The Definition of Pediatric Sepsis has evolved over the years, and the prediction, detection and response
to pediatric sepsis and its consequences for multiple organ failure are extremely important. In this session, we
will review the data and process for development and publication of the Novel Phoenix Pediatric Sepsis
Criteria, using broad registries of data and an artificial intelligence (AI) informed analysis to generate and
validate the published criteria. Note that data from both typical resource rich and global resource constrained
environments were included in the analytic dataset. This unique process of consensus Sepsis definitions and
criteria development for research will be presented and discussed.

This presentation is from the SCCM 2024 Congress that included a review of the most recent literature covering topics in neurocritical care. These included publications regarding acute ischemic stroke (AIS), intracerebral hemorrhage (ICH), subarachnoid hemorrhage (SAH), and subdural hematomas (SDH).

The audience take aways will include: lower blood pressure targets are not necessarily optimal after mechanical thrombectomy for AIS; early recurrent AIS stroke prophylaxis with a direct oral anticoagulant (DOAC) for atrial fibrillation is safe compared to delayed DOAC initiation; intravenous thrombolysis can be safe in AIS patients taking a DOAC; tirofiban may improve outcomes after AIS; bundled goal-directed care after ICH may improve outcomes; drainage of cerebrospinal fluid by lumbar drain after SAH may improve outcomes; evacuation of an acute SDH by craniotomy has similar outcomes compared to decompressive craniectomy; and dexamethasone is not a safe alternative to burr hole evacuation of a chronic SDH.

Ongoing research continues to refine the management of critically ill patients with neurological injuries.

We are at the precipice of an artificial intelligence or AI revolution. AI is unique because it is boundless when it comes to applications. It is poised to upend every aspect of our lives, from healthcare, to education, to law enforcement, AI can theoretically augment every human task that involves thinking, and we shall see every decision-making infiltrated by AI. But the biggest difference of AI from other innovations lies in how it is developed. AI is inequitable by design because its raw materials are the digital exhaust of health systems that we know are discriminatory.
AI has the potential to transform critical care medicine worldwide. bearing promises of increased accuracy, efficiency, and cost-effectiveness, in areas as diverse as clinical prediction and prognostication, disease phenotyping, and treatment optimization, among others. But even with high quality data from the intensive care units from across the world, the social patterning of the data generation process can still produce AI that is bound to preserve and even scale existing disparities in care with resulting inequities in patient outcomes. For example, medical devices that measure physiologic signals and biologic markers are developed with subjects that often don’t fully represent those who are most vulnerable to the disease processes in question. Non-random variations in testing and monitoring are bound to adversely affect accurate depictions of disease trajectory in a manner that cannot be compensated for by current imputation methodologies. The promise/hype of AI will translate into huge dividends only if the intensive care community works with computer scientists, social scientists, patients and their caregivers in understanding the backstories of the data and designing an equity-focused curation and analytics pipeline.

Evidence informed Extubation criteria and clinical pathways, and the risks/benefits of extubation readiness
tests for pediatric patients have been debated for many years. The evidence basis, pros and cons, for
extubation readiness testing and clinical pathways will be presented and discussed. Clinical pathway
compliance decrease morbidity, duration of mechanical ventilation and ICU length of stay, but often require
ancillary staffing and support. In addition, special circumstances and unique patient populations may not
benefit to the same extent as the general ICU population. Appropriate evaluation and planning will be
discussed and debated, with welcome input from attendees on their clinical practice and challenges.

Post-cardiac arrest syndrome, including global ischemia-reperfusion injuries, during cardiac arrest and cardiopulmonary resuscitation leads to the brain and multiple organ damages. Post-cardiac arrest care has evolving recently with efforts on improving the outcomes for the cardiac arrest patients. Chain of survival for cardiac arrest includes the concepts of post-cardiac arrest care and recovery in the post-cardiac arrest period. Structured interventions and systemic approaches have been proposed in treating cardiac arrest patients. Targeted temperature management (TTM) has been proposed to improve neurological outcomes. Concepts of optimal targeted temperature has been challenged in recent trials. High quality post-resuscitation bundle cares have become the most issue in the current post-cardiac arrest care. Rehabilitation and recovery are mentioned to increase the chance of bringing cardiac arrest patients to normal social life. The challenges to implement high quality post-cardiac arrest exist. It needs novel interventions and solutions in the future.

This presentation will introduce the selection of intravenous fluids for resuscitation in various critical conditions based on the latest clinical practice guidelines. It will also discuss the pros and cons of different intravenous fluids based on their composition.

