Intrapleural Airflow Signal Processing to Predict Duration of Pulmonary Air Leak: A Preliminary Analysis
Daniel French, Sebastien Gilbert.
University of Ottawa, Ottawa, ON, Canada.
OBJECTIVE: The ability to accurately predict patients that will develop a prolonged air leak (PAL) in the early postoperative period is very useful in optimizing patient management, discharge planning and use of hospital resources. Digital drainage systems allow airflow signals to be captured and analyzed. The objective of this study is to analyze the characteristics of the airflow signal in the early postoperative period to identify patients that will develop a prolonged air leak.
METHODS: The airflow signals of patient who underwent anatomic lung resections were prospectively captured using digital drainage systems and retrospectively analyzed. The absence of an air leak was defined as no airflow signal greater than 20 mL/minute for 8 hours. The volume of airflow in the first 12 hours was computed by calculating the area under the airflow curve. The mean volume of airflow for each class of signal was compared. A receiver operator characteristic (ROC) curve was plotted and sensitivity and specified were computed for multiple volume thresholds.
RESULTS: Of the 67 patients included in the analysis, 43 (64%) patients never developed an air leak, 16 (24%) patients had an air leak that resolved within 5 days, and 8 (12%) patients had a prolonged air leak. The ROC curve is shown as Figure 1. The area under the curve (AUC) of the ROC is 0.94 (95% CI: 0.88 - 0.99). Using different thresholds, the sensitivity and specificity were computed as follows: 1055 mL (sensitivity=1, specificity=0.73), 5530 mL (sensitivity=1, specificity=0.85) and 10,706 mL (sensitivity=0.75, specificity=0.88).
CONCLUSIONS: The volume of airflow measured in the first 12 hours after a pulmonary resection can predict patients that will develop a prolonged air leak. Further investigations are needed to validate these findings on a larger set of airflow signals and should include analysis of other airflow signal characteristics that can predict duration of air leak after pulmonary resection.
Back to 2016 Annual Meeting Thoracic Track