Adding a New Country to epicR

Introduction

EPIC (Evaluation Platform in COPD) was originally developed for the Canadian healthcare context. However, the model’s architecture supports adaptation to different countries and healthcare systems through jurisdiction-specific configuration files. This vignette explains how to add a new country to epicR.

Overview of the Configuration System

epicR uses JSON configuration files to store country-specific parameters. These files contain all the epidemiological, demographic, economic, and healthcare system parameters needed to run the model for a specific jurisdiction.

Current Configuration Files

• inst/config/config_canada.json - Fully configured Canadian parameters
• inst/config/config_us.json - US template with placeholders (requires configuration)

Step-by-Step Guide to Adding a New Country

Step 1: Create the Configuration File

Create a new JSON configuration file for your country in the inst/config/ directory:

# Example: Adding Germany
file_name <- "inst/config/config_germany.json"

Step 2: Copy the Template Structure

Start by copying the structure from an existing configuration file. You can use the US template as a starting point:

# Read the US template
us_config <- jsonlite::fromJSON("inst/config/config_us.json")

# Modify for Germany
germany_config <- us_config
germany_config$jurisdiction <- "germany"

# Save the template
jsonlite::write_json(germany_config, "inst/config/config_germany.json", 
                     pretty = TRUE, auto_unbox = TRUE)

Step 3: Parameter Categories to Configure

The configuration file contains several major parameter categories that need country-specific data:

3.1 Global Parameters

{
  "global_parameters": {
    "age0": 40,
    "time_horizon": 20,
    "discount_cost": 0.03,
    "discount_qaly": 0.03,
    "closed_cohort": 0
  }
}

Required data: - Discount rates for costs and QALYs (country-specific economic guidelines)

3.2 Demographics and Population

{
  "agent": {
    "p_female": 0.51,
    "height_0_betas": [...],
    "weight_0_betas": [...],
    "p_prevalence_age": [...],
    "p_bgd_by_sex": {
      "male": [...],
      "female": [...]
    }
  }
}

Required data sources: - Population demographics: National statistics office - Age pyramid: Current population by age group - Life tables: Age and sex-specific mortality rates - Anthropometric data: Height and weight distributions by age and sex

3.3 Smoking Patterns

{
  "smoking": {
    "logit_p_current_smoker_0_betas": [...],
    "minimum_smoking_prevalence": 0.12,
    "mortality_factor_current": [...],
    "mortality_factor_former": [...]
  }
}

Required data sources: - Smoking prevalence: National health surveys - Smoking-related mortality: Meta-analyses or national studies - Smoking cessation rates: Longitudinal studies

3.4 COPD Epidemiology

{
  "COPD": {
    "logit_p_COPD_betas_by_sex": {
      "male": [...],
      "female": [...]
    },
    "ln_h_COPD_betas_by_sex": {
      "male": [...],
      "female": [...]
    }
  }
}

Required data sources: - COPD prevalence: Spirometry-based population studies - COPD incidence: Longitudinal cohort studies - Risk factors: Smoking, age, sex associations

3.5 Healthcare Costs

{
  "cost": {
    "bg_cost_by_stage": [...],
    "exac_dcost": [...],
    "cost_gp_visit": 85.50,
    "cost_outpatient_diagnosis": 125.50,
    "cost_smoking_cessation": 485.25
  }
}

Required data sources: - Healthcare unit costs: National fee schedules or health economics studies - COPD treatment costs: Health administrative data or costing studies - Currency: Convert to local currency or standardize to USD/EUR

3.6 Healthcare Utilization

{
  "outpatient": {
    "ln_rate_gpvisits_COPD_by_sex": {
      "male": [...],
      "female": [...]
    }
  }
}

Required data sources: - Healthcare utilization patterns: Administrative health data - GP visit rates: Primary care databases

Step 4: Data Collection Strategy

Essential Data Sources

1. National Statistics Offices
   • Population demographics
   • Life tables
   • Health surveys
2. Health Administrative Databases
   • Healthcare utilization
   • Treatment costs
   • Disease prevalence
3. Published Literature
   • Disease-specific studies
   • Health economics evaluations
   • Epidemiological studies
4. International Databases
   • WHO Global Health Observatory
   • OECD Health Statistics
   • Global Burden of Disease Study

