Antibiotics are commonly used in the hospital settings, especially in the pediatric intensive care unit. Some reasons for the administration of antibiotics are the risk of infection because of critical conditions, chronic comorbid conditions, surgical procedures, and exposure to invasive procedures and tools as a port of entry for various kinds of microorganisms [1, 2]. Moreover, patients with critical conditions, especially children, may have impaired immune systems, which facilitate the occurrence and spread of infection [3]. Without the intervention of appropriate antibiotics [4], the condition of the patients may worsen to the point of organ damage or even death.

Grohskopf et al. [5] stated that as many as 71% of PICU patients received one or more types of antibiotics. Similar results were reported by Briassoulis et al. [6] who did a cohort study on the use of antibiotics. They found that as many as 67.2% of PICU patients received antibiotics on the first day of treatment and 80.5% received at least one antibiotic during PICU treatment. In addition, Ding et al. [7] reported that in three PICUs in China, the use of antibiotics reached 95%, with 30% of them using a combination of more than one antibiotic.

The use of antibiotics in PICU patients is often empirically based upon vital signs and laboratory results. Toltizs et al. [8] evaluated the use of antibiotics in children admitted to the PICU with fever (axillary temperature > 38.3°C). They found that only 3.3% did not receive parenteral antibiotics. Broad-spectrum antibiotics are often used as empirical therapy, with third-generation cephalosporins, vancomycin, and second generation cephalosporins most often used, according to Grohskopf et al. [5]. Ding et al. [7] found that they often used the second and third generation cephalosporins as empirical therapy. However, Tjekyan [9] reported that the three broad-spectrum antibiotics most often used were ampicillin (53.4%), ceftriaxone (31.1%), and meropenem (7.3%). Additionally, in Indonesia, the most commonly used antibiotics in the Cipto Mangunkusumo Hospital (RSCM) PICU were cefotaxime (30.1%), amikacin (14.46%), and piperacillin-tazobactam (12.10%) [10]. Unfortunately, the use of antibiotics is still largely inappropriate. Hadi et al. [11] conducted a study in two teaching hospitals and reported that 84% of patients received antibiotic therapy, of whom 60% received treatment that was unsuitable or without indication.

In addition, studies have mentioned the negative effects of antibiotic resistance on patient clinical outcomes. Hence, we conducted a retrospective study in one center to determine the appropriateness of antibiotic use, microbe patterning, and possible associations with patient clinical outcomes. The suitability of antibiotic use was based on the 12-step prevention of antimicrobial resistance by the Centers for Disease Control (CDC) [12]. The primary objective of this study was to determine the relationship between appropriate antibiotic use and the mortality rate of children treated in the PICU. The secondary objective of this study was to identify the degree of appropriate antibiotic use in the PICU, microbe patterns, and antibiotic resistance. We also aimed to assess the relationships between microbe pattern and patient mortality, inadequate empirical antibiotic administration and patient mortality, and antibiotic use patterns and microbe patterns. A deeper understanding of antibiotic resistance data will contribute to rational selection of antibiotics for PICU patients.

This retrospective cohort was done in the PICU, Sanglah General Hospital, Denpasar from 2015 to 2017, using patients’ medical records. Informed consent was obtained from subjects’ parents or guardians.

All children admitted to the PICU of Sanglah Hospital during the study period were screened for study inclusion. Inclusion criteria were children aged 28 days-12 years when treated at the PICU, Sanglah Hospital, who had undergone blood culture examination (growth of bacteria and antibiotic sensitivity test). Patients with congenital abnormalities, incomplete medical record data (missing blood culture or sensitivity test results), or those who underwent non-standard blood culture screening procedures were excluded from the study.

Age was defined as the chronological age at the time of treatment in the PICU, expressed in months. The subjects were aged 28 days to 12 years. The 12-year age limit was used because Sanglah Hospital policy stipulates that patients’ ≤12 years of age are to be treated in the PICU (except for neonates, who should be treated in the neonatal intensive care unit).

Antibiotic sensitivity tests were determined using the Kirby-Bauer disc diffusion technique, with interpretation by the National Committee for Clinical Laboratory Standards (NCCLS) is called as the Clinical and Laboratory Standards Institute (CLSI) [4]. The results of the bacterial sensitivity tests were based on the germicidal value of various types of antibiotics, and the assessment was performed by a competent expert [4, 13].

Use of antibiotics was defined as the administration of one or more types of antibiotics during PICU treatment. The use of antibiotics was classified as 1) empirical (antibiotic therapy for signs of infection, but no microbe-sensitivity test results yet), 2) definitive (antibiotic therapy for pathogenic pathogens), or 3) prophylactic (antibiotic therapy for patients without signs of infection, but have immune-compromised conditions, anatomical defects, indwelling device, or planned surgery [14, 15].

