Clinical and economic impact of antibiotic resistance in developing countries: A systematic review and meta-analysis

1Antimicrobial resistance (AMR) is the ability of bacteria, parasites, viruses and fungi to grow and spread in the presence of antimicrobial medicines that are normally active against them. AMR occurs via a range of resistance mechanisms, such as a modified antimicrobial target, enzymatic hydrolysis/degradation, efflux and impermeability. This resistance is mediated by diverse resistance genes that evolve as a result of antimicrobial selection pressure exerted by the appropriate and/or inappropriate use of antimicrobial medicines, and is aggravated by the void of new antimicrobial agents in the current therapeutic pipeline. AMR increases health-care costs, length of stay in hospitals, morbidity and mortality in both developed and developing countries. A recent report estimated that 10 million deaths will be attributed to AMR by 2050, and 100 trillion USD of the world’s economic outputs will be lost if substantive efforts are not made to contain this threat.

2The World Health Organization (WHO) published the first global surveillance report on antibiotic resistance (ABR) in 2014 to show the clinical impact of resistant bacteria in WHO regions across the world. This reported shown that five out of the six WHO regions had more than 50% resistance to third generation cephalosporins and fluoroquinolones in Escherichia coli and methicillin resistance in Staphylococcus aureus in hospital settings. Similarly, more than 50% resistance to third generation cephalosporins and carbapenems was reported in Klebsiella pneumoniae. The report attributed 45% of deaths in both Africa and South-East Asia to multi-drug resistant (MDR) bacteria. It further revealed that K. pneumoniae resistant to third generation cephalosporins was associated with elevated deaths in Africa (77%), the Eastern Mediterranean region (50%), South East Asia (81%) and Western Pacific region (72%).

3Several resistant bacteria have been increasingly involved in infectious diseases in humans, specifically, Enterococcus spp, S. aureus, K. pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter spp. They are collectively termed ESKAPE and recently gained further global attention by being listed by the WHO as priority antibiotic-resistant bacteria to guide research, discovery, and development of new antibiotics. The particularity of these bacteria is their ability to develop high level resistance to multiple drugs, thereby limiting therapeutic options and increasing morbidity and mortality. Numerous studies have confirmed that ESKAPE bacteria and their resistant clones, are actively transmitted in hospitals and communities in both developed and developing countries. The threat posed by these resistant bacteria is however exacerbated in developing countries due to sub-optimal hygiene conditions, poor infection, prevention and control measures, lack of surveillance and the dearth antimicrobial stewardship programs. Reports have shown high isolation rates of methicillin resistant S. aureus (MRSA) in healthcare settings in Cameroon (72%), South Africa (52%), Ethiopia (42.8%), Nigeria (29.6%), Kenya (27.7%), Ivory Cost (16.8%) and Morocco (14.4%). In 2008, the prevalence of nosocomial acquired and MDR infections due to Enterobacteriaceae isolated from blood cultures were 57.1% and 15.4% respectively, in South Africa. Likewise, rapid increases in the rates of infections due to carbapenemase-producing K. pneumonia, metallo-beta-lactamase-producing A. baumannii (MBL-AB), metallo-beta-lactamase-producing P. aeruginosa (MBL-PA), and extended-spectrum beta-lactamase (ESBL) producing Enterobacter spp. have been reported across the world. In Saudi Arabia, the rate of P. aeruginosa producing carbapenemase was 33%, of which 27% were MBL-producers, while in India, a 22.4% prevalence of P. aeruginosa producing MBLs was reported in tertiary care hospitals.

4MDR-ESKAPE bacteria have been reported in hospital acquired infections (HAI), particularly in intensive care units (ICUs) where immune-compromised patients suffering from some non-communicable diseases (NCDs) including diabetes, cancers, chronic lung, cardiovascular and kidney diseases were highly affected. The emergence and spread of these highly resistant bacteria in hospital care settings could thus have negative health repercussions and be an obstacle for the treatment of infections of patients with these NCDs.

5Despite the evidenced threat posed by ABR, information on its clinical and economic impact is limited in developing countries, and thus impede appropriate interventions for its containment. Heightened awareness of policy-makers, health care workers, and the general population about the risks associated with ABR is essential to preserve antibiotics for future generations. Hence, the aim of this study was to analyze the published literature on the clinical and economic impact of ABR in developing countries, in order to inform containment strategies such as antimicrobial stewardship programs and infection prevention and control measures in these nations.

6The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were followed.

7This systematic review and meta-analysis was based on published reports, and was therefore exempt from ethical approval.

