Dataset Information

Bayesian estimation of the prevalence of antimicrobial resistance: a mathematical modelling study.

ABSTRACT:

Background

Estimates of the prevalence of antimicrobial resistance (AMR) underpin effective antimicrobial stewardship, infection prevention and control, and optimal deployment of antimicrobial agents. Typically, the prevalence of AMR is determined from real-world antimicrobial susceptibility data that are time delimited, sparse, and often biased, potentially resulting in harmful and wasteful decision-making. Frequentist methods are resource intensive because they rely on large datasets.

Objectives

To determine whether a Bayesian approach could present a more reliable and more resource-efficient way to estimate population prevalence of AMR than traditional frequentist methods.

Methods

Retrospectively collected, open-source, real-world pseudonymized healthcare data were used to develop a Bayesian approach for estimating the prevalence of AMR by combination with prior AMR information from a contextualized review of literature. Iterative random sampling and cross-validation were used to assess the predictive accuracy and potential resource efficiency of the Bayesian approach compared with a standard frequentist approach.

Results

Bayesian estimation of AMR prevalence made fewer extreme estimation errors than a frequentist estimation approach [n = 74 (6.4%) versus n = 136 (11.8%)] and required fewer observed antimicrobial susceptibility results per pathogen on average [mean = 28.8 (SD = 22.1) versus mean = 34.4 (SD = 30.1)] to avoid any extreme estimation errors in 50 iterations of the cross-validation. The Bayesian approach was maximally effective and efficient for drug-pathogen combinations where the actual prevalence of resistance was not close to 0% or 100%.

Conclusions

Bayesian estimation of the prevalence of AMR could provide a simple, resource-efficient approach to better inform population infection management where uncertainty about AMR prevalence is high.

SUBMITTER: Howard A

PROVIDER: S-EPMC11368424 | biostudies-literature | 2024 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Bayesian estimation of the prevalence of antimicrobial resistance: a mathematical modelling study.

Howard Alex A Green Peter L PL Velluva Anoop A Gerada Alessandro A Hughes David M DM Brookfield Charlotte C Hope William W Buchan Iain I

The Journal of antimicrobial chemotherapy 20240901 9

<h4>Background</h4>Estimates of the prevalence of antimicrobial resistance (AMR) underpin effective antimicrobial stewardship, infection prevention and control, and optimal deployment of antimicrobial agents. Typically, the prevalence of AMR is determined from real-world antimicrobial susceptibility data that are time delimited, sparse, and often biased, potentially resulting in harmful and wasteful decision-making. Frequentist methods are resource intensive because they rely on large datasets.< ...[more]

PMID: 39051678

Similar Datasets

Project description:BackgroundThe World Health Organization recommends changing the first-line antimicrobial treatment for gonorrhoea when ≥ 5% of Neisseria gonorrhoeae cases fail treatment or are resistant. Susceptibility to ceftriaxone, the last remaining treatment option has been decreasing in many countries. We used antimicrobial resistance surveillance data and developed mathematical models to project the time to reach the 5% threshold for resistance to first-line antimicrobials used for N. gonorrhoeae.MethodsWe used data from the Gonococcal Resistance to Antimicrobials Surveillance Programme (GRASP) in England and Wales from 2000-2018 about minimum inhibitory concentrations (MIC) for ciprofloxacin, azithromycin, cefixime and ceftriaxone and antimicrobial treatment in two groups, heterosexual men and women (HMW) and men who have sex with men (MSM). We developed two susceptible-infected-susceptible models to fit these data and produce projections of the proportion of resistance until 2030. The single-step model represents the situation in which a single mutation results in antimicrobial resistance. In the multi-step model, the sequential accumulation of resistance mutations is reflected by changes in the MIC distribution.ResultsThe single-step model described resistance to ciprofloxacin well. Both single-step and multi-step models could describe azithromycin and cefixime resistance, with projected resistance levels higher with the multi-step than the single step model. For ceftriaxone, with very few observed cases of full resistance, the multi-step model was needed to describe long-term dynamics of resistance. Extrapolating from the observed upward drift in MIC values, the multi-step model projected ≥ 5% resistance to ceftriaxone could be reached by 2030, based on treatment pressure alone. Ceftriaxone resistance was projected to rise to 13.2% (95% credible interval [CrI]: 0.7-44.8%) among HMW and 19.6% (95%CrI: 2.6-54.4%) among MSM by 2030.ConclusionsNew first-line antimicrobials for gonorrhoea treatment are needed. In the meantime, public health authorities should strengthen surveillance for AMR in N. gonorrhoeae and implement strategies for continued antimicrobial stewardship. Our models show the utility of long-term representative surveillance of gonococcal antimicrobial susceptibility data and can be adapted for use in, and for comparison with, other countries.

