All Publications

  • The interplay of policy, behavior, and socioeconomic conditions in early COVID-19 epidemiology in Georgia. medRxiv : the preprint server for health sciences Harris, M. J., Tessier-Lavigne, E., Mordecai, E. A. 2021


    To investigate the impact of local public health orders, behavior, and population factors on early epidemic dynamics, we investigated variation among counties in the U.S. state of Georgia. We conducted regressions to identify predictors of (1) local public health orders, (2) mobility as a proxy for behavior, and (3) epidemiological outcomes (i.e., cases and deaths). We used an event study to determine whether social distancing and shelter-in-place orders caused a change in mobility. Counties at greater risk for large early outbreaks (i.e., larger populations and earlier first cases) were more likely to introduce local public health orders. Social distancing orders gradually reduced mobility by 19% ten days after their introduction, and lower mobility was associated with fewer cases and deaths. Air pollution and population size were predictors of cases and deaths, while larger elderly or Black population were predictors of lower mobility and greater cases, suggesting self-protective behavior in vulnerable populations. Early epidemiological outcomes reflected responses to policy orders and existing health and socioeconomic disparities related to disease vulnerability and ability to socially distance. Teasing apart the impact of behavior changes and population factors is difficult because the epidemic is embedded in a complex social system with multiple potential feedbacks.

    View details for DOI 10.1101/2021.03.24.21254256

    View details for PubMedID 33791739

    View details for PubMedCentralID PMC8010771

  • Real-time, interactive website for US-county-level COVID-19 event risk assessment. Nature human behaviour Chande, A., Lee, S., Harris, M., Nguyen, Q., Beckett, S. J., Hilley, T., Andris, C., Weitz, J. S. 2020


    Large events and gatherings, particularly those taking place indoors, have been linked to multitransmission events that have accelerated the pandemic spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). To provide real-time, geolocalized risk information, we developed an interactive online dashboard that estimates the risk that at least one individual with SARS-CoV-2 is present in gatherings of different sizes in the United States. The website combines documented case reports at the county level with ascertainment bias information obtained via population-wide serological surveys to estimate real-time circulating, per-capita infection rates. These rates are updated daily as a means to visualize the risk associated with gatherings, including county maps and state-level plots. The website provides data-driven information to help individuals and policy makers make prudent decisions (for example, increasing mask-wearing compliance and avoiding larger gatherings) that could help control the spread of SARS-CoV-2, particularly in hard-hit regions.

    View details for DOI 10.1038/s41562-020-01000-9

    View details for PubMedID 33168955

  • Early warning signals of malaria resurgence in Kericho, Kenya. Biology letters Harris, M. J., Hay, S. I., Drake, J. M. 2020; 16 (3): 20190713


    Campaigns to eliminate infectious diseases could be greatly aided by methods for providing early warning signals of resurgence. Theory predicts that as a disease transmission system undergoes a transition from stability at the disease-free equilibrium to sustained transmission, it will exhibit characteristic behaviours known as critical slowing down, referring to the speed at which fluctuations in the number of cases are dampened, for instance the extinction of a local transmission chain after infection from an imported case. These phenomena include increases in several summary statistics, including lag-1 autocorrelation, variance and the first difference of variance. Here, we report the first empirical test of this prediction during the resurgence of malaria in Kericho, Kenya. For 10 summary statistics, we measured the approach to criticality in a rolling window to quantify the size of effect and directions. Nine of the statistics increased as predicted and variance, the first difference of variance, autocovariance, lag-1 autocorrelation and decay time returned early warning signals of critical slowing down based on permutation tests. These results show that time series of disease incidence collected through ordinary surveillance activities may exhibit characteristic signatures prior to an outbreak, a phenomenon that may be quite general among infectious disease systems.

    View details for DOI 10.1098/rsbl.2019.0713

    View details for PubMedID 32183637

  • Climate drives spatial variation in Zika epidemics in Latin America. Proceedings. Biological sciences Harris, M. n., Caldwell, J. M., Mordecai, E. A. 2019; 286 (1909): 20191578


    Between 2015 and 2017, Zika virus spread rapidly through populations in the Americas with no prior exposure to the disease. Although climate is a known determinant of many Aedes-transmitted diseases, it is currently unclear whether climate was a major driver of the Zika epidemic and how climate might have differentially impacted outbreak intensity across locations within Latin America. Here, we estimated force of infection for Zika over time and across provinces in Latin America using a time-varying susceptible-infectious-recovered model. Climate factors explained less than 5% of the variation in weekly transmission intensity in a spatio-temporal model of force of infection by province over time, suggesting that week to week transmission within provinces may be too stochastic to predict. By contrast, climate and population factors were highly predictive of spatial variation in the presence and intensity of Zika transmission among provinces, with pseudo-R2 values between 0.33 and 0.60. Temperature, temperature range, rainfall and population size were the most important predictors of where Zika transmission occurred, while rainfall, relative humidity and a nonlinear effect of temperature were the best predictors of Zika intensity and burden. Surprisingly, force of infection was greatest in locations with temperatures near 24°C, much lower than previous estimates from mechanistic models, potentially suggesting that existing vector control programmes and/or prior exposure to other mosquito-borne diseases may have limited transmission in locations most suitable for Aedes aegypti, the main vector of Zika, dengue and chikungunya viruses in Latin America.

    View details for DOI 10.1098/rspb.2019.1578

    View details for PubMedID 31455188