Theoretical Framework and Conceptual Model

Social Ecological Theory Foundation

This research adopts a multi-level social ecological model as its theoretical foundation, which emphasizes that the development and transmission of antimicrobial resistance (AMR) is determined by interconnected factors at multiple levels:

  1. Individual level: Knowledge, attitudes, and practices (KAP)
  2. Interpersonal level: Information dissemination and norm formation in family and community networks
  3. Organizational level: Healthcare delivery systems and farm production structures
  4. Community level: Collective norms and resource accessibility
  5. Policy level: Regulatory environment and market structures

This theoretical framework enables us to move beyond single behavioral or knowledge factors and examine the complex nature of AMR as a social ecological phenomenon.

One Health Conceptual Framework

This research employs the One Health conceptual framework, which recognizes the intrinsic connections between human health, animal health, and environmental health. In the context of AMR, this means:

  1. Cross-boundary transmission pathways: Resistant genes and bacteria can travel between humans, animals, and the environment through multiple routes including food chains, direct contact, and environmental media
  2. Shared selection pressures: Human and animal antibiotic use create and maintain common resistance selection pressures
  3. System interactions: Structural interactions exist between human healthcare and animal production systems
  4. Common determinants: Socioeconomic factors, knowledge dissemination, and cultural norms simultaneously influence antibiotic use in both human and animal health sectors

Multidisciplinary Integration Framework

This research adopts a multidisciplinary integration framework, combining:

  1. Sociological perspective: Focusing on how social structures and cultural norms shape antibiotic use behaviors
  2. Behavioral science perspective: Analyzing how knowledge, attitudes, and external factors influence decision-making processes
  3. Spatial epidemiology perspective: Examining how geographical and environmental factors affect AMR risk
  4. Economic perspective: Analyzing how costs, benefits, and incentives drive antibiotic use decisions
  5. Systems science perspective: Understanding feedback loops and non-linear relationships in AMR transmission

This integration framework ensures we can comprehensively grasp all dimensions of AMR as a complex public health challenge.

Cultural Adaptation Considerations

Conducting research in the Vietnamese cultural and social context requires special considerations:

  1. Collectivist orientation: Vietnamese culture emphasizes community and family decision-making rather than purely individual choices
  2. Medical pluralism: A healthcare environment where traditional and modern medicine coexist
  3. Power distance: The impact of hierarchical structures on information access and decision-making
  4. Cultural significance of agriculture: The cultural and social status of agriculture and livestock in Vietnamese rural life
  5. Regional differences: Different traditions in medical practices and agricultural methods in northern, central, and southern regions

These cultural factors will be incorporated into the research design and interpretive framework to ensure cultural sensitivity and applicability of the analysis.

Research Question 1: Patterns of Antibiotic Use Prevalence in Vietnamese Farms and Households and Their Comparative Analysis

1.1 Research Hypotheses

  • H1.1: Preventive and growth-promoting antibiotic use is common in farm animals, with prevalence rates higher than corresponding use in human medicine
  • H1.2: Different animal species (chickens, pigs, cattle) have systematically different patterns of antibiotic use, reflecting the structural needs of different production systems
  • H1.3: Farm size and antibiotic use practices have a non-linear relationship, with medium-sized farms potentially having the highest rates of inappropriate use
  • H1.4: Household human antibiotic use patterns and farm animal antibiotic use in the same community show isomorphism, influenced by common information sources and social norms

1.2 Statistical Methods

  • Descriptive analysis: Calculate prevalence rates of human and farm antibiotic use, distribution of usage purposes, and drug type distributions
    • Feasibility: Farm data variables f71-f75 (antibiotic use purposes) and f76-f82 (dosage information) are sufficient to support this analysis
    • Innovation: Adopt standardized intensity indicators (such as thousand animal daily doses, TID-DDD) to make human and animal antibiotic use directly comparable
  • χ² tests and Fisher’s exact tests: Compare differences in antibiotic use patterns between different animal species
    • Feasibility: Variable f94 (key animal species) can be used for grouping, and f240/f404/f568 (antibiotic use in different animal species) support comparative analysis
    • Sample size consideration: For low-frequency events, Fisher’s exact tests will be used to accommodate potential small sample situations
  • Segmented regression models: Evaluate the non-linear relationship between farm size and antibiotic use
    • Feasibility: Variable f52 (farm area) and variables f71-f75 (antibiotic use) can be used for this analysis
    • Innovation: Unlike simple correlation analysis, segmented regression can reveal thresholds and non-linear patterns in scale effects
  • Multilevel structural equation models: Analyze community-level isomorphism effects
    • Feasibility: Depends on community matching through geographic identifiers, sample size may be a limiting factor
    • Alternative approach: If sample size is insufficient, will shift to using generalized linear mixed effects models with community random effects
    • Innovation: Incorporate social network theory to test the mediating effect of community norms

1.3 Data Visualization Methods

  • Two-layer stacked bar charts: Compare the distribution of antibiotic use purposes between humans and different animal species
    • Feasibility: Data structures of h7.10-7.16 and f71-f75 support this type of visualization
    • Innovation: Add confidence interval bands to visually present estimation uncertainty
  • Heatmaps with cluster analysis: Comparison of different types of antibiotics’ usage frequency in humans and animals
    • Feasibility: Variables related to specific antibiotic use (drug list in h7.10 and f83-f91) support this visualization
    • Innovation: Use hierarchical clustering algorithms to identify patterns of antibiotic use combinations, rather than simple side-by-side display
  • LOESS smoothing curves: Relationship between farm size, number of animals, and antibiotic use intensity
    • Feasibility: Variables f52 (farm area), f37-f46 (number of animals), and antibiotic use indicators support this visualization
    • Innovation: Use LOESS smoothing techniques to show non-parametric trends, capturing more complex patterns than simple linear relationships
  • Community maps and spatial association graphs: Spatial association of human and animal antibiotic use at the community level
    • Feasibility: Geographic identifiers support basic spatial visualization, although precise maps may require additional geographic data
    • Innovation: Display local spatial autocorrelation indicators to identify spatial clustering areas of usage patterns

1.4 Variables Needed

  • Household antibiotic use: h7.10 (recent treatments used), h7.17-7.23 (antibiotic use history)
  • Farm antibiotic use: f66-f67 (whether antibiotics are used), f71-f75 (antibiotic use purposes), f76-f82 (dosage information), f83-f93 (specific antibiotic use)
  • Animal species: f94 (key animal species), f240/f404/f568 (antibiotic use in each type of animal)
  • Farm characteristics: f37-f46 (numbers of various animals), f52 (farm area)
  • Geographic identifiers: h1.1-1.3 and f1.1-1.3 (locations)

1.5 Relationship to Doctoral Thesis

This research question directly supports the One Health framework of the doctoral thesis by comparing antibiotic use patterns in humans and animals, establishing connections between human health and animal health sectors. This cross-sector comparison is fundamental to understanding antimicrobial resistance as a One Health issue and provides key parameters for subsequent transmission dynamics modeling.

