Clinically significant levels of resistance to various antibiotic classes have developed in most Gram-negative bacteria, though there is marked geographic variation in the prevalence of resistance [1]. Historically, many forms of antibiotic resistant organisms emerged in and were disseminated from healthcare facilities in high-income settings, but in more recent times, hospitals in low and middle-income settings appear to have been the origin and amplifiers of several forms of resistance [2–4]. Weak antibiotic stewardship and limited infection control resources promote the spread of antibiotic resistant organisms in hospitals, especially where there are no microbiological diagnostic services for the identification of resistance [5].

In the family of Enterobacteriaceae, organisms such as Escherichia coli and Klebsiella pneumoniae can produce extended-spectrum β-lactamase (ESBL) enzymes, which confer resistance to third generation cephalosporins and most penicillins but are inhibited by agents such as clavulanate. These types of resistant bacteria are increasingly implicated as causes of both community- and hospital-acquired infections [6, 7] and are associated with higher levels of mortality than non-ESBL producers [8]. Across sub-Saharan Africa, a limited number of studies have described high levels of ESBL faecal carriage in children in association with hospital facilities [9–13]. Gentamicin is an aminoglycoside antibiotic that has been used for the treatment of Gram-negative infections for more than 40 years [14]. Its low-cost and broad spectrum of activity means that it is widely used for treatment of severe infections in many low-income countries.

The gastrointestinal tract is the principal ecological niche for Enterobacteriaceae. In this environment, inter- and intra-species exchange of resistance genes can occur and under appropriate selective pressure, resistant species can rapidly emerge and dominate [15]. While faecal carriage of antibiotic resistant bacteria is not an immediate threat to a healthy individual, it poses two risks—firstly, when auto-infection of a sterile body-site subsequently occurs (for example, hospital-acquired bacteraemia [16]) treatment is substantially more difficult and secondly, there may be transmission of the resistant organism to other individuals [17].

In order to describe levels of carriage of Gram-negative antibiotic resistance amongst hospitalised children in Zimbabwe, we aimed to measure 1) the prevalence of carriage of ESBL producing and gentamicin-resistant enteric bacteria in paediatric patients at the time of admission to hospital and 2) the risk of acquisition for the same types of bacteria during the corresponding time of hospitalization. We also sought to evaluate different methods for screening for the presence of these resistant organisms.

Our study investigated enteric carriage of ESBL-producing and gentamicin resistant enteric bacteria in a sample of 164 hospitalised children drawn from two tertiary public-sector hospitals in Harare, Zimbabwe over a 3 month period. The main findings were high rates of carriage of ESBL and gentamicin resistant bacteria on admission to hospital and high rates of acquisition for the same resistance types during the inpatient period.

An initial sample was collected within 24 h of hospital admission for each patient, so these samples are likely to represent the carriage status prior to the current hospital admission. Other studies of ESBL prevalence around the time of admission to hospital have found carriage rates of 33% (133/408) in Guinea Bissau [9], 34% (37/110) in Gabon [12] and 22% (54/244) in Madagascar [11]. Amongst neonates in Tanzania, 25% (32/126) were found to have ESBL carriage, predominantly Klebsiella pneumoniae [13]. In a study conducted in a renutrition centre in Niger [10], ESBL carriage was 31% (17/55 patients) on admission and by time of discharge a further 15/16 children had acquired carriage. Our study, showed a higher prevalence of carriage than those previously described levels, though our methods of screening were probably more sensitive than the direct plating of rectal swabs used elsewhere. We are not aware of any previous African studies specifically investigating carriage of gentamicin resistance in hospitalised children. Carriage of resistance amongst hospital admissions is liable to be higher than carriage in the general community, as sick individuals probably have exposure to various healthcare-related risk factors prior to admission, so we cannot compare these results to truly community-based studies. However, the extremely high levels of carriage rates found suggest that circulation of ESBL and gentamicin resistance probably occurs in the community in Harare. Although we found an association between ESBL carriage and inpatient death, we were not able to assess for the confounding effects and hence it remains uncertain whether this is a causal association.

