The Positive Predictive Value of Ventilator Alarms in a Pediatric Intensive Care Setting

Background: Clinical alarms are critical in the early recognition and timely intervention to changes in the patient’s vital signs. However, frequent occurrences of non-actionable alarms place patients at increased risk of harm due to alarm fatigue. Earlier studies suggest ventilator alarms contribute significantly to the total number of non-actionable alarm events in the intensive care setting. Ventilator alarms and their positive predictive value (PPV) are not well studied. Understanding this is fundamental to effective management of ventilator alarms. Methods: A retrospective review was performed on ventilator alarm occurrences for two common types of ventilators used in the pediatric intensive care (PICU) and pediatric cardio-thoracic intensive care (CTICU) units between June 1, 2017 and November 31, 2017, at a major metropolitan children’s hospital. Data collected included vital signs associated with alarm data for the Carefusion Avea and Maquet Servo-I ventilators. Analysis was performed using Matlab to identify the PPV of ventilator alarms related to changes in ETCO2, SpO2 and interventions by increasing FIO2 levels. Results: The PPV of eleven unique types of ventilator alarms were identified in the PICU and CTICU during 2,091 days of mechanical ventilation. Overall PPV of ventilator alarms was 27% when using oxygenation intervention indices. The PPV was greatest for High Ve and Apnea alarms at 60% and 62% respectively. Overall, ventilator alarm PPV based on ETCO2 changes was 34%. PPVs varied significantly between the different types of alarms. High Ve, high VT and apnea alarms were most predictive of ETCO2 change at 55%, 43% and 40% PPVs respectively. Conclusions: The PPV of ventilator alarms vary significantly between different types of alarms. The PPV for most types of ventilator alarms is statistically significant despite little clinical change in ETCO2 or SpO2. Some alarms are two to three times more effective in predicting physiological change.