Hasil untuk "Medical emergencies. Critical care. Intensive care. First aid"

Alexandre De Masi, Sergio Manzano, Johan N. Siebert et al.

Artificial intelligence systems that record voice and video during pediatric emergencies are emerging as human-computer interaction (HCI) technologies with direct implications for clinical work, promising improvements in documentation, team performance, and post-event debriefing. Yet the perspectives of those most affected, including clinicians, parents, and child patients, remain largely absent from the design and governance of these technologies. This position paper argues that this has direct consequences for the legitimacy and effectiveness of these systems. We examine four areas where these missing perspectives prove consequential (consent, emotional impact, surveillance dynamics, and participatory governance) and propose four positions for reorienting AI recording in pediatric emergency care toward stakeholder-centered HCI inquiry.

Emmett B. Kendall, Jonathan P. Williams, Curtis B. Storlie et al.

Detection of occult hemorrhage (i.e., internal bleeding) in patients in intensive care units (ICUs) can pose significant challenges for critical care workers. Because blood loss may not always be clinically apparent, clinicians rely on monitoring vital signs for specific trends indicative of a hemorrhage event. The inherent difficulties of diagnosing such an event can lead to late intervention by clinicians which has catastrophic consequences. Therefore, a methodology for early detection of hemorrhage has wide utility. We develop a Bayesian regime switching model (RSM) that analyzes trends in patients' vitals and labs to provide a probabilistic assessment of the underlying physiological state that a patient is in at any given time. This article is motivated by a comprehensive dataset we curated from Mayo Clinic of 33,924 real ICU patient encounters. Longitudinal response measurements are modeled as a vector autoregressive process conditional on all latent states up to the current time point, and the latent states follow a Markov process. We present a novel Bayesian sampling routine to learn the posterior probability distribution of the latent physiological states, as well as develop an approach to account for pre-ICU-admission physiological changes. A simulation and real case study illustrate the effectiveness of our approach.

Arwin Nemani, Schahryar Kananian, Annabelle Starck et al.

Abstract Background Refugees and asylum seekers encounter numerous post-migration living difficulties (PMLDs) that can substantially affect their mental health. However, the role of PMLDs remains insufficiently explored, particularly in clinical refugee populations. This study aimed to identify subgroups based on patterns of PMLD by examining their relationship with depressive symptoms and determining which stressors function as key bridges. Methods This study reports a secondary analysis of baseline data from the ReTreat trial. Data were collected from 141 refugees and asylum seekers enrolled in a multicentre randomized controlled trial of a culturally adapted CBT program in Germany. Participants completed measures of depressive symptoms (PHQ-9) and post-migration stressors (27-item checklist). Latent Profile Analysis (LPA) was used to identify distinct burden profiles. Exploratory Factor Analysis (EFA) examined the dimensionality of PMLDs. Network analysis was conducted to investigate symptom–stressor connectivity. Results Three latent profiles emerged: Class 1 showed elevated distress across all domains; Class 2 was characterized by family separation and homesickness; and Class 3 exhibited minimal post-migration stress. EFA of PMLDS supported a four-factor solution: institutional/legal stressors, structural hardship, health/service access, and emotional/family-related strain. Depressive symptoms differed significantly across profiles, with highest scores in the high burden group (Class 1). Network analysis identified institutional/legal and emotional/family-related stressors as central bridge nodes linking PMLDs to depressive symptoms. Conclusions PMLDs are multidimensional and heterogeneously distributed among forcibly displaced individuals. Legal insecurity and emotional strain are particularly influential in connecting environmental hardship to depressive symptoms. Trial registration This study uses baseline data from a registered randomized controlled trial (DRKS00021536).

Basri Lenjani, Agron Dogjani, Luljeta Abdullahu et al.

