Fig 1.
Schematic overview of transmission dynamics assumed in this study.
The population is structured into two demes: ’hospital’ and ’community’, denoted with red and blue colours, respectively. Demes are connected through patient admission and discharge, illustrated with type-change events along the lineages of the transmission tree. Furthermore, transmission events and sampling events can occur within subpopulations. The sampled transmission tree is presented in the middle of the figure. For the corresponding full transmission tree, see S1 Fig.
Table 1.
Key differences in phylodynamic model assumptions across simulation scenarios. Scenario abbreviations are as follows: HDT = hospital-driven transmission, ET = equal transmission, and CDT = community-driven transmission. All scenarios were simulated using the same values for the rate of becoming non-infectious (), hospital admission rate (
), and discharge rate (
). For phylodynamic inference, either a two-deme model (bdmm) or a one-deme model (bdsky) was applied. Genomic evolution in all scenarios was simulated using the HKY substitution model with κ = 4.04 and nucleotide frequencies A: 0.34, C: 0.16, G: 0.16, and T: 0.34. For each scenario, parameters of the sequence evolution model were inferred using default prior distributions during Bayesian analysis. Results for the main scenarios are provided in Tables 2, 3, 4, 5. Results for HDT (a) sensitivity 1 and 2 are provided in S1 Table.
Fig 2.
Examples of sampled transmission trees simulated based on hospital-driven transmission rates (scenarios HDT (a)–(c)).
Branch colors indicate the community (blue) and hospital (red). Tip labels denote the deme from which the sample was collected, with ’C’ indicating community and ’H’ indicating hospital. In all panels, the x-axis represents time, here with the unit being years. All trees shown were generated under the assumption that 0.1% of infected individuals from the community and 20% from the hospital were sampled.
Fig 3.
Examples of sampled transmission trees simulated based on equal and community-based transmission rates (scenarios ET and CDT, respectively).
Branch colors indicate the community (blue) and hospital (red). Tip labels denote the deme from which the sample was collected, with ‘C’ indicating community and ‘H’ indicating hospital. In both panels, the x-axis represents time, here with the unit being years. Both trees shown were generated under the assumption that 0.1% of infected individuals from the community and 20% from the hospital were sampled.
Table 2.
Summary of the Bayesian posterior inference results from each simulation scenario assuming a community sampling rate of . Scenario abbreviations are as follows: HDT = hospital-driven transmission, ET = equal transmission, and CDT = community-driven transmission. The column labelled
indicates the number of replicates (out of 100) in which all parameters had an effective sample size (ESS) of at least 200. Epidemiological parameters not listed in the third column (e.g., admission and discharge rates as well as sampling proportions) were fixed to their true values. The ‘Relative error‘ column reflects the average relative absolute deviation, calculated as
/truth. The ‘Relative bias‘ column shows the average relative deviation of the median estimate from the true value, calculated as
. Relative 95% highest posterior density (HPD) widths are computed as (upperbound – lowerbound)/truth. The ‘95% HPD accuracy‘ column indicates the number of replicates in which the 95% HPD interval included the true value for each parameter.
Table 3.
Summary of the Bayesian posterior inference results from each simulation scenario assuming a community sampling rate of . Scenario abbreviations are as follows: HDT = hospital-driven transmission, ET = equal transmission, and CDT = community-driven transmission. The column labelled
indicates the number of replicates (out of 100) in which all parameters had an effective sample size (ESS) of at least 200. Epidemiological parameters not listed in the third column (e.g., admission and discharge rates as well as sampling proportions) were fixed to their true values. The ‘Relative error‘ column reflects the average relative absolute deviation, calculated as
/truth. The ‘Relative bias‘ column shows the average relative deviation of the median estimate from the true value, calculated as
. Relative 95% highest posterior density (HPD) widths are computed as
. The ‘95% HPD accuracy’ column indicates the number of replicates in which the 95% HPD interval included the true value for each parameter.
