Chest drainage insights
Time to read: 1 min.
Dr. Gilbert’s particular interest is in applying process engineering, automation, and machine learning principles to harness the massive volume and variety of patient data that underlie an episode of care but currently go untapped. Pairing these principles and data with underutilised digital tools can foster the development of decision support systems for delivery of safe, efficient, high-quality care.
He reviewed two recent publications from his group that use chest tube management as an example of a classification task that can be used to achieve reliable predictions from data collected digitally, in a sliding window over time, during patient monitoring.
In the first study, Dr. Gilbert and colleagues constructed, tested, and validated a random forest classifier to identify patients who could have chest tubes removed safely and timely during follow-up, based on longitudinal fluid drainage data collected by the Thopaz+ system.1 Their results showed that reliable predictions could be achieved early in the process and that the classifier ensured that chest tubes that needed to be maintained were not removed at the cost of potentially maintaining a few unnecessarily.
The goal of the second study was to use digital data from the Thopaz+ system to derive optimal parenchymal air-leak resolution criteria that minimize duration of chest tube drainage without increasing complications.2 The investigators collected airflow data prospectively from 400 patients and averaged it in 10-minute intervals, and analysed air leak duration and air leak recurrence, frequency, and volume. The analysis was used to identify the optimal criteria based on patient safety (low frequency and volume of air leak recurrences), and efficiency (shortest initial air leak duration). Authors concluded that a postoperative air leak can be deemed resolved if it remains at < 50 mL/minute for 8 consecutive hours. Chest tube removal when air leak reaches this level was optimised for both safety and efficiency.
Together, these studies exemplify how digitally collected data can be combined with advanced analytical techniques to derive information for decision-support.
1. Klement W, Gilbert S, Resende VF et al. The validation of chest tube management after lung resection surgery using a random forest classifier. International Journal of Data Science and Analytics. 2022; 13: 251-263. doi: 10.1007/s41060-021-00296-8
2. Alayche M, Choueiry J, Mekdachi A et al. Determining optimal air leak resolution criteria when using digital pleural drainage device after lung resection. JTCVS Open. 2024; 18: 360-368. doi: 10.1016/j.xjon.2024.01.016
Articles that may be of interest