8A systematic search was carried out independently by RF and LF, in Medline via PubMed and Web of Sciences from January 2000 to December 09, 2016, using a combination of boolean operators (AND/OR), Medical Subject Heading (MeSH) and pre-defined keywords. Only published after 2000 were considered to ensure that the analysis focuses on contextual literature that depict current resistance patterns, infection rates, prevention measures, and clinical practice guidelines. Peer-reviewed papers in English and French on the clinical and/or economic impacts of ABR in developing countries were retrieved and independently evaluated for eligibility by RF and LF based on titles and abstracts (Table 1). Thereafter, the full-texts of eligible papers were assessed according to pre-defined inclusion and exclusion criteria (Table 1), with inconsistencies and disagreements being resolved by consensus. Efforts were made to contact the authors when data was missing and full-texts could not be retrieved, and a hand search was conducted in the reference list of all selected papers.

9Papers were managed using EndNote (version X7.7.1, Thomson Reuters) and the data from eligible papers was abstracted independently by two authors (RF and LF) using a standardized data extraction spreadsheet in Excel® (Microsoft® Office Excel 2016). Relevant data from papers included countries, WHO regions, World Bank classification, publication year, type of study, participant characteristics (number of participant, diseases, age), hospital’ ward, bacteria, follow-up period, length of stay in hospital, mortality related to resistant bacteria, and, costs as described in Table 2.

10Meta-analyses were undertaken using Comprehensive Meta-analysis software (Biostat, Inc., New Jersey, USA) version 3 for Windows, to determine overall mortality risk associated with resistance. Sub-group analyses for mortality were conducted for the data by WHO region, World Bank classification, countries, group of bacteria, and bacterial species where there were three or more studies that could be combined. Forest plots were performed to assess the significance of the results and generated using 95% confidence intervals (CIs). Analyses were undertaken across sub-groups for the selected outcome and the results presented as odds ratios. Studies were weighted in favor of those with narrower confidence intervals (more precise results), and the random-effects method was used to provide more confident data considering heterogeneity within and between reports. The I-square (I2) statistic with cut-off values of 25, 50 and 75% was used to assess low, moderate and high heterogeneity respectively, and a p-value of <0.05 was considered statistically significant. Publication bias was evaluated using the funnel plot and statistical egger’s test.

11Quality assessment was performed independently by RF and LF using the Newcastle-Ottawa quality assessment scale (NOS) for each study included in the systematic review and meta-analysis. NOS assesses methodological quality, based on three-dimensional criteria and included (i) selected population, (ii) comparability of groups, and (iii) outcome/exposure of interest. Studies were scored using a scale with a possible maximum of eight points where a score ≥ 6 indicated high-quality studies, a score between 3–6 as moderate and a score ≤ 3 as low quality.

12The systematic search conducted in the two electronic databases generated 27 033 papers. A total of 24 057 papers were screened for probable inclusion according to titles and abstracts after de-duplication. Of these, the full texts of 92 eligible papers were fully evaluated based on predefined inclusion and exclusion criteria. One article was added following a hand-search in the reference lists of included papers. Forty studies were finally eligible for the qualitative and quantitative analysis (Fig 1), of which 18 were of high quality, while 15 and seven were moderate and low quality respectively.

13The majority of data analyzed were obtained from single center studies conducted in 11 countries. Thirty percent (n = 12) of the observational studies on ABR were conducted in hospitals and communities in Thailand, the rest were performed in 10 different-countries namely Brazil (n = 7; 17.5%), China (n = 6; 15%), Turkey (n = 5; 12.5%), Colombia (n = 2; 5%), Malaysia (n = 2; 5%), India (n = 2; 5%), Mexico (n = 1; 2.5%), Jordan (n = 1; 2.5%), Palestine (n = 1; 2.5%), and Senegal (n = 1; 2.5%) (Table 2 and Fig 2).

14Fourteen studies investigated the impact of ABR on mortality, two reported its impact on morbidity only (Table 2) while 24 considered both morbidity and mortality concomitantly. Eight studies reported on the economic consequences of ABR (Table 3). A. baumannii (n = 14; 35%), K. pneumoniae (n = 6; 15%), S. aureus (n = 5; 12.5%), P. aeruginosa (n = 4; 10%) represented the main pathogens reported with ICUs being the principal hospital ward concerned (Tables 2 and 3).

15Pooled estimates revealed 90% prevalence (95%CI, 2.852–3.557; p = 0.000) of mortality attributable to infections in developing countries with greater mortality associated with ABR at an odds ratio (OR) 2.828 (95%CI, 2.231–3.584; p = 0.000) (Fig 3A).

