(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Implications of reducing antibiotic treatment duration for antimicrobial resistance in hospital settings: A modelling study and meta-analysis [1]

['Yin Mo', 'Centre For Tropical Medicine', 'Global Health', 'Nuffield Department Of Medicine', 'University Of Oxford', 'Oxford', 'United Kingdom', 'Mahidol-Oxford Tropical Medicine Research Unit', 'Faculty Of Tropical Medicine', 'Mahidol University']

Date: 2023-06

Modelling dynamics of colonising bacteria in response to antibiotic treatment

Three stochastic agent-based models with increasing complexity and biological realism were constructed. From here on, they will be referred to as simple exclusive colonisation model, co-colonisation model, and within-host growth model (Fig 1).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Host states and transitions in the exclusive colonisation, co-colonisation, and within-host growth models. These flow diagrams describe within-host changes in resistant and sensitive bacteria carried by each individual patient in response to antibiotic consumption, bacterial growth, and transmission events. In the 3 models, all patients were assumed to be always colonised with bacteria. In the exclusive colonisation model, patients can be colonised with resistant bacteria (R), and high or low levels of sensitive bacteria (S or s), but not both resistant and sensitive at the same time due to complete bacterial interference. The co-colonisation model makes a more realistic assumption that patients may be colonised with both resistant and sensitive bacteria at the same time with different levels of abundance. Similarly, in the within-host growth model, patients carry both resistant and sensitive bacteria but their combined populations cannot exceed a maximum carrying capacity. Blue, orange, and pink squares/circles represent the host states when the host is carrying susceptible, a mix of susceptible and resistant, and resistant bacteria, respectively. The red arrows highlight antibiotic killing of resistant bacteria. All transmission and decolonisation events apply to resistant bacteria. https://doi.org/10.1371/journal.pmed.1004013.g001

In all models, we simulated individual patients (agents) in a single hospital ward environment. The patients’ colonisation status could change due to (i) differential growth/killing rates of resistant and sensitive bacterial populations within a patient; (ii) transmission events between patients; and (iii) loss of resistance carriage (Fig 1). Because the population size under consideration is small and local fade-out events (when resistant populations reach zero) are potentially important, we considered stochastic implementations of these models (Fig 2 and Section S1.1 in S1 Text), i.e., changes in colonisation status took place with a probability randomly drawn from a distribution.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Example output from the co-colonisation model. This example simulation shows two 10-bedded wards with the same transmission risk of resistant bacteria, patient lengths of stay, proportion of patients who required antibiotic treatment, and proportion of patients who were resistance carriers upon admission. The left column panels represent the ward where antibiotic durations were short (mean of 5 days), while the right column panels represent the ward where antibiotic durations were long (mean of 15 days). Top row panels: Antibiotic treatment types and duration for one single iteration. Each square in the graphs represents one patient day. Light and dark green squares indicate one day of antibiotic effective only against susceptible organisms or both susceptible and resistant organisms. Vertical blue lines represent new admissions to the ward. Second row panels: The carriage status for each patient on a particular day for one single iteration. The increasingly darker shades of orange and green indicate an increasing amount of resistant and susceptible colonising bacteria carried by each patient, respectively. Third row panels: The number of antimicrobial resistance carriers in each ward per day for 40 iterations. Fourth row panel: The cumulative number of resistance carriers in each ward over 60 observation days for 40 iterations. In the third and fourth row panels, each line represents output from one iteration. The thick lines are outputs from the single example iteration illustrated in the top 2 panels. https://doi.org/10.1371/journal.pmed.1004013.g002

Within the same ward, antibiotics (administered via both intravenous and oral routes) were prescribed with the same mean duration on or during admission. Exact duration for each antibiotic course was drawn from a uniform distribution (range for short duration 3 to 7 days, range for long duration 14 to 21 days). We considered 2 scenarios. In the first, the resistant phenotype was assumed to be susceptible to one of the administered antibiotics, e.g., carbapenems against third-generation cephalosporin resistance. In the second scenario, there was no effective antibiotic available against the resistant phenotype in the colonising bacteria, e.g., colistin or carbapenem resistance. Subsequently, we refer to these 2 scenarios as “administered antibiotics have activity against resistant and susceptible organisms” and “administered antibiotics have activity only against susceptible organisms.”

The models assumed that antibiotics act within-host by selectively promoting the growth of existing resistant bacteria in the gut [17,18], and between-host by predisposing the treated individual to become more likely to transmit resistant bacteria as a consequence of a higher bacterial load (Fig 1) [19]. Spontaneous decolonisation of resistant bacteria was assumed to only occur when a patient was not receiving antibiotics to which the colonising bacteria were resistant.

In the exclusive colonisation model, patients could be carriers of either susceptible or resistant bacteria but not of both at the same time [20]. In this model, for a carrier of susceptible bacteria to become a resistance carrier required transmission of resistant bacteria from another carrier in the ward, i.e., we did not consider de novo resistance emergence or horizontal transfer from other components of the flora. In the other 2 more complex models, patients could carry both susceptible and resistant bacteria simultaneously. In the co-colonisation model, these abundances were dichotomised into high and low, and only bacteria carried at a high level were able to be transmitted to other patients [21]. In the within-host growth model, bacterial population sizes could vary continuously but had to be above a threshold in order to be capable of transmission to others [22]. We also assumed a total carrying capacity for gram-negative bacteria [23] and note that newly admitted patients may carry fewer bacteria than the carrying capacity as we would expect some patients were prescribed antibiotics prior to admission. The mean total carrying capacity for each iteration was drawn from a lognormal distribution (Table A in S1 Text). In the latter 2 models, an individual could become a resistance carrier from within-host selection of preexisting resistant bacteria through antibiotic consumption.

While the modelling framework we have adopted is quite general, here we focus on parameter space that are appropriate for gram-negative bacteria as they are the commonest cause of urinary tract and bloodstream infections and frequently associated with multidrug-resistant hospital-acquired infections. The predominant mechanisms for resistance dissemination in gram-negative bacteria colonising the gut are horizontal transfer of mobile genetic elements and clonal expansion [24].

We first explored the models by varying pairs of parameters over a grid of values while holding the other parameter values constant. This was followed by global sensitivity analysis using Latin Hypercube sampling and Partial Rank Correlation Coefficient (S1.3 Section in S1 Text). Real-world parameter values were obtained from the literature and used in these explorations (Table A in S1 Text). For those parameter values not found in the literature, we explored the largest reasonable range of parameter values. In all results shown below, each simulation was produced from at least 50,000 iterations (100 repeats of 500 unique sets of parameters selected within the ranges shown in Table A in S1 Text) over 300 days. The computer code is publicly available (https://github.com/moyinNUHS/abxduration_abm).

The models were used to assess how changing duration of antibiotic treatment affected the risk of resistance colonisation at both individual and population levels. Three types of outcomes were considered: resistance carriage among (i) treated patients and, therefore, directly affected by different treatment durations; (ii) overall resistance carriage within a ward population that consisted of treated and untreated patients, i.e., indirectly affected by treatment duration received by those treated patients; and (iii) patients who were not carriers of resistant bacteria when admitted to the ward. To evaluate reducing antibiotic duration as an intervention, all the model outputs were assessed as absolute difference between the short and long wards.

All simulations were performed with R version 4.0.4 (2021-02-15) [25] using packages msm [26], pse [27], and spartan [28]. Code review was done using unit tests for each function’s base scenarios, which can be found under /unit_tests in the source code.

[END]
---
[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1004013

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.