Project description:BackgroundThe existing estimate of the global burden of latent TB infection (LTBI) as "one-third" of the world population is nearly 20 y old. Given the importance of controlling LTBI as part of the End TB Strategy for eliminating TB by 2050, changes in demography and scientific understanding, and progress in TB control, it is important to re-assess the global burden of LTBI.Methods and findingsWe constructed trends in annual risk in infection (ARI) for countries between 1934 and 2014 using a combination of direct estimates of ARI from LTBI surveys (131 surveys from 1950 to 2011) and indirect estimates of ARI calculated from World Health Organisation (WHO) estimates of smear positive TB prevalence from 1990 to 2014. Gaussian process regression was used to generate ARIs for country-years without data and to represent uncertainty. Estimated ARI time-series were applied to the demography in each country to calculate the number and proportions of individuals infected, recently infected (infected within 2 y), and recently infected with isoniazid (INH)-resistant strains. Resulting estimates were aggregated by WHO region. We estimated the contribution of existing infections to TB incidence in 2035 and 2050. In 2014, the global burden of LTBI was 23.0% (95% uncertainty interval [UI]: 20.4%-26.4%), amounting to approximately 1.7 billion people. WHO South-East Asia, Western-Pacific, and Africa regions had the highest prevalence and accounted for around 80% of those with LTBI. Prevalence of recent infection was 0.8% (95% UI: 0.7%-0.9%) of the global population, amounting to 55.5 (95% UI: 48.2-63.8) million individuals currently at high risk of TB disease, of which 10.9% (95% UI:10.2%-11.8%) was isoniazid-resistant. Current LTBI alone, assuming no additional infections from 2015 onwards, would be expected to generate TB incidences in the region of 16.5 per 100,000 per year in 2035 and 8.3 per 100,000 per year in 2050. Limitations included the quantity and methodological heterogeneity of direct ARI data, and limited evidence to inform on potential clearance of LTBI.ConclusionsWe estimate that approximately 1.7 billion individuals were latently infected with Mycobacterium tuberculosis (M.tb) globally in 2014, just under a quarter of the global population. Investment in new tools to improve diagnosis and treatment of those with LTBI at risk of progressing to disease is urgently needed to address this latent reservoir if the 2050 target of eliminating TB is to be reached.

