A significant proportion of bacteria causing neonatal sepsis is multidrug-resistant (MDR), not only to the usual empirical regimens, but also to second- and third-line antibiotics in many neonatal intensive care units (NICUs) [
6,
7,
8]. A point prevalence study of 226 hospitals in 41 countries from all over the world showed that 40% of the isolated pathogens in neonatal sepsis were resistant to first-line antibiotics that were usually a combination of ampicillin/amoxicillin/benzylpenicillin and aminoglycoside [
7]. Another study from four hospitals in Southwest China found that early-onset sepsis (EOS) was most caused by
Escherichia coli, while the main isolates of late-onset sepsis (LOS) were
Klebsiella pneumoniae and
Escherichia coli. Almost 65% of
Escherichia coli isolates and 78% of LOS isolates were MDR [
8]. A cohort study of three tertiary care centers in India found methicillin resistance in 61% of coagulase-negative staphylococci and 38% of
Staphylococcus aureus isolates [
9]. Another study of 183 neonates in India found that almost all isolates of
Pseudomonas aeruginosa were multidrug-resistant [
10].
2. The Unique Characteristics of the Neonatal Population and NICUs
Antimicrobial therapy is a common practice in the everyday routine of NICUs and antibiotics are the most prescribed drugs in NICUs for many reasons. Firstly, neonates, and specifically preterm ones, are more vulnerable to infections and their survival relies on antibiotics.
Moreover, there are many diagnostic challenges in NICUs. Neonatal sepsis can present with non-specific clinical signs, such as apnea, respiratory distress, tachycardia and temperature instability rendering the diagnosis difficult, while on the other hand, the absence of clinical findings does not exclude infection [
13]. Many other non-infectious causes can have similar signs and symptoms as sepsis [
14]. Consequently, neonates with respiratory distress syndrome, or transient tachypnea of the newborn, are treated with empirical antibiotic therapy until the blood culture results are negative [
15,
16]. Furthermore, laboratory tests have a low specificity and sensitivity for bacteremia in the early stages. Blood cultures can be falsely negative due to difficulties in obtaining sufficient sample volume, low bacteremia levels and the intrapartum use of antibiotics [
17].
The risk factors for developing multidrug-resistant pathogens are also different in NICUs compared to adult and pediatric ICUs. Gestational age below 32 weeks, birth weight below 1500 g and exposure to vancomycin or carbapenems were found to be independent risk factors for methicillin-resistant
Staphylococcus aureus (MRSA) in NICUs [
18,
19,
20]. Moreover, neonates that were transferred to a tertiary NICU after birth, or were born to mothers colonized with MRSA, also carry an increased risk [
18,
19,
20].
3. Prevention and Control of Infection
A traditional, simple and important way to reduce the spread of resistant bacteria is to reduce the number of infections, since fewer infections means fewer prescribed drugs. Simple and effective methods for infection control, such as safe delivery, avoidance of unnecessary invasive procedures, sanitation, proper cleaning of NICU equipment, hand hygiene before and after interaction with patients or their environment and restricted entry to NICUs are the milestones of prevention and should be strictly adhered to. Hand hygiene is crucial for the restriction of horizontal transfer of MDR microorganisms [
21]. Although hand hygiene has long been proven as a very effective measure, a systematic review showed that compliance among healthcare providers is lower than internationally set targets [
22]. Vertical transfer of MDR organisms refers to the transmission of such pathogens during labor, rendering safe delivery practices very important [
21]. Prompt identification and isolation of MDR-colonized neonates is crucial in the setting of NICUs [
11].
Central-line-related infections are also an important problem in NICUs. A prospective registry of 6215 very low birth weight infants found that central line catheters were commonly used in hospitalized infants and were associated with an increased risk of late-onset sepsis [
23]. However, central-line-associated bloodstream infections were found to be preventable after proper intervention, as shown both in adult and neonatal ICUs [
24,
25]. Since line insertion is often mandatory in premature and ill neonates, several prevention strategies have been studied and proved effective in preventing associated infections. Such secondary prevention strategies include proper hand hygiene, appropriate maintenance and handling of central lines, clinician education and prompt removal of unnecessary central lines. The care bundle approach was found to be the most effective one in decreasing central-line-associated infections in various age groups [
25,
26].
Antifungal prophylaxis is also administered in many NICUs, especially if there is a high incidence of invasive candidiasis. According to the Clinical Practice Guidelines for the Management of Candidiasis in 2016, neonates with a birth weight below 1000 g could be administered intravenous or oral fluconazole twice weekly for 6 weeks. Alternatively, for neonates with a birth weight below 1500 g, oral nystamycin can be used [
27].