1.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides a comprehensive picture of antibiotic selection pressure, helping to explain the distribution patterns of resistant genes in different environments (human/animal)
  • Contribution to Chapter 5: Provides key input parameters for transmission dynamics models, such as antibiotic use rates, patterns, and intensity in different host populations

1.7 Cultural Adaptation Considerations

  • Farm classification will consider Vietnam’s unique agricultural structure, such as the distinction between small-scale mixed family farms and commercialized single-species farms
  • Purpose analysis will consider the cultural significance of antibiotic use in traditional livestock practices
  • Community norm analysis will consider the collectivist nature of Vietnamese rural communities
  • Concepts of “disease” and “prevention” in the questionnaire will be interpreted through Vietnamese language and cultural background

1.8 Quality Control and Limitation Management

  • Data quality assessment: Use internal consistency tests and logic checks to evaluate the quality of key variables
  • Missing data handling: For key variables with >5% missing, use multiple imputation methods and conduct sensitivity analysis
  • Measurement error handling: Assess the reliability of self-reported data through cross-validation questions
  • Potential limitations: Acknowledge that cross-sectional design limits causal inference; farm samples may be biased toward formalized farms, affecting representativeness

Research Question 2: Comparative Analysis of Antibiotic Knowledge, Attitudes, and Practices (KAP) Between Farm Owners and General Household Members

2.1 Research Hypotheses

  • H2.1: Farm owners and general household members have systematically different perceptions of antibiotics, characterized by high familiarity with animal antibiotics but insufficient awareness of human health impacts among farm owners
  • H2.2: The relationship between farm owners’ education level, experience, knowledge, and rational antibiotic use is moderated by market orientation
  • H2.3: The consistency between antibiotic knowledge and actual use behavior is lower in farm owners than in general household members, reflecting a structural contradiction between production pressure and health knowledge
  • H2.4: The association between knowledge improvement and behavior change in groups exposed to correct antibiotic use campaigns is moderated by education level and social capital

2.2 Statistical Methods

  • Propensity score matching t-tests: Compare knowledge scores between farm owners and general household members
    • Feasibility: Variables h4.3-4.5 (antibiotic knowledge) and f53-f65 (product identification) provide comparable knowledge measurements
    • Innovation: Use propensity score matching to control for confounding factors, improving comparison validity
    • Sample size consideration: Power analysis for main effect detection indicates at least 120 people per group are needed to achieve 80% power (α=0.05, d=0.4)
  • Stratified multiple regression models: Influence of farm owner characteristics on antibiotic knowledge levels
    • Feasibility: Variables f2.4 (education), f11 (birth year/age), f97/f267/f431 (experience) and other independent variables are sufficient to build the model
    • Innovation: Stratify analysis by market orientation degree to test moderation effects
    • Alternative approach: If stratified sample size is insufficient, will use interaction terms instead of stratification
  • Structural equation modeling: Comparative analysis of knowledge-attitude-behavior chains
    • Feasibility: Based on available variables, simplified SEM models can be constructed
    • Innovation: Test whether the knowledge→attitude→behavior mediation pathway differs between farm owners and general household members
    • Sample size consideration: Complex SEM models require larger sample sizes; if insufficient, model structure will be simplified
  • Moderated mediation analysis: Pathway analysis of how campaign exposure influences behavior through knowledge
    • Feasibility: Variables h8.2-8.6 (campaign exposure) support this analysis
    • Innovation: Simultaneously examine the moderation effects of education and social capital, revealing conditional limitations of intervention effectiveness

2.3 Data Visualization Methods

  • Radar charts with confidence regions: Comparison of farm owners and household members across different knowledge dimensions
    • Feasibility: Knowledge measurements in h4.3-4.5 and f53-f65 support multidimensional comparison
    • Innovation: Add Bootstrap confidence regions to visually present estimation uncertainty
  • Heatmaps combined with dendrograms: Three-dimensional relationship between education level, knowledge domains, and groups (farm owners/non-farm owners)
    • Feasibility: Education variables (h2.4/f2.4) and knowledge scores support this visualization
    • Innovation: Use hierarchical clustering dendrograms to identify knowledge structure patterns, combined with heatmaps for presentation
  • Interactive forest plots: Comparison of effect sizes for knowledge-behavior consistency across different background factors
    • Feasibility: Can be created based on regression analysis results
    • Innovation: Add interactive functionality allowing filtering and comparison of effects by different characteristics
  • Path diagram visualization: Mediation effects in the knowledge-attitude-behavior chain
    • Feasibility: Can be created based on SEM or mediation analysis results
    • Innovation: Contrastively display different path coefficients and explained variance between farm owners and general household members

2.4 Variables Needed

  • Antibiotic knowledge: h4.3-4.5 (antibiotic identification and indications), f53-f65 (product identification and AMR awareness)
  • Farm owner characteristics: f2.4 (education), f10 (gender), f11 (birth year), f97/f267/f431 (farming experience)
  • General population characteristics: h2.1 (gender), h2.2 (date of birth), h2.4 (education)
  • Antibiotic use behavior: h7.10-7.23 (household antibiotic use), f66-f93 (farm antibiotic use)
  • Campaign exposure: h8.2-8.6 (exposure to antibiotic use campaigns)
  • Social capital: h3.19-3.28 (social capital and community participation)

2.5 Relationship to Doctoral Thesis

This research question explores the intersection of human health and animal health knowledge systems, revealing differences in antibiotic knowledge across different populations and their impact on usage behavior. This comparative analysis supports the cross-sectoral perspective of the doctoral thesis, helping to understand the key role of knowledge in AMR management.