16The subgroup analyses were performed by World Bank classification, WHO region, country, group of bacteria and bacterial species. Fig 3B presents a forest plot of mortality due to AMR categorized per World Bank classification. The risk of mortality due to resistant bacteria was high in upper middle-income countries (OR 2.769, 95% CIs, 2.142–3.579; p = 0.000), with studies from lower-middle and low-income nations not being evaluated due to insufficient data.

17Four out of the six WHO regions were included in the analysis, with three showing a high risk of mortality (Fig 3C). High statistical significance was observed in the Americas (OR 2.126, 95% CIs; 1.546–2.925; p = 0.000), South East Asia (OR 3.754, 95% CIs; 2.333–6.041; p = 0.000) and the Western Pacific (OR 3.746, 95% CIs; 2.463–5.697; p = 0.000) (Fig 3C). Results from Europe were not statistically significant and insufficient reports precluded analysis in Africa.

18Subgroup analyses per country showed high statistical significance (OR 2.665, 95%CIs; 2.074–3.425, p = 0.000) (Fig 3D) in favor of mortality. Brazil, China and Thailand, had statistically significant risk of mortality with OR being 1.825 (95%CIs; 1.239–2.689; p = 0.002), 3.746 (95%CIs; 2.463–5.697; p = 0.000), 3.928 (95%CIs; 2.116–7.293; p = 0.000) respectively, in contrast to Turkey, which was not statistically significant (Fig 3D). In other countries, the number of reports was insufficient (less than three) to perform the meta-analysis.

19Studies were categorized into three groups of bacteria namely ESKAPE, non-ESKAPE, and mixed (both ESKAPE and non-ESKAPE). The ESKAPE group was associated with the highest risk of mortality with a high statistical significance (OR 3.217; 95%CIs; 2.395–4.321; p = 0.001) (Fig 3E). Although, the non-ESKAPE group was not associated with the risk of mortality (OR 1.167; 95%CIs; 0.385–3.534; p = 0.000), when combined with ESKAPE within a study, it became statistically significant (OR 2.634; 95%CIs; 1.858–3.734; p = 0.000) (Fig 3E).

20High risk of mortality due to antibiotic-resistant A. baumannii was observed with high statistical significance (OR 4.636; 95%CIs; 2.954–7.277; p = 0.000), followed by S. aureus (OR 2.842; 95%CIs; 1.868–4.324; p = 0.000). P. aeruginosa (OR 2.076; 95%CIs; 0.833–5.177; p = 0.117) and K. pneumoniae (OR 2.026; 95%CIs; 0.733–5.598; p = 0.173) were not significantly associated with mortality (Fig 3F).

21AMR is a global public health threat that affects human health, particularly hospitalized patients, and has substantive health and financial consequences. This study analyzed the published literature on the clinical and economic implications of ABR in developing countries from 40 eligible studies. Antibiotic-resistant bacteria were associated with increased mortality (OR 2.8341, 95%CIs; 2.2180–3.6213; P = 0.000), consistent with several reports in both developed and developing countries. The main ward involved was the ICU, possibly due to the heavy use of antibiotics and hence the selection pressure for ABR development and prevalence in these units). This concurred with studies from Mexico, Brazil, China, Thailand, France and Serbia, that reported high mortality due to antibiotic-resistant bacteria in ICUs. The study further showed that ABR research is neglected in developing countries with only one report from low-income (Senegal), two from lower-income (Palestine and Jordan), and 37 from upper-middle income nations (Table 1 and Fig 2). Developing countries are thus far behind high resource settings in the fight against AMR and that requiring considerable efforts to reduce its consequences. Three WHO regions, i.e., the Americas, South East Asia and the Western Pacific region showed the highest risk of mortality associated with MRSA and K. pneumoniae resistant to third generation cephalosporins. Our results concurred with the 2014’s WHO report, which showed a significant increase of mortality due to antibiotic-resistant K. pneumoniae and S. aureus in hospitals particularly in ICU across WHO regions. Resistance levels could be explained by the practices of self-medication and the purchase of antibiotics over-the-counter common in these settings. Policies and regulations promoting rational antibiotic use are also minimal or non-existent. Additionally, limitations in managing nosocomial infections, sub-optimal infection control measures, unsafe water, poor hygienic conditions, lack of knowledge and inadequately trained personnel might also be associated with the prevailing resistance in these regions. Comprehensive studies are needed to provide accurate and reliable data to inform decision-makers about the danger of ABR in developing countries and suggest a way forward for the alleviation of the resulting implications.