Trauma-related airway difficulties in trauma patients are critical challenges requiring immediate and effective management to ensure survival. These emergencies often stem from direct airway
injuries, cervical spine trauma, or compromised ventilation due to thoracic trauma. Rapid assessment and intervention are crucial, utilizing techniques such as endotracheal intubation, cricothyrotomy, or tracheostomy based on the injury's severity and nature. Advanced airway management protocols and training for healthcare providers are essential to improve outcomes. Continuous monitoring and supportive care in trauma patients are vital to address complications and maintain airway patency throughout the recovery process.

The integration of deep learning (DL) technologies into trauma care might transform diagnostic and decision-making processes. In this presentation, I will discuss the application of DL models for trauma imaging, including rib fractures, pelvic fractures, abdominal sonography, abdominal computed tomography (CT), and a bleeding prediction model for intensive care unit (ICU) patients.

Our DL models for rib and pelvic fractures, based on X-rays, have demonstrated improved diagnostic accuracy and sensitivity compared to traditional radiological assessments. The DL models can achieve higher accuracy rates in detecting subtle rib fractures, particularly in cases where conventional methods may overlook minor fractures. Similarly, we also highlighted the role of DL in enhancing pelvic fracture detection, reducing misdiagnosis and expediting treatment decisions.

For abdominal trauma, DL-enhanced ultrasonography and CT scanning play a critical role in detecting solid organ injuries, such as liver, spleen or kidney lacerations. The exsting DL algorithms could increase the sensitivity of detecting visceral injuries, even when clinical signs are absent. I also demostrate the feasibility to apply DL model for CT scanning to detect the solid organ injuries.

Moreover, our ICU bleeding prediction model analyzes physiological and laborary data, predicting bleeding risk with a high degree of accuracy. This study demonstrated that AI-driven prediction models significantly reduce mortality by providing early intervention opportunities.

Through these concepts, I will demonstrate how DL models are advancing acute care surgery, leading to improved diagnostic accuracy, expedited decision-making, and optimized patient management in trauma settings

The respiratory management of immunocompromised patients requires selecting appropriate ventilation
strategies based on their specific conditions and risk factors. These patients often have weakened
immune systems due to treatments such as chemotherapy, organ transplantation, or long-term steroid use.
Therefore, when acute respiratory failure (ARF) occurs, it is crucial to choose suitable respiratory support
strategies. The main ventilation strategies include Non-Invasive Ventilation (NIV) and High-Flow Nasal
Cannula (HFNC). NIV is recommended as the first-line treatment for ARF in immunocompromised patients
because it can reduce intubation and mortality rates. However, the application of NIV remains controversial
in some cases, especially when patients exhibit strong inspiratory efforts and high tidal volumes, which may
worsen lung injury.
Studies have shown that when immunocompromised patients require Invasive Mechanical Ventilation (IMV),
the mortality rate can reach 50% to 70%. Early use of NIV can reduce intubation rates and ICU mortality to
some extent, but if the patient's condition is severe and intubation cannot be avoided, delayed intubation may
increase the risk of death. Therefore, early identification of patients who need IMV and timely intubation is
key to avoiding unnecessary delays and high mortality rates.
During the COVID-19 pandemic, managing immunocompromised patients has become more complex,
requiring special consideration of viral infection risks and the choice of respiratory support strategies.
For these patients, selecting the appropriate ventilation strategy is particularly important to minimize
complications and improve survival rates. Overall, ventilation management for immunocompromised
patients requires individualized treatment plans, adjusting respiratory support strategies according to
the patient's condition and risk factors. Early identification and timely intervention are crucial for improving
outcomes. More research is needed in the future to determine the best oxygenation strategies and
management plans to reduce mortality in these high-risk patients.

The human body hosts over 10,000 different microbe species, forming complex communities known as microbiota. The aggregate of these microbial species, along with their genetic information and functions, is referred to as the microbiome. Initiated in 2007, the Human Microbiome Project (HMP) aims to enhance our understanding of human–microbe interactions and clarify the mechanisms of host–microbiome interactions under specific conditions. Alterations in microbiomes from different bodily regions can influence human development and lead to abnormal development and disease. Over 90% of diseases can be traced back to microbiota dysbiosis. While the respiratory microbiome has been less explored compared to the gut microbiome, its dysbiosis is speculated to play a significant role in ventilator-associated pneumonia (VAP) and acute respiratory distress syndrome (ARDS). Despite limited studies and small cohorts, there is evidence of overlap between bacterial populations in the lower respiratory and oropharyngeal tracts. A comprehensive understanding of microbial interplay and host-microbiome interactions could guide better preventive and therapeutic strategies for ARDS and VAP. We herein review and summarizes current knowledge and potential clinical applications of respiratory microbiome research in mechanically ventilated patients.

