Gross national income and antibiotic resistance in invasive isolates: analysis of the top-ranked antibiotic-resistant bacteria on the 2017 WHO priority list

Abstract

Objectives

To assess the association between country income status and national prevalence of invasive infections caused by the top-ranked bacteria on the WHO priority list: carbapenem-resistant (CR) Acinetobacter spp., Klebsiella spp. and Pseudomonas aeruginosa; third-generation cephalosporin-resistant (3GCR) Escherichia coli and Klebsiella spp.; and MRSA and vancomycin-resistant Enterococcus faecium (VR E. faecium).

Methods

Active surveillance systems providing yearly prevalence data from 2012 onwards for the selected bacteria were included. The gross national income (GNI) per capita was used as the indicator for income status of each country and was log transformed to account for non-linearity. The association between antibiotic prevalence data and GNI per capita was investigated individually for each bacterium through linear regression.

Results

Surveillance data were available from 67 countries: 38 (57%) were high income, 16 (24%) upper-middle income, 11 (16%) lower-middle income and two (3%) low income countries. The regression showed significant inverse association (P<0.0001) between resistance prevalence of invasive infections and GNI per capita. The highest rate of increase per unit decrease in log GNI per capita was observed in 3GCR Klebsiella spp. (22.5%, 95% CI 18.2%–26.7%), CR Acinetobacter spp. (19.2% 95% CI 11.3%–27.1%) and 3GCR E. coli (15.3%, 95% CI 11.6%–19.1%). The rate of increase per unit decrease in log GNI per capita was lower in MRSA (9.5%, 95% CI 5.2%–13.7%).

Conclusions

The prevalence of invasive infections caused by the WHO top-ranked antibiotic-resistant bacteria is inversely associated with GNI per capita at the global level. Public health interventions designed to limit the burden of antimicrobial resistance should also consider determinants of poverty and inequality, especially in lower-middle income and low income countries.

Introduction

The association between the spread of infectious diseases and socioeconomic factors is well established. The term ‘social medicine’ dates back to the mid-19th century, when it was used for the first time by the German pathologist Robert Virchow during a typhus outbreak.¹ Instead of prescribing a medical solution for the ‘disastrous disease’, he proposed a revolutionary programme of social reconstruction, promoting a more equal distribution of economic resources and education.^2,3 Following his initial statements, the relevance of ‘social medicine’ in the field of infectious diseases has grown progressively, making the assessment of different dimensions of poverty a crucial step in the development of infection prevention and control strategies.^4–6

In the field of antimicrobial resistance (AMR), universally recognized as an emergent global threat, several attempts to explore the link between AMR spread and societal inequality have been made. A recent systematic review examined the social aspects of AMR and described an association between the most common dimensions of poverty (i.e. income, housing condition, social deprivation, lack of education) and colonization/infection with antibiotic-resistant bacteria. Nineteen studies were included, mostly conducted in high-resource settings with high heterogeneity of methodology. The authors concluded that, although further research to prove this association is needed, targeting social determinants of poverty should be regarded as a crucial step towards AMR prevention.³

When moving from individual to country-specific determinants of socioeconomic status, a few authors proposed several national economic indicators as the main independent variables. A study conducted in 15 European countries found a positive correlation between a standardized income inequality score [Standardized World Income Inequality Database (SWIID) score] and prevalence of AMR in Enterococcus faecalis, Enterococcus faecium, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, Staphylococcus aureus and Streptococcus pneumoniae.⁷ Similarly, Alvarez-Uria et al.,⁸ in collaboration with the Center for Disease Dynamics, Economics & Policy (CDDEP) working group, analysed the relationship between 2013–14 AMR surveillance data from 45 countries and gross national income (GNI) per capita. The analysis demonstrated an inverse association between income status and prevalence of MRSA, third-generation cephalosporin-resistant (3GCR) E. coli and Klebsiella spp. The relationship between several anthropological and socioeconomic factors and AMR was recently evaluated by Collignon et al.⁹ through a multivariable model: improved governance and infrastructure showed a significant inverse association with AMR.