22Resistant ESKAPE bacteria including carbapenem-resistant A. baumannii, MBL- producing P. aeruginosa, ESBL-producing K. pneumoniae, and MRSA represented the most common resistant bacteria associated with increased mortality. These bacteria were the main cause of morbidity and mortality in bloodstream infections in hospital settings, with a high statistical significance (OR 2.978, 95%CIs; 2.362–3.753; p = 0.000) (Fig 3F). This concurred with the WHO Global Antimicrobial Surveillance System (GLASS), which recognized A. baumannii, K. pneumoniae, and S. aureus, as priority pathogens in blood specimens and list them together with P. aeruginosa as priority antibiotic resistant-bacteria for research and development in 2017.

23According to the meta-analysis, MDR-ESKAPE were associated with a greater risk of mortality than mono-drug (including imipenem, methicillin, and linezolid) resistant bacteria, with a high statistical significance (OR 2.846, 95% CIs; 1.744–4.643; p = 0.000; versus OR 2.301; 95%CIs; 1.718–3.082; p = 0.000; Table 3). Moreover, when comparing the mortality risk between resistant- and susceptible-ESKAPE pathogens (Table 3), results showed that carbapenem-resistant A. baumannii (CRAB) were associated with higher mortality risk than susceptible strains with a high statistical significance. The pooled estimate of mortality rate ranged from 15.9 to 91.7% (p = 0.001), consistent with a report from Taiwan, where a significant increase of mortality from 14% to 46% (p = 0.0001) was associated with carbapenem-resistant-A. baumannii implicated in HAIs during 2003–2008.

24Although the mortality attributable to ESKAPE pathogens is indisputable compared to non-ESKAPE pathogens, we observed that when these two groups infected patients concomitantly, they were associated with a long length of hospital stay (LOS) and a higher mortality. This concurred with studies from Senegal, Turkey and China which have reported high LOS and death due to MDR-A. baumannii, ESBL-producing Enterobacteriaceae and MRSA, respectively.

25Eight studies reported that ABR increased health care costs with resistant ESKAPE bacteria being the main causative agents associated with high hospital costs (Table 4). Four out of the eight revealed that length of stay had an impact on hospital costs. LOS was also a risk factor for acquisition of nosocomial infections, and thereby increased mortality. Overall, health-care costs in all studies for case and control groups were 8,107.375 USD versus 5,469.487 USD respectively. Two studies indicated health care costs >10 000 USD in Thailand and Colombia while one report showed cost ≥ 35 000 USD in Turkey. In contrast, three studies reported overall hospital costs ≤ 1000 USD, with one below 250 USD in Senegal. These differences are attributed to the diverse socio-economic characteristics of the countries concerned.

26In terms of the limitations of the study, several papers were not included in the meta-analysis because they did not provide sufficient information regarding clinical and/or economic impact of ABR in developing countries. We were unable to present the genomic characteristics of antibiotic-resistant bacteria due to the scarcity of data. In addition, we did not focus on antibiotic classes and resistance patterns due to the lack of standard methods for identification and interpretation in developing countries. Moderate heterogeneity (I2 = 58.88%, p = 0.000) was reported, which could be due to various external factors, such as different type of studies (retrospective, retrospective cohort, retrospective case-control, prospective cohort, prospective case-control, etc.), diverse populations (adult, children, neonates), infection prevention and control measures and antimicrobial stewardship practices. Moreover, minor publication bias was observed in the funnel plot (Fig 4) which could possibly be attributed to the lack of reports from lower-middle and low-income countries. We tried to limit the influence of heterogeneity and publication bias in our statistical analysis by using the random effects model that considers differences within and between studies, as well as by including articles in different languages (English and French).

27The key findings of this study confirm that ABR, particularly antibiotic-resistant ESKAPE pathogens are associated with a high risk of mortality and greater economic costs. Developing countries need to optimize their management of communicable and non-communicable diseases, implement infection, prevention and control (IPC) measures, as well as antimicrobial stewardship programs (ASP) in both hospital and community settings to reduce morbidity, mortality and the costs associated with ABR. Furthermore, optimization of rational antibiotic use at regional and national levels, is essential to ensure a high quality and effective of therapeutic options. Substantial and sustainable efforts to develop rapid diagnostics, new antibiotics and vaccines are required. An international platform for global real-time surveillance and monitoring of antimicrobial resistance could advance containment of this threat.