HAL Université des Antilles
Not a member yet
25497 research outputs found
Sort by
Epidemiology of pain, delirium, psychiatric disorders, discomfort and sedation-analgesia management in the intensive care unit: a one-day nationwide study
No full textInternational audienceBackgroundThe administration of sedatives and analgesics in intensive care units (ICUs) has evolved significantly over the past 20 years, shifting from deep to light sedation strategies to minimize adverse effects. Despite this shift, substantial variability persists in sedation-analgesia practices. This study aimed to provide an updated national overview of sedation-analgesia management with a focus on discomfort assessment practices, including pain, delirium, anxiety, thirst, mood, and sleep disorders.MethodsThis was a one-day, multicenter, cross-sectional study conducted in French ICUs. Data were collected from all adult patients hospitalized in the ICU on the study day. A Unit-level survey documented ICU characteristics and sedation-analgesia protocols. Patient-level data included sedation levels, pain scores, and assessments of discomfort conditions. Statistical analyses were performed using descriptive methods and multilevel logistic regression.ResultsAmong 258 French ICUs contacted, 128 units (50%) participated, enrolling 2,063 patients. Most ICUs were university-affiliated (54%) and mixed medical-surgical (58%); 63% had a written protocol for sedation-analgesia. Sedation and pain were assessed in 96% and 91% of ICUs, respectively. Light or no sedation was observed in 84% of patients, while 15% were deeply sedated – 63% of whom were misaligned with usual indications. Pain assessment was performed at rest in 90% of patients and during care in 62%. Pain prevalence increased with lighter sedation levels and during care. Hypnotics were used in 31% of patients, Mainly propofol and midazolam. Discomfort was reported in 44% of patients, mainly anxiety, sleep disorders, and thirst. Written protocols for sedation and analgesia were not associated with sedation depth, drug use, or delirium screening, but were linked to more frequent pain assessment at rest. Multivariable analyses showed that higher SOFA scores were associated with deep or misaligned deep sedation. The presence of a written protocol for sedation and analgesia reduced the risk of unassessed pain but was not associated with deep or misaligned deep sedation.ConclusionThe shift toward lighter sedation has been successfully achieved; however, a broad spectrum of stressful symptoms persists, including pain, anxiety, thirst, and sleep disruption. These findings underscore the need for more effective strategies to optimize pain and overall patient comfort in non-deeply sedated ICU patients
Histoire de la descendance indienne en Guadeloupe
No full textInternational audienceHistoire de la descendance indienne en Guadeloupe est d’abord une œuvre, enquête et reconquête, véritable plongée dans les contributions féminines et émotions de quatre-vingt familles guadeloupéennes d’origine indienne. Cet ouvrage tend à mettre en lumière les apports et les résistances de grands-mères et mères du XIXe au XXe siècle. Inspirée de la démarche de co-écriture d’Ernest Moutoussamy, cette étude vise à restaurer les mémoires des aïeules indiennes, indo-guadeloupéennes, guadeloupéennes d’origine indienne, à rétablir l’identité indienne dans la culture créole guadeloupéenne : une identité partagée et une histoire créolisée. A travers deux types de figures – celles des pionnières et des gardiennes (ou garantes) – l’étude explore la transmission matrilinéaire intergénérationnelle. Ces femmes ont joué un rôle crucial, fondamental mais silencieux dans la préservation de leurs racines ancestrales. Par le biais des récits de vie et des méthodes de co-écriture, ce projet rassemble des mots, des portraits, des souvenirs couvrant la période de 1855 à nos jours. Ces femmes, devenues « pionnières », puis « gardiennes », ont contribué à bâtir des foyers malgré les interdits et les tensions dans la société coloniale. Oeuvrant pour l’éducation des enfants, elles ont permis l’ascension sociale des leurs. Toutes ont au fil du siècle lutté contre l’assimilation progressive, transcendé les barrières sociales et se sont dévouée à l’émancipation de leurs enfants. Cet ouvrage souligne l’importance de la mémoire matricielle, ancré dans les récits de l’enfance ; il perpétue et préserve ces trésors fragiles pour les générations futures
Pressurised aquaria for the study of deep-water Corals
No full textInternational audienceA set of four identical pressurised mesocosms is presented, aiming at long-term incubations of deep-water corals, as deep as 3000 m. Care was also taken to enable practical boarding of these instruments on oceanographic ships. Four-month incubations of two scleractinian species (D. pertusum and M. oculata) originating from 800 m depth were achieved at the laboratory (including 3 months at 8 MPa pressure). Two of the aquaria were also operated during a 2-week cruise in the Bay of Biscay, focusing on the same species. Specific requirements for long-term studies are exposed and discussed, emphasizing resistance to corrosion and the possibility to feed fauna without decompression. The first-time long-term incubation of deep-water corals at in situ pressure opens perspectives for future studies, including investigations on deeper corals not yet accessible to laboratory experiments. The use of pressurised mesocosms may be of particular importance, considering the predicted consequences of ocean warming and acidification on the bathymetric distribution of reef-forming deep-water scleractinians
Using game-based research approaches to gauge children's perceptions: insights from a food education project
