Automated oxygen control in the preterm infant: automation yes, but we need intelligence

Artificial intelligence (AI), the mimicry of human intelligence with computer systems, has found its way into so many fields of human endeavour and is finding its way into ours.1 Already well entrenched in several healthcare disciplines, there would appear to be great potential for AI and lesser forms of automation in neonatology and neonatal intensive care, and examples already exist in the areas of clinical decision support (eg, risk prediction from large and/or complex datasets2) and machine-oriented repetitive operations (eg, tidal volume targeting3 and temperature control4). To this latter list can also be added automated control of inspired oxygen concentration in preterm infants, numerous algorithms for which have been developed over the years, with a number of them now incorporated in contemporary neonatal ventilators and studied in preclinical5 or clinical6–10 settings. Common findings are that these devices afford a greater proportion of time with oxygen saturation (SpO2) in the target range, with less time in the extremes of hypoxaemia and hyperoxaemia. Seemingly against this tide of technological advance, the evolution of respiratory care in preterm infants in recent decades has seen the persistence/resurgence of one ‘low-tech’ form of mechanical respiratory support, continuous positive airway pressure (CPAP) and the emergence of another, nasal high flow (HF).11 Within a few years, enthusiasm for nasal HF has burgeoned, and it has been used in increasing numbers of preterm infants, at a variety of stages of their journey through the neonatal intensive care unit (NICU). Many aspects of the nasal HF package are attractive; it is simple and easy to use, induces less nasal trauma and can be widely applied, including in low-resource settings. Given the potential value of automated oxygen control for preterm infants, and the increasing uptake of nasal HF, it makes sense to link the …