Fig 1.
Both boxes shaded blue represent an infectious patient.
Table 1.
Transmission model parameters and priors.
For parameters constrained to be positive, normal priors were truncated at zero. In addition to the listed parameters, augmented data included the unobserved transmission tree, consisting of infection times and sources denoted T and the set of unobserved recovery times denoted R.
Table 2.
Probability of infection by a given source patient or background source.
Fig 2.
Each box represents an infected patient (blue: patient j, yellow: patient i). Time runs from the bottom of the plot to the top. Samples are shown as circles. t0 denotes the time of the first sample, t1 the time of the second sample. t is the time between the samples, and u the time between the first sample and the common ancestor of the samples.
Table 3.
Genetic model parameters and priors.
For parameters constrained to be positive, normal priors were truncated at zero.
Table 4.
Simulated scenarios for a hospital with 4 wards, serving a population of 6000 patients at risk of admission with a daily probability of admission of 0.002 and a mean length of stay of 5 days. Simulation run for 365 days. Sampling distribution parameters, μ = 10, size = 5; Recovery distribution parameters, μ = 90, size = 3. Assuming 1 mutation per genome per year, Npop = 20000 and Ne = 22.5. All patients assumed to start susceptible.
Fig 3.
Source type and individual source attribution in simulated data.
The left-hand columns (panels A and C) display the accuracy of inferences about the source type, i.e. distinguishing between hospital background, community background, ward, spore and hospital-wide transmission. The right-hand columns (panels B and D) display the accuracy of inferences about the specific patient source of each infection (for background source types this is an unknown source). Panels A and B display inferences from the full transmission model. Panels C and D display the accuracy of predictions running an inference model without using genetic data (circles) or based on a heuristic rule (squares). CI, credibility interval.
Fig 4.
Comparison of source type attribution using a stochastic transmission model and heuristic rules.
Fig 5.
Proportion of cases acquired from each transmission route, by sequence type.
Panels A to C show the estimated proportion of cases acquired from each of the 3 major source types, known patients (A, blue), hospital background sources (B, green) and the community (C, gold). Panels D to F subdivide the cases shown in panel A, into those acquired by direct ward contact (D), ward-based spores (E) and hospital-wide transmission (F). Note the y-axis scale differs between the upper panels and lower panels, to allow for comparison of the relative contribution of ward, spore and hospital-wide routes to transmission from other cases, and hospital vs. community for background transmission.
Fig 6.
Inpatient C. difficile acquisition by specialty.
The numbers at the bottom of each bar indicate the total number of CDI cases in each category.
Fig 7.
Inpatient C. difficile acquisition by hospital.
Fig 8.
Temporal trends in C. difficile inpatient acquisition.
Q, quarter; CDI, C. difficile infection.