This repository contains files that can be used to reconstruct birth-year specific probabilities of imprinting to influenza A, but subtype or by group, following the methods of (Gostic et al. 2016).

Imprinting probabilities are specific to the country and year of observation:

• COUNTRY - For years 1997-present, we pull data from WHO Flu Mart to estimate the fraction of influenza A infections caused by subyptes H1N1 and H3N2. We use country-specific data whenever there are 30 or more influenza A samples available in the country and year of interest. If sample sizes are insufficient, we pull in data from the country's WHO region.
• YEAR OF OBSERVATION - We asume children <13 years of age can be naive to influenza. The year of observation affects which birth years can be naive. In birth years >13 years of age, we assume that everyone has been infected at least one, and normalize so that subtype-specific imprinting probabilities sum to one (Gostic et al. 2016).

Data

Fraction of influenza A infections caused by A/H1N1, A/H2N2, and A/H3N2 over time

• From 1918-1976, only one influenza A subtype circulated at a time, and pandemic years mark the transitions between subtypes. We assume that H1N1 caused 100% of influenza A cases from 1918-1956, that H2N2 causes 100% of influenza A cases from 1957-1967, and that H3N2 caused 100% of influenza A cases from 1968-1976.

• From 1977-present, A/H3N2 and A/H1N1 have both caused seasonal epidemics. The fraction of influenza cases caused by either subtype varies from year to year. Data avilability has improved over time:

  • From 1977-1996, data on the relative dominance of A/H1N1 and A/H3N2 are limited. We use United States-specific data from Table 1 of Thompson et al. JAMA, 2003 to estimate the fraction of flu A cases caused by each subtype. These data are stored in processed-data/Thompson_data.csv.
  • From 1997-present, we use country or region-specific surveillance data from WHO Flu Mart. We use country-specific data when more than 30 flu A specimens were reported in the year of interest, and region-specific data otherwise.

Relative intensity of influenza circulation over time

• TO WRITE

Workflow

All code runs in the /R/ directory.

• get_imprinting_probabilities() is the main function. It inputs a vector of observation years, a vector of country names, and the maximum year it outputs a matrix of birth year-specific imprinting probabilities for each country and observation year. script_calculate_imprinting_probs.R is a script that shows example calls to get_imprinting_probabilities()

Code

• calculation_funs.R contains functions used to calculate imprinting probabilities.
• data_import_funs.R contains functions used to import country or region-specific data on the fraction of infections caused by difference influenza A subtypes over time.

Developer notes

• raw-data/ contains influenza surveillance data for each WHO region, downloaded from WHO Flu Mart.
• processed-data/ contains data on the scaled annual intensity of influenza circulation (Gostic et al. 2016), data from Table 1 of Thompson et al. JAMA, 2003, and a table of WHO regions and countries obtained here.

To do

[x] Update readme

[x] Finish downloading data for all WHO regions

[] Check country and region names

[] Unit tests

[] System tests

[] Plan for systematic data updates (can we automate this?)

[] Plan output filestructure and format.

[] Update Intensities in data and readme.