2.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides knowledge gaps and possible incorrect usage practices, helping to explain differences in resistance patterns
  • Contribution to Chapter 5: Provides parameters for knowledge-based intervention modeling; quantifies the potential impact of knowledge change on behavior

2.7 Cultural Adaptation Considerations

  • Knowledge measurement will consider cultural understanding of antibiotic-related terminology in Vietnamese
  • Attitude measurement will consider respect for authority (such as medical professionals) in Vietnamese culture
  • Knowledge-behavior relationship analysis will consider collective decision-making rather than purely individual decision-making cultural background
  • Campaign effect analysis will consider special channels and patterns of information dissemination in Vietnamese rural areas

2.8 Quality Control and Limitation Management

  • Measurement equivalence verification: Ensure knowledge measurement items in household and farm questionnaires have cross-group equivalence
  • Social desirability bias control: Use indirect question techniques to reduce response bias for sensitive questions
  • Knowledge measurement reliability analysis: Calculate internal consistency coefficients (such as Cronbach’s α) to assess the reliability of knowledge scales
  • Potential limitations: Acknowledge that self-reported knowledge may be influenced by social desirability bias; antibiotic identification may be affected by packaging and brand familiarity

Research Question 3: How Socioeconomic Factors Simultaneously Influence Antibiotic Use Decisions in Households and Farms

3.1 Research Hypotheses

  • H3.1: The socioeconomic gradient manifests differently in household and farm antibiotic use, with low socioeconomic status associated with overuse in households but with underuse in farms
  • H3.2: The degree of market orientation of farms is uniquely associated with antibiotic use types, with highly market-oriented farms tending to use preventive and growth-promoting antibiotics
  • H3.3: Economic shocks (such as COVID-19) have structurally different impact pathways on antibiotic access decisions in households and farms
  • H3.4: The impact of education, moderated by market and social network structures, produces differentiated effects on antibiotic use in households and farms

3.2 Statistical Methods

  • Multiple component principal component analysis: Construct socioeconomic indicators for households and farms
    • Feasibility: Variables h3.2-3.18 (household assets, housing, etc.) and farm-related economic indicators (f52-farm size, etc.) are sufficient to support this analysis
    • Innovation: Use multiple component PCA methods to handle mixtures of categorical and continuous variables, more suitable for complex socioeconomic indicators than standard PCA
  • Paired sample quantile regression: Impact of socioeconomic factors on antibiotic use
    • Feasibility: To be implemented in household-farm pairs that can be matched
    • Innovation: Quantile regression analyzes the differentiated impact of socioeconomic factors at different levels of antibiotic use, beyond mean effects
    • Alternative approach: If matched samples are insufficient, will shift to community-level aggregated analysis
  • Counterfactual mediation analysis: Evaluate how knowledge mediates the relationship between socioeconomic status and antibiotic use
    • Feasibility: Although limited by cross-sectional data, mediation analysis based on counterfactual framework provides more robust estimates
    • Innovation: Use counterfactual methods to strengthen causal inference, better handling confounding factors than traditional mediation analysis
  • Latent class analysis: Identify farm types by market orientation degree and antibiotic use patterns
    • Feasibility: Feasible based on f146/f311/f475 (whether for income) and other relevant indicators
    • Innovation: Use data-driven methods to identify potential farm types, rather than preset categories

3.3 Data Visualization Methods

  • Socioeconomic gradient comparison graphs: Trends of antibiotic use rates across socioeconomic quintiles in households and farms
    • Feasibility: Socioeconomic indicators constructed using multiple component PCA can support this visualization
    • Innovation: Use parallel coordinate systems to simultaneously display multiple antibiotic use indicators across socioeconomic status
  • Sankey diagrams: Flow of socioeconomic status→knowledge→use behavior and its differences between households/farms
    • Feasibility: Can be created based on mediation analysis results
    • Innovation: Intuitively display path effect sizes and their differential distribution across different contexts
  • Decision tree maps: Impact pathways of socioeconomic and market factors on antibiotic use
    • Feasibility: Can be created based on decision tree or random forest model results
    • Innovation: Provide intuitive representation of decision paths, helping identify key branch points and thresholds
  • Counterfactual prediction graphs: Simulate the potential impact of different socioeconomic interventions on antibiotic use
    • Feasibility: Can create hypothetical scenario predictions based on regression model results
    • Innovation: Use counterfactual scenarios to visualize potential intervention effects, supporting policy discussions

3.4 Variables Needed

  • Socioeconomic indicators-household: h3.2-3.4 (housing materials), h3.10-3.11 (assets), h3.18 (fuel type)
  • Socioeconomic indicators-farm: f52 (farm area), f146/f311/f475 (whether for income), f147-f155 (farm challenges)
  • Education level: h2.4 (household education), f2.4 (farm owner education)
  • Antibiotic use: h7.10-7.23 (household antibiotic use), f66-f93 (farm antibiotic use)
  • Antibiotic knowledge: h4.3-4.5 (antibiotic knowledge), f53-f65 (product identification)
  • COVID-19 impact: h5.1-5.3 (COVID impact), f109-f116/f279-f286/f443-f450 (farm COVID impact)

3.5 Relationship to Doctoral Thesis

This research question explores how socioeconomic factors as key social determinants influence antibiotic use in human and animal sectors. This cross-sectoral socioeconomic analysis reveals common structural drivers, supporting the social ecological framework of the doctoral thesis, emphasizing the importance of contextual factors in antibiotic use decisions.

