Experience with developing antibiotic stewardship programs in Serbia: potential model for other Balkan countries?

A retrospective drug utilization (DU) and surveillance study was conducted among the pediatric and pediatric surgery clinics (125 beds) of a tertiary care institution in Serbia, the Clinical Centre Kragujevac hospital, which overall has 1274 beds. The number of beds available for children (patients aged 0–18 years) did not significantly change during the study period, with the number of bed days oscillating between 28,576 and 36,171 per year in the pediatric wards.

In the first part of the study, we undertook an analysis combining antimicrobial utilization data with cumulative resistance in 2010 versus 2014.

Antibiotic utilization data were measured using the Anatomical Therapeutic Chemical Classification (ATC)/defined daily dose (DDD) methodology, with DDDs typically accepted for recording medicine use for comparative purposes [20–24]. We are aware that DDDs are normally assigned based on their use in adults, and for medicines approved for use in children, dose recommendations will differ based on children's age and body weight. In addition, many medicines used in children are typically not approved for such use, and documentation regarding dosing regimens are generally unavailable. Consequently, the WHO International Working Group for Drug Statistics Methodology concluded that pediatric DDDs are impossible to assign, and prescribed daily dosages and indications in pediatric population should be used if available [25]. However, if this is difficult, the Working Group suggested that DDDs should be used as a measuring tool for overall comparisons [25].

Antibiotic dispensing data were collected from the hospital pharmacy on all antibiotics prescribed for systemic use among children from 2010 to 2014. In order to concentrate on the most prescribed antibiotics, the data on their utilization were limited to those antibiotics which comprised 90% of the total utilization expressed in DDDs/100 patient days [23,26,27]. At the same time, we followed the percentages of bacterial isolates resistant and susceptible to the same antibiotics that were used in the study site. Utilization and resistance rates were combined. Cumulative resistance rates were calculated for those microbes naturally susceptible to each antibiotic from those antibiotics comprising 90% of the total antibiotic utilization as a percentage of resistant, intermediate or susceptible strains from the total number of strains analyzed. Resistant and intermediate strain data were subsequently combined. This is part of ongoing antimicrobial stewardship programs in the hospital.

We are aware that AS programs do differ in their content from hospital to hospital, and from country to country, to reduce infections and colonization with antibiotic-resistant bacteria within hospitals [15,28,29]. We are also aware that there are limited data available on their implementation and effectiveness among pediatric patients [30]. In the Clinical Center Kragujevac, AS was composed of the following elements: establishing a drug and therapeutics committee; issuing antibiotic prescribing policies and hospital formularies via the drug and therapeutics committee; biannual analysis of resistance patterns among the isolates from the central intensive care unit (ICU) and its distribution to all clinicians; preauthorization of reserve antibiotics dispensing by a clinical pharmacology specialist; issuing local guidelines for antibiotic prophylaxis and empiric treatment; and consulting clinical pharmacologists and infectious diseases specialists when prescribing antibiotics to complex patients. The AS was fully implemented in this hospital at the beginning of 2014, and is ongoing.

Data on AMR were obtained from hospital microbiological laboratory. This included more than 90,000 uniform Excel files from 2010 to 2014 containing information, for instance, regarding the clinic/ward where the specimens were collected and in which material, for example, sputum, blood or liquor, which bacteria were isolated, what antibiotics were used to test potential bacteria resistance and what the results were. Using Excel macros, data involving children were extracted for each year. The data were subsequently filtered to improve the understanding in tables and graphs. Duplicate samples with the same isolates from the same patients were taken into account as one sample.

We included only species with at least ≥30 isolates tested. Under certain circumstances, when we did not have ≥30 isolates, we combined the isolates form two consecutive years into the calculations [27]. All sources of potential microbes, including pus, sputum, wounds, blood and urine, were analyzed together since we wished to represent the total volume of resistant pathogens circulating in the hospital. Microbiology reports define resistant, intermediate or sensitive strains according to the Clinical and Laboratory Standard Institute (CLSI) standards [31]. Antimicrobial susceptibility systems employed by the microbiology department have not changed through the years.

The costs per DDD were also calculated to help with future guidance. Cost data was recorded in local currency (RSD) to avoid problems with currency conversions. This included developments in the procurement system, including centralized procurement, to help reduce costs [32].

