Several studies of pediatric antibiotic consumption have been performed in Saudi Arabia. Balkhy et al.  conducted a prospective surveillance study over 33 months in two PICUs and one neonatal intensive care unit (NICU) . They reported that the consumption of cephalosporins (237.3 DOT per 1000 patient-days) and carbapenems (85.8 DOT per 1000 patient-days) was higher in PICU patients in Saudi Arabia than the levels reported in the USA and Canada. Al Matar et al.  conducted a point prevalence study on antibiotic consumption using data obtained in 26 hospitals in Saudi Arabia in both adults and children and found that 24% of patients admitted to ICUs received antibiotics, with third-generation cephalosporins the most commonly prescribed agents. AlAwdah et al.  reported in a pilot ASP study that 72% of admissions to a tertiary care PICU were prescribed antibiotics. None of these studies addressed the appropriateness of antibiotics, and only Balkhy et al.  used the metric DOT per 1000 patient-days to report the consumption of antibiotics. The DOT and DOT per 1000 patient-days are the metrics of choice for measuring antibiotic consumption in the pediatric population as recommended by the National Healthcare Safety Network and CDC [26, 30].
To the best of our knowledge, this is the first report to assess antibiotic appropriateness in pediatric patients admitted to intensive care units in Saudi Arabia. We calculated 3009 total DOTs and 848.5 DOTs per 1000 patient-days, indicating a rising increment with regard for the 708.3 DOTs per 1000 patient-days previously reported between 2012 and 2015 . This difference may be attributable to seasonal variation because prescription rates increased during winter months in our study period . The current antibiotic consumption places in our center are at the higher end of the ranges reported in Switzerland, Canada and the USA, where 550, 750 and 1000 DOTs per 1000 patient-days, respectively, have been reported [2, 4, 32].
On initiation, empiric antibiotics accounted for almost two-thirds of antibiotic prescriptions. This frequency is higher than the international rates reported in Canada and European countries (approximately 40%) and similar to those reported other countries, such as Indonesia (85%) [1, 33, 34]. The high rate of empiric prescriptions may be influenced by a higher burden of infections in PICUs in Saudi Arabia. Almuneef et al. [35, 36] found that the rates of hospital-acquired infections, including central line-related bloodstream and ventilator-associated pneumonia, were higher than those of the National Nosocomial Infection Surveillance system. Another factor may be the difficulty of differentiating between bacterial and viral infections, as reflected in the fact that 26 courses were started empirically in patients diagnosed with bronchiolitis. On review of non-adherence to CDC step 4 (target the pathogen), the majority of such cases were found to be due to inappropriate use of empiric antibiotics. This suggests a need for a better understanding of common pathogens and local susceptibility patterns, better utilization of diagnostics, and guidelines for the initiation of empiric antibiotics.
The most frequently used antibiotics were ceftriaxone, vancomycin, ceftazidime and cefazolin. Antibiotic choices and top indications for antibiotics in this cohort (pneumonia, sepsis, and surgical prophylaxis) were similar to those in the US and Canada [1, 37]. Forty-eight percent of courses were deemed nonadherent to at least one of the five CDC steps, resulting in a total of 677 inappropriate DOTs (22% from a total of 3009 DOTs). This is lower than the rates reported in Canada by Blinova et al.  (62%) and in Turkey by Ceyhan et al.  (50%).
Cefazolin was the most inappropriate antibiotic, with a total of 151 inappropriate DOTs and 20% of total inappropriate DOTs for all antibiotics. The main purpose of cefazolin use in our institute is perioperative prophylaxis, with 22% of antibiotics initiated for this purpose. High inappropriate use is largely due to a prolonged duration of postoperative prophylaxis, which had a median duration of 3 days (IQR 2 to 5). This may explain why the rate of inappropriate antibiotics was higher in surgical patients than in medical patients (P = 0.001). The misconception of prolonging prophylaxis for specific procedures or the presence of drains may contribute to this finding. Guidelines for perioperative prophylaxis recommend durations not extending beyond 24 hours from initiation irrespective of the surgical procedure or the presence of drains or catheters [23, 24].
A lack of reassessment or action at 72 hours by the treating team was found to be associated with inappropriate antibiotic use (P = 0.028). In view of this finding, conducting an antibiotic ‘time out’ would be a suitable intervention. A time out is an ASP intervention recommended by the CDC that standardizes the review of clinical, laboratory and microbiological data at a predefined time (48–72 hours post antibiotic initiation) to identify opportunities for discontinuing or deescalating empiric antibiotics . Adams et al.  conducted a quality improvement trial over an 8-month period to evaluate three target antibiotics (vancomycin, meropenem, and piperacillin/tazobactam) and found that the DOT per 1000 days was reduced for both vancomycin and meropenem.
This is the first study to address the appropriateness of antibiotic consumption among PICU patients in Saudi Arabia. We included antibacterial consumption, indications for antibiotic initiation as parameters and, most importantly, assessed the appropriateness of antibiotics by two independent infectious disease physicians based on the CDC 12 step classification. We further quantified and categorized inappropriate antibiotic use, highlighting areas for improvement and allowing for targeted rectification.
The study has several limitations, and data were collected retrospectively; thus, some information influencing decision making could have been missed. In addition, this is a single center report, which limits its generalizability. Decisions regarding antibiotic appropriateness are prone to bias and the subjective interpretation of data despite attempts to unify understanding of the CDC steps and consensus decision-making in more complex cases. This may explain why patients considered to be at a higher risk (i.e., with respiratory comorbidities and younger age) had a higher likelihood of being judged by the reviewing ID physicians to be on appropriate antibiotics. Additionally, due to the retrospective nature of the study, the accuracy of diagnoses was not confirmed, and the authors accepted all diagnoses made by treating physicians. This meant that we could not assess non-adherence to step 8 (treat infection but not contamination or colonization). In a prospective approach, the necessary information regarding diagnoses could be gathered in real-time. ID consultation decreased the likelihood of inappropriate use and should be utilized in addition to other ASP initiatives.