Project description:BackgroundVector-borne diseases are important public health issues and, consequently, in silico models that simulate them can be useful. The susceptible-infected-recovered (SIR) model simulates the population dynamics of an epidemic and can be easily adapted to vector-borne diseases, whereas the Hardy-Weinberg model simulates allele frequencies and can be used to study insecticide resistance evolution. The aim of the present study is to develop a coupled system that unifies both models, therefore enabling the analysis of the effects of vector population genetics on the population dynamics of an epidemic.MethodsOur model consists of an ordinary differential equation system. We considered the populations of susceptible, infected and recovered humans, as well as susceptible and infected vectors. Concerning these vectors, we considered a pair of alleles, with complete dominance interaction that determined the rate of mortality induced by insecticides. Thus, we were able to separate the vectors according to the genotype. We performed three numerical simulations of the model. In simulation one, both alleles conferred the same mortality rate values, therefore there was no resistant strain. In simulations two and three, the recessive and dominant alleles, respectively, conferred a lower mortality.ResultsOur numerical results show that the genetic composition of the vector population affects the dynamics of human diseases. We found that the absolute number of vectors and the proportion of infected vectors are smaller when there is no resistant strain, whilst the ratio of infected people is larger in the presence of insecticide-resistant vectors. The dynamics observed for infected humans in all simulations has a very similar shape to real epidemiological data.ConclusionThe population genetics of vectors can affect epidemiological dynamics, and the presence of insecticide-resistant strains can increase the number of infected people. Based on the present results, the model is a basis for development of other models and for investigating population dynamics.

Project description:BackgroundGonorrhoea is a highly prevalent sexually transmitted infection and an urgent public health concern because of increasing antibiotic resistance in Neisseria gonorrhoeae. Only ceftriaxone remains as the recommended treatment in the USA. With the prospect of new anti-gonococcal antibiotics being approved, we aimed to evaluate how to deploy a new drug to maximise its clinically useful lifespan.MethodsWe used a compartmental model of gonorrhoea transmission in a US population of men who have sex with men (MSM) to compare strategies for introducing a new antibiotic for gonorrhoea treatment. The MSM population was stratified into three sexual activity groups (low, intermediate, and high) characterised by annual rates of partner change. The four introduction strategies tested were: (1) random 50-50 allocation, where each treatment-seeking infected individual had a 50% probability of receiving either drug A (current drug; a ceftriaxone-like antibiotic) or drug B (a new antibiotic), effective at time 0; (2) combination therapy of both the current drug and the new antibiotic; (3) reserve strategy, by which the new antibiotic was held in reserve until the current therapy reached a 5% threshold prevalence of resistance; and (4) gradual switch, or the gradual introduction of the new drug until random 50-50 allocation was reached. The primary outcome of interest was the time until 5% prevalence of resistance to each of the drugs (the new drug and the current ceftriaxone-like antibiotic); sensitivity of the primary outcome to the properties of the new antibiotic, specifically the probability of resistance emergence after treatment and the fitness costs of resistance, was explored. Secondary outcomes included the time to a 1% resistance threshold for each drug, as well as population-level prevalence, mean and range annual incidence, and the cumulative number of incident gonococcal infections.FindingsUnder baseline model conditions, a 5% prevalence of resistance to each of drugs A and B was reached within 13·9 years with the reserve strategy, 18·2 years with the gradual switch strategy, 19·2 years with the random 50-50 allocation strategy, and 19·9 years with the combination therapy strategy. The reserve strategy was consistently inferior for mitigating antibiotic resistance under the parameter space explored and was increasingly outperformed by the other strategies as the probability of de novo resistance emergence decreased and as the fitness costs associated with resistance increased. Combination therapy tended to prolong the development of antibiotic resistance and minimise the number of annual gonococcal infections (under baseline model conditions, mean number of incident infections per year 178 641 [range 177 998-181 731] with combination therapy, 180 084 [178 011-184 405] with the reserve strategy).InterpretationOur study argues for rapid introduction of new anti-gonococcal antibiotics, recognising that the feasibility of each strategy must incorporate cost, safety, and other practical concerns. The analyses should be revisited once robust estimates of key parameters-ie, the likelihood of emergence of resistance and fitness costs of resistance for the new antibiotic-are available.FundingUS Centers for Disease Control and Prevention, National Institute of Allergy and Infectious Diseases.

Dataset Information

Bayesian estimation of the prevalence of antimicrobial resistance: a mathematical modelling study.

Background

Objectives

Methods

Results

Conclusions

Publications

Bayesian estimation of the prevalence of antimicrobial resistance: a mathematical modelling study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets