Social contact patterns relevant for infectious disease transmission in Cambodia

Abstract

Social mixing patterns are key determinants of infectious disease transmission. Mathematical models parameterised with empirical data from contact pattern surveys have played an important role in understanding epidemic dynamics and informing control strategies, including for SARS-CoV-2. However, there is a paucity of data on social mixing patterns in many settings. We conducted a community-based survey in Cambodia in 2012 to characterise mixing patterns and generate setting-specific contact matrices according to age and urban/rural populations. Data were collected using a diary-based approach from 2016 participants, selected by stratified random sampling. Contact patterns were highly age-assortative, with clear intergenerational mixing between household members. Both home and school were high-intensity contact settings, with 27.7% of contacts occurring at home with non-household members. Social mixing patterns differed between rural and urban residents; rural participants tended to have more intergenerational mixing, and a higher number of contacts outside of home, work or school. Participants had low spatial mobility, with 88% of contacts occurring within 1 km of the participants' homes. These data broaden the evidence-base on social mixing patterns in low and middle-income countries and Southeast Asia, and highlight within-country heterogeneities which may be important to consider when modelling the dynamics of pathogens transmitted via close contact.

Figure 1

Map of the six study districts within Cambodia. Orange = urban, green = rural. (1) Chamka Morn, Phnom Penh municipality; (2) Kandal Stung, Kandal Province; (3) Krong Kracheh, Kratie Province; (4) Snuol, Kratie Province; (5) Krong Kampot, Kampot Province; (6) Chhuk, Kampot Province. The borders of included provinces are marked with heavy black lines. Within each selected district, three sub-administrative divisions (‘sangkats’—urban; ‘communes’—rural) were randomly selected.

Figure 2

Contact matrices stratified by urban and rural (participant residence locations, (A–F), and by social setting (B–F). Each cell in a given matrix shows the mean number of contacts made between two age groups. All matrices are weighted by day of the week (weekday or weekend) and employment status (non-working age, employed, or unemployed). Matrices were further adjusted to account for recipricocity of contacts such that the total number of contacts that age group (i) made with age group (j), were equal to the total number of contacts age group (j) made with age group (i) (c_ijN_i = c_jiN_j; where N is the population size of an age group). Home (hh) = contacts at home with household members; Home(non-hh) = contacts at home with non-household members. Contact matrices which included only contacts that were physical or over 15 min in duration (which accounted for 81.2% of all diary-reported contacts) exhibited the same patterns of age-specific mixing (Fig. S1A–F).

Figure 3

The proportion of contacts reported in each setting according to participant age (A), urban or rural residence (B), sex (C), day of the week (D), and occupational group (E). ‘Other’ settings refer to contacts made in settings other than home, school, work, or multiple. ‘Multiple’ refers to multiple settings—for example, home and school.

Figure 4

Intensity of contacts, as measured by duration (A–C) and proportion physical (D–F). Results shown by participant age, social setting (HH at home with household members, S school, M multiple settings, W work, L leisure, HN at home with non-household members, O other settings, T transport) and urban/rural residence.

Figure 5

Frequency of travel outside of each administrative division by age group and urban/rural residence.