Data Quality Considerations

• Representativeness: Ensure data represents the target population
• Recency: Use the most recent available data
• Consistency: Maintain consistent definitions across parameters
• Validation: Cross-check with multiple sources when possible

Step 5: Parameter Estimation

When direct data is not available, parameters can be estimated using:

5.1 Regression Models

# Example: Estimating COPD prevalence coefficients
# Using logistic regression on survey data
model <- glm(copd ~ age + sex + smoking_status + pack_years, 
             data = survey_data, family = binomial())
coefficients(model)

5.2 Literature Meta-analysis

• Systematic review of published studies
• Meta-analytic techniques to pool estimates
• Adjustment for population differences

5.3 Calibration

• Use model calibration to match observed outcomes
• Iterative adjustment of parameters
• Validation against external data sources

Step 6: Testing the New Configuration

Once you have populated the configuration file:

6.1 Test Loading

# Test that the configuration loads without errors
library(epicR)
input <- get_input(jurisdiction = "germany")

6.2 Run Basic Simulation

# Run a small simulation to check for errors
results <- simulate(
  jurisdiction = "germany",
  n_agents = 10000,
  time_horizon = 5
)
print(results$basic)

6.3 Validate Outputs

• Compare population demographics to national statistics
• Check COPD prevalence against published estimates
• Validate healthcare utilization patterns

Step 7: Documentation

Create documentation for your new country configuration:

7.1 Data Sources Documentation

# Create a data sources file
data_sources <- list(
  demographics = "German Federal Statistical Office, 2023",
  copd_prevalence = "BOLD Study Germany, 2022",
  healthcare_costs = "German Health Economics Association, 2023"
)

7.2 Parameter Assumptions

Document any assumptions or approximations made during parameter estimation.

Example: Adding Germany

Here’s a simplified example of adding Germany to epicR:

# 1. Create base configuration
germany_config <- list(
  jurisdiction = "germany",
  global_parameters = list(
    age0 = 40,
    time_horizon = 20,
    discount_cost = 0.03,  # German health economics guidelines
    discount_qaly = 0.03,
    closed_cohort = 0
  ),
  agent = list(
    p_female = 0.507,  # German Federal Statistical Office 2023
    # ... other parameters
  ),
  cost = list(
    cost_gp_visit = 25.50,  # German fee schedule 2023
    cost_outpatient_diagnosis = 85.40,
    # ... other costs
  )
  # ... other parameter categories
)

# 2. Save configuration
jsonlite::write_json(germany_config, 
                     "inst/config/config_germany.json", 
                     pretty = TRUE, auto_unbox = TRUE)

# 3. Test the configuration
library(epicR)
input <- get_input(jurisdiction = "germany")

Quality Assurance

Validation Checklist

• All placeholder values replaced with actual data
• Data sources documented
• Parameter ranges are clinically plausible
• Model outputs match expected population patterns
• Healthcare costs reflect local context
• Currency and units are consistent

Common Issues and Solutions

1. Missing Data: Use proxy data from similar countries or systematic reviews
2. Currency Conversion: Use purchasing power parity adjustments
3. Different Healthcare Systems: Adapt utilization patterns to local context
4. Data Quality: Document limitations and uncertainty

Contributing Your Configuration

If you’ve successfully created a configuration for a new country:

1. Validate thoroughly using local data
2. Document data sources and methods
3. Consider contributing to the epicR project
4. Share with the community for peer review

Conclusion

Adding a new country to epicR requires substantial data collection and parameter estimation, but the modular configuration system makes this process systematic and reproducible. The key is to ensure high-quality, country-specific data while maintaining the model’s scientific rigor.

For questions or assistance with adding a new country, consider: - Reviewing published adaptations of EPIC - Consulting with local health economists - Engaging with the epicR development community - Collaborating with researchers who have local data access