No full textInternational audienceThis paper provides insight into how gaming can be used to explore the children's experience of food, and to envision possible future transitions in their eating habits. The case study presented is part of the ERMES project, an interdisciplinary initiative funded by the French National Research Agency (ANR-23-CE36-0009), focused on primary school children aged 9-10 in the Îlede-France region. The project aims to develop scientific understanding on how children engage with and relay "food messages" in their daily lives, and how these messages influence their attitudes and eating habits. The paper presents and discusses the use of research game-based workshops to capture children's attitudes, preferences, and understanding of nutrition in a playful and engaging manner. These activities not only facilitate data collection but also foster trust and active participation among children. Additionally, the case study highlights the challenges of conducting action-oriented research in school settings, including minimizing researcher influence and addressing scientific bias. Findings suggest that game-based approaches are effective in exploring complex topics like sustainable eating, offering insights into children's peer culture and food-related behaviors. However, the study emphasizes that game-based activities should complement, rather than replace, traditional methods such as observations and surveys. By integrating games into research, the study provides a novel framework for understanding children's food perceptions, contributing to the development of effective food education strategies
"Mayotte et la Réunion: de la cohésion numérique à la cohésion territoriale, ou l'émergence du concept de "communauté d'intérêts" dans le droit de l'Union européenne". dans LE NUMERIQUE ET LES OUTRE-MER DE L'UNION EUROPEENNE - test ocdhal PR 19/01
No full textLa Traduction à Mayotte et à la Réunion du concept de "communauté d'intérêts" sous l'angle de la politique de cohésion de l'Union européenne dans le secteur numérique
LUMA: A Benchmark Dataset for Learning from Uncertain and Multimodal Data
No full textInternational audienceMultimodal Deep Learning enhances decision-making by integrating diverse information sources, such as texts, images, audio, and videos. To develop trustworthy multimodal approaches, it is essential to understand how uncertainty impacts these models. We propose LUMA, a unique benchmark dataset, featuring audio, image, and textual data from 50 classes, for learning from uncertain and multimodal data. It extends the well-known CIFAR 10/100 dataset with audio samples extracted from three audio corpora, and text data generated using the Gemma-7B Large Language Model (LLM). The LUMA dataset enables the controlled injection of varying types and degrees of uncertainty to achieve and tailor specific experiments and benchmarking initiatives. LUMA is also available as a Python package including the functions for generating multiple variants of the dataset with controlling the diversity of the data, the amount of noise for each modality, and adding out-of-distribution samples. A baseline pre-trained model is also provided alongside three uncertainty quantification methods: Monte-Carlo Dropout, Deep Ensemble, and Reliable Conflictive Multi-View Learning. This comprehensive dataset and its benchmarking tools are intended to promote and support the development, evaluation, and benchmarking of trustworthy and robust multimodal deep learning approaches. We anticipate that the LUMA dataset will help the ICLR community to design more trustworthy and robust machine learning approaches for safety critical applications
Spatio-temporal distribution and environmental determinants of dengue vectors in Phnom Penh, Cambodia
No full textInternational audienceDengue fever, one of the most widespread vector-borne diseases globally, is mainly transmitted by Aedes aegypti and Ae. albopictus mosquitoes. In Cambodia, dengue has been a recurrent public health challenge, with major outbreaks occurring in 1995, 2007, 2012, and 2019. The latter epidemic severely impacted the capital, Phnom Penh, yet the spatial and temporal dynamics of the two key vector species had not been studied in this urban context. This study aimed to investigate how the distribution of Ae. aegypti and Ae. albopictus is organized in the urban and peri-urban landscapes of Phnom Penh. Ovitraps were deployed every two months over a year in 40 pagodas randomly selected across Phnom Penh, chosen to ensure future replicability of the study. The larvae collected were reared to adulthood for accurate species identification. High-resolution satellite imagery (SPOT7) and daily rainfall data were used to analyze the surrounding environments through remote sensing techniques. The results revealed distinct spatio-temporal patterns for each species: Ae. albopictus was associated with peri-urban areas rich in vegetation and water, while Ae. aegypti predominated in highly urbanized and construction-dense environments. Spatial analysis using buffer zones (250 m, 500 m, 1000 m) around trapping sites confirmed that the use of pagodas as proxies for urban sampling is effective. These findings highlight the importance of monitoring these vector species, particularly as Phnom Penh continues to undergo rapid environmental transformation. The identification of simple, remotely sensed environmental indicators offer a valuable tool for predicting future outbreaks and guiding targeted vector control strategies. This study also provides a replicable methodological framework to assess the impact of urbanization and climate change on dengue vector distribution in Phnom Penh and similar urban settings
Climate Change Discourse and Coastal Erosion in New Caledonia: an Analysis of Press Coverage in Les Nouvelles Calédoniennes
No full textInternational audienc
Dialogue entre science et société civile au REVOSIMA : le groupe de travail « Médiation scientifique »
No full textNational audienc
F-LoRA-QA: Finetuning LLaMA Models with Low-Rank Adaptation for French Botanical Question Generation and Answering
No full textInternational audienceDespite recent advances in large language models (LLMs), most question-answering (QA) systems remain English-centric and poorly suited to domain-specific scientific texts. This linguistic and domain bias poses a major challenge in botany, where a substantial portion of knowledge is documented in French. We introduce F-LoRA-QA, a fine-tuned LLaMA-based pipeline for French botanical QA, leveraging Low-Rank Adaptation (LoRA) for efficient domain adaptation. We construct a specialized dataset of 16,962 question-answer pairs extracted from scientific flora descriptions and fine-tune LLaMA models to retrieve structured knowledge from unstructured botanical texts. Expert-based evaluation confirms the linguistic quality and domain relevance of the generated responses. Compared to baseline LLaMA models, F-LoRA-QA achieves a four-fold improvement in BLEU, a 70% ROUGE-1 F1 gain, a 16.8% increase in BERTScore F1, and an Exact Match improvement from 2.01% to 23.57%. These results demonstrate the effectiveness of adapting LLMs to low-resource scientific domains and highlight the potential of our approach for automated trait extraction and biodiversity data structuring