Introduction: A traffic accident is when only material damage is caused to the vehicle and track environment, and there are no casualties. RTA represents a significant risk for morbidity and mortality in Kosovo, of which head injury and multiple-site injury increase injury severity. The anatomical site, mechanism of injury, time to reach an initial health facility, time of the day, patient condition at ED, type of treatment given, GCS at admission, and days spent in the hospital were among independent predictors of management outcome. Targeted approaches to improving the care of the injured victims may improve outcomes. Thus, the clinician should consider the clinical presentation of RTA and give due attention to the identified contributing factors in managing it. Law enforcers should also emphasize the identified types and mechanisms of accidents. The PubMed database was utilized for article selection, and papers were obtained and reviewed. The ATLS protocol has been developed to manage trauma patients systematically so as not to miss any condition that may kill the patient. Conclusions: Triage is essential in managing accidental situations and strengthening the primary, secondary, and tertiary health systems. To design clinical guidelines, algorithms, and triage protocols at the three levels of health care, all healthcare professionals should be educated and trained with continuing courses in triage, communication, and Basic Life Support -AED, ACLS, PHTLS, BTLS, and ATLS.

Kaibin Liu, Di Qian, Dongsheng Zhang et al.

Abstract Background Early treatment and prevention are the keys to reducing the mortality of VTE in patients with thoracic trauma. This study aimed to develop and validate an automatic prediction model based on machine learning for VTE risk screening in patients with thoracic trauma. Methods In this national multicenter retrospective study, the clinical data of chest trauma patients hospitalized in 33 hospitals in China from October 2020 to September 2021 were collected for model training and testing. The data of patients with thoracic trauma at Shanghai Sixth People’s Hospital from October 2021 to September 2022 were included for further verification. The performance of the model was measured mainly by the area under the receiver operating characteristic curve (AUROC) and the mean accuracy (mAP), and the sensitivity, specificity, positive predictive value, and negative predictive value were also measured. Results A total of 3116 patients were included in the training and validation of the model. External validation was performed in 408 patients. The random forest (RF) model was selected as the final model, with an AUROC of 0·879 (95% CI 0·856–0·902) in the test dataset. In the external validation, the AUROC was 0.83 (95% CI 0.794–0.866), the specificity was 0.756 (95% CI 0.713–0.799), the sensitivity was 0.821 (95% CI 0.692–0.923), the negative predictive value was 0.976 (95% CI 0.958–0.993), and the positive likelihood ratio was 3.364. Conclusions This model can be used to quickly screen for the risk of VTE in patients with thoracic trauma. More than 90% of unnecessary VTE tests can be avoided, which can help clinicians target interventions to high-risk groups and ensure resource optimization. Although further validation and improvement are needed, this study has considerable clinical value.

Elina Kurkurina, Craig Rothenberg, Katherine Couturier et al.

Introduction: Clear communication is essential for emergency medical services (EMS) clinicians to assess a situation and make appropriate transport decisions. When barriers are present that impede communication between emergency responders and patients, EMS clinicians report difficulty navigating these encounters. As communication barriers potentially delay definitive care, it remains unclear the amount of time that EMS clinicians spend on scene during these encounters and how often they result in non-transport. In this study we sought to characterize the association between the presence of communication barriers, time spent on scene, and non-transport. Methods: We conducted an observational analysis using 2022 data from the ESO Data Collaborative, a deidentified national prehospital electronic health record dataset. Encounters were restricted to 9-1-1 responses in which the responding ambulance was first on scene, the patient was alive, ≥ 18 year of age, and able to refuse transport. The primary outcomes were time on scene and non-transport. We used logistic regression models to estimate non-transport by communication barrier (including non-English language preference, speech disability, deaf or hard of hearing, and blind or low vision) and control for key patient and encounter characteristics. Results: Of 3,477,008 EMS responses, 233,084 (6.7%) resulted in non-transport and 99,263 (2.9%) had a communication barrier identified. Among encounters with a communication barrier identified, EMS clinicians spent more time on scene with patients who were not transported (21.0 minutes) compared to patients who were transported for definitive care (15.9 minutes). Compared to those without an identified barrier, encounters with a patient who had a non-English language preference (odds ratio [OR] 0.51, confidence interval [CI] 0.49–0.53, P < .001), patients who had a speech disability (OR 0.36, CI 0.33–0.40, P < .001), were deaf or hard of hearing (OR 0.71, CI 0.66–0.76, P < .001), or were blind or had low vision (OR 0.80, CI 0.69–0.92, P < .001) were less likely to result in non-transport, with non-transport rates of 3.6%, 1.9%, 4.0%, and 4.4% respectively. Conclusion: Encounters with communication barriers were less likely to end in non-transport. When communication barriers were identified, EMS clinicians spent 32% (5.1 minutes) longer on scene on encounters that resulted in non-transport, showing that EMS clinicians may be dedicating additional time and resources caring for this population.

