Supplementary Materials

Estimating epidemiologic dynamics from cross-sectional viral load distributions

James A. Hay, Lee Kennedy-Shaffer, Sanjat Kanjilal, Niall J. Lennon, Stacey B. Gabriel, Marc Lipsitch, Michael J. Mina

Materials/Methods, Supplementary Text, Tables, Figures, and/or References

Download Supplement
  • Materials and Methods
  • Figs. S1 to S19
  • Table S1
  • Captions for Movies S1 to S3
  • Captions for Data S1 and S2
  • References
MDAR Reproducibility Checklist
Data S1
Ct values and collection dates for samples obtained from four long-term care facilities in Massachusetts. Variables included are anonymized facility identifier, collection date and week, RT-qPCR result, anonymized unique identifier, and RP (control), N2 and N1 Ct values.
Data S2
ORF1ab Ct values and collection dates for samples obtained from the Brigham & Women's Hospital, Massachusetts.

Images, Video, and Other Media

Movie S1
Multiple cross-sections of cycle threshold (Ct) values can be combined to improve the estimation of the epidemic trajectory over time. Animation of epidemic trajectory estimation (Bottom) using the Gaussian process GP model fit repeatedly to weekly samples of observed cycle threshold (Ct) value data (Top) in an ongoing simulated epidemic. 2000 samples were randomly taken from the population each week. The red line indicates the true daily per capita incidence of the simulated data. The blue line and ribbons show the posterior median, 50% (dark blue) and 95% (light blue) credible intervals for the estimated daily per capita incidence curve. Dashed vertical lines show time of sample collection. Incidence was estimated back to 35 days prior to the first sample time.
Movie S2
Estimated epidemic trajectories using multiple cross-sections of cycle threshold (Ct) values at different weekly sample sizes. Using the same simulation and Gaussian process (GP) model as in Movie S1, the animation shows the results estimating the underlying per capita incidence curve each week using no data (prior only), 50, 200, 500, 1000 or 2000 Ct values (including negative samples) obtained each week through randomly sampling the population. Dashed vertical lines show time of sample collection. The blue line and ribbons show the posterior median, 50% (dark blue) and 95% (light blue) credible intervals for the estimated daily per capita incidence curve. Incidence was estimated back to 35 days prior to the first sample time.
Movie S3
Multiple cross-sections of cycle threshold (Ct) values can estimate the underlying incidence curve when sampling is initiated partway through the epidemic. Using the same simulation, sampling scheme and Gaussian process (GP) model as in Movie S1, but with sampling commencing part way through the epidemic. (Top) cycle threshold (Ct) data used for the simulation. (Bottom) the red line indicates the true daily per capita incidence of the simulated data. The blue line and ribbons show the posterior median, 50% (dark blue) and 95% (light blue) credible intervals for the estimated daily per capita incidence curve. Dashed vertical lines show time of sample collection. Incidence was estimated back to 35 days prior to the first sample time.