Explore the vast range of titles available.

PurposePatients receiving venoarterial extracorporeal membrane oxygenation (VA-ECMO) frequently develop arterial hyperoxaemia, which may be harmful. However, lower oxygen saturation targets may also lead to harmful episodes of hypoxaemia.MethodsIn this registry-embedded, multicentre trial, we randomly assigned adult patients receiving VA-ECMO in an intensive care unit (ICU) to either a conservative (target SaO2 92–96%) or to a liberal oxygen strategy (target SaO2 97–100%) through controlled oxygen administration via the ventilator and ECMO gas blender. The primary outcome was the number of ICU-free days to day 28. Secondary outcomes included ICU-free days to day 60, mortality, ECMO and ventilation duration, ICU and hospital lengths of stay, and functional outcomes at 6 months.ResultsFrom September 2019 through June 2023, 934 patients who received VA-ECMO were reported to the EXCEL registry, of whom 300 (192 cardiogenic shock, 108 refractory cardiac arrest) were recruited. We randomised 149 to a conservative and 151 to a liberal oxygen strategy. The median number of ICU-free days to day 28 was similar in both groups (conservative: 0 days [interquartile range (IQR) 0–13.7] versus liberal: 0 days [IQR 0–13.3], median treatment effect: 0 days [95% confidence interval (CI) – 3.1 to 3.1]). Mortality at day 28 (59/149 [39.6%] vs 59/151 [39.1%]) and at day 60 (64/149 [43%] vs 62/151 [41.1%] were similar in conservative and liberal groups, as were all other secondary outcomes and adverse events. The conservative group experienced 44 (29.5%) major protocol deviations compared to 2 (1.3%) in the liberal oxygen group (P < 0.001).ConclusionsIn adults receiving VA-ECMO in ICU, a conservative compared to a liberal oxygen strategy, did not affect the number of ICU-free days to day 28.

The impossible task paradigm has been extensively used to study the looking back behaviour in dogs. This behaviour is commonly considered a social problem-solving strategy: dogs facing an unsolvable task, soon give up and look back at the experimenter to ask for help. We aimed to test if the looking back in an impossible task does indeed represent a social problem-solving strategy. We used a modified version of the classic impossible task, in which the subjects simultaneously faced three possible and one impossible trials. Additionally, subjects were tested in four different conditions: social condition (with an unknown experimenter); asocial condition (subject alone); 'dummy' human condition (with a ‘dummy’ human); object condition (with a big sheet of cardboard). Finally, we compared two populations of dogs differing in their experience of receiving help from humans: 20 pet dogs tested in their houses and 31 free-ranging dogs tested in Morocco. We found that the pet dogs and free-ranging dogs had similar persistence in interacting with the impossible task in all conditions. Moreover, subjects looked back with similar latencies at the human, at the dummy human and at the object. Overall, pet dogs looked back longer at the human than free-ranging dogs. This could be interpreted as pet dogs being more attracted to humans and/or having a stronger association between humans and food than free-ranging dogs. Concluding, the looking back in an impossible task does not represent a problem-solving strategy. This behaviour seems rather linked to the subject’s persistence, to the salience of the stimuli presented, and potentially to the past reinforcement history.

Background: Data errors are a common challenge in machine learning (ML) projects and generally cause significant performance degradation in ML-enabled software systems. To ensure early detection of erroneous data and avoid training ML models using bad data, research and industrial practice suggest incorporating a data validation process and tool in ML system development process. Aim: The study investigates the adoption of a data validation process and tool in industrial ML projects. The data validation process demands significant engineering resources for tool development and maintenance. Thus, it is important to identify the best practices for their adoption especially by development teams that are in the early phases of deploying ML-enabled software systems. Method: Action research was conducted at a large-software intensive organization in telecommunications, specifically within the analytics R\\&D organization for an ML use case of classifying faults from returned hardware telecommunication devices. Results: Based on the evaluation results and learning from our action research, we identified three best practices, three benefits, and two barriers to adopting the data validation process and tool in ML projects. We also propose a data validation framework (DVF) for systematizing the adoption of a data validation process. Conclusions: The results show that adopting a data validation process and tool in ML projects is an effective approach of testing ML-enabled software systems. It requires having an overview of the level of data (feature, dataset, cross-dataset, data stream) at which certain data quality tests can be applied.