Obstruction sleep breathing disorder, obesity
Sleep is a naturally occurring periodic, reversible state of cognitive and sensory disengagement from the external environment, from which a person can be aroused by sensory stimulation. Sleep is essential for physiologic rest and emotional well-being Chemically-induced sedation is physiologically different to normal sleep and lacks restorative effects. Sleep patterns and architecture in the critically ill is usually abnormal. There are multiple causes of disrupted sleep in ICU.
I focus on the complication of sleep breathing disorder and obesity in ICU. Obstructive sleep apnea is common in patients admitted to the ICU. Obstructive sleep apnea was associated with a reduction in both ICU and hospital mortality.
Journal of Critical Care Volume 30, Issue 1. February 2015, Pages 178-18
The prevalence of obesity is increasing worldwide and is around 20% in ICU patients. Obesity is strongly linked with chronic diseases such as type 2 diabetes, hypertension, cardiovascular diseases, dyslipidemia, non-alcoholic fatty liver disease, chronic kidney disease, obstructive sleep apnea and hypoventilation syndrome, mood disorders and physical disabilities. Intensive Care Med (2019) 45:757–769
We analysis the four endotypic traits, namely arousal threshold, upper airway collapsibility, loop gain, and upper airway muscle compensation, were estimated using the PUP method, specifically through a validated cloud-based Python application (PUPpy) (E. Finnsson et al., 2021).
Endotyping has the potential to open novel therapeutic avenues. BiPAP or CPAP is highly efficacious in treating OSA with hypoventilation, regardless of the underlying pathophysiology of OSA. Therefore, endotypes could help choose the personalised alternatives to BiPAP of CPAP. Novel pharmacological therapies targeted at specific endotypes have shown promise but are not yet fully evaluated. thorax-2023-22003

The rapid growth of Generative AI (GAI) is revolutionizing emergency and critical care, offering
unprecedented opportunities and challenges. As GAI technologies, including machine learning
(ML) and large language models (LLMs), advance, they enable more efficient, accurate, and
personalized patient care. In this speech, we will explore the vast potential of GAI in enhancing
diagnostic accuracy, predicting patient outcomes, and optimizing treatment strategies in highstakes
environments.
Opportunities abound in leveraging GAI to improve decision-making processes, streamline
workflows, and enhance the integration of complex data. However, the challenges of deploying
these technologies in clinical settings are equally significant, including ethical considerations,
data privacy, and the need for continuous model validation and fine-tuning.
The transition from traditional ML workflows to LLM-driven approaches will be highlighted,
showcasing how LLMs can refine clinical reasoning through advanced prompting techniques,
Retrieval-Augmented Generation (RAG), and fine-tuning. These methods allow for more
nuanced and context-aware AI-driven support, ultimately improving patient outcomes in
emergency and critical care.
This talk aims to provide a comprehensive overview of GAI's current state, its future in
healthcare, and the practical steps needed to harness its full potential while addressing the
inherent challenges.

Key to Seamless Integration: Introducing generative AI into healthcare systems must be painless, reducing resistance and difficulty during the implementation process to ensure smooth adoption.
Generative AI Proposal Criteria: When evaluating a proposal, ask three questions first: Does it improve efficiency? Does it reduce errors? Does it mitigate legal risks?
Limitations and Challenges of generative AI: LLMs (Large Language Models) excel at handling data, but in practical applications, they may produce correct yet impractical outputs. For instance, an anorexic patient’s BMI of 14.8 is an objective fact, but directly recording it in medical notes might negatively impact the patient's emotional state.
Product Demonstration and Application: Medical documents without personal data or privacy-sensitive information are uploaded to the cloud, enabling a chatbot to respond to user inquiries. These tools allow frontline healthcare workers to quickly access accurate information while ensuring patient privacy is protected.
Growth Process of Doctors, Residents, and Medical Students: Effective management of personalized knowledge is crucial. Using tools like Obsidian, medical professionals can convert notes into bi-directional links, creating a knowledge graph that enhances learning outcomes.
Personalized Knowledge Management and Tool Application: Daily handwritten notes from healthcare professionals can be converted into Markdown syntax for use in knowledge management tools, improving work efficiency and knowledge integration.