether vaccination spreads, declines, or produces partial uptake.
  • The model family always includes disease-free and endemic equilibria, but payoff specifications determine their stability and whether additional behavioral equilibria exist.
  • Phenomenological information-dependent uptake models arise as a special case of imitation-based models when imitation is fast.

9. Behavior-vaccination dynamics in structured, networked populations

Network structure and behavioral feedback reshape vaccination and epidemic outcomes, producing cost-dependent trade-offs and topology-specific effects. These models also show that subsidy design and vaccine quality can reverse which intervention or network performs best.

  • Most networked vaccination models assume homogeneous mixing, limiting their representation of structured interactions.
  • Impact of interaction network topology: Networks promote vaccination at low cost but suppress it above a threshold, with stronger effects in more heterogeneous networks.
  • Impact of imitation dynamics: Peer pressure increases vaccination when costs are small but causes coverage to plummet beyond a critical cost, narrowing coexistence as α increases.
  • Impact of interaction network topology: On scale-free networks, initially pro-vaccination individuals can constrain epidemics better than on Erdős–Rényi graphs because middle- and high-degree nodes vaccinate more.
  • Impact of interaction network topology: Temporarily effective vaccination may generate oscillations, but disease spread can ultimately be controlled on complex networks.
  • Impact of interaction network topology: Imperfect vaccines can reverse the topology advantage: scale-free networks outperform Erdős–Rényi graphs for perfect vaccines, whereas Erdős–Rényi graphs can perform better for imperfect vaccines.
  • The reviewed findings enrich understanding of behavior–vaccination dynamics and can inform vaccination-strategy design.
  • External incentive programs: Free subsidies generally produce higher vaccination coverage than partial-offset subsidies, whereas myopic updating instead favors partial-offset subsidies for herd-immunity formation and socioeconomic benefit.

10. Measuring and assessing vaccination levels: traditional approaches

Vaccination and immunization levels are important for public health and modeling but remain difficult to measure accurately. Traditional administrative records, surveys, and biomarkers each face practical or interpretive limitations.

  • Measuring vaccination coverage: Vaccination coverage measurement is important for understanding vaccine-preventable disease dynamics, yet its magnitude, distribution, and efficacy are difficult to assess.
  • Administrative records: Administrative vaccination records can be distorted by discontinuous registration, migration, duplicated records, and weak registry systems in low- and middle-income countries.
  • Survey methods: Surveys such as DHS, MICS, EPI cluster surveys, and Lot Quality Assurance Sampling provide established alternative estimates of vaccination coverage.
  • Biomarker surveys: Biomarker surveys cannot usually distinguish vaccine-derived antibodies from natural infection, and waning antibodies prevent reliable dose estimation.
  • Vaccine efficacy: Vaccine efficacy varies widely; influenza estimates range from 8% to 93% depending on vaccine type, year, and study population.
  • Indirect estimation: Under random mixing, case distributions among vaccinated and unvaccinated people can help estimate vaccine effectiveness or vaccination coverage when the other quantity is known.
  • Interpreting case distributions: A rising fraction of cases among unvaccinated people can indicate increasing vaccination coverage, while the reverse may accompany increasing vaccine hesitancy.
  • Consumer technologies, including internet and mobile devices, can supplement traditional vaccination and immunization measurement efforts.

11. Digital epidemiology

Digital epidemiology supplies timely, high-resolution behavioral and contact data for coupled behavior-disease models, while network analyses reveal heterogeneous interactions and vaccination-related correlations. These approaches can improve surveillance and inform targeted control, but digital platforms complement rather than replace clinical and traditional systems.

  • Digital data sources: Digital epidemiology uses social, mobile, GPS, and RFID data to capture epidemiologically relevant behavior and contacts at fine spatial and temporal resolution.These sources can include health discussions, symptom searches, mobility, and proximity information.
  • Participatory surveillance: Participatory platforms such as FluNearYou and InfluenzaNet crowdsource near-real-time outbreak monitoring across North America and Europe.They complement conventional influenza-like-illness surveillance by engaging users in data collection.
  • Participatory surveillance: User-reported vaccination status enables geographically resolved vaccination-rate estimates and can support vaccination-efficacy estimation and model refinement.These measurements complement rates obtained through traditional approaches.
  • Contact measurement: Wearable devices objectively measure time-varying contact networks with heterogeneous activity, contact duration, and inter-contact intervals.Traditional surveys and diaries face scalability and self-reporting biases, while wearables can capture contacts those methods miss.
  • Network structure: Simple activity-driven networks omit empirical correlations such as modularity, transitivity, assortativity, and community structure.A high-school study found vaccination assortativity above chance, and simulations suggested these contact features can increase outbreak size and probability.
  • Network-based control: Vaccinating a very small fraction of the most active nodes can halt contagion, whereas egocentric sampling is less efficient than targeted vaccination but more efficient than random vaccination.The review reports that vaccinating the top 0.04% of nodes was sufficient in one model, while local observation enabled partial-information targeting.

12. Conclusion, discussion and future research

The review concludes that coupling vaccination behavior with disease dynamics produces richer emergent behavior than models that ignore behavior, while network structure and digital data expand the field’s scope. It identifies empirical validation, realistic uncertainty, interdisciplinary coordination, and multilayer-network theory as priorities for future research.

  • Conclusions: Coupled behavior-disease models generate emergent dynamics, including oscillations in outbreaks and vaccine coverage, that may be absent from corresponding compartmental models.The review emphasizes that incorporating vaccination behavior produces more dynamical outcomes overall.
  • Future research: Future mean-field research should test behavioral theories across countries and infections and incorporate stochasticity while preserving high-level population predictions.The review presents these as directions for extending parsimonious models and improving empirical understanding.
  • Network models: Network behavior-disease models show that local structure, stochasticity, and adaptation to local epidemiological conditions shape vaccination and outbreak outcomes.Reported factors include imitation, topology, incentives, protection measures, and social clustering.
  • Network models: Social clusters can produce a large outbreak probability even when empirical networks have extremely high vaccination coverage.The review also discusses temporal, multilayer, and metapopulation networks as settings requiring further attention.
  • Open problems: Multilayer vaccination theory remains at an early stage because centrality indices lack uniform definitions and different layers may support distinct dynamical processes.The review calls for interdisciplinary work on multilayer and metapopulation settings.
  • Empirical validation: Coupled-model evaluation remains constrained by uncertainty in vaccination coverage and variability in vaccine efficacy and effectiveness, especially in low-income settings.The review notes that theoretical studies have rarely incorporated these real-world issues.
  • Digital epidemiology: Digital data sources can support comprehensive, changing contact networks and real-time tracking of movement and behavior during epidemics or interventions.The review describes digital epidemiology as promising but requiring new methods for collecting, storing, and analyzing increasingly large heterogeneous datasets.
  • Interdisciplinary collaboration: Interdisciplinary progress is hindered by inconsistent terminology across behavioral epidemiology, economic epidemiology, and coupled disease-behavior systems.The review proposes workshops and broader coordination to reduce language barriers.

Appendix

The appendix provides a reference table of abbreviations used throughout the report.

  • Abbreviations: Table A1 lists abbreviations for phrases used in the report.It serves as a terminology reference for the main text.
Loading 1608.09010v3…