Marnie K. McLean, MS, Bradley J. Petek, MD, Lidija McGrath, MD et al.

The purpose of this narrative review was to summarize perinatal exercise guidelines and associations of perinatal physical activity and/or exercise with cardiovascular disease (CVD) risk. Observational studies, randomized controlled trials, systematic reviews, and meta-analyses were included. Gaps in literature and suggestions for future studies were identified. Despite concordant international guidelines, data to support nuanced activity advice for some subgroups are limited. Perinatal physical activity and exercise are consistently recommended to combat traditional CVD risk factors during the perinatal period, like excessive gestational weight gain, high blood pressure, and high blood glucose. Physical activity and exercise appear to improve nontraditional risk factors such as poor sleep and depression. Data are emerging regarding associations with some pregnancy-specific factors, such as placental characteristics. Further research investigating associations with pregnancy-specific CVD risk factors and associations in the longer term, as well as data to support uptake, adherence, and resistance exercise prescription is warranted.

Nattikarn Meelarp, Wachira Wongtanasarasin

OBJECTIVES: Various studies have delved into its incidence and risk factors, but a comprehensive meta-analysis exploring this life-threatening complication during emergent endotracheal intubation has been lacking. This study quantitatively assesses the global incidence and associated risk factors of peri-intubation cardiac arrest (PICA). METHODS: We conducted a systematic literature search on PubMed, Embase, Web of Science, and Cochrane Library from inception to October 28, 2024. Two independent authors searched, reviewed, and evaluated selected studies. Any peer-reviewed published studies reporting the incidence of PICA among adults (≥18 years) outside of the operating theater were included. Studies reporting incidence within heterogeneous populations or from overlapping groups were excluded. The primary outcome focused on determining the global incidence of PICA, while the secondary outcome addressed associated risk factors. A random-effects model was used to aggregate overall incidence rates. Subgroup analysis and meta-regression were conducted to examine PICA incidence in different locations and with the study’s sample size. The publication bias was assessed via Egger’s test and visualization of the funnel plot. The risk of bias was evaluated using the Joanna Briggs Institute Critical Appraisal Checklist. RESULTS: Fifteen articles met the inclusion criteria for the meta-analysis. PICA incidence varied from 0.5% to 23.3%. The estimated pooled incidence was 2.7% (95% confidence interval [CI]: 1.9–3.6) across PICA in the emergency department (ED) (2.5%, 95% CI: 1.4–3.7) and outside of the ED (2.9%, 95% CI: 2.2–3.6). Egger’s test yielded P = 0.009, indicating potential publication bias due to small-study effects, as suggested by the funnel plot. Meta-regression analysis revealed higher incidence in studies with smaller populations. Notably, preintubation hypotension, hypoxemia, and body mass index were found to be the most associated risk factors for PICA. Additionally, there was significant variability in PICA definitions, ranging from immediate to occurrences within 60 min after intubation. CONCLUSION: PICA occurrences during emergent endotracheal intubation reached up to 3%, showing a similar rate both within and outside the ED. While limitations such as heterogeneity and potential bias exist, these findings underscore the imperative for prospective research. Prospective studies are warranted to further delineate this critical aspect of emergent intubation.

Alemitu Bezabih, Shadi Nourriz, Anne-Marie Snider et al.

