Epidemic-induced local awareness behavior inferred from surveys and genetic sequence data

TL;DR

The proposed containment score was correlated positively with policy stringency during the COVID-19 pandemic, and validated as a proxy for local awareness in simulation experiments, and brought important insight into the field of awareness modeling through the analysis of large-scale genetic sequence data.

Abstract

Behavior-disease models suggest that pandemics can be contained cost-effectively if individuals take preventive actions when disease prevalence rises among their close contacts. However, assessing local awareness behavior in real-world datasets remains a challenge. Through the analysis of mutation patterns in clinical genetic sequence data, we propose an efficient approach to quantify the impact of local awareness by identifying superspreading events and assigning containment scores to them. We validate the proposed containment score as a proxy for local awareness in simulation experiments, and find that it was correlated positively with policy stringency during the COVID-19 pandemic. Finally, we observe a temporary drop in the containment score during the Omicron wave in the United Kingdom, matching a survey experiment we carried out in Hungary during the corresponding period of the pandemic. Our findings bring important insight into the field of awareness modeling through the analysis of large-scale genetic sequence data, one of the most promising data sources in epidemics research.

Figures (14)

• Figure 1: (a) The MASZK Hungarian telephone survey, with 1000 participants in each of the 9 months, shows that the mean local awareness score (in blue) remains relatively constant throughout the recording period, except during the Omicron wave, when the score drops. The government-imposed preventive measures (mask wearing, in yellow, and social distancing, in green) show a different temporal pattern. For all survey results, we show the mean response for each month, with confidence intervals calculated under the assumption of normality. The daily number of cases (with a rolling-mean of 7 days, normalized by 170) are shown in red. Source data are available in Supplementary Data 1. (b) Our proposed pipelines to generate synthetic (blue) and process real genetic sequence data (grey) to compute collision clusters, superspreading events (SSEs), and finally Event Containment Scores (ECSs) -- a proxy measure for local awareness behavior.
• Figure 2: (a) Bar plot showing the number of SARS-CoV-2 genetic sequences collected in Belgium and shared through the GISAID platform over time for the Delta and the early Omicron variants, with the dashed lines marking the weeks when a new variant became dominant. The solid red line depicts the number of reported cases. (b) Visualization of the size of 7 collision clusters in Belgium over time. Within these 7 clusters, our proposed thresholding approach detected 4 superspreading events shown with square markers (often at the beginning of a cluster). The color of the squares marks the sign of the containment scores.
• Figure 3: Event Containment Scores and their median values (MECS) computed on genetic sequence data generated from simulated epidemics on synthetic and real networks as a function of (a) the local, (b) the global awareness function parameter, (c) the infection probability and (d) the subsampling probability. For each set of parameters, we simulated $n=200$ independent epidemic processes with different random seeds. Colored intervals show the 25th and 75th percentiles of the ECS values, while black intervals indicate confidence intervals for the median, computed using a normal approximation. Source data are available in Supplementary Data 2. When not stated otherwise, all parameters are set to be their default values $\alpha_l=0$, $\alpha_g=0$, $\beta_0=0.15$ and $p=1$. We observe positive MECS values in case of local awareness, and noisy MECS values near zero if the subsampling probability is low.
• Figure 4: Event Containment Scores (ECS, blue) and Containment Health Index (CHI, green) in European countries with at least 15 detected superspreading events in the (a) Delta, (b) Omicron BA.1 variants, and (c) when all Omicron variants are merged. Bar plots and black dots mark median values. The number of ECS values corresponding to each Median ECS (MECS) value is shown in Supplementary Table B.1. Colored intervals show the 25th and 75th percentiles of the distribution, while black intervals indicate confidence intervals for the median, computed using a normal approximation. Country-variant pairs with a confidence interval larger than 3 around the MECS values are filtered out. Grey background signifies a statistically significant correlation between MECS and the median CHI values (Table \ref{['tab:sprearman']}).
• Figure 5: (a) Median Event Containment Scores (MECS) during the Delta and the Omicron BA.1 variants as computed in Figure \ref{['fig:results_comb']}. Datapoints below the dashed ($x=y$) line hint at drops in local awareness during Omicron BA.1 variant. (b) Containment Health Index (CHI) during the Delta and the Omicron BA.1 variants as computed in Figure \ref{['fig:results_comb']}. (c) MECS values computed biweekly with a 4-week sliding window in the UK for the Delta, Omicron BA.1 and BA.2 variants. Confidence intervals were computed using a normal approximation, and datapoints with a confidence interval larger than 2 are filtered out. We observe a drop in MECS in December 2021 - January 2022 during the Omicron BA.1 wave.