The main limitation of the current evidence resides in the non-homogeneous representativeness of surveillance data from countries with lower economic resources. In very recent years, thanks to the effort of several international stakeholders and organizations, surveillance coverage in lower-middle income countries (LMICs)/low income countries (LICs) has significantly improved, providing a more comprehensive insight into the global distribution of AMR.

The aim of this study was to collect the most recent data on prevalence of AMR from surveillance networks and to test the hypothesis of an association between lower income status and AMR prevalence in invasive infections. For our purpose, we selected the top-ranked antibiotic-resistant bacteria on the 2017 WHO priority list for research and development of new antibiotics:¹⁰ carbapenem-resistant (CR) Acinetobacter spp., 3GCR E. coli, vancomycin-resistant Enterococcus faecium (VR E. faecium), CR Klebsiella spp., 3GCR Klebsiella spp., CR P. aeruginosa and MRSA.

Materials and methods

Data sources

We conducted a comprehensive web search of surveillance systems reporting data on prevalence of the selected bacterial phenotypes. The whole search was performed by using the Google search engine (www.google.com) by combining the terms ‘surveillance system’, ‘surveillance network’, ‘antimicrobial resistance’ and ‘antibiotic resistance’ with the name of each country of the world. Boolean logic was used to widen or narrow the search. We additionally consulted the websites of the following institutes of public health and scientific societies: WHO, CDC, ECDC, CDDEP and ESCMID. If no data relative to a given country were retrieved, we contacted representatives from the ministry of health or from national infectious diseases institutions in that country, asking specifically whether any surveillance network was active. The search was restricted to surveillance networks reporting data in the English language and it was carried out between September 2017 and April 2018 following the Campbell Collaboration Method Guide 2017.¹¹

Definition and eligibility criteria

Surveillance networks were included if data on prevalence of resistance were provided for at least one of the selected bacteria for at least 1 year from 2012 onwards. Prevalence of resistance was defined as the number of resistant isolates divided by the total number of tested isolates. Inclusion was limited to invasive isolates from blood and CSF and to surveillance systems whose AMR data were clearly based on validated clinical breakpoints (defined by EUCAST or CLSI guidance). Surveillance systems providing AMR data exclusively on animals, food and/or the environment were excluded from the search. The GNI per capita in US dollars was used as the indicator for the income status of each country. Income was defined in four categories according to the World Bank classification:¹² high income countries (HICs), upper-middle income countries (UMICs), LMICs and LICs.

Data extraction

Data from eligible surveillance systems were extracted by two authors independently (A. S., E. C.). Inconsistencies were discussed and addressed by consensus or involving a third reviewer (E. T.). The following data were extracted and entered into a predefined database: pathogen’s name and species, total number of resistant isolates and total number of tested isolates, year, type of population (paediatric and/or adult population), type and numbers of participating laboratories, and type of hospital or institution providing the samples. For each country, both the income category and the GNI per capita were extracted from the 2017 World Bank classification datasets.^12,13

Data analysis and synthesis

Country-specific prevalence rates were computed and CIs calculated using the Wilson score method.¹⁴ Association between prevalence of resistance and income was studied individually for each antibiotic-resistant bacterium with linear regression using variance-weighted least squares accounting for the number of isolates.

GNI per capita in US dollars was log transformed to account for non-linearity. A P value <0.05 was considered significant. Resistance data encompassing <30 isolates and/or coming from a single institution were deemed as scarcely representative of the country pattern of AMR prevalence and therefore excluded from the analysis. All statistical analyses were carried out using STATA version 14.2.