Despite a growing need for spiritual care in the US, it is often under-served, inaccessible, or misunderstood, while almost no prior work in CSCW/HCI research has engaged with professional chaplains and spiritual care providers. This interdisciplinary study aims to develop a foundational understanding of how spiritual care may (or may not) be expanded into online spaces -- especially focusing on anonymous, asynchronous, and text-based online communities. We conducted an exploratory mixed-methods study with chaplains (N=22) involving interviews and user testing sessions centered around Reddit support communities to understand participants' perspectives on technology and their ideations about the role of chaplaincy in prospective Online Spiritual Care Communities (OSCCs). Our Grounded Theory Method analysis highlighted benefits of OSCCs including: meeting patients where they are at; accessibility and scalability; and facilitating patient-initiated care. Chaplains highlighted how their presence in OSCCs could help with shaping peer interactions, moderation, synchronous chats for group care, and redirecting to external resources, while also raising important feasibility concerns, risks, and needs for future design and research. We used an existing taxonomy of chaplaincy techniques to show that some spiritual care strategies may be amenable to online spaces, yet we also exposed the limitations of technology to fully mediate spiritual care and the need to develop new online chaplaincy interventions. Based on these findings, we contribute the model of a ``Care Loop'' between institutionally-based formal care and platform-based community care to expand access and drive greater awareness and utilization of spiritual care. We also contribute design implications to guide future work in online spiritual care.

Drago Plecko, Paul Secombe, Andrea Clarke et al.

The new era of large-scale data collection and analysis presents an opportunity for diagnosing and understanding the causes of health inequities. In this study, we describe a framework for systematically analyzing health disparities using causal inference. The framework is illustrated by investigating racial and ethnic disparities in intensive care unit (ICU) outcome between majority and minority groups in Australia (Indigenous vs. Non-Indigenous) and the United States (African-American vs. White). We demonstrate that commonly used statistical measures for quantifying inequity are insufficient, and focus on attributing the observed disparity to the causal mechanisms that generate it. We find that minority patients are younger at admission, have worse chronic health, are more likely to be admitted for urgent and non-elective reasons, and have higher illness severity. At the same time, however, we find a protective direct effect of belonging to a minority group, with minority patients showing improved survival compared to their majority counterparts, with all other variables kept equal. We demonstrate that this protective effect is related to the increased probability of being admitted to ICU, with minority patients having an increased risk of ICU admission. We also find that minority patients, while showing improved survival, are more likely to be readmitted to ICU. Thus, due to worse access to primary health care, minority patients are more likely to end up in ICU for preventable conditions, causing a reduction in the mortality rates and creating an effect that appears to be protective. Since the baseline risk of ICU admission may serve as proxy for lack of access to primary care, we developed the Indigenous Intensive Care Equity (IICE) Radar, a monitoring system for tracking the over-utilization of ICU resources by the Indigenous population of Australia across geographical areas.

Federico Pirola, Fabio Stella, Marco Grzegorczyk

Dynamic Bayesian networks (DBNs) are increasingly used in healthcare due to their ability to model complex temporal relationships in patient data while maintaining interpretability, an essential feature for clinical decision-making. However, existing approaches to handling missing data in longitudinal clinical datasets are largely derived from static Bayesian networks literature, failing to properly account for the temporal nature of the data. This gap limits the ability to quantify uncertainty over time, which is particularly critical in settings such as intensive care, where understanding the temporal dynamics is fundamental for model trustworthiness and applicability across diverse patient groups. Despite the potential of DBNs, a full Bayesian framework that integrates missing data handling remains underdeveloped. In this work, we propose a novel Gibbs sampling-based method for learning DBNs from incomplete data. Our method treats each missing value as an unknown parameter following a Gaussian distribution. At each iteration, the unobserved values are sampled from their full conditional distributions, allowing for principled imputation and uncertainty estimation. We evaluate our method on both simulated datasets and real-world intensive care data from critically ill patients. Compared to standard model-agnostic techniques such as MICE, our Bayesian approach demonstrates superior reconstruction accuracy and convergence properties. These results highlight the clinical relevance of incorporating full Bayesian inference in temporal models, providing more reliable imputations and offering deeper insight into model behavior. Our approach supports safer and more informed clinical decision-making, particularly in settings where missing data are frequent and potentially impactful.

Andrew Cho, Jason M. Woo, Brian Shi et al.