The appropriateness of antibiotic use was an assessment of the precision of antibiotics based on a 12-step recommendation by the CDC [12]. The assessment focused on two points, that were the diagnosis and efficient handling of infection and the wise use of an antibiotic. Use of antibiotic therapy was considered appropriate if it did not deviate from institutional therapy guidelines and the CDC step assessments. The antibiotic conformity was evaluated twice, i.e., at initiation and 2 days after initiation. Incompatibilities in antibiotic therapy were expressed when one of the following criteria was met: 1) there was no strong justification for initiation of antibiotics; 2) empirical therapy was continued for more than 3 days, unless there was a documented, strong reason for continuing empirical antibiotic therapy; 3) Empirical therapy was continued or antibiotics were not replaced, despite conflicting microbe-sensitivity test results; 4) the choice of definitive therapy was inconsistent with the sensitivity test, or 5) prophylactic therapy was resumed 24 hours after surgery [16].

Specifically, empirical antibiotics were adequate if the antibiotics in isolated cultures were sensitive in vitro, and empirically administered within 24 hours after the culture was taken. The microbe pattern was the proportion and number of microbes found based on the blood culture results for each patient. We recorded microbial type (gram-negative or positive) and sensitivity to the antibiotic.

Patients treated during a predetermined time range were included in the study. All patients’ medical records were screened. The research sample data included general characteristics: name, address, age, biological sex, nutritional status, duration of PICU hospitalization, a primary underlying disease in the PICU, infection marker, comorbidity (malignancy/HIV), and patient outcome. In addition, blood culture results in the antibiotic susceptibility test results were also obtained from the medical records.

From patients’ medical record, antibiotic usage and antibiotic appropriateness were evaluated and recorded, in accordance with the above-described criteria. The determination of the suitability of antibiotics was evaluated progressively from the development sheets from the day the patient was first admitted to the PICU. As a guide to determine the appropriate use of antibiotics, we used a checklist adapted from Stocker et al. [15] with necessary modifications. The checklist asks for clinical indications and possible infection at the start of antibiotic therapy. Then, this checklist was used to evaluate therapy at 48 hours and 5 days after antibiotic discontinuation, when the blood culture results were available.

The minimum required sample size was calculated by the large sample formula to test the difference of two proportions [17]. By setting the rate of type I error at 5%, a type II error at 20%, and the proportion of appropriate antibiotic therapy (not considered a risk factor) according to the literature to 18%, 24 and relative risks considered significant at 2, the required sample size was estimated to be 95 patients.

Za = error rate of type I, set at 5%, equal to 1.64

Zb = value of type II error is set at 20%, equal to 0.84

P2 = proportion of patients receiving appropriate antibiotic therapy, set at 0.18 [15].

RR = relative risk considered significant, set at 2.

p1 = p2 x RR [17], is set at 0:36

Subjects’ data were collected and processed by 2007 Microsoft Excel software, then analyzed with SPSS 16.0 software. Results are presented as the mean and Standard Deviation (SD) for continuous and median (continuous interquartile) for continuous non-distributed continuous data. Absolute numbers (percentages) indicated categorical or nominal data.

A Chi-square test was used for bivariate analysis to assess for associations between the suitability of the antibiotics and patient mortality, microbial patterns and mortality, as well as the use of antibiotics and microbial patterns. The two-tailed Fisher Exact test was used when the expected frequency was less than 5. A p value of < 0.05 was considered to be statistically significant. The study was approved by the Ethics Committee of Udayana University Medical School, Sanglah Hospital, Denpasar.

From the 692 subjects included in the PICU register at Sanglah Hospital Denpasar, 95 (13.7%) patients had complete medical records and were included in the study. We separated the subjects’ based on the mean length of stay by local insurance in the PICU. As shown in Table 1 with the 95 subjects, 14 had less than 4 days treatment in the PICU and male subjects were dominant and represented 11 of the 14 (78.6%). It was similar for the subjects whose treatment was for more than 4 days, and the males were also dominant with 49 (60.5%). The largest categories of subjects were in the normal body weight category (35.7% and 48.1%). Bacterial growth was seen in a total of 60 patients and were separated into 9 patients with a length of stay less than 4 days and 51 patients with more than 4 days (Table 1).

The appropriate use of antibiotics during initiation and continuation is displayed in Table 2. We found that 80% of subjects used appropriate antibiotic during initiation and 70.5% of subjects used antibiotics appropriately during continuation (Table 2). In this study, we also found that 40% of subjects died and we found an association between mortality and the appropriate use of antibiotic during initiation (p = 0.021; RR = 0.542; CI = 0.341-0.9861) (Table 3). Meanwhile, we did not find an association between mortality and the appropriate use of antibiotic during the continuation (p = 0.408; RR = 0.855; CI = 0.578-1.264) (Table 4).

In this study, we also looked for an association between mortality and the microbial gram status (Table 5) and there was no significant association (p = 0.090; RR = 0.584; CI = 0.311-1.110). The length of stay variable was found to be associated with the compatibility of antibiotic use, although the association was not very strong in the population (p = 0.049; RR = 3.661; CI = 0.869-15.428). Not only was compatibility associated with length of stay, but also the appropriate continuation variable (p = 0.047; RR = 1.196; CI = 1.043-1.373). However, no other variable was associated with the length of stay, including mortality (p = 0.813; RR = 1.125; CI = 0.424-2.985), appropriate use of antibiotic during initiation (p = 0.563; RR = 1.500; CI = 0.366-6.144), and microbial gram status (p = 0.971; RR = 1.023; CI = 0.304-3.438) (Table 6).