In the second part of the study, we looked at AMR patterns in more detail for five consecutive years to determine specific cumulative resistance and utilization rates for the predominant bacterial population to provide future guidance. For these microbes, cumulative resistance rates were calculated for the most used antibiotics from the antibiotic panel used in the sensitivity tests with the same methodology described above.

Data from clinical and surgical units included in the study are provided in one group as the first step of introducing a new surveillance system in hospital settings with limited resources. This methodology can be further modified by the type of medical specialty (medical vs surgical, ICU and non-ICU) since the nature of any surveillance system is to disentangle the epidemiology of antibiotic use and define areas at risk of antibiotic misuse.

No ethical approval was sought since this study collected anonymized aggregated data. This is in line with other studies employing similar methodologies [22,33–35], and is the current situation in Serbia.

Developing effective antibiotic policies in hospitals depends on the surveillance of current resistance patterns, coupled with an understanding of current antibiotic utilization patterns, to guide empiric use while sensitivity analyses are being undertaken. Hence, it should become mandatory for hospitals to establish efficient surveillance systems along with monitoring current antibiotic utilization to improve their appropriate use. This can be helped by instigating measures, such as the WHONET Software program, to monitor local resistance patterns; however, we believe only one hospital in Serbia is currently using this program [37].

From the antibiotic DU90% profile in 2010 versus 2014, we can conclude that the cumulative antimicrobial resistance is similar and relatively high (Figures 1 & 2 & Supplementary Tables 1 & 2), with high cumulative resistance rates among the most widely used antibiotics from 2010.

Overall, total antibiotic use decreased appreciably from 2010 to 2014 (Supplementary Tables 1 & 2); however, there was high prescribing of the reserved antibiotic ceftriaxone. In addition, sensitivity tests were not available for 4 out of the 15 most commonly used antibiotics in 2014, pointing to a lack of integration between microbiology and routine clinical care in the hospital. We will now be reviewing this within the hospital to develop policies to improve future prescribing, and this will be part of future studies. This can also act as an exemplar for other hospitals in Serbia to improve future antibiotic prescribing.

One of the main findings of this study is the high resistance of S. pneumoniae to azithromycin (Figure 3). This maybe because there has been high total antibiotic consumption in Serbia versus other European countries in recent years [38]. In addition, antibiotic use among children in primary care in Serbia has been extremely high, with frequent prescribing for indications with little or no benefit from antibiotics such as upper respiratory tract infections [27]. This is important since previous antibiotic exposure in primary care is related to high antibiotic resistance in hospitals [39].

The debate continues in literature with regard to the impact of macrolide resistance on the outcome of pneumococcal pneumonia. Antibiotic-resistant strains increase the severity of illness and make it more difficult to treat these infections effectively. Several cases of macrolide treatment failure have been documented in the literature [40,41], and with high levels of macrolide resistance in this children's hospital, we believe these antibiotics should not be routinely used as empiric therapy of S. pneumoniae while AMR rates remain high. Probable reason for the increase in the resistance rate of azithromycin and amoxicillin in 2014, despite a decrease in their intrahospital use, is the extensive and unjustified prescribing of these antibiotics in primary care. On the other hand, the antibiotics used only in hospitals, such as ceftriaxone, had decreased both in usage and resistance rates. The main promoter of resistance to antibiotics is likely inappropriate use of these antibiotics in ambulatory care especially for wrong indications, which include potential viral infections, and this should be placed in focus in future interventions within the healthcare system in Serbia and other countries. A reasonable alternative to macrolides may be levofloxacin, which is already being used for the treatment of community-acquired pneumonia in children in clinical trials [42,43]. Although quinolones may have adverse effects on the cartilage of great joints in children, recent studies with long-term follow-up showed that this risk was overestimated and its use in hospitalized children with such serious infections as community-acquired pneumonia is justified [44]. Consequently, we have begun implementing this recommendation in this children's hospital, alongside educational and other measures, to reduce inappropriate macrolide use, and will be following this up in future studies to further guide prescribing in this and other similar hospitals in Serbia.

Additionally, the Ministry of Health in Serbia is now planning to introduce in a national immunization program – the pneumococcal conjugate vaccine – in all infants and young children. This may also be critical in reducing nasopharyngeal carriage and limiting the dissemination of drug-resistant strains [45–47]. We will also be monitoring this impact in the future.

