Epidemiology, antimicrobial resistance, and incremental medical costs of hospital-acquired infections in the Intensive Care Unit: 2-year experience from Serbia CURRENT STATUS: POSTED

Background: Hospital-acquired infections are a major complication of hospital treatment. The growing presence of multidrug-resistant pathogens contributes to increased mortality and costs, particularly in intensive care units where patients are predisposed to numerous risk factors. Comprehensive data about hospital-acquired infections from Serbian intensive care units is scarce. The aim of this study was to determine the presence of hospital-acquired infections among intensive care unit patients and look into the patterns of antimicrobial resistance, risk factors, and incremental costs of diagnosis and antimicrobial treatment. Methods: This retrospective study included 355 patients over a two-year period. Etiology, antimicrobial resistance patterns, and incremental costs of diagnosis and antimicrobial treatment were examined. Risk factors for infection acquisition, as well as length of stay, were statistically analyzed using Pearson’s chi-square tests and logistic regression analysis. Results: At least one hospital-acquired infection was identified in 32.7% of patients. A total of 204 infection episodes were documented, the most common type being urinary tract infections (36.3%). Clostridium difficile , Klebsiella spp. , and Acinetobacter baumanii were the most common isolates. Antimicrobial resistance rates < 20% were observed for linezolid (0%), colistin (9%), and tigecycline (14%). Resistance rates > 50% were seen in all other tested antibiotic agents. Mortality rates were not higher in patients who acquired only one hospital-acquired infection (p=0.09), but were significantly higher for patients in whom more than one episode occurred (p=0.038). Length of stay > 20 days carried a 7.5-fold increase in odds of acquiring an infection (CI 4.4-12.7, p<0.001), whereas length of stay > 30 days carried a 10-fold increase (CI 5.5-16.1, p<0.001). During the study period, over 37,000 EUR was incrementally spent on diagnosis and antimicrobial treatment for hospital-acquired infections. Conclusion: Our results suggest a high prevalence of hospital-acquired infections and very high antimicrobial resistance rates compared to most

According to the World Health Organization (WHO), "Healthcare-associated infections represent the most common adverse event among hospitalized patients" [1]. This is particularly true in the Intensive Care Units (ICUs) [2]. The latest research shows over 2.5 million hospital-acquired infection (HAI) episodes occur every year in Europe, with more than 90,000 deaths attributed to the six most common types: healthcare-associated pneumonia (HAP), healthcare-associated urinary tract infection (HA-UTI), surgical site infection (SSI), healthcare-associated Clostridium difficile (HA-CDI), healthcareassociated neonatal sepsis, and healthcare-associated bloodstream infection (HA-BSI) [3]. Acquisition of HAIs carries a substantial increase in mortality and cost of treatment [4], particularly if the isolates are multidrug-or extensively drug-resistant (MDR/XDR) [5,6]. Incremental costs range from 6700 EUR to almost 30 000 EUR per episode and cause a significant loss of Disability-Adjusted Life Years (DALYs) [3,5]. Some authors suggest that evidence-based strategies for surveillance and prevention could eliminate up to 70% of all HAIs [1,7], which should serve as an incentive for optimizing healthcare policies. The European Centre for Disease and Control (ECDC) has successfully implemented such efforts, leveraging advances in digital technology [8,9].
In Serbia, isolated single-center reports from ICUs have been published and looked at the neonatal population [10], risk factors for acquisition in trauma patients [11], and antimicrobial resistance patterns of specific isolates [12]. However, virtually no studies have looked at the overall prevalence of HAIs in adult ICUs in our country.
The primary goal of our study was to investigate the rates of HAIs in the ICU over a two year period and assess the risk factors, etiologies and antimicrobial resistance patterns. The secondary goal was to calculate, for the first time in Serbia, incremental costs of diagnosis and treatment.

Study population
This 2-year retrospective study was conducted in the 15-bed ICU of the 150-bed University Teaching Hospital for Infectious and Tropical Diseases, Clinical Centre of Serbia, with a patient:nurse ratio of 4:1. From September 2016 to September 2018, healthcare records of all admitted and discharged patients were evaluated (n=495). Exclusion criteria were length of stay (LOS) < 48 hours (n=76) and incomplete patient record data (n=64).

