Antimicrobial resistance prevalence in bloodstream infection in 29 European countries by age and sex: An observational study

Abstract

Background: Antibiotic usage, contact with high transmission healthcare settings as well as changes in immune system function all vary by a patient's age and sex. Yet, most analyses of antimicrobial resistance (AMR) ignore demographic indicators and provide only country-level resistance prevalence values. This study aimed to address this knowledge gap by quantifying how resistance prevalence and incidence of bloodstream infection (BSI) varied by age and sex across bacteria and antibiotics in Europe.

Methods and findings: We used patient-level data collected as part of routine surveillance between 2015 and 2019 on BSIs in 29 European countries from the European Antimicrobial Resistance Surveillance Network (EARS-Net). A total of 6,862,577 susceptibility results from isolates with age, sex, and spatial information from 944,520 individuals were used to characterise resistance prevalence patterns for 38 different bacterial species and antibiotic combinations, and 47% of these susceptibility results were from females, with a similar age distribution in both sexes (mean of 66 years old). A total of 349,448 isolates from 2019 with age and sex metadata were used to calculate incidence. We fit Bayesian multilevel regression models by country, laboratory code, sex, age, and year of sample to quantify resistant prevalence and provide estimates of country-, bacteria-, and drug-family effect variation. We explore our results in greater depths for 2 of the most clinically important bacteria-antibiotic combinations (aminopenicillin resistance in Escherichia coli and methicillin resistance in Staphylococcus aureus) and present a simplifying indicative index of the difference in predicted resistance between old (aged 100) and young (aged 1). At the European level, we find distinct patterns in resistance prevalence by age. Trends often vary more within an antibiotic family, such as fluroquinolones, than within a bacterial species, such as Pseudomonas aeruginosa. Clear resistance increases by age for methicillin-resistant Staphylococcus aureus (MRSA) contrast with a peak in resistance to several antibiotics at approximately 30 years of age for P. aeruginosa. For most bacterial species, there was a u-shaped pattern of infection incidence with age, which was higher in males. An important exception was E. coli, for which there was an elevated incidence in females between the ages of 15 and 40. At the country-level, subnational differences account for a large amount of resistance variation (approximately 38%), and there are a range of functional forms for the associations between age and resistance prevalence. For MRSA, age trends were mostly positive, with 72% (n = 21) of countries seeing an increased resistance between males aged 1 and 100 years and a greater change in resistance in males. This compares to age trends for aminopenicillin resistance in E. coli which were mostly negative (males: 93% (n = 27) of countries see decreased resistance between those aged 1 and 100 years) with a smaller change in resistance in females. A change in resistance prevalence between those aged 1 and 100 years ranged up to 0.51 (median, 95% quantile of model simulated prevalence using posterior parameter ranges 0.48, 0.55 in males) for MRSA in one country but varied between 0.16 (95% quantile 0.12, 0.21 in females) to -0.27 (95% quantile -0.4, -0.15 in males) across individual countries for aminopenicillin resistance in E. coli. Limitations include potential bias due to the nature of routine surveillance and dependency of results on model structure.

Conclusions: In this study, we found that the prevalence of resistance in BSIs in Europe varies substantially by bacteria and antibiotic over the age and sex of the patient shedding new light on gaps in our understanding of AMR epidemiology. Future work is needed to determine the drivers of these associations in order to more effectively target transmission and antibiotic stewardship interventions.

Fig 1. Trends in resistance prevalence in BSIs vary by antibiotic, bacteria, and demographic factors across Europe.

The proportion of isolates from BSIs tested (y axis) that are resistant to each antibiotic (panel) within drug families and AWaRe groupings (for beta-lactams) (colour) for each bacteria (row) is shown for all European data over 2015–2019 by age (x axis) for the 4 gram-negative (A+B) and 4 gram-positive bacteria (C+D). Data is shown as points with number of samples indicated by size of point. Shaded areas are 95% confidence intervals around the LOESS fit line by sex (linetype). AWaRe groupings were used here to better distinguish clinically important subsets within the beta-lactam family. Blank panels indicate where no data was available. BSI, bloodstream infection; LOESS, locally estimated scatterplot smoothing.

Fig 2. Incidence of BSIs per 100,000 population in 2019 across European countries for 8 bacterial pathogens.

To demonstrate trends clearly, the incidence is split to show (A) patterns in the first 50 years of life by sex and bacteria and then (B) shows the lifelong trends split by sex (panel) and bacteria (colour). Shaded areas are 95% confidence intervals using an LOESS fit. Infections in individuals younger than 0 are excluded, and those aged 80 and older are pooled into the 80-yr data point. The y axis is on a log scale (base 10). BSI, bloodstream infection; LOESS, locally estimated scatterplot smoothing.

Fig 3

Model parameters and example model predictions for MRSA (left) and aminopenicillin resistance in E. coli (right). (A, B) Model parameters. (m) indicates that the parameter is the coefficient for males. (C, D) Data (points) and model predictions (lines) with 95% credible intervals (ribbons) for males for the most extreme (left and right panels) and the middle country (middle panel) estimated age slope. Each country has 2 lines, depicting the predictions for the most extreme laboratories in the country. Data sample size (shown by dot size) is grouped across years and laboratories. MRSA primarily indicates oxacillin or cefoxitin resistance, but other markers are accepted for oxacillin, if oxacillin was not reported. See protocol for details [41]. Country labels are a random anonymised 3-letter code used for this study only but consistent across all analyses. MRSA, methicillin-resistant Staphylococcus aureus.

Fig 4. Change in model-predicted proportion resistant between older (age 100) and younger (age 1) patients.

This index is shown for each country and sex for MRSA (A) and aminopenicillin resistance in E. coli (B), with the point indicating the median and the error bars the 95% quantiles across model predictions. Country labels are random anonymised 3-letter code used for this study only but consistent across all analyses. MRSA, methicillin-resistant Staphylococcus aureus.

Fig 5. Change in model-predicted proportion of samples resistant between older (age 100) and younger (age 1) patients with BSIs.

The index is split by sex (panels) and Gram stain test result. Each point indicates the median index for each individual antibiotic coloured by drug family with the error bars showing the 95% quantiles across model predictions. The dashed line indicates 0 (i.e., no difference in model-predicted difference between resistance prevalence at age 1 vs. 100). The anonymised country-level results can be seen in Section 5 in S2 Appendix. BSI, bloodstream infection.