Ramos et al. showed that the prevalence of resistance is country specific and reflects differences in the availability and consumption of antibiotics. In countries with a long tradition of surveillance programs such as Australia, Slovakia and Sweden, there have been reductions in the prescription and use of antibiotics [48]. In our study, ESBL-rate E. coli decreased with decreased antibiotic consumption in 2010 versus 2014; however, this was still high at 28% in 2014. For an antibiotic to be considered as first-line empirical treatment for urinary tract infections, resistance should not exceed 20% in the most likely infecting strains [49,50]. According to these criteria, all third generation cephalosporins used in our hospital are currently above this with reported resistance rates from 25 to 28% (Figure 5). The use of antimicrobials for which the uropathogen has shown resistance can lead to serious consequences [51], particularly for patients with pyelonephritis. Consequently, in this situation, healthcare providers should consider empiric treatment with carbapenem or amikacin or another agent found to be consistently active on the basis of the local antibiogram. This policy has again already been introduced into our hospital, and again we will be monitoring the situation to provide guidance to others.

Encouragingly, the rate of methicillin-resistant S. aureus among the pediatric population in KC Kragujevac is relatively low compared with reports from other countries [52], and it should be kept this way through the continuous monitoring of local S. aureus susceptibility patterns.

Since cloxacillin must be administered frequently (i.e., four-times daily in children), and cloxacillin is sometimes associated with severe phlebitis, cefazolin is a reasonable substitute for the empiric management of bacteremia with methicillin-sensitive S. aureus. With such high levels of resistance in our hospital, clindamycin should only be used for infections caused by methicillin-sensitive S. aureus if the sensitivity of an isolate was confirmed, and the D-test for erythromycin-induced resistance was negative. Such recommendations have also now been introduced in our hospital.

In recent years, the costs of antibiotics have been falling. This can be partially explained by a reform in the procurement system for hospital medicines in Serbia. Until 2013, Serbia had a system whereby each hospital procured their own medicines individually, choosing their own suppliers and brands for medicines to address particular diseases. Rather than competing on price, suppliers typically competed on the level of ‘rebate’ offered to a hospital to increase the monies available to the hospital to purchase other goods and services. These rebates were often as much as a third of the total purchase price. The new procurement system was introduced in 2014 to address this leading to central procurement and an appreciable drop in prices such as prices for ceftriaxone (Table 1) [32].

We acknowledge that there are typically a low number of tissue specimens analyzed in Serbian hospital, which may have influenced interpretation of the results. General limitations of DDDs are common for all aggregated data, exacerbated in the case of pediatric patients. This together with and the limitations of the aggregate microbiology data are described by Goryachkina et al. [26]. Calculations of costs and DU in children based on adult DDDs can also not be used for estimates of prevalence, so the only aspects we could follow were time trends and comparisons between the groups. Despite these limitations, we believe our findings are robust and are already influencing prescribing in this hospital and wider.

Overall, we believe our findings show that increased local efforts to enhance surveillance for AMR are necessary to inform treatment decisions, especially empiric use. The approach to local infection control should be multifaceted and should include microbiology and utilization assessments together with measures to improve local hygiene such as improved hand-hygiene in wards [53–56]. The development of pro-active AS programs is a way forward [13–15,29,57–61] as well as consulting microbiologists and clinical pharmacologists when prescribing antibiotics for complex clinical cases [62]. Furthermore, joint efforts should be made to enhance appropriate antibiotic prescribing and dispensing in all primary, secondary and tertiary care settings in Serbia which treat children. This again will be the subject of future research projects together with increasing efforts in Serbia to reduce the illegal purchase of antibiotics in Serbia [38] as well as improve physician education surrounding antibiotic use.

Finally, we hope this original research in Serbia has implications for other central and eastern European countries struggling to enhance their appropriate use of antibiotics.

To view the supplementary data that accompany this paper please visit the journal website at: Supplementary Material

All authors have been active participants in the research. We would like to thank R Kuzmanovic for IT support regarding microbial resistance data.

The write-up of the paper was in part supported by a grant from the Karolinska. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

The authors state that they have obtained appropriate institutional review board approval or have followed the principles outlined in the Declaration of Helsinki for all human or animal experimental investigations. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved.