Inference and control of the nosocomial transmission of methicillin-resistant Staphylococcus aureus

  1. Sen Pei  Is a corresponding author
  2. Flaviano Morone
  3. Fredrik Liljeros
  4. Hernán Makse
  5. Jeffrey L Shaman  Is a corresponding author
  1. Columbia University, United States
  2. City College of New York, United States
  3. Stockholm University, Sweden
5 figures, 1 video, 3 tables and 2 additional files

Figures

Figure 1 with 2 supplements
Observed incidence of UK EMRSA-15 and the agent-based MRSA transmission model.

(A) Incidence of UK EMRSA-15 every 4 weeks (red crosses) and cumulative cases (blue curve). (B) The raster plot for infections in 114 infected wards. Color indicates number of observed infections …

https://doi.org/10.7554/eLife.40977.003
Figure 1—source data 1

Numerical data represented in Figure 1.

The data set includes: (1) incidence data in Figure 1A; (2) ward number, infection number, total visit, average stay and ward capacity visualized in Figure 1B–C.

https://doi.org/10.7554/eLife.40977.006
Figure 1—figure supplement 1
Association between infection numbers and patient-days per ward.

(A) Number of wards with a given number of total patient-days in the hospitalization records. Inset shows the scatter plot of the number of positive cases and patient-days per ward. (B) Average …

https://doi.org/10.7554/eLife.40977.004
Figure 1—figure supplement 2
Patient traffic in the study Swedish hospitals.

(A) Daily patient overlap ratio Q during 300 days. Inset provides a zoom-in plot of the first 30 days. (B) The number of total patients and new patients (with respect to patients present the …

https://doi.org/10.7554/eLife.40977.005
Figure 2 with 5 supplements
Inference of model parameters for a synthetic outbreak.

(A) Distributions of the posterior parameters β (top), I0 (middle) and C0 (bottom) (300 ensemble members) for 20 iterations of inference in one realization of the IF algorithm. Orange Tukey boxes …

https://doi.org/10.7554/eLife.40977.008
Figure 2—source data 1

Numerical data represented in Figure 2.

The data set includes: (1) distributions of β, I0 and C0 in Figure 2A; (2) distributions of inferred incidence and actual observation in the synthetic outbreak in Figure 2B; (3) distributions of inferred colonization and actual colonization in the synthetic outbreak in Figure 2C; (4) distributions of inferred transmitted incidence and actual transmitted incidence in the synthetic outbreak in Figure 2D; (5) distributions of inferred imported incidence and actual imported incidence in the synthetic outbreak in Figure 2E.

https://doi.org/10.7554/eLife.40977.014
Figure 2—figure supplement 1
Evaluation of the goodness of fit in Figure 2B.

(A) Distribution of the log likelihood calculated from surrogate data (blue bars). We generated 1000 synthetic outbreaks using the inferred parameters, from which we approximated the distribution of …

https://doi.org/10.7554/eLife.40977.009
Figure 2—figure supplement 2
Synthetic test of IF for an outbreak in which the majority of infections are imported.

(A) Parameters used in the synthetic outbreak (inference targets) are marked by red lines. The posterior parameter distributions (300 ensemble members) for different iterations in one realization of …

https://doi.org/10.7554/eLife.40977.010
Figure 2—figure supplement 3
Evaluation of the goodness of fit in Figure 2—figure supplement 2B.

Same analysis as in Figure 2—figure supplement 1. Comparisons of the log likelihood (A), RMSE (B), R2 (C), and Pearson correlation coefficient (D) obtained from the inference in Figure 2—figure …

https://doi.org/10.7554/eLife.40977.011
Figure 2—figure supplement 4
Synthetic test of IF for observations every 4 weeks.

Parameters are set as in Figure 2. (A) Parameters used in the synthetic outbreak (inference targets) are marked by red lines. The posterior parameter distributions (300 ensemble members) for …

https://doi.org/10.7554/eLife.40977.012
Figure 2—figure supplement 5
Evaluation of the goodness of fit in Figure 2—figure supplement 4B.

Same analysis as in Figure 2—figure supplement 1. Comparisons of the log likelihood (A), RMSE (B), R2 (C), and Pearson correlation coefficient (D) obtained from the inference in Figure 2—figure …

https://doi.org/10.7554/eLife.40977.013
Figure 3 with 3 supplements
Inference of the nosocomial transmission of UK EMRSA-15.

(A) Inferred distributions of the MLEs for key parameters β, I0 and C0 over 6 years, obtained from 100 independent realizations of the IF algorithm. (B) Observed incidence every 4 weeks (red …

https://doi.org/10.7554/eLife.40977.015
Figure 3—source data 1

Numerical data represented in Figure 3.

The data set includes: (1) distributions of inferred parameters in Figure 3A; (2) distributions of inferred incidence and actual observation in the real-world outbreak in Figure 3B; (3) distribution of the number of infected wards obtained from inference in Figure 3C; (4) observed and inferred distributions of the number of infections per ward in Figure 3D; (5) distributions of inferred nosocomial transmitted and imported cases in Figure 3E.

https://doi.org/10.7554/eLife.40977.019
Figure 3—figure supplement 1
Distributions of posterior parameters (300 ensemble members) in 20 iterations for different years in one realization of the IF algorithm.

In each year, we performed 20 iterations using the IF. Boxes show the inferred distributions of 300 posterior parameters (box: interquartile; whisker: 95% CI).

https://doi.org/10.7554/eLife.40977.016
Figure 3—figure supplement 2
Evaluation of the goodness of fit in Figure 3B.