3.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides socioeconomic background, helping to explain the social gradient in resistant gene distribution; identifies high-risk socioeconomic groups
  • Contribution to Chapter 5: Provides parameters for the impact of socioeconomic differences on transmission dynamics; supports modeling of interventions targeting specific socioeconomic groups

3.7 Cultural Adaptation Considerations

  • Socioeconomic indicator construction will consider unique wealth markers in Vietnamese rural and peri-urban areas
  • Market orientation analysis will consider the special historical and cultural background of Vietnamese agricultural transformation
  • Economic shock impact analysis will consider the tradition of mutual assistance in Vietnamese families and community networks
  • Education effect analysis will consider the relationship between the Vietnamese educational system and rural knowledge inheritance

3.8 Quality Control and Limitation Management

  • Socioeconomic indicator validation: Compare with Vietnam’s National Bureau of Statistics socioeconomic classifications to ensure the validity of constructed indicators
  • Reverse causality control: Use instrumental variable or directed acyclic graph (DAG) methods to handle potential reverse causality
  • Sensitivity analysis: Test result robustness using different socioeconomic indicator construction methods
  • Potential limitations: Acknowledge that the dynamic nature of socioeconomic status cannot be fully captured in cross-sectional research; self-reported farm economic data may not be precise

Research Question 4: How Farm Biosecurity Measures and Management Practices Influence Antibiotic Use and Potential Antimicrobial Resistance Risk

4.1 Research Hypotheses

  • H4.1: Farm biosecurity level and preventive antibiotic use show a non-monotonic relationship, with medium levels potentially associated with the highest preventive use
  • H4.2: Different types of biosecurity measures have heterogeneous impacts on antibiotic use, with isolation measures potentially more effective than general cleaning measures in reducing antibiotic use
  • H4.3: Waste management practices form regional network effects, affecting potential AMR exposure in surrounding households and farms
  • H4.4: Farm animal stocking density, feed type, and management intensification collectively constitute structural conditions affecting antibiotic use

4.2 Statistical Methods

  • Latent variable modeling: Create latent variables for biosecurity level and antibiotic use types
    • Feasibility: Farm questionnaire contains rich biosecurity-related variables (f179-f201) that can be used for latent variable construction
    • Innovation: Use latent variable methods to handle multidimensional concepts, better capturing complex structures than simple indicators
  • Generalized additive models (GAM): Non-linear impact of biosecurity measures on antibiotic use
    • Feasibility: Constructed biosecurity indicators and antibiotic use variables (f66-f93) can be used for GAM analysis
    • Innovation: GAM can capture complex non-linear relationships, beyond the linear relationships assumed in traditional regression
    • Alternative approach: If model convergence issues arise, will use segmented linear regression as an alternative
  • Network autocorrelation models: Spatial spillover effects of waste management practices
    • Feasibility: Depends on geographic data and waste management variables (f206-f218)
    • Innovation: Integrate spatial econometrics methods to assess the externalities of farm management practices
  • Structural equation mixture models: Comprehensive effect analysis of management practices
    • Feasibility: Multiple management practice variables support this analysis, but complex models may be limited by sample size
    • Innovation: Simultaneously handle latent variables and random effects, capturing complex relationships between variables
    • Alternative approach: If sample size is insufficient, will simplify to basic multilevel models

4.3 Data Visualization Methods

  • Non-parametric smoothing curves: Non-linear relationship between biosecurity level and antibiotic use
    • Feasibility: Constructed composite indicators support this visualization
    • Innovation: Use thin plate spline smoothing techniques to show complex non-linear relationships with confidence bands
  • Multi-layer heatmaps: Biosecurity measures and antibiotic use patterns of different farm types
    • Feasibility: Farm classification based on latent class analysis can be used for heatmap display
    • Innovation: Use hierarchical heatmaps to visualize multidimensional patterns, showing associations between different categorical variables
  • Spatial spillover effect graphs: Spatial network effects of waste management practices
    • Feasibility: Variables f206-f218 (manure treatment) and geographic data support this visualization
    • Innovation: Use spatial network analysis methods to visualize the spatial impact range of management practices
  • Radar charts combined with density distributions: Comprehensive comparison of antibiotic use patterns across different management practice combinations
    • Feasibility: Can be implemented by combining management practice indicators and antibiotic use data
    • Innovation: Overlay density distributions on radar chart axes, showing data variability rather than just point estimates

4.4 Variables Needed

  • Biosecurity measures: f186-f192 (biosecurity), f193-f201 (visitor measures), f179-f185 (housing characteristics)
  • Waste management: f206-f218 (manure treatment), f213-f218 (liquid waste)
  • Animal density: f37-f46 (animal numbers), f179/f345/f512 (housing area)
  • Farm management: f158-f185 (feeding practices), f178-f185 (cleaning and disinfection)
  • Antibiotic use: f66-f93 (antibiotic use), f71-f75 (use purposes)
  • Geographic data: f1.1-1.3 (location information)

4.5 Relationship to Doctoral Thesis

This research question focuses on farm management practices, which are key determinants of AMR risk in the animal health sector. By analyzing how biosecurity measures influence antibiotic use, it reveals the critical balance between infection prevention and antibiotic use reduction, providing important insights for AMR management under the One Health approach.

4.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides background on selection pressure variation in farm environments; helps explain differences in resistant gene distribution under different management practices
  • Contribution to Chapter 5: Provides parameters for modeling interventions targeting farm management practices; quantifies the potential impact of biosecurity improvements on antibiotic use

4.7 Cultural Adaptation Considerations

  • Biosecurity concepts will be recontextualized based on Vietnamese agricultural background, considering local farm structure characteristics
  • Waste management analysis will consider traditional practices and cultural acceptability in rural areas
  • Management practice analysis will consider agricultural tradition differences across Vietnamese regions
  • Intervention recommendations will consider local resource constraints and cultural acceptability

4.8 Quality Control and Limitation Management

  • Biosecurity measurement validation: Validate self-reported biosecurity measures’ accuracy through field observation of a subsample
  • Spatial data quality assessment: Evaluate the precision and reliability of geographic coordinates to ensure the validity of spatial analysis
  • Multicollinearity check between variables: Use variance inflation factors (VIF) to detect and address potential multicollinearity between management practice variables
  • Potential limitations: Acknowledge that lack of longitudinal data limits assessment of the impact of management practice changes; self-reported management practices may have social desirability bias

Research Question 5: Knowledge Transmission and Behavioral Influence Between Households and Farms in Antibiotic Use

5.1 Research Hypotheses

  • H5.1: Individuals who are both household members and farm owners/workers play a knowledge bridge role between the two domains, forming unique knowledge integration patterns
  • H5.2: Farm owners’ antibiotic decision frameworks influence human medical antibiotic use in their households, with this influence moderated by knowledge, economic, and cultural factors
  • H5.3: The diversity of health information sources has an inverted U-shaped relationship with antibiotic knowledge quality and integration level, with medium diversity potentially optimal
  • H5.4: Community-level knowledge transmission network structures influence the spread speed and adoption rate of antibiotic use norms

