Using health administrative databases to identify mechanical ventilation in transported pediatric critically ill patients

RATIONALE: Accurate identification of children who received mechanical ventilation (MV) is important for operational decisions related to resource allocation and to provide valid population-level evaluations of critically ill children. OBJECTIVES: The object of this study was to validate health administrative database codes for the identification of MV in children. METHODS: Algorithms composed of hospital-based Canadian Classification of Health Interventions (CCI) and physician billing codes were validated against a reference dataset composed of critically ill pediatric patients transported to the pediatric critical care unit at the Hospital for Sick Children in Toronto, Canada between 2004 and 2012. MEASUREMENTS: Descriptive statistics, sensitivity, specificity and positive and negative predictive values (PPV and NPV, respectively) were obtained. MAIN RESULTS: Of 611 patients, 75% received MV, and of these, 94% received invasive MV only. For all types of MV (invasive and noninvasive), CCI and billing codes had a sensitivity of 85.6% and 87.8% and a specificity of 67.8% and 41.4%, respectively. The combination of CCI and billing codes yielded a sensitivity of 98% and a specificity of 37.5% for MV. Invasive MV CCI codes had a sensitivity of 86.3% and specificity of 71.5%. When only noninvasive MV CCI codes were tested, the sensitivity was 40.2% and the specificity was 97.1%. The PPV was highest using the CCI codes alone for any MV (89%) and for invasive MV (88%). CONCLUSIONS: The combination of CCI and billing codes yielded an excellent sensitivity for MV; however, there is a risk of overestimating MV as specificity was low for all algorithms