Collection of multiple follow-up samples during the inpatient period allowed us to observe acquisition of carriage of resistance over time. An important question is whether these in-hospital acquisitions are occurring via endogenous (arising from undetectably low levels already present in the gastrointestinal tract) or exogenous (arising from cross-infection from an external source) routes. This distinction has important implications for antibiotic stewardship and infection control measures. For endogenous routes of acquisition, antibiotic stewardship activities focussing on reducing the selective pressure from antibiotic prescribing are important. For exogenous acquisition, measures to minimise cross-infection, such as improved hand hygiene, are priorities. Whilst we cannot be certain, we feel that both routes are likely to be contributing to the high levels of nosocomial acquisition seen here. Our screening methods included an enrichment step so that we should have been able to detect small quantities of resistant bacteria, making endogenous acquisition less likely, though a single faecal specimen is not a representative sample of the estimated 10¹⁴ cells in the intestinal microbiome. Enrichment broths with supplementary antibiotics have been used elsewhere to aid detection of resistant organisms in faecal samples [18]. Our clinical experience in these hospitals suggests there are substantial challenges with over-crowding, limited access to hand-hygiene facilities and high patient:staff ratios. With regard to hand-hygiene, at the time the study was conducted, water supplies in these wards were sporadic and alcohol-based hand rub (ABHR) was not widely available. Concurrent work conducted in the same wards as a separate part of this pilot study showed that hands of adult carers were frequently contaminated with Gram-negative antibiotic resistant bacteria [19]. These are all factors that might be expected to contribute to exogenous routes of transmission.

In line with findings from other studies, we found that most ESBL-producing bacteria from these children were also resistant to several other clinically-relevant antibiotic classes [9–13]. This represents a substantial clinical challenge for empirical treatment of suspected hospital-acquired infections—ciprofloxacin (for severe urinary sepsis) and chloramphenicol (for bacteraemia and meningitis) represent important and widely used second-line treatment options in many African hospitals. The level of carbapenem resistance amongst resistant isolates was very low, but as these agents are not currently routinely available through public-sector pharmacies in Zimbabwe, this information is of limited practical value for local clinicians.

Due to rapid turnover of patients, we were only able to collect follow-up samples in a moderate number of patients; 73 patients had only a single sample collected. Whilst this does not affect estimates for the time of admission, it limits the ability of the study to detect new acquisition of carriage—in reality, probably even more children became carriers of resistance during admission than we detected. Ideally, an additional sample on the day of discharge would have been collected for all patients. Our simple screening and identification methods meant that we could not distinguish the simultaneous carriage of multiple different resistant organisms. Again, this means that the acquisition risk described here may underestimate the true in-hospital acquisition risk. We did not make use of molecular techniques for identification of resistance mechanisms or for investigating transmission of resistance. In this preliminary study, we were not able to evaluate clinical impacts of antibiotic resistance, either during hospital inpatient stay or after discharge. Whilst we did find some evidence that carriage of ESBL resistance was associated with inpatient mortality, investigating whether carriage of resistance was followed by clinical infection with the same organisms was beyond the scope of the study.

Future research is needed to better understand and tackle these high levels of antibiotic resistance. This will take many forms. We did not collect information on risk factors associated with carriage of resistance—this would be a valuable first step in understanding why this is occurring. In other settings, prior use of antibiotics and recent hospital admission have been associated with carriage of resistance [20]—we do not know if this is the case in Zimbabwe. More work is needed to identify the enzyme types and genes responsible for the forms of resistance that we detected—this will enable us to start to understand the dynamics of resistance transmission. Screening for drug-resistance on admission and cohorting or isolating patients can result in improved patient outcomes [21]—but no modelling work has ever been conducted to explore the value of this in the setting of African hospitals. We do not know how the acquisition of antibiotic resistance in hospitals relates to resistance in the community in Zimbabwe [22], nor do we know the wider impact of antibiotic resistance in terms of mortality, morbidity and costs in sub-Saharan African countries, though we suspect these to be substantial [23]. Given that many African hospitals have limited access to carbapenems, infections caused by ESBL-producers might have a substantially greater impact than has been seen in high-income settings [8].

At a more practical level, immediate action can be taken in a number of areas such as improving infection control practices, especially hand hygiene, which has been shown to reduce infections caused by antibiotic-resistant Gram-negative bacteria in high-income settings [24]. The use of alcohol-based hand rub as part of the WHO 5 moments for hand hygiene approach is a low-cost technology that could be much more widely adopted in African hospitals. Ideally, a rigorous evaluation of the implementation of such new practices would contribute to the evidence-base for these measures in low-income settings, which is currently extremely limited [25].

The rising level of antimicrobial resistance is of great concern worldwide. These high rates of ESBL and gentamicin resistance carriage found amongst paediatric inpatients in Zimbabwe are particularly worrisome. These bacteria pose risks of multi-drug resistant infections and transmission to other hospital inpatients. The high levels of carriage at the time of hospital admissions suggest that they are already in widespread circulation in the general community. Relevant infection control measures such as access to alcohol-based hand rub should be urgently scaled up in these hospitals and beyond, so that children who arrive at hospital without carriage of antibiotic resistant bacteria can remain free of these for the duration of their admissions.