Definitions
Each episode of HAI was defined using the criteria of the ECDC -occurring ≥ 48h after admission, with onset from day 3 onwards [8]. Each subtype of HAI was defined using the same ECDC criteria and abbreviations, and were classified as follows: bloodstream infection (BS), urinary tract infection (UTI), catheter-associated UTI (CAUTI), hospital-acquired Clostridium difficile infection (GI-CDI), pneumonia (PN), device-associated infection (DAI), hospital-acquired central nervous system infection (CNS-HAI), and skin infection (SST-SKIN) [8]. Infections of more than one site in the same patient were reported as independent events unless the same pathogen was isolated concurrently.
Two incremental direct medical costs were defined: (1) the costs of microbiological testing related to the diagnosis of HAI; and (2) the costs of antibiotics indicated for HAI treatment. To standardize the representation of costs, all expenses were converted from the national currency (Serbian dinar -RSD) to EUR using local conversion rates for each year [13], and adjusted for annual inflation rates using standardized converters [14].

Data collection
Basic demographic (gender, age) and clinical characteristics (primary diagnosis, comorbidities, risk factors, antimicrobial use within 48h upon admission) were collected, LOS and clinical outcome. Each HAI was documented using the standardized ECDC criteria for HAI documentation [8], including the etiology, microbiological method of confirmation, date of onset, antimicrobial susceptibility testing, as well as the choice of treatment and duration. Antimicrobial susceptibility testing was conducted using the European Committee on Antimicrobial Susceptibility Testing (EUCAST) standards [15], and all isolates identified as "intermediate" (I) were classified as "sensitive" (S). Multidrug-resistant pathogens were confirmed using the susceptibility criteria by Magiorakos et al. [16].

Statistical analysis
The Statistical Package for Social Sciences (SPSS) software version 23 (IBMCorp. Released 2015. IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp.) was used for analysis of patient data. Basic epidemiological indicators (incidence and mortality rates) were calculated. For normally distributed data mean and standard deviation were applied, whilst median and interquartile range (IQR) were used for data that did not exhibit normal distribution. Pearson`s chi-square test was appropriate to test the difference between categorical data, and Mann-Whitney U test was adequate for numerical data without normal distribution. A logistic regression analysis was performed to predict the probability of acquiring a HAI and derive odds ratios (ORs), followed by adjustment for potential confounding factors (age on admission, gender, number of comorbidities, and LOS).
The study was approved by the Ethics committee of the Clinical Centre of Serbia. All data that were included in the analysis were previously anonymized.

Results
A total of 355 patients were included in the study, of which 190 (53.5%) were male and 165 (46.5%) were female. The mean age was 63.14 years (SD=16.5). Main causes of admission included primary CNS infections and sepsis (63.5% and 16.9%, respectively). Comorbidities and risk factors are shown in Table 1. In the study period, 116 patients were diagnosed with at least one HAI, an incidence rate of 32.7%, and a total of 204 HAI episodes were documented. Presence of an indwelling urinary catheter (p=0.001), nasogastric tube (p<0.001), and a CVC (p=0.001) were all associated with the development of HAI, as were endotracheal intubation (p=0.001) and mechanical ventilation > 48 hours (p<0.001). HAI was common among patients with cardiovascular diseases and diabetes mellitus (p=0.01 in both cases), and among those who received penicillin in the 48h upon admission (p=0.008). Older age was also associated with an increased risk of acquiring HAI (66.63 vs. 61.42 years, p=0.003).
The median LOS was 20 days (IQR 21), and was significantly longer in patients with at least 1 HAI compared to patients who did not acquire a HAI (34.5 vs 16 days, p<0.001). Using logistic regression analysis (Table 4), each ICU day incrementally increased the odds of acquiring a HAI by 7.6%, even when adjusted for age, sex, and number of comorbidities. LOS > 20 days carried a 7.5fold increase in odds of acquiring HAI (CI 4.4-12.7, p<0.001). Furthermore, LOS > 30 days carried a 10-fold increase in odds of acquiring a HAI (CI 5.5-16.1, p<0.001).
The overall mortality rate was 39.4% (140 patients), and was not significantly higher for patients who acquired a single HAI (p=0.09). A statistically significant increase in mortality, however, was observed in patients who acquired > 1 HAI (p=0.038).
During the study period, a total of 37,583 EUR was incrementally spent on the four most common types of infections, of which 5,804 EUR on microbial testing and 31,779 EUR on antimicrobial treatment (Table 5). Statistically significant differences in the costs for microbiological diagnosis and antimicrobial treatment for the two 12-month periods were not observed (p=0.764 and p=0.904 respectively).