We generated the surrogate data (1000 synthetic outbreaks) using the inferred parameters, and compared the log likelihood (A), RMSE (B), R2 (C), and Pearson correlation coefficient (D) obtained from …

https://doi.org/10.7554/eLife.40977.017
Figure 3—figure supplement 3
Classification of patients using days from admission to infection.

We classified the patients testing positive according to the time intervals between their hospital admission and confirmation date: those with 2 days (48 hr) were regarded as imported cases (199 …

https://doi.org/10.7554/eLife.40977.018
Figure 4 with 1 supplement
Inference of asymptomatic colonization in Swedish hospitals.

(A) Inferred distributions of colonized patients through time. (B) The distribution of colonization probability for each individual in hospital at T = 40 (week 160) calculated from 104 model …

https://doi.org/10.7554/eLife.40977.023
Figure 4—source data 1

Numerical data represented in Figure 4.

The data set includes: (1) distributions of inferred colonization in Figure 4A; (2) distribution of colonization probability in Figure 4B; (3) individuals’ colonization probability visualized in Figure 4C.

https://doi.org/10.7554/eLife.40977.025
Figure 4—figure supplement 1
(A) The KS statistic for different lower bounds of power-law behavior.

(B) Comparison of the KS statistic between 104 synthetic samples and observed data. The vertical line indicates the KS statistic for the observed data.

https://doi.org/10.7554/eLife.40977.024
Retrospective control experiment in Swedish hospitals.

The cumulative cases of colonization (A) and infection (B) after decolonizing patients with a hazard of colonization higher than a specified decolonization threshold. Simulations were performed with …

https://doi.org/10.7554/eLife.40977.026
Figure 5—source data 1

Numerical data represented in Figure 5.

The data set includes: (1) distributions of colonization for each decolonization threshold in Figure 5A; (2) distributions of infection for each decolonization threshold in Figure 5B; (3) colonization number for each control strategy in Figure 5C; (4) infection number for each control strategy in Figure 5D.

https://doi.org/10.7554/eLife.40977.027

Videos

Video 1
One realization of the agent-based model simulation.

We visualize a single realization of the agent-based model during a one-year period. The grey nodes represent susceptible people, green nodes represent colonized individuals, and red nodes highlight …

https://doi.org/10.7554/eLife.40977.022

Tables

Table 1
Parameter ranges used in the agent-based transmission model.
https://doi.org/10.7554/eLife.40977.007
ParameterDescriptionRangeUnit
αSpontaneous decolonization rate[1/525, 1/175]per day
pInfection progress rate[0.1α, 0.3α]per day
μRecovery rate with treatment[1/120, 1/20]per day
βTransmission rate in hospitals[0, 0.01]per day
I0Infection importation rate[0, 0.001]per admission
C0Colonization importation rate[0, 0.1]per admission
  1. Sources for parameter ranges – α: (Cooper et al., 2004a; Bootsma et al., 2006; Eveillard et al., 2006; Wang et al., 2013; Macal et al., 2014; Jarynowski and Liljeros, 2015); p: (Kajita et al., 2007; Jarynowski and Liljeros, 2015); μ: (D'Agata et al., 2009; Wang et al., 2013); β: Prior; I0: Prior; C0: Prior, (Hidron et al., 2005; Eveillard et al., 2006; Jarvis et al., 2012). For each individual, the infection progress rate p is drawn after α is specified.

Table 2
Inferred parameters and 95% CIs across 6 years using the actual diagnostic data.
https://doi.org/10.7554/eLife.40977.020
Inferred parameters and 95% CIs
YearβI0C0
I2.16,[1.83,2.60]×10-33.67,[3.28,4.06]×10-58.61,[7.92,9.47]×10-3
II2.87,[2.48,3.44]×10-31.27,[1.13,1.45]×10-41.68,[1.40,1.98]×10-2
III4.71,[4.29,5.13]×10-36.19,[5.31,7.48]×10-53.03,[2.36,3.62]×10-2
IV2.91,[2.47,3.44]×10-32.31,[1.93,2.64]×10-42.53,[1.85,3.26]×10-2
V3.18,[2.61,3.79]×10-41.62,[1.29,2.04]×10-42.08,[1.51,2.63]×10-2
VI2.16,[1.83,2.60]×10-35.31,[4.27,6.30]×10-59.57,[7.72,12.43]×10-3
Table 2—source data 1

Numerical data represented in Table 2.

Results are obtained from 100 independent realizations of the IF algorithm.

https://doi.org/10.7554/eLife.40977.021
Table 3
Inferred parameters and 95% CIs for three synthetic tests.
https://doi.org/10.7554/eLife.40977.029
βI0C0
Actual9×10-32×10-37.5×10-2
Inference (weekly)9.00,[8.07,9.68]×10-31.91,[1.38,2.54]×10-37.18,[5.84,8.70]×10-2
Actual6×10-32×10-37.5×10-2
Inference (weekly)5.54,[4.17,5.80]×10-32.11,[1.52,2.55]×10-37.05,[5.79,8.11]×10-2
Actual9×10-32×10-37.5×10-2
Inference (monthly)9.00,[8.17,9.66]×10-31.99,[1.21,2.64]×10-37.14,[5.99,9.04]×10-2

Additional files

Source code 1

The Matlab code for parameter inference in a synthetic MRSA outbreak simulated in an example time-varying contact network.

https://doi.org/10.7554/eLife.40977.030
Transparent reporting form
https://doi.org/10.7554/eLife.40977.031

Download links