Results

The systematic web search retrieved a total of nine surveillance systems reporting data on 38 countries. Thirteen surveillance systems, encompassing data from 29 countries, were added after searching international institutions’ websites. No additional surveillance data were gathered after having contacted representatives of the ministry of health and infectious diseases institutions. The flow chart of the selection process is displayed in Figure S1 (available as Supplementary data at JAC Online). Overall, AMR prevalence data were extracted from 22 surveillance systems reporting data from 67 countries. The largest proportion of data was from HICs (57%, 38 countries), with 24% (16) from UMICs, 16% (11) from LMICs and 3% (2) from LICs. The surveillance coverage, in accordance with the World Bank classification, reached 47% in HICs (38 out of 81 countries) and 29% in UMICs (16 out of 56 countries), whereas the surveillance coverage was markedly lower in LMICs and LICs, reaching 23% (11 out of 47 countries) and 6% (two out of 34 countries), respectively. Details on the geographical distribution of the surveillance systems stratified by income category are displayed in Figure 1. LMICs/LICs with at least one active surveillance system were: Egypt, Georgia, Ghana, India, Kenya, Kosovo, Malawi, Pakistan, Philippines, Tunisia, Vietnam, Zambia and Zimbabwe. In four countries (Kenya, Malawi, Mexico, Pakistan) data reporting was limited to a single laboratory. The included surveillance systems and their main characteristics are listed in Table S1.

Figure 1.

Geographical distribution of the included surveillance systems stratified by income category. Income categories were extracted from the World Bank classification website.¹² This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The most frequently tracked bacterium was E. coli (65/67 countries, 97%), followed by Klebsiella spp. (63/67 countries, 94%), S. aureus (59/67 countries, 88%), Acinetobacter spp. (58/67 countries, 87%), P. aeruginosa (56/67 countries, 84%) and E. faecium (54/67 countries, 81%).

Sixty-three countries provided data for testing the association between income status and AMR prevalence data. The regression analysis showed a statistically significant inverse relationship between AMR prevalence data and GNI per capita for all bacterial phenotypes, except for VR E. faecium, for which the increase was found to be statistically non-significant (Table 1). The highest rate of increase per unit decrease in log GNI per capita was observed in 3GCR Klebsiella spp. (22.5%, 95% CI 18.2%–26.7%; P<0.0001), followed by CR Acinetobacter spp. (19.2%, 95% CI 11.3%–27.1%; P<0.0001) and 3GCR E. coli (15.3%, 95% CI 11.6%–19.1%; P<0.0001). The rate of increase per unit decrease in log GNI per capita was slightly lower in Gram-positive bacteria (MRSA 9.5%, 95% CI 5.2%–13.7%; P<0.0001; VR E. faecium 1%, 95% CI −6.3% to 8.3%; P=0.78). Results of the linear regression analysis are detailed in Table 1. The inverse association between national GNI per capita and AMR is graphically represented for 3GCR Klebsiella spp., CR Acinetobacter spp. and MRSA in Figure 2.

Figure 2.

Linear regression analysis displaying the association between prevalence of 3GCR Klebsiella spp. (a), CR Acinetobacter spp. (b) and MRSA (c) invasive isolates and GNI per capita. The solid line represents the estimated increase in resistance; the dashed lines represent the 95% CI.

Table 1.

Linear regression analysis showing the association between prevalence of antibiotic-resistant bacteria (%) and national GNI per capita (in US dollars)

Bacterial phenotype	Increase in prevalence of resistance (%) with 95% CI per unit log decrease in GNI per capita (US dollars)
	estimate (%)	95% CI	P value	R2 (%)
CR Acinetobacter spp.	19.2	11.3–27.1	<0.0001	−34.1
CR Klebsiella spp.	6.1	2.9–9.2	<0.0001	−28.5
CR P. aeruginosa	9.8	4.8–14.9	<0.0001	−26.1
MRSA	9.5	5.2–13.7	<0.0001	−27.2
3GCR E. coli	15.3	11.6–19.1	<0.0001	−61.5
3GCR Klebsiella spp.	22.5	18.2–26.7	<0.0001	−69.0
VR E. faecium	1.0	−6.3–8.3	0.78	—

Discussion

Our study shows a significant association between the lower national GNI per capita and the increased AMR prevalence for invasive infections caused by the top-ranked priority pathogens on the WHO priority list.¹⁰ By adding data from a further 18 countries and by updating prevalence data from 2011 onwards, our findings confirm and strengthen the association already highlighted by Alvarez-Uria and the CDDEP working group in 2016.⁸ For the first time we demonstrate a significant inverse association between income status and prevalence of invasive infections due to carbapenem resistance in Acinetobacter spp., P. aeruginosa and Klebsiella spp.

There are several contributing factors potentially explaining the relationship between increased AMR prevalence and lower income status. The most relevant is suboptimal hygiene conditions, inadequate sanitation and reduced availability of clean water.⁸ Our findings firmly support this hypothesis, since the strongest association between AMR and lower economies was found for 3GCR Enterobacteriaceae, whose transmission mechanism typically involves the faecal–oral route.^8,9,15

The climate represents an additional factor impacting on the AMR distribution. After having analysed 1.6 million clinically relevant isolates with different resistance patterns in the time frame 2013–15, MacFadden et al.¹⁶ demonstrated that an increase of 10°C across US regions was associated with an increase in AMR of 4.2%, 2.2% and 2.7% for E. coli, K. pneumoniae and S. aureus, respectively. More recently, this positive association has been confirmed at a global scale.⁹ Higher temperature might enhance environmental bacterial growth and ease transmission from food and other environmental sources to animals and to humans. Thus, the increased transmission might promote the selection of resistant strains in countries with higher temperatures.¹⁶

Resource-poor countries suffer also from unregulated access to antibiotics with a higher proportion of inappropriate prescriptions,¹⁷ inadequate treatment duration and dosing, medication sharing, or use of expired medications.^18–21 As for carbapenem resistance, both the high consumption of carbapenems²² and the lack of defined infection control policies and limited resources in the hospital setting could contribute to the greater spread of CR Gram-negative bacteria.²³

The relevance of anthropological and socioeconomic factors in determining AMR rates has been recently assessed by Collignon et al.,⁹ whose conclusion slightly differs from ours. Using 2008–14 surveillance and literature data from 103 countries, the authors created two indices expressing country AMR, while national income status was expressed as gross domestic product (GDP) per capita. In the regression analysis, a significant positive association was observed between GDP per capita and the first AMR index, defined as the worldwide average prevalence of fluoroquinolone-resistant (FQR) and 3GCR E. coli. Despite the same positive trend, statistical significance was absent when including the second AMR index, defined as the combined average prevalence of E. coli and Klebsiella spp. resistant to fluoroquinolone, third-generation cephalosporins, and carbapenems and MRSA.⁹ Several aspects make comparison with our results difficult. First, the AMR prevalence data used for generating the AMR indices were collected from different sources, including less recent data. Second, the AMR indices, incorporating heterogeneous antibiotic-resistant phenotypes, could mask the relationship between GDP per capita and single antibiotic-resistant phenotype. Third, the indices were both calculated including FQR E. coli, whilst our data extraction focused on more emergent threats, such as carbapenem resistance among Klebsiella, Acinetobacter and Pseudomonas.

Our study provides an updated picture of the current coverage of global surveillance focusing only on the most recent AMR data collected from 2012 onwards. According to our analysis, recent data are available from 22 surveillance networks accounting for a total of 67 countries. Gaps in surveillance have been detected at a global level. Fewer than 50% of HICs were covered by active surveillance and very scattered data were collected from several LMICs and LICs, especially in Africa, South-East Asia and Eastern Mediterranean areas. So far, only 13 LMICs/LICs are actively surveyed: Egypt, Georgia, Ghana, India, Kenya, Kosovo, Malawi, Pakistan, Philippines, Tunisia, Vietnam, Zambia and Zimbabwe. It is worth noting that in lower-resource settings microbiological diagnostics are often inadequate.^24,25 Laboratories frequently do not meet the basic requirements owing to lack of equipment and undertrained personnel.^26,27 These drawbacks strongly limit the ability to identify microorganisms and to conduct antimicrobial susceptibility testing, making surveillance difficult to perform.²⁸

Since surveillance is one of the key points in the fight against AMR, major stakeholders and organizations have recognized the urgent need to support LMICs and LICs in defining actionable interventions aimed at reinforcing the surveillance and implementing several public health initiatives.