Discussion
The primary aim of our study was to obtain a deeper insight into the epidemiology of HAIs and guide the next steps in terms of surveillance and hospital policies. As one of the rare studies looking at incidence rates and profiles of HAIs in adult ICU patients in Serbia, the incidence rate of 32.7% is in accordance with isolated European studies from Slovenia (35.7%) [17], and Poland (27.6%) [18], but still much higher than the overall rates of 19.4% in Europe [19].
The use of invasive devices and mechanical ventilation associated with the development of HAI in ours study is consistent with other reports that evaluated risk factors for HAI [20,21]. Further prospective analyses are, however, required to discern the relationship between the use of invasive devises and etiologies of HAIs.
Contrary to the findings in large European studies, where approximately 20% of all HAIs were UTIs [19,21], this type of infection comprised more than a third of HAIs in our sample. These findings mandate a deeper assessment of current practices and patient protocols.
The antimicrobial resistance patterns and the percentage of MDR strains observed reiterates the concerning results from other Serbian studies [12,22]. Overall resistance rates of 88% for cephalosporins, 85% and 75% for gentamycin and amikacin, 92% for ciprofloxacin, and 56% for carbapenems are higher compared to most European countries [19]. Such high resistance rates support the trend increasing prevalence MDR and XDR strains [23]. The emergence of colistinresistant Klebsiella spp. is particularly troubling, as 1 in 4 of our isolates exhibited such characteristics. As effective therapeutic options narrow for HAIs, the use of colistin will increase, but may invariably lead to higher resistance rates [24]. For this reason, it is imperative to reserve colistin until antimicrobial susceptibility patterns mandate its use.
The retrospective nature of data collection, as well as numerous biases that arise when HAIs are associated prolonged LOS [25,26], impeded us from determining the true relationship between LOS and HAIs. We have shown, however, that an extended LOS at our facility carries significantly higher odds of acquiring a HAI, regardless of patient age, gender, or accompanying comorbidity. Mortality rates were not higher if patients acquired a single HAI, contrary to other reports [27]. However, acquisition of > 1 HAI was associated with a higher mortality rate. Such findings necessitate a revision of existing healthcare policies and highlight the importance of HAI prevention. This is the first study from Serbia which has described healthcare-economic data regarding HAIs.
Incremental medical costs -microbiological testing and antimicrobial treatment, were obtained and revealed valuable insights into the burden of HAI in Serbia. Although we were not able to estimate an in-depth burden of HAIs by type and possibly by antimicrobial resistance patterns [4,28], several small-scale studies such as ours were done in other countries and achieved to show a substantial burden of HAI [29][30][31][32]. Our intent is to use this data as the starting point to leverage large-scale prospective studies that would effectively show the actual costs of HAIs in our country.
Our study had several important limitations. First, a small patient sample gathered from a single center did not allow us to draw more generalized conclusions. The small sample also prevented us to perform appropriate patient matching, which is often the study of choice when comparing incremental costs of HAIs [33]. Second, the retrospective nature of the study was a limitation. Risk factors

Acknowledgments
We thank the nursing and supporting staff of the ICU of the University Teaching Hospital for Infectious and Tropical Disease for their patience and help.

Funding
No funding was received for this study.

Availability of data and material
The datasets that were obtained and used for this study are available from the corresponding author on request.

Authors' contributions
AD was involved in the conceptualization of the study, its design, and original draft preparation; BM was involved in the review and editing of the manuscript; IM was involved in analysis of data and review of the manuscript; SDJ was involved in acquisition and interpretation of microbiological and financial data ; AC was involved in interpretation of data and its analysis; GS was involved in   Cephalosporins -ceftriaxone, ceftazidime, and cefepime tested; Carbapenems -Imipenem and meropenem tested.