5.2 Statistical Methods

  • Mixed methods social network analysis: Identify knowledge flow patterns and key nodes
    • Feasibility: Requires clear household-farm matching mechanisms, may need supplementary social network data collection
    • Innovation: Combine quantitative network parameters and qualitative knowledge flow analysis, beyond simple statistical associations
    • Alternative approach: If complete network data is not available, will use ego-centric network analysis based on h3.21 (health communication network)
  • Cross-domain knowledge integration index: Develop new indicators to measure individual integration of knowledge across human-animal domains
    • Feasibility: Based on knowledge items in h4.3-4.5 and f53-f65
    • Innovation: Create new measurement tools to assess individual ability to integrate cross-domain knowledge
  • Polynomial response surface analysis: Test non-linear relationships between information source diversity and knowledge quality
    • Feasibility: h4.2 (information sources) and knowledge scores support this analysis
    • Innovation: Use response surface methods to explore optimal diversity levels, beyond simple linear or quadratic relationships
  • Agent-Based modeling simulation: Simulate information diffusion in community knowledge transmission networks
    • Feasibility: Simplified models parameterized based on survey data are feasible
    • Innovation: Use complex systems methods to simulate knowledge transmission dynamics in social networks
    • Alternative approach: If time or technical resources are limited, will shift to using simplified mathematical diffusion models

5.3 Data Visualization Methods

  • Bipartite network graphs: Visualization of household-farm knowledge flow networks
    • Feasibility: Can create simplified network visualizations based on limited network data
    • Innovation: Use bipartite networks to show association structures between people and information sources
  • Knowledge integration heatmaps: Comparison of cross-domain knowledge integration patterns across different populations
    • Feasibility: Can create integration pattern visualizations based on knowledge measurement items
    • Innovation: Combine hierarchical clustering with heatmaps to identify and display knowledge integration pattern types
  • Information ecosystem graphs: Visualization of knowledge sources and flows within communities
    • Feasibility: Based on information source and knowledge measurement data
    • Innovation: Adopt an ecosystem perspective to show the diversity and interrelationships of information sources
  • Simulation animations: Temporal dynamics visualization of knowledge transmission processes
    • Feasibility: Based on outputs from Agent-Based models or mathematical diffusion models
    • Innovation: Dynamically display knowledge transmission processes, helping understand temporal dimension changes

5.4 Variables Needed

  • Knowledge scores: h4.3-4.5 (antibiotic knowledge), f53-f65 (product identification)
  • Information sources: h4.2 (health information), h8.8 (AMR information), corresponding farm variables
  • Social networks: h3.21 (health communication network), h3.19-3.28 (social capital)
  • Antibiotic use: h7.10-7.23 (household antibiotic use), f66-f93 (farm antibiotic use)
  • Geographic identifiers: h1.1-1.3 and f1.1-1.3 (for community matching)
  • Demographics: h2.1-2.5 (household), f2.1-2.5 (farm owner)

5.5 Relationship to Doctoral Thesis

This research question explores knowledge transmission and behavioral influence between human and animal health sectors, which are core elements of the One Health approach. By analyzing knowledge flows between the two sectors, it reveals opportunities to strengthen cross-sectoral collaboration to address AMR challenges, directly supporting the integrated perspective of the doctoral thesis.

5.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides knowledge transmission networks that can be compared with gene transmission networks; helps explain cross-sectoral similarities in resistance patterns
  • Contribution to Chapter 5: Provides parameters for knowledge transmission models; supports the design and evaluation of cross-sectoral knowledge interventions

5.7 Cultural Adaptation Considerations

  • Social network analysis will consider the unique social structures and hierarchical relationships in Vietnamese rural communities
  • Knowledge transmission analysis will consider the coexistence and integration of traditional and modern information channels in Vietnam
  • Cross-domain knowledge research will consider traditional views of human-animal relationships in Vietnamese culture
  • Intervention recommendations will consider local community leadership structures and decision-making patterns

5.8 Quality Control and Limitation Management

  • Social network data supplementation strategy: If existing data is insufficient, design small-scale supplementary surveys to collect key social network data
  • Model validation methods: Use historical data or small-scale validation studies to test the predictive ability of knowledge transmission models
  • Self-reported network data validation: Use best practices in network data collection to reduce recall bias
  • Potential limitations: Acknowledge challenges in collecting complete social network data; simplified assumptions in simulation models; potential inability to obtain complete bidirectional knowledge flow data due to time constraints

Research Question 6: Seasonal and Temporal Patterns of Farm Antibiotic Use and Their Relationship with Disease Burden

6.1 Research Hypotheses

  • H6.1: Farm antibiotic use exhibits distinctive seasonal patterns that reflect the combined influence of local climatic conditions, production cycles, and market demand
  • H6.2: Different animal species (chickens, pigs, cattle) show systematically different seasonal patterns in antibiotic use, reflecting their different physiological and production system sensitivities to seasonal changes
  • H6.3: Market price fluctuations and festival demands form major external drivers affecting the temporal patterns of farm antibiotic use decisions
  • H6.4: The impact of the COVID-19 pandemic on farm antibiotic use varies by farm type and product market, forming differentiated adaptation patterns

6.2 Statistical Methods

  • Temporal pattern decomposition analysis: Extract seasonal patterns based on retrospective reports
    • Feasibility: Although the data is not a true time series, basic time patterns can be reconstructed through retrospective questions
    • Innovation: Apply time series decomposition techniques to extract seasonal, trend, and irregular components
    • Alternative approach: If temporal data quality is insufficient, will shift to using simple seasonal comparison analysis
  • Seasonal index models: Compare seasonal patterns across different animal species
    • Feasibility: Can be constructed based on reported antibiotic use at different time points
    • Innovation: Calculate seasonal indices to quantify relative use intensity across different seasons
  • Panel data analysis: Dynamic relationship between disease burden and antibiotic use
    • Feasibility: Depends on retrospective data for different time points, quality may be limited
    • Innovation: Handle compounded variation in time and cross-sectional dimensions
    • Alternative approach: If data doesn’t support panel analysis, will use stratified cross-sectional analysis
  • Difference-in-differences analysis: Evaluate the differentiated impact of COVID-19 on different farm types
    • Feasibility: Variables f109-f116/f279-f286/f443-f450 (COVID impact) support this analysis
    • Innovation: Use quasi-experimental design methods to estimate causal effects
    • Sample size consideration: Requires sufficiently large subgroup samples to achieve adequate statistical power

6.3 Data Visualization Methods

  • Seasonal heatmaps with climate overlay: Relationship between antibiotic use intensity across different seasons and animal species with climatic factors
    • Feasibility: Can be created based on seasonal use patterns and meteorological data
    • Innovation: Integrate climate data to show associations between environmental factors and antibiotic use
  • Interrupted time series visualization: Changes in antibiotic use trends before and after COVID-19
    • Feasibility: Can create simplified time trends based on retrospective data
    • Innovation: Visually display the impact of interventions or external shocks on trends
  • Pulse charts: Temporal association between market price fluctuations and antibiotic use
    • Feasibility: Requires supplementary market price data, obtainable through secondary data
    • Innovation: Use pulse charts to show relationships between discrete events and continuous processes
  • Dynamic system diagrams: Visualization of feedback loops among disease-antibiotic use-market demand
    • Feasibility: Can create conceptual system diagrams based on survey data and supplementary information
    • Innovation: Use system dynamics visual language to display dynamic feedback relationships

6.4 Variables Needed

  • Antibiotic use temporal patterns: Temporal information in f66-f93
  • Disease occurrence: Farm-reported disease patterns (if available)
  • COVID-19 impact: f109-f116/f279-f286/f443-f450 (COVID impact on farms)
  • Animal species: f94 (key animal species), f240/f404/f568 (antibiotic use in different animal types)
  • Market factors: f146/f311/f475 (whether for income), relevant market orientation indicators
  • Seasonal information: Seasonal use patterns extracted from the questionnaire (if available)
  • Supplementary data: Regional climate data, market price data (to be obtained from external sources)

6.5 Relationship to Doctoral Thesis

This research question explores the temporal dynamics of antibiotic use, particularly in farm environments, providing important background for understanding temporal variation in AMR transmission. By analyzing seasonal patterns and the impact of external shocks (such as COVID-19), it reveals the dynamic nature of livestock antibiotic use, supporting the doctoral thesis’s comprehensive understanding of transmission dynamics.

6.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides background on temporal variation in antibiotic selection pressure; helps explain seasonal differences in resistance patterns (if any)
  • Contribution to Chapter 5: Provides parameters for temporal variables in transmission dynamics models; supports scenario modeling considering seasonality and external shocks

6.7 Cultural Adaptation Considerations

  • Seasonal analysis will consider Vietnam’s unique climate patterns and agricultural seasons
  • Festival demand analysis will consider the cultural importance and consumption patterns of Vietnamese traditional festivals
  • COVID-19 impact analysis will consider Vietnam-specific pandemic response measures and socioeconomic impacts
  • Temporal concepts and recall will consider time perception methods in Vietnamese rural communities

6.8 Quality Control and Limitation Management

  • Retrospective temporal data validation: Use important event anchoring techniques to improve recall data accuracy
  • External data integration: Carefully match meteorological and market price data with survey timeframes
  • Temporal measurement consistency check: Assess internal consistency of reports at different time points
  • Potential limitations: Acknowledge limitations of retrospective temporal data; lack of detailed disease surveillance data; counterfactual challenges in COVID-19 impact assessment

Research Question 7: Comprehensive Predictive Models of Antibiotic Use and AMR Risk: Integrating Household and Farm Factors

7.1 Research Hypotheses

  • H7.1: Models integrating household and farm factors have significantly superior predictive capability for inappropriate antibiotic use compared to single-domain models, especially in One Health risk hotspots
  • H7.2: Factors at community, farm, and household levels form unique interaction patterns that predict high-risk scenarios better than main effects
  • H7.3: Key leverage points in the social-ecological system are located at the meso level (e.g., community norms and knowledge transmission networks) rather than the micro level (individual knowledge)
  • H7.4: Coordinated interventions combining biosecurity improvements and antibiotic stewardship produce positive spillover effects for both farms and surrounding households

7.2 Statistical Methods

  • Ensemble machine learning models: Integrate random forests, gradient boosting, and deep learning methods
    • Feasibility: While sample size may limit deep learning applications, basic ensemble methods are feasible
    • Innovation: Use model stacking techniques to enhance predictive ability, beyond single algorithms
    • Sample size consideration: Employ cross-validation and regularization methods to mitigate limitations of small samples on complex models
  • Bayesian network models: Capture conditional dependency relationships between variables
    • Feasibility: Complex Bayesian networks may be limited by sample size, but structure learning algorithms can adjust complexity
    • Innovation: Probabilistic reasoning to simulate intervention effects, supporting more nuanced decision analysis
    • Alternative approach: If computational complexity is too high, will use simplified directed acyclic graph methods
  • Cross-validation and nested validation: Robust evaluation of model predictive performance
    • Feasibility: Standard model evaluation methods, completely feasible
    • Innovation: Use nested cross-validation to avoid model selection bias, improving evaluation reliability
  • Causal trees: Evaluate potential causal effects of different intervention strategies
    • Feasibility: Depends on reasonable causal assumptions and sufficient background variables
    • Innovation: Use causal machine learning methods to estimate heterogeneous treatment effects
    • Alternative approach: If causal assumptions are difficult to verify, will shift to using traditional predictive scenario analysis

7.3 Data Visualization Methods

  • Variable importance network graphs: Display interconnected importance networks of household and farm predictors
    • Feasibility: Based on variable importance and interaction importance metrics from machine learning models
    • Innovation: Use network structures to show interaction importance between variables, rather than simple univariate importance bar charts
  • Spatial risk prediction heatmaps: Spatial distribution of integrated household and farm risk factors
    • Feasibility: Can be created based on geographic identifiers and prediction model outputs
    • Innovation: Use spatial interpolation techniques to create continuous risk surfaces, integrating multi-source predictors
  • Decision path visualization: Key decision paths for high-risk scenarios
    • Feasibility: Based on decision tree or Bayesian network model outputs
    • Innovation: Use interactive visualizations to allow exploration of risk outcomes along different decision paths
  • Intervention simulation dashboards: Comparison of predicted effects across different intervention strategy combinations
    • Feasibility: Based on counterfactual simulations or predictive model scenario analyses
    • Innovation: Develop interactive dashboards allowing adjustment of intervention parameters, real-time viewing of predicted impacts

7.4 Variables Needed

  • Household factors: Key household variables from previous analyses
    • Socioeconomic indicators (h3.2-3.18)
    • Knowledge levels (h4.3-4.5)
    • Healthcare seeking behaviors (h7.5-7.9)
    • Antibiotic use (h7.10-7.23)
  • Farm factors: Key farm variables from previous analyses
    • Farm characteristics (f37-f52)
    • Biosecurity measures (f186-f201)
    • Antibiotic use (f66-f93)
    • Management practices (f158-f185)
  • Community variables: Variables that can be aggregated at the community level
    • Geographic identifiers (h1.1-1.3, f1.1-1.3)
    • Social capital indicators (h3.19-3.28)
    • Supplementary geographic data (e.g., population density, healthcare facility accessibility)

7.5 Relationship to Doctoral Thesis

This research question integrates key findings from all previous studies, building comprehensive predictive models that directly support the One Health framework of the doctoral thesis. By integrating household and farm data, identifying key risk factors and intervention points, it provides a scientific basis for evidence-based AMR management strategies.

7.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Provides risk predictions that can guide genomic sample analysis; supports explanatory frameworks for genotype-phenotype associations
  • Contribution to Chapter 5: Provides key parameters for transmission dynamics models; supports the design and evaluation of comprehensive intervention strategies; provides a basis for risk stratification

7.7 Cultural Adaptation Considerations

  • Risk definitions and priorities will consider risk perception and priority concerns in the Vietnamese context
  • Intervention simulations will consider local resource constraints and implementation feasibility
  • Predictive model interpretation will consider the needs of local decision-makers and stakeholders
  • Visualization design will consider local data literacy levels and cultural understanding habits

7.8 Quality Control and Limitation Management

  • Model validation strategy: Use independent validation samples or temporal validation (if subsequent data is available) to assess model robustness
  • Uncertainty quantification: Use bootstrap or Monte Carlo simulation to quantify prediction uncertainty ranges
  • Model interpretation tools: Use SHAP values and partial dependence plots to improve interpretability of complex models
  • Potential limitations: Acknowledge limitations of small samples for complex models; inherent challenges of causal inference; limitations of model generalization to other regions of Vietnam

Research Question 8: Spatial Dimensions of the Household-Farm-Environment Interface: Geographic Distribution of Antibiotic Use and AMR Risk

8.1 Research Hypotheses

  • H8.1: Household and farm antibiotic use form spatially heterogeneous clusters, reflecting the influence of local knowledge transmission networks and social-ecological systems
  • H8.2: The spatial distribution of farm density and types creates unique risk geographies, with high-density areas forming AMR “hotspots” but relationships may vary by management practices
  • H8.3: Environmental media (water sources, waste management systems) form network connectivity effects, creating AMR transmission pathways beyond direct contact
  • H8.4: Spatial accessibility (such as distance to medical facilities, veterinary services, and pharmacies) forms structural geographic conditions for antibiotic access and use

8.2 Statistical Methods

  • Geographically weighted regression: Regression analysis considering spatial heterogeneity and non-stationarity
    • Feasibility: Basic spatial regression can be performed based on geographic identifiers, but precision is limited by the granularity of geographic data
    • Innovation: Capture spatial variability of effects, rather than assuming spatial consistency
    • Alternative approach: If the number of spatial units is insufficient, will use spatial variability exploration methods instead of formal GWR
  • Spatial autocorrelation and heterogeneity detection: Assess spatial clustering and heterogeneity of antibiotic use
    • Feasibility: Basic spatial statistical methods, can be implemented with limited geographic data
    • Innovation: Use local indicators (LISA) to identify hotspots and cold spots, beyond global indicators
  • Network flow analysis: Evaluate potential transmission pathways through environmental media
    • Feasibility: Requires environmental and hydrological data, may need supplementary geographic data
    • Innovation: Use flow network analysis methods to simulate environmental transmission pathways
    • Alternative approach: If environmental data is insufficient, will use simplified distance functions as approximations
  • Spatial accessibility analysis: Evaluate spatial accessibility of medical and veterinary services
    • Feasibility: Requires data on locations of medical facilities, may need supplementary collection
    • Innovation: Use Enhanced Two-Step Floating Catchment Area (E2SFCA) method, considering distance decay and supply-demand relationships
    • Sample size consideration: Spatial precision is limited by the scale of the study area and density of sampling points

8.3 Data Visualization Methods

  • Spatial autocorrelation heatmaps: Visualization of spatial clustering of antibiotic use and risk
    • Feasibility: Based on geographic identifiers and calculated local spatial autocorrelation indicators
    • Innovation: Use differentiated color schemes to distinguish different types of spatial clustering (high-high, low-low, spatial outliers)
  • Bivariate spatial association maps: Spatial relationships between household and farm antibiotic use
    • Feasibility: Requires matched household and farm geographic data
    • Innovation: Simultaneously display association patterns of two spatial processes, revealing cross-sectoral spatial dependencies
  • Environmental medium network maps: Spatial representation of environmental transmission pathways
    • Feasibility: Requires environmental data, may need supplementary collection
    • Innovation: Combine topographic, hydrological, and waste management data to visualize potential transmission networks
  • Accessibility isochrone maps: Spatial distribution of accessibility to medical and veterinary services
    • Feasibility: Requires location data for service points and road network information
    • Innovation: Integrate accessibility of multiple service types into a comprehensive accessibility index

8.4 Variables Needed

  • Geographic identifiers: h1.1-1.3 (household locations), f1.1-1.3 (farm locations)
  • Antibiotic use indicators: h7.10-7.23 (households), f66-f93 (farms)
  • Environmental factors: h3.13-3.15 (water sources), f206-f218 (waste management)
  • Farm density: Regional farm density calculated based on geographic data
  • Community characteristics: Aggregated socioeconomic and knowledge indicators
  • Supplementary data: Distribution of medical facilities, road networks, topography and hydrology (to be obtained from external sources)

8.5 Relationship to Doctoral Thesis

This research question introduces a spatial dimension, exploring the geographic distribution of antibiotic use and AMR risk, revealing spatial dynamics at the human-animal-environment interface. This spatial analysis directly supports the One Health perspective of the doctoral thesis, particularly regarding the spatial aspects of transmission dynamics, providing geographic guidance for targeted interventions.