The extensive use of antibiotics has been closely linked to the subsequent problem of antibiotic resistance [18]. Antibiotic resistance occurs as a result of continuous use of antibiotics over a longer time period. In addition, the empirical use of antibiotics, not based on bacterial sensitivity results, also contributes to antibiotic resistance [7]. To date, antibiotic resistance remains a threat, particularly in PICU patients [14]. Erbay et al. [3] found that antibiotic use was not appropriate in 47.3% of the cases (OR = 3.8; 95% CI: 1.1-13.1), as the empirical use of antibiotics often does not conform to definitive antibiotic therapy (corresponding to culture and sensitivity test results). Isolation of antibiotic-resistant bacteria is more often found in patients who received prior antibiotic therapy, with the highest prevalence of antibiotic-resistant bacteria in hospitals with the highest antibiotic use [14]. In addition, Sritippayawan et al. [13] reported that two weeks of broad-spectrum antibiotic therapy increased the risk of bacterial infection by Multidrug-Resistant (MDR) bacteria by 9.7 times (95% CI: 1.8 to 53.4). Additionally, Soroush et al. [20] reported resistant Acinetobacter baumannii in patients previously given broad-spectrum antibiotics.

For years, studies have shown increased infection rates by antibiotic-resistant microbes and their corresponding negative effects. Kapoor et al. [21] showed that 100% of Acinetobacter baumannii cases were resistant to ampicillin-sulbactam, ceftriaxone, cefotaxime, cefipime, ceftazidime, and levofloxacin. Soroush et al. [20] reported a decrease from 2001 to 2007 in the sensitivity of Acinetobacter baumannii to ceftriaxone (43.7% vs. 16.2%, respectively), ceftazidime (50% vs. 14.7%, respectively), ampicillin (18.7% vs. 10.4%, respectively), gentamicin (50% vs. 34.3%, respectively), and amikacin (81.2% vs. 41.5%, respectively). In addition, Johnson [22] noted an increase in the proportion of isolates of Methicillin-resistant Staphylococcus Aureus from blood samples in Wales and the UK, where the proportion of methicillin-resistant Staphylococcus aureus (MRSA) in 1992 was <5%, which significantly increased in 2001 to more than 40%. Similarly, gram-negative bacteria, which are confirmed carbapenemase producers, increased dramatically from 2005 to 2013 (Table 7).

Cosgrove [23] reported the negative effects of antibiotic-resistant bacterial infection to-include an increase in the mortality of patients infected with MRSA compared to those infected with methicillin-susceptible Staphylococcus aureus (MSSA) (OR = 3.4; P = 0.003). The duration of hospitalization was also elevated in MRSA-infected patients (mean 29.1 days) and compared with MSSA-infected patients (mean 13.2 days). In addition, patients with MRSA required higher treatment costs (US $118,414) than those with MSSA (US $73,165). Similarly, infection with Enterobacter resistant to third-generation cephalosporins increased the risk of death by 5.09 times (P = 0.01), increased the duration of hospitalization by 1.47 times (P < 0.001), and increased treatment costs up to 1.5-fold (P < 0.001). In addition, Roberts et al. [24] reported that in 2009, the maintenance costs because of antibiotic-resistant infections ranged from US $18,588 to $29,069 in a teaching hospital in Chicago. The duration of hospitalization for these patients was also longer, ranging from 6.4-12.7 days. The mortality from this study caused by gram positive 41.6%, gram negative 58.3% and 3 from gram positive was from MRSI category.

Therefore, to decrease the high rate of antibiotic resistance, which in recent years has gained worldwide attention, antibiotic use should be justifiable, based on microbe-sensitivity data available in the treatment room, especially in the PICU [10]. To reduce inappropriate antibiotic use, the CDC issued a 12-step recommendation to limit resistance to antimicrobials, educate clinicians about antibiotic resistance, and provide a variety of strategies to change clinical practice, including the prescription of antibiotics [12]. This recommendation has been widely adopted, according to the previous studies [15, 16, 25]. Stocker et al. [15] mentioned that the 12-step CDC Guideline is useful for evaluating the therapeutic use of antibiotics in PICU patients, even in settings with limited funding and resources. The study also suggested that the proportion of conformity of antibiotic therapy in patients with negative culture results increased, ranging from only 18% before implementation to 74% after implementation. Similarly, the proportion of empirical antibiotic use for <3 days increased from 18% to 35%, and the precision of definitive therapy increased from 58% to 83%.

In conclusion, an appropriate continuation of antibiotic therapy is a significant predictor of length of PICU patient stay, based on the bivariate analysis. However, multivariate analysis revealed that appropriate continuation of antibiotics was not a significant predictor.