Under-resourced or rural hospitals have limited access to medical specialists and healthcare professionals, which can negatively impact patient outcomes in sepsis. To address this gap, we developed the MATEC (Multi-AI Agent Team Care) framework, which integrates a team of specialized AI agents for sepsis care. The sepsis AI agent team includes five doctor agents, four health professional agents, and a risk prediction model agent, with an additional 33 doctor agents available for consultations. Ten attending physicians at a teaching hospital evaluated this framework, spending approximately 40 minutes on the web-based MATEC application and participating in the 5-point Likert scale survey (rated from 1-unfavorable to 5-favorable). The physicians found the MATEC framework very useful (Median=4, P=0.01), and very accurate (Median=4, P<0.01). This pilot study demonstrates that a Multi-AI Agent Team Care framework (MATEC) can potentially be useful in assisting medical professionals, particularly in under-resourced hospital settings.

Ram Nirula, Rebecca Empey, Hyunkyu Ko

Background Morning postoperative labs are often obtained for emergency general surgery (EGS) patients. Studies in other surgical fields indicate that routine postoperative day 1 (POD1) labs are sometimes being performed excessively and do not require intervention. The purpose of this study is to identify predictors indicating the need for POD1 labs in EGS patients based on likelihood of intervention.Methods This is a retrospective review of non-critically ill EGS patients from 2022 to 2023 who received POD1 morning labs. The odds of having an abnormal result and likelihood of intervention were measured through multivariate logistic regression accounting for patient characteristics and procedure. Least absolute shrinkage and selection operator (LASSO) regression analysis was performed to determine significant predictors of an abnormal result and intervention.Results 502 EGS patients were included. LASSO revealed that procedure duration, fever, lysis of adhesions, preoperative systolic blood pressure &lt;90 mm Hg, older age, heart failure, operative blood loss, chronic kidney disease, and anticoagulation use were independent predictors for any abnormal result (area under the receiver operation curve (AUC)=0.785). Independent predictors of intervention were procedure duration, older age, higher estimated blood loss (EBL), anticoagulant use, and lysis of adhesions (AUC=0.704). Procedures &gt;400 min carried an 84.3% chance of an abnormal lab requiring intervention. EBL &gt;200 mL carried a 75.5% chance of an abnormal lab requiring intervention.Conclusion POD1 labs for non-critically ill EGS patient rarely require intervention and can be safely omitted. Labs should be considered for longer procedures, higher EBLs, older patients, those on anticoagulation, or after lysis of adhesions.

Charles W. Stube, Alexander S. Ljungberg, Jason A. Borton et al.

Introduction: There are several options for receiving acute care besides emergency departments (ED), such as primary care physician (PCP) offices, urgent care centers (UCC), and telehealth services. It is unknown whether these alternative modes of care have decreased the number of ED visits for patients or whether they are considered before visiting the ED. A comprehensive study considering all potential methods of care is needed to address the evolving landscape of healthcare. Our goal was to identify any factors or barriers that may have influenced a patient’s choice to visit the ED as opposed to a UCC, PCP, another local ED, or use telehealth services. Methods: We surveyed ED patients between three hospital sites in the greater Buffalo, NY, area. The survey consisted of questions regarding the patients’ reasons and rationale for choosing the ED over the alternative care options. The study also involved a health record review of the patients’ diagnoses, tests/procedures, consults, and final disposition after completion of the survey. Results: Of the 590 patients consented and surveyed, 152 (25.7%) considered seeking care at a UCC, 18 (3.1%) considered telehealth services, and 146 (24.7%) attempted to contact their PCP. On the recommendation of their PCP, patients presented to the ED 110 (20.7%) times and on the recommendation of the clinician at the UCC 54 (9.2%) times. Patients’ perceived seriousness of their condition was the most common reason for their selected mode of transport to the ED and reason for choosing the ED as opposed to alternative care sites (PCP, UCC, telehealth). Based on criteria for an avoidable ED visit, 83 (14.1%) ED patients met these criteria. Conclusion: Individuals prioritize the perceived severity of their condition when deciding where to seek emergency care. While some considered alternatives (PCP, UCC, telehealth services), uncertainties about their condition and recommendations from other clinicians led many to opt for ED care. Our findings suggest a potential gap in understanding the severity of symptoms and determining the most suitable place to seek medical care for these particular conditions.

