During the study period, 497 patients were admitted to the PICU; these accounted for 3009 patient days. After excluding 238 patients based on the study criteria, 259 patients were included in the review. Table 2 summarizes the cohort demographics and characteristics. One hundred thirty-two (51%) patients were male, and the median age was 23 months (IQR: 6-66.5 months). The median PICU length of stay was 3 days (IQR: 1–6 days). One hundred and seventy-two (66%) of the admitted patients were medical admissions, while 87 (32%) were either surgical, trauma, or burn admissions.
Overall, 259 children received 274 antibiotic courses, resulting in 2553 DOTs during the study period. The median duration of an antibiotic course was 4 days (range, 3–7). The indications for antibiotics at initiation were empiric (n = 187, 68%), prophylactic (n = 61, 22%), or definitive therapy (n = 6, 9%). The reasons for antibiotic initiation were infections of the respiratory system (n = 115, 42%), surgical procedures (n = 74, 27%), bloodstream infections (n = 65, 24%), central nervous system infections (n = 7, 3%), skin and soft tissue infections (n = 5, 2%), and other systemic involvement (gastrointestinal, cardiovascular, renal) (n = 8, 3%). The most common clinical indications for antibiotic initiation were community-acquired pneumonia (20.5%), community-acquired sepsis (10.4%), bronchiolitis (10%), hospital-acquired sepsis (9.3%), and neurosurgical procedures, e.g., drain insertion or tumor resection (8.9%). Appropriate cultures were collected prior to antibiotic initiation in 205 courses. Of these, an infection was isolated in 60 cultures (29.3%). An MDRO was isolated in 27 cultures (45% of positive cultures). The ID service was consulted in 40 of 274 antibiotic courses (equivalent to 14.6%).
In our center, ceftriaxone, vancomycin, ceftazidime, and cefazolin were the most frequently used antibiotics at 164.8, 150.5, 91.7, and 83.1 DOTs per 1000 patient-days, respectively (Fig. 1). Cefazolin, vancomycin, ceftazidime, and ceftriaxone were the most inappropriately used antibiotic therapies, with 50.2, 40.9, 39.5, and 24.3 inappropriate DOTs per 1000 patient-days, respectively. Figure 2 shows inappropriate and appropriate antibiotic use by indications. Compared to other clinical indications, surgical prophylaxis for neurosurgical procedures and gastrointestinal surgery were positively associated with an increased likelihood of inappropriate antibiotic use (P = 0.001) (Fig. 2).
Out of 274 courses, 133 (48%) were found to be nonadherent to at least 1 of the five CDC steps evaluated, leading to 677/3009 (22.5%) inappropriate DOTs. Thirty-one percent of the courses did not target the pathogen (step 4), 30.2% did not include antimicrobial control (step 6), 16.2% inappropriately began with broad spectrum coverage (step 9), and 22.3% did not stop antibiotics when an infection was considered cleared or unlikely (step 10). No antibiotic courses were found to be violating step 8. On reviewing courses that did not target the pathogen and were nonadherent to step 4 (56 courses), we found that 66% of these were due to an inappropriate empiric choice, 27% were due to inappropriate de-escalation to definitive therapy, and 4 cases had an inappropriate empiric and inappropriate definitive choice.
Subgroup analyzes are demonstrated in Table 2. The duration of antibiotic courses was significantly higher in children with inappropriate antibiotic use than in children who had appropriate antibiotic use (median: 9.0 vs. 7.0, P = 0.001). Surgical, trauma, and burn admissions each had a higher percentage of inappropriate antibiotic use than was found in medical patients (67.4% vs. 40.1%, P = 0.001). Cefazolin and vancomycin were associated with a higher percentage of inappropriate judgment of antibiotic use than was found for other antibiotic agents (P = 0.001). The presence of respiratory comorbidity was associated with a lower percentage of inappropriate antibiotic use (P = 0.028). No action taken regarding antibiotics after 72 hours was associated with a higher percentage of inappropriate antibiotic use than was found for other antibiotic decisions (stopped: 40.8%, continued with planned duration: 43.8%, changed based on clinical condition and culture results: 54.5%, no action taken: 75%, P = 0.001). The rest of the sample characteristics were not significantly associated with inappropriate antibiotic use.
Adjusted odds ratios are provided in Table 3. Age was positively associated with inappropriate judgment of antibiotic use (adjusted odds ratio (aOR): 1.012, 95% confidence interval (CI): 1.002 to 1.022, P = 0.016). Compared to ceftriaxone, vancomycin and cefazolin significantly increased the risk of inappropriate judgment of antibiotic use (aOR: 3.781, 95% CI: 1.265 to 11.299, P = 0.017 and aOR: 13.560, CI: 1.170 to 157.105, P = 0.037, respectively). Compared to cases with infectious disease consultation, in cases where an ID consultation was not carried out, the likelihood of inappropriate antibiotic use was higher (aOR: 7.407, CI: 2.078 to 26.405, P = 0.002). The Hosmer and Lemeshow test indicate that this model fits the data well (P = 0.868).