contactmatrix {hhh4contacts}

POLYMOD Contact Matrices for Germany

Description

The function contactmatrix retrieves various social contact matrices for Germany from the POLYMOD survey (Mossong et al., 2008). Such a matrix contains the average numbers of reported contacts by participant age group. The original age groups (5-year intervals) can be joined together by the grouping argument, which first sums over contact groups (columns) and then averages over the corresponding participant groups (rows) using the corresponding age distribution as weights.

Usage

contactmatrix(
  which = c("corrected", "mossong", "reciprocal"),
  type = c("all", "physical"),
  grouping = c(1, 2, 2, 4, 4, 2),
  normalize = FALSE
)

contactmatrix_mossong

contactmatrix_mossong_physical

contactmatrix_POLYMOD

contactmatrix_POLYMOD_physical

contactmatrix_wallinga

contactmatrix_wallinga_physical

Arguments

which

character string indicating which contact matrix to return. "mossong" uses the average numbers of reported contacts as published in Table S5 of Mossong et al. (2008), available as contactmatrix_mossong or contactmatrix_mossong_physical. "corrected" (from contactmatrix_POLYMOD or contactmatrix_POLYMOD_physical) fixes an error in these numbers related to the age group 70+ (see the Examples) and is the default. If which="reciprocal" (corresponding to contactmatrix_wallinga or contactmatrix_wallinga_physical as used by Meyer and Held, 2017), the returned social contact matrix fulfils reciprocity of contacts with respect to the age distribution of Berlin, pop2011, via the method of Wallinga et al. (2006).

type

a character string to select the type of contacts to use: either "all" contacts, i.e., count both physical and pure conversational contacts, or only "physical" contacts.

grouping

specification of how to aggregate groups using aggregateC, i.e., a named list of vectors of group labels, or an integer vector of sequential group sizes. The "agedistri" attribute of the contact matrix is used as weights. If grouping=NULL, aggregation is skipped and the original (5-year) age groups are returned. The default setting produces the six age groups of Meyer and Held (2017).

normalize

a logical indicating whether to normalize the matrix such that each row sums to 1.

Format

The dataset contactmatrix_POLYMOD and its variants are all square numeric matrices with 15 rows (participants) and 15 columns (contacts), labelled with the corresponding age groups. There is an attribute "agedistri", a named numeric vector of length 15, which for the “_mossong_” and “_POLYMOD_” variants gives the age distribution of the German POLYMOD sample, and for the _wallinga_ variants gives the age distribution of Berlin, i.e., prop.table(colSums(pop2011)).

Value

a square numeric matrix containing the average numbers of contact persons recorded per day per survey participant in Germany, potentially averaged over multiple row (participant) age groups and aggregated over the corresponding column (contact) age groups.

Author(s)

Sebastian Meyer

Source

contactmatrix_mossong and contactmatrix_mossong_physical are taken from the Supporting Information in Mossong et al. (2008): the matrices from Table S5 (8.2), and the attached age distribution from Table S2 (3.2).

The corrected versions contactmatrix_POLYMOD and contactmatrix_POLYMOD_physical were constructed from the raw POLYMOD data initially made available at https://www.researchgate.net/publication/232701632_POLYMOD_contact_survey_for_researchers (a reformatted and better documented version is nowadays available at doi:10.5281/zenodo.1043437). The reciprocal contact matrices contactmatrix_wallinga and contactmatrix_wallinga_physical were estimated from these raw data via the method of Wallinga et al. (2006).

References

Meyer S and Held L (2017): Incorporating social contact data in spatio-temporal models for infectious disease spread. Biostatistics, 18 (2), 338-351. doi:10.1093/biostatistics/kxw051

Mossong et al. (2008): Social contacts and mixing patterns relevant to the spread of infectious diseases. PLoS Medicine, 5 (3), e74. doi:10.1371/journal.pmed.0050074

Wallinga J, Teunis P and Kretzschmar M (2006): Using data on social contacts to estimate age-specific transmission parameters for respiratory-spread infectious agents. American Journal of Epidemiology, 164 (10), 936-944. doi:10.1093/aje/kwj317

Examples

## contact matrix reported in Mossong et al (2008, Table S5)
(C_original <- contactmatrix(which = "mossong", grouping = NULL))
## this simply returns the dataset 'contactmatrix_mossong'
stopifnot(identical(C_original, contactmatrix_mossong))

## with corrected numbers for the 70+ age group (the default)
C_corrected <- contactmatrix(which = "corrected", grouping = NULL)
## this simply returns the dataset 'contactmatrix_POLYMOD'
stopifnot(identical(C_corrected, contactmatrix_POLYMOD))

## check for differences
C_original == round(C_corrected, 2)
## compare entries of last row and last column
round(rbind(original = C_original[15,], corrected = C_corrected[15,]), 2)
round(cbind(original = C_original[,15], corrected = C_corrected[,15]), 2)

## contact matrix estimated to be reciprocal on the population level
C_reciprocal <- contactmatrix(which = "reciprocal", grouping = NULL)
## this simply returns the dataset 'contactmatrix_wallinga'
## (without its "overdisp" attribute)
stopifnot(all.equal(C_reciprocal, contactmatrix_wallinga, check.attributes=FALSE))

## check reciprocity
agedistriBE <- attr(C_reciprocal, "agedistri")
stopifnot(identical(agedistriBE, prop.table(colSums(pop2011))))
stopifnot(isSymmetric(C_reciprocal * agedistriBE, check.attributes=FALSE))

## visually compare raw to reciprocal contact matrix
if (require("gridExtra"))
    grid.arrange(plotC(C_corrected, main = "raw"),
                 plotC(C_reciprocal, main = "reciprocal"),
                 nrow = 1)

## select physical contacts and aggregate into 5 age groups
contactmatrix(type = "physical", grouping = c(1, 2, 7, 3, 2))

## the default 6 age groups, normalized to a transition matrix
contactmatrix(normalize = TRUE)

## reciprocity also holds for this grouping
(C6 <- contactmatrix(which = "reciprocal"))
stopifnot(isSymmetric(C6 * attr(C6, "agedistri"), check.attributes=FALSE))