Assessing prediction accuracy for outcomes of ventilator-associated events and infections in critically ill children: a prospective cohort study.

OBJECTIVES: To assess the prediction accuracy of the 2008 US Centers for Disease Control and Prevention (CDC) definitions for ventilator-associated pneumonia (VAP)/ventilator-associated tracheobronchitis (VAT), 2013 CDC definitions for ventilator-associated events (VAE) and a new VAE algorithm in the paediatric (Ped) population, the Ped-VAE.

METHODS: We performed a prospective 13-month cohort study at a multidisciplinary paediatric intensive care unit (PICU). Primary endpoints were duration of ventilation episode, PICU or hospitalization length of stay from episode and episode mortality. Episodes without VAE (or VAP/VAT) served as comparison groups.

RESULTS: One hundred eight episodes of ventilation (99 children) with 2554 ventilator-days were assessed. In episodes not meeting 2008 CDC definitions, a median of 6 ventilator-days (PICU stay 11 days) was documented (with eight deaths), not significantly different from episodes not meeting VAE or Ped-VAE definitions. Using 2008 CDC criteria, 11 (10.2%) respiratory infections (eight tracheobronchitis) were identified, seven VAEs using 2013 CDC criteria (6.4%) and 29 (26.8%) using Ped-VAE criteria (relative risk vs. 2008 CDC criteria 2.58; 95% confidence interval 1.36-4.91). In contrast with their comparison groups, episodes meeting 2008 CDC criteria did not significantly predict outcomes, whereas VAEs (only four possible VAPs) were associated with significantly more ventilation and PICU length of stay (12-day/8-day increase) and sevenfold increase in mortality. Ped-VAE did not increase mortality, but it was associated with 4-day increase in ventilation and PICU length of stay, with ten possible VAPs, and atelectasis (9/12) as the main paediatric ventilator-associated condition.

CONCLUSIONS: The 2008 CDC criteria did not predict outcomes, whereas VAE only identified very severe events. The Ped-VAE algorithm had more accuracy predicting outcomes by characterizing lower oxygenation changes and identifying hypoxaemia severity, a major driver of management.