4. Diagnosis of Neonatal Sepsis
NICUs face many diagnostic challenges, as already described. Accurate diagnosis of sepsis and prompt identification of high-risk infants that need antibiotic therapy is an important first step in limiting the inappropriate use of antibiotics. The main challenge is that neonatal sepsis can present with various nonspecific signs and symptoms, while many of these findings overlap widely with those of other noninfectious causes [
14]. As a result, cultures are often obtained, and empirical antimicrobial therapy is initiated with a low threshold.
The gold standard for diagnosing sepsis is blood cultures. Ideally, two blood cultures of sufficient blood volume (1–2 mL) should be collected with a sterile technique, before antimicrobial therapy initiation [
28,
29].
Biomarkers, such as C-reactive protein (CRP), procalcitonin (PCT), white blood cell count and erythrocyte sedimentation rate, can serve as useful adjunctive tools in combination with clinical symptoms and signs in the diagnosis of infection. A systematic review and meta-analysis of 22 cohort studies including 2255 infants, found that CRP was not helpful at the initial evaluation of possible LOS, and was insufficient alone to guide diagnostic and treatment decisions [
30]. However, the combination of PCT and CRP or presepsin alone improved the accuracy of diagnosis of neonatal sepsis, as shown in a meta-analysis of 28 studies enrolling 2661 neonatal patients [
31]. Randomized controlled trials have tested the accuracy and efficacy of PCT. The use of serial PCT values, in the decision tree process, led to a shorter duration of antibiotic therapy in EOS [
32,
33]. Interleukin-6 (IL-6) is another potentially useful marker that has been studied in neonatal sepsis. A systematic review of 3276 newborns found that IL-6 is a good early diagnostic marker of EOS, with higher accuracy when used in preterm neonates and when measured in the first 48 h [
34]. The sensitivity had a wide range of 42.1–100%, and a specificity of 43–100% (with median values of 83 and 83.3%, respectively). A meta-analysis of 694 neonates showed higher sensitivity and specificity of IL-6 (pooled sensitivity 85% and specificity 88%) in infants with premature rupture of membranes [
35].
Another area of active research is the development and validation of sepsis calculators to better predict EOS. In 2014, Escobar G et al. published a retrospective nested case-control study of 600,000 neonates from 14 hospitals from 1993 to 2007, and developed a risk classification scheme for EOS. This calculator, which is perhaps the most promising and widely used, applies to neonates 34-week-old and older and is a method of estimating the risk of EOS. This calculator uses both maternal risk factors, as well as factors from the clinical examination of the newborn, and the neonates are stratified into three risk groups: treat empirically, observe and evaluate, and continued observation [
36,
37].
The accuracy of this calculator, provided by the Kaiser Permanente Division of Research, has been tested in many studies. A meta-analysis of 13 studies, analyzing a total of 175,752 newborns, showed that the use of this calculator limited the use of antibiotic therapy for suspected EOS without missing cases of EOS, while evidence regarding safety was limited [
38]. The number of missed cases of EOS was similar, whether the EOS calculator was used or conventional management strategies were applied [
38]. A retrospective study from our center found that the adoption of the online sepsis risk calculator would have significantly reduced antibiotic usage, however, a significant proportion of the cases would have been missed [
39]. Therefore, the calculator can be applied to late preterm and term neonates with EOS, but it is not useful for preterm neonates.
For LOS, there is a score, namely, the RALIS score [
40,
41], that is based on vital signs, including heart rate, respiratory rate, body temperature, oxygen saturation below 85%, bradycardia and body weight, monitored 12 times a day. The RALIS score has the advantage of being a non-invasive tool and uses clinical data that are routinely collected in NICUs [
41]. This score has been tested in a prospective study of 118 very low birth weight infants and was found to have reasonable sensitivity (75%) and specificity (81%), and its negative predictive value was very high (95%). However, it had a low positive predictive value of almost 39% [
41]. A smaller retrospective study of 73 preterm infants found higher sensitivity (82%) for the RALIS score, but low specificity (44%) and a better positive predictive value (67%) [
40]. In both studies, the RALIS alert preceded clinical suspicion by almost 3 days. The RALIS score is a promising tool when used in association with clinical judgement, laboratory values and specific infection biomarkers.
Except for the RALIS score, there is a variety of predictive scores for LOS that has been studied in the literature. The available scores were based exclusively on clinical variables, laboratory variables or risk factors, or were based on combinations of these variables. Sofouli et al. conducted a systematic review of the available scores and concluded that the majority of them may assist in early diagnosis, but almost all had a limited diagnostic accuracy [
42]. The scores that were found to have high sensitivity and negative predictive value were those based on clinical, laboratory and risk factors, such as the NOSEP-NEW-I and NOSEP-NEW-II scores [
42,
43], and on clinical, laboratory variables and management decisions [
42,
44,
45].
