Characteristics, outcome and risk factors for mortality of pediatric patients with ICU-acquired candidemia in India: a multicenter prospective study.

BACKGROUND The epidemiology, clinical profile and outcome of pediatric candidemia varies considerably by age, healthcare settings and prevalent Candida species. Despite these differences, few comprehensive studies are undertaken. This nationwide study addresses this knowledge gap. METHODS 487 children who contracted ICU-acquired candidemia at 23 Indian tertiary care centers were assessed for 398 variables spanning demography, clinical characteristics, microbiology, treatment and outcome. RESULTS Both neonates (5.0 days; range=3.0-9.5) and non-neonatal children (7.0 days; range=3.0-13.0) developed candidemia early after ICU admission. Majority of neonates were premature (63.7%) with low birth weight (57.1%). Perinatal asphyxia (7.3%), pneumonia (8.2%), congenital heart disease (8.4%) and invasive procedures were common comorbidities, and antibiotic use (94.1%) was widespread. C. tropicalis (24.7%) and C. albicans (20.7%) dominated both age-groups. Antifungal treatment (66.5%) and removal of central catheters (44.8%) lagged behind. Overall resistance was low, however, emergence of resistant C. krusei and C. auris needs attention. The 30-day crude mortality was 27.8% (neonates) and 29.4% (non-neonates). Logistic regression identified admission to public sector ICUs (OR=5.64), mechanical ventilation (OR=2.82), corticosteroid therapy (OR=8.89), and antifungal therapy (OR=0.22) as independent predictors of 30-day crude mortality in neonates. Similarly, admission to public sector ICUs (OR=3.62), mechanical ventilation (OR=3.13), exposure to carbapenems (OR=2.18), and azole antifungal therapy (OR=0.48) were independent predictors for non-neonates. CONCLUSIONS Our findings reveal a distinct epidemiology, including early infection with a different spectrum of Candida species, calling for appropriate intervention strategies to reduce candidemia morbidity and mortality. Independent factors identified in our regression models can help tackle these challenges.