Quantifying population contact patterns in the United States during the COVID-19 pandemic

Abstract

SARS-CoV-2 is transmitted primarily through close, person-to-person interactions. Physical distancing policies can control the spread of SARS-CoV-2 by reducing the amount of these interactions in a population. Here, we report results from four waves of contact surveys designed to quantify the impact of these policies during the COVID-19 pandemic in the United States. We surveyed 9,743 respondents between March 22 and September 26, 2020. We find that interpersonal contact has been dramatically reduced in the US, with an 82% (95%CI: 80%-83%) reduction in the average number of daily contacts observed during the first wave compared to pre-pandemic levels. However, we find increases in contact rates over the subsequent waves. We also find that certain demographic groups, including people under 45 and males, have significantly higher contact rates than the rest of the population. Tracking these changes can provide rapid assessments of the impact of physical distancing policies and help to identify at-risk populations.

Fig. 1. Reported interpersonal contact across four survey waves.

a, b Histograms of reported number of contacts (a) and non-household contacts (b) among respondents for each wave. Reported contacts are topcoded at 10 in these plots. The vertical lines show the median number of contacts. c, d Estimated average number of non-household contacts each person reported to have taken place by contact’s relationship (c) and location (d) based on n = 29,880 reports about detailed contacts (“Methods”). Uncertainty estimates are 95% intervals derived from the bootstrap. Each point shows estimated average numbers of non-household contacts in each category, per person. For example, Panel c shows that the average respondent reported almost 0.8 non-household contacts with family members in Wave 2. Avg. average.

Fig. 2. Conditional effect plots showing the predicted mean number of non-household contacts and 95% posterior credible intervals for several covariates.

Predicted mean number of non-household contacts is shown for (a) day of the week; (b) household size; (c) race/ethnicity; (d) age/sex group; and (e) geography. Predictions come from a negative binomial model fit to reported numbers of non-household contacts made by n = 9743 survey respondents. Colors are used in panels c–e to show estimated interactions. Covariate values not being manipulated in each panel are set to values for a white female aged 35–44 from the national sample who lives in a two-person household during a weekday in wave 3 (“Methods”). Uncertainty bars show 95% posterior credible intervals. Supplementary Fig. 2 shows the same predictions for an analogous model fit to all contacts.

Fig. 3. Comparison of age-structured contact matrices with baseline.

a–d Age-structured contact matrices from the four BICS waves after adjusting for the age distribution of survey respondents and the reciprocal nature of contacts; lighter colors indicate higher number of average daily contacts. e–h Difference in the average number of contacts between the 2015 study and the four BICS waves; lighter colors indicate a larger absolute difference between the 2015 study and the BICS data. i–l Average number of reported contacts for each respondent age group for the BICS data (darker color) compared to the 2015 study (lighter color), along with 95% confidence intervals derived from the bootstrap. The BICS estimates are based on n = 3163 in Wave 0, n = 7473 in Wave 1, n = 7842 in Wave 2, and n = 11,402 in Wave 3 reported contacts; The 2015 study estimates are based on n = 5944 reported contacts. Top row shows BICS Wave 0; second row shows BICS Wave 1; third row shows BICS Wave 2; and bottom row shows BICS Wave 3.

Fig. 4. Implied R₀ estimates for each wave.

The implied R₀ from the BICS contact matrices for each wave relative to two baseline contact matrices from the 2015 study and the UK POLYMOD study, and assuming a baseline R₀ value drawn from a normal distribution with mean 2.5 and standard deviation of 0.54. Circles indicate R₀ estimates calculated from age-structured contact matrices for all reported contacts (n = 3163 in Wave 0, n = 7473 in Wave 1, n = 7842 in Wave 2 and n = 11, 402 in Wave 3); diamonds indicate R₀ estimates calculated from age-structured contact matrices for contacts where no mask usage was reported (n = 5777 in Wave 1, n = 5818 in Wave 2 and n = 7583 in Wave 3). Ninety-five percent confidence intervals were derived from the bootstrap. FB: 2015 Facebook survey.

Fig. 5. Characteristics of survey respondents.

We use calibration weights to improve the representativeness of our sample. Each facet shows the unweighted (red) and calibration weighted (blue) composition of survey respondents for a given covariate.