8.6 Connection to Subsequent Chapters

  • Contribution to Chapter 4: Guides the geographic distribution strategy for genomic samples; supports analysis and interpretation of spatial distribution of resistant genes
  • Contribution to Chapter 5: Provides parameters for spatially explicit transmission models; supports the design of geographically targeted interventions; quantifies the impact of spatial heterogeneity on transmission dynamics

8.7 Cultural Adaptation Considerations

  • Spatial analysis will consider the specificity of residential patterns in Vietnamese rural and peri-urban areas
  • Environmental transmission pathway analysis will consider cultural practices in local water resource use
  • Accessibility analysis will consider Vietnam-specific transportation and healthcare service access patterns
  • Risk hotspot interpretation will consider local community understanding of spatial clustering concepts

8.8 Quality Control and Limitation Management

  • Spatial data quality assessment: Evaluate the accuracy and completeness of geographic coordinates
  • Spatial scale sensitivity analysis: Repeat analysis at different spatial aggregation levels to check result robustness
  • Spatial autocorrelation significance testing: Use appropriate permutation tests to assess statistical significance of spatial patterns
  • Potential limitations: Acknowledge limitations of spatial data granularity; geographic samples may be insufficient to capture fine spatial variation; lack of detailed environmental data may affect environmental transmission pathway analysis

Data Integration and Analysis Feasibility Considerations

Data Integration Strategies

  1. Geographic identifier matching
    • Method: Use h1.1-1.3 and f1.1-1.3 (district, commune, village identifiers) for spatial matching
    • Feasibility assessment: While exact matching of each household and farm may not be possible, community-level matching is feasible
    • Technical implementation: Use spatial join and geographic buffer analysis techniques for matching
    • Quality control: Design matching validation procedures to assess the reliability of different matching strategies
  2. Community aggregation methods
    • Method: Aggregate household and farm data at the community level, creating community-level indicators
    • Feasibility assessment: Completely feasible, suitable for exploring community-level associations
    • Technical implementation: Use appropriate weighting schemes to ensure representativeness; address aggregation issues for communities of different sizes
    • Innovation: Develop advanced aggregation indicators that consider intra-community heterogeneity, rather than simple means
  3. Creating composite indicators
    • Method: Construct cross-dataset composite indicators (such as antibiotic use intensity, knowledge levels)
    • Feasibility assessment: Feasible, but requires careful consideration of indicator construction rationality and comparability
    • Technical implementation: Adopt standardization and weighting schemes to ensure indicator validity; conduct sensitivity analysis to assess robustness
    • Theoretical framework: Use social ecological models to guide indicator construction, ensuring consistency between theory and measurement
  4. Multilevel data structure handling
    • Method: Consider the hierarchical structure of households/farms nested within communities
    • Feasibility assessment: Basic multilevel models are feasible, but complex models may be limited by sample size
    • Technical implementation: Use appropriate random effects structures to capture within-group correlations; assess the balance between model complexity and sample size
    • Alternative approach: If sample size is insufficient, use robust standard errors and confounder adjustment instead of full multilevel models

Methodological Challenges and Solutions

  1. Causal inference limitations
    • Challenge: Cross-sectional design limits causal relationship inference
    • Solution: Use directed acyclic graphs (DAGs) to clarify causal assumptions and possible confounding pathways
    • Technical implementation: Adopt propensity score methods, instrumental variables, and marginal structural models to strengthen causal inference
    • Transparency strategy: Clearly distinguish between association analysis and causal interpretation, maintaining appropriate caution in causal claims
  2. Matching household and farm data
    • Challenge: May lack direct identifiers linking households and farms
    • Solution: Develop multi-level matching algorithms combining geographic, social network, and demographic characteristics
    • Technical implementation: Use fuzzy matching techniques to handle imperfect matches; construct a matching quality scoring system
    • Data supplementation: Design small-scale follow-up surveys to collect key linking information to fill matching gaps
  3. Spatial data precision
    • Challenge: Geographic identifiers may not be precise enough for detailed spatial analysis
    • Solution: Use multi-resolution spatial analysis methods, making inferences at reliable spatial scales
    • Technical implementation: Apply geographic weighting techniques to adjust for insufficient spatial precision; use auxiliary geographic data to enhance spatial analysis
    • Result interpretation: Clarify the applicable scale of spatial analysis, avoiding over-interpretation of fine-grained spatial patterns
  4. Sample size limitations for complex models
    • Challenge: Sample size may be insufficient to support very complex multivariate or multilevel models
    • Solution: Use Bayesian frameworks to handle small sample situations; adopt regularization techniques to reduce overfitting risk
    • Technical implementation: Apply bias correction and prior information to enhance small sample inference; conduct statistical power analysis to guide analysis strategy
    • Model simplification: Develop theory-driven parameterization constraints to reduce the number of parameters that need to be estimated
  5. Recall bias
    • Challenge: Self-reported data may be affected by recall bias
    • Solution: Use multiple information sources to triangulate key variables; develop recall assistance techniques
    • Technical implementation: Design internal consistency checks to identify possible bias patterns; use anchoring events to improve retrospective reporting accuracy
    • Sensitivity analysis: Assess result robustness under different recall bias scenarios

Resource Requirements and Time Planning

  1. Data management and processing
    • Requirements: R/RStudio environment, spatial analysis software (such as QGIS), high-performance computing resources (cloud computing platform)
    • Estimated workload: Data cleaning and integration (2-3 weeks), preliminary exploratory analysis (1-2 weeks)
    • Technical challenges: Handling large spatial datasets; standardizing farm and household data formats; integrating multi-source data
    • Solutions: Use database management systems to optimize big data processing; develop automated data cleaning workflows
  2. Analysis implementation timeline
    • Phase 1: Descriptive and basic inferential analysis (4 weeks)
      • Basic characteristic analysis of household and farm data
      • Preliminary association exploration and hypothesis refinement
    • Phase 2: Advanced analysis (8 weeks)
      • Multivariate and multilevel modeling
      • Spatial analysis and network analysis
      • Machine learning model development
    • Phase 3: Comprehensive modeling and integration (6 weeks)
      • Cross-dataset model integration
      • Predictive model validation and optimization
      • Policy scenario simulation
    • Phase 4: Result interpretation and thesis writing (6 weeks)
      • Result visualization and interpretation
      • Integration with theoretical framework
      • Thesis writing and revision