The Global Antibiotic Resistance Partnership (GARP; https://cddep.org/partners/global-antibiotic-resistance-partnership/garp-network/) is a project supported by the CDDEP started in 2009 with the purpose of implementing AMR surveillance in lower-income countries. GARP has sustained the creation of multisectoral working groups aimed at documenting antibiotic use and AMR in the human and animal sector in the national context, and successively at developing evidence-based proposals to slow the spread of AMR and to preserve antibiotic effectiveness. So far, 15 countries are successfully involved and surveillance data are now publicly available for 6 of them. ReAct (https://www.reactgroup.org/), set up in 2005, is an international independent global network advocating engagement in the field of AMR by collaborating with a broad range of organizations, individuals and stakeholders. ReAct supports, together with partners from WHO, GARP and other organizations, the establishment of comprehensive national action plans, highlighting the strengthening of surveillance as an essential component of the national plans. ReAct is currently working in 12 LMICs/LICs.

The Fleming Fund (https://www.flemingfund.org/), a UK aid programme working in partnership with WHO, the Food and Agriculture Organization and the World Organisation for Animal Health, is funding a wide range of key initiatives and activities with the aim of implementing sustainable AMR surveillance systems in LMICs/LICs. By 2022, the Fleming Fund will have helped 24 LICs and LMICs across the world to establish a sustainable national AMR surveillance programme.

In October 2015 WHO launched the Global Antibiotic Resistance Surveillance System (GLASS; http://www.who.int/glass/en/), as a part of the global action plan for AMR.²⁹ GLASS is the first global collaborative effort to standardize the AMR surveillance for eight selected antibiotic-resistant bacteria: Acinetobacter spp., E. coli, K. pneumoniae, Neisseria gonorrhoeae, Salmonella spp., Shigella spp., S. aureus and S. pneumoniae. During the early implementation phase (2015–19), a total of 50 countries have been enrolled, of which 11 and 6 are LMICs and LICs, respectively.³⁰ In Africa, seven countries (Burkina Faso, Ethiopia, Kenya, Mauritius, Tanzania, South Africa and Zimbabwe) have had their national action plan approved, while a further five countries (Ghana, Malawi, Nigeria, Uganda and Zambia) are awaiting approval. The major benefit targeted by GLASS is the switch from a surveillance approach based uniquely on microbiological data (isolate-based data) to a network that includes additionally epidemiological, clinical and population-level data (e.g. morbidity, mortality and costs).³¹

Our study has the main limitation of including data extracted from national AMR surveillance, which might not be always representative of the true resistance status in the whole country. Moreover, we considered only online publicly available data in the English language, which might have favoured the inclusion of countries from the European and American regions. However, some specific considerations were taken into account in the attempt to maximize the reliability of our results. First, we restricted inclusion to surveillance systems clearly reporting resistance data based on validated breakpoints. Second, we extracted most of the data from the CDDEP ResistanceMap repository, which provides harmonized resistance definitions and allows inter-country comparisons. Third, the association between resistance rates and GNI per capita was evaluated with linear regression using variance-weighted least squares. Thus, each country contributed to the analysis as a single unit and higher weights were attributed to countries with higher numbers of tested isolates.

Taking into account all these considerations, despite the scattered and heterogeneous data distribution, our study shows a strongly significant inverse association between income status and invasive infections due to MRSA, 3GCR Klebsiella spp. and E. coli, CR Acinetobacter spp., Klebsiella spp. and P. aeruginosa. Public health officers and policy makers must take determinants of poverty and inequality into account when designing and implementing interventions and infection control policies targeting AMR, especially in LMICs/LICs.

Acknowledgements

We thank Ruth Joanna Davis for editorial support and Alessandro Tavelli for the graphical editing.

Funding

The contribution of A. S. was partially supported by the EPI-Net COMBACTE-MAGNET project (Innovative Medicines Initiative Joint Undertaking, grant agreement number 115737), resources of which include financial contribution from the European Union Seventh Framework Programme (FP7/2007–2013) and European Federation of Pharmaceutical Industries and Associations companies in-kind contribution.

Transparency declarations

None to declare. The corresponding author had access to all data and responsibility for the decision to submit for publication.