Yerkin Abdildin, Karina Tapinova, Assel Nemerenova et al.

Background This meta-analysis aims to evaluate the effects of ketamine in critically ill intensive care unit (ICU) patients. Methods The search for randomized controlled trials (RCTs) in PubMed, Scopus, and the Cochrane Library was performed initially in January but was repeated in December 2023. Included studies compared ketamine with other traditional agents used in the ICU. We synthesized evidence using RevMan v5.4, presenting the results as forest plots, and used trial sequential analysis (TSA) software v. 0.9.5.10 Beta, presenting results as TSA plots. Our outcomes were mortality, pain, opioid and midazolam requirements, delirium rates, and ICU length of stay. Results Twelve RCTs involving 805 ICU patients (ketamine group, 398; control group, 407) were included in the meta-analysis. The ketamine group was not superior to the control group in terms of mortality, pain, mean and cumulative opioid consumption, midazolam consumption, and ICU length of stay. However, the model favored the ketamine group over the control group in delirium rate. This result is significant in terms of conventional boundaries (alpha=5%) but is not robust in TSA. The applicability of the findings is limited by the small number of patients pooled for each outcome. Conclusions No differences were found between ketamine and control groups regarding any outcome except delirium rate, where the model favored the ketamine group over the control group. However, this result is not robust as sensitivity analysis and trial sequential analysis suggest that more RCTs should be conducted in the future.

Modan Tailleur, Vincent Lostanlen, Jean-Philippe Rivière et al.

Oxygenators, alarm devices, and footsteps are some of the most common sound sources in a hospital. Detecting them has scientific value for environmental psychology but comes with challenges of its own: namely, privacy preservation and limited labeled data. In this paper, we address these two challenges via a combination of edge computing and cloud computing. For privacy preservation, we have designed an acoustic sensor which computes third-octave spectrograms on the fly instead of recording audio waveforms. For sample-efficient machine learning, we have repurposed a pretrained audio neural network (PANN) via spectral transcoding and label space adaptation. A small-scale study in a neonatological intensive care unit (NICU) confirms that the time series of detected events align with another modality of measurement: i.e., electronic badges for parents and healthcare professionals. Hence, this paper demonstrates the feasibility of polyphonic machine listening in a hospital ward while guaranteeing privacy by design.

Elham Estiri, Hossein Mirinejad

Fluid administration, also called fluid resuscitation, is a medical treatment to restore the lost blood volume and optimize cardiac functions in critical care scenarios such as burn, hemorrhage, and septic shock. Automated fluid administration systems (AFAS), a potential means to improve the treatment, employ computational control algorithms to automatically adjust optimal fluid infusion dosages by targeting physiological variables (e.g., blood volume or blood pressure). Most of the existing AFAS control algorithms are model-based approaches, and their performance is highly dependent on the model accuracy, making them less desirable in real-world care of critically ill patients due to complexity and variability of modeling patients physiological states. This work presents a novel model-free reinforcement learning (RL) approach for the control of fluid infusion dosages in AFAS systems. The proposed RL agent learns to adjust the blood volume to a desired value by choosing the optimal infusion dosages using a Q-learning algorithm. The RL agent learns the optimal actions by interacting with the environment (without having the knowledge of system dynamics). The proposed methodology (i) overcomes the need for a precise mathematical model in AFAS systems and (ii) provides a robust performance in rejecting clinical noises and reaching desired hemodynamic states, as will be shown by simulation results.

Hyewon Jeong, Siddharth Nayak, Taylor Killian et al.

