共找到 20 条结果
The aim of this study is to determine the predictive power of the National Early Warning Score in identifying the outcomes of