5. Fighting Biofilm Infections
Biofilm formation is a severe health concern in NICUs and a persistent threat, as it increases both morbidity and mortality. Bacterial biofilms are immobile microbial communities with diverse bacterial cell colonies in groups, which colonize and grow on the surfaces of almost all medical devices or prostheses [
82]. The development of biofilm has many stages, including attachment of macro and micro molecules to surfaces, colonization and proliferation of bacterial pathogens, the release of extracellular polymeric substances and biofilm maturation and detachment [
83,
84,
85].
It has been estimated that most bacterial infections in humans are correlated with biofilms and about 50% of the nosocomial infections are indwelling-device associated [
86]. Neonates hospitalized in NICUs are at a high risk of biofilm infections, since they often need central lines, endotracheal tubes, umbilical arterial and vein catheters and peripheral central lines for a prolonged duration. Bacterial biofilms are highly resistant to antibiotic treatments and immune responses [
87]. Biofilms provide protection to bacteria from the pH, nutrient deficiency and mechanical forces, and also prevent antibiotic and host immune cells from gaining access to their bacteria [
83,
84,
88]. Therefore, biofilms contribute to persistent and resistant chronic infections.
Bacteria that have been associated with biofilm infections in central venous catheters are
Enterococcus faecalis,
Pseudomonas aeruginosa,
Staphylococcus aureus and
Staphylococcus epidermidis, while numerous other bacteria have been associated with biofilm infections in other adherent surfaces [
83].
Although it is well known that antibiotic treatment is currently the most important and effective measure of control for microbial infections, antibiotic treatments are almost impossible to eradicate biofilm infections [
86]. The development of intact biofilms is critical for the spreading and persistence of bacterial infections in the host [
84]. Therefore, biofilms give the power to bacteria to become resistant to antibiotics.
As already described, biofilm infections are chronic and can have exacerbations. Antimicrobial therapy can be helpful in mitigating an exacerbation, but cannot fully eradicate the bacteria in a biofilm [
86]. Since biofilms are multimicrobial, antibiotic therapy should cover against Gram-positive, Gram-negative and fungal microorganisms, as well as resistant bacteria. The above facts render treatment of a biofilm infection really challenging. Antibiotics are more effective when used in the early stages of biofilm formation, probably because the cells are not completely embedded into biofilm communities [
83,
89,
90]. The antibiotic regimens should have a wide spectrum and employ different modes of action. Removal of the foreign body or replacement of the central venous catheter is almost always mandatory to successfully control a biofilm infection [
86]. A short course of intravenous antibiotic treatment is also important to eradicate the bacteria that were released into the bloodstream.
There are many novel approaches under development for the prevention and treatment of biofilm infections. Specific agents may inhibit the initial biofilm attachment, result in the removal of biofilms or inhibit biofilm formation by quorum quenching (the communication process among cells) [
91]. Initial biofilm attachment can be inhibited by altering either the chemical or physical properties of the biomaterials. For example, antibiotics, biocides and ion coatings modify the surface of devices and inhibit biofilm formation [
92]. Moreover, hydrophilic polymers, such as hyaluronic acid and poly N-vinylpyrrolidone, alter the hydrophobicity of polymeric materials and reduce the adhesion of microbes [
93,
94]. This occurs in the same way by altering the physical properties of devices, various superhydrophobic surfaces, hydrogel coatings and heparin coatings [
95,
96]. Another studied field is the novel ways to remove biofilms. Matrix-degrading enzymes have the purpose of dissociating a biofilm matrix and allowing antimicrobials to act more drastically [
97]. Furthermore, free fatty acids, specific amino acids and nitric oxide generators have been shown to induce biofilm dispersal against several bacteria [
98,
99]. Quorum sensing is crucial in biofilm formation, therefore several strategies have been studied to inhibit this system of cellular communication in biofilm [
91]. Molecules that degrade, inhibit or antagonize quorum sensing signals have been studied, as well as molecules that inhibit signal transduction or transportation [
91]. These novel approaches could lead to anti-biofilm therapies in the future, that are much more effective than current antibiotic treatments [
83,
84,
86,
100].
Therefore, accurate diagnosis of sepsis is of paramount importance in order to restrict antibiotic administration. A combination of clinical signs and symptoms, properly drawn blood cultures and serum biomarkers can help promptly identify suspected sepsis. The sepsis calculators are also adjunct useful tools. Serum biomarkers are also useful in limiting the duration of antibiotic therapy.