Sepsis is a life-threatening condition defined by end-organ dysfunction due to a dysregulated host response to infection. Although the Surviving Sepsis Campaign has launched and has been releasing sepsis treatment guidelines to unify and normalize the care for sepsis patients, it has been reported in numerous studies that disparities in care exist across the trajectory of patient stay in the emergency department and intensive care unit. Here, we apply a number of reinforcement learning techniques including behavioral cloning, imitation learning, and inverse reinforcement learning, to learn the optimal policy in the management of septic patient subgroups using expert demonstrations. Then we estimate the counterfactual optimal policies by applying the model to another subset of unseen medical populations and identify the difference in cure by comparing it to the real policy. Our data comes from the sepsis cohort of MIMIC-IV and the clinical data warehouses of the Mass General Brigham healthcare system. The ultimate objective of this work is to use the optimal learned policy function to estimate the counterfactual treatment policy and identify deviations across sub-populations of interest. We hope this approach would help us identify any disparities in care and also changes in cure in response to the publication of national sepsis treatment guidelines.

Subhash Nerella, Ziyuan Guan, Andrea Davidson et al.

Intensive Care Units (ICU) provide close supervision and continuous care to patients with life-threatening conditions. However, continuous patient assessment in the ICU is still limited due to time constraints and the workload on healthcare providers. Existing patient assessments in the ICU such as pain or mobility assessment are mostly sporadic and administered manually, thus introducing the potential for human errors. Developing Artificial intelligence (AI) tools that can augment human assessments in the ICU can be beneficial for providing more objective and granular monitoring capabilities. For example, capturing the variations in a patient's facial cues related to pain or agitation can help in adjusting pain-related medications or detecting agitation-inducing conditions such as delirium. Additionally, subtle changes in visual cues during or prior to adverse clinical events could potentially aid in continuous patient monitoring when combined with high-resolution physiological signals and Electronic Health Record (EHR) data. In this paper, we examined the association between visual cues and patient condition including acuity status, acute brain dysfunction, and pain. We leveraged our AU-ICU dataset with 107,064 frames collected in the ICU annotated with facial action units (AUs) labels by trained annotators. We developed a new "masked loss computation" technique that addresses the data imbalance problem by maximizing data resource utilization. We trained the model using our AU-ICU dataset in conjunction with three external datasets to detect 18 AUs. The SWIN Transformer model achieved 0.57 mean F1-score and 0.89 mean accuracy on the test set. Additionally, we performed AU inference on 634,054 frames to evaluate the association between facial AUs and clinically important patient conditions such as acuity status, acute brain dysfunction, and pain.

Miguel Contreras, Brandon Silva, Benjamin Shickel et al.

The acuity state of patients in the intensive care unit (ICU) can quickly change from stable to unstable. Early detection of deteriorating conditions can result in providing timely interventions and improved survival rates. In this study, we propose APRICOT-M (Acuity Prediction in Intensive Care Unit-Mamba), a 150k-parameter state space-based neural network to predict acuity state, transitions, and the need for life-sustaining therapies in real-time in ICU patients. The model uses data obtained in the prior four hours in the ICU and patient information obtained at admission to predict the acuity outcomes in the next four hours. We validated APRICOT-M externally on data from hospitals not used in development (75,668 patients from 147 hospitals), temporally on data from a period not used in development (12,927 patients from one hospital from 2018-2019), and prospectively on data collected in real-time (215 patients from one hospital from 2021-2023) using three large datasets: the University of Florida Health (UFH) dataset, the electronic ICU Collaborative Research Database (eICU), and the Medical Information Mart for Intensive Care (MIMIC)-IV. The area under the receiver operating characteristic curve (AUROC) of APRICOT-M for mortality (external 0.94-0.95, temporal 0.97-0.98, prospective 0.96-1.00) and acuity (external 0.95-0.95, temporal 0.97-0.97, prospective 0.96-0.96) shows comparable results to state-of-the-art models. Furthermore, APRICOT-M can predict transitions to instability (external 0.81-0.82, temporal 0.77-0.78, prospective 0.68-0.75) and need for life-sustaining therapies, including mechanical ventilation (external 0.82-0.83, temporal 0.87-0.88, prospective 0.67-0.76), and vasopressors (external 0.81-0.82, temporal 0.73-0.75, prospective 0.66-0.74). This tool allows for real-time acuity monitoring in critically ill patients and can help clinicians make timely interventions.