Content area
Full text
Background
Before and during invasive airway management, trauma patients are at high risk of hypoxia due to primary lung injury, hypovolemia, insufficient respiratory drive, lack of airway protection, or airway injury [1]. Therefore, counter-strategies are an important part of emergency airway management [2]. Basically, several methods are available: oxygen mask, bag-valve mask, high flow oxygen therapy or non-invasive ventilation (NIV) with positive end expiratory pressure. Whereas oxygen and bag-valve masks are inexpensive and easy to use, the amount of deliverable oxygen is limited and assisted respiratory support can be technically challenging [3, 4]. In contrast, NIV, especially in a pressure support mode, not only improves alveolar recruitment and therefore oxygenation and denitrogenation, but can also increase minute ventilation. This makes NIV a favored method of choice in hypercapnic respiratory failure [5, 6]. Baillard et al. demonstrated that, for the intubation of hypoxemic patients, preoxygenation with NIV is more effective at avoiding arterial oxyhemoglobin desaturation than a non-rebreather bag-valve mask [7]. In the emergency department and in the intensive care unit, Gibbs et al. recently reported not only a lower rate of cardiac arrest but also a halving of desaturation with NIV for preoxygenation compared to a simple oxygen mask. Importantly, the incidence of aspiration was not increased [8]. However, in prehospital emergency medicine, on-site invasive medical treatment might be challenging and available resources at the scene are limited. Furthermore, patients with medical emergencies often exhibit an altered state of consciousness or are at risk of aspiration, both of which are contraindications for NIV. Therefore, the current German guideline on prehospital airway management considers NIV only potentially superior in preoxygenation [2]. However, current data on preoxygenation methods for prehospital invasive airway management in Germany are lacking.
In recent years, studies using machine learning have given new insights in prehospital emergency care. Basically, machine learning can be applied to four different problems: earlier disease identification (for example forecasting resuscitation during transport), disease evolution prediction (e.g. success of resuscitation), disease phenotyping (like sepsis patterns) and guiding clinical decisions (e.g. airway management in trauma patients) [9, 10, 11, 12–13]. In addition to a classical direct statistical attribute comparison, the algorithms of machine learning can gain deeper insights by addressing complex attribute dependencies [14].
The aim of the study is...