Table 4.
Summary of the Bayesian posterior inference results from each simulation scenario assuming a community sampling rate of . Scenario abbreviations are as follows: HDT = hospital-driven transmission, ET = equal transmission, and CDT = community-driven transmission. The column labelled
indicates the number of replicates (out of 100) in which all parameters had an effective sample size (ESS) of at least 200. Epidemiological parameters not listed in the third column (e.g., admission and discharge rates as well as sampling proportions) were fixed to their true values. The ‘Relative error‘ column reflects the average relative absolute deviation, calculated as
/truth. The ‘Relative bias‘ column shows the average relative deviation of the median estimate from the true value, calculated as
. Relative 95% highest posterior density (HPD) widths are computed as
. The ‘95% HPD accuracy‘ column indicates the number of replicates in which the 95% HPD interval included the true value for each parameter. For scenario HDT (b) replicates 3136 and 3189 as well as for scenario HDT (c) replicates 3112 and 3152 were excluded (for details, see S3 Text).
Fig 4.
Estimated transmission rates for scenarios simulated under the assumption of hospital-driven transmission (HDT) and a community sampling proportion of SC = 0.001.
In scenarios HDT (a), HDT (b), and HDT (c) the hospital transmission rates were assumed to be lower, equal, or higher, respectively, than the assumed rate at which an infected individual is discharged from the hospital (i.e.,
=
= 45). Each bar represents the 95% HPD interval from a simulation replicate with a point denoting the median estimate. Horizontal lines indicate the true transmission rates within the community (blue) and the hospital (red). For clarity, simulation replicates are displayed in ascending order based on their mean values.
Fig 5.
Estimated transmission rates for scenario simulated under the assumption of equal transmission (ET) and a community sampling proportion of sC = 0.001.
Each bar represents the 95% HPD interval from a simulation replicate with a point denoting the median estimate. Horizontal lines indicate the true transmission rates within the community (blue) and the hospital (red). For clarity, simulation replicates are displayed in ascending order based on their mean values.
Fig 6.
Estimated transmission rates for scenario simulated under the assumption of community-driven transmission (CDT) and a community sampling proportion of sC = 0.001.
Each bar represents the 95% HPD interval from a simulation replicate with a point denoting the median estimate. Horizontal lines indicate the true transmission rates within the community (blue) and the hospital (red). For clarity, simulation replicates are displayed in ascending order based on their mean values.
Table 5.
Summary of the Bayesian posterior inference results from scenarios, where samples were obtained exclusively from the hospital. Scenario abbreviations are as follows: HDT = hospital-driven transmission, ET = equal transmission, and CDT = community-driven transmission. The column labelled indicates the number of replicates (out of 100) in which all parameters had an effective sample size (ESS) of at least 200. Epidemiological parameters not listed in the third column (e.g., admission and discharge rates) were fixed to their true values. Sampling proportions were fixed to their true values in all scenarios except CDT”. The ‘Relative error‘ column reflects the average relative absolute deviation, calculated as
/truth. The ‘Relative bias’ column shows the average relative deviation of the median estimate from the true value, calculated as
. Relative 95% highest posterior density (HPD) widths are computed as
. The ‘95% HPD accuracy’ column indicates the number of replicates in which the 95% HPD interval included the true value for each parameter.
Fig 7.
Estimated transmission rates for scenarios simulated under the assumption of sampling hospital cases only, i.e., scenarios HDT (a)’, ET, CDT’, and CDT”.
In scenario CDT’ for the inference the hospital sampling proportion was fixed to its true value (), whereas in scenario CDT”
was estimated. Each bar represents the 95% HPD interval from a simulation replicate with a point denoting the median estimate. Horizontal lines represent the true transmission rates. In scenario CDT”, the estimated hospital transmission rate aligns more closely with the true community rate (blue line) rather than the actual hospital rate (red line). For clarity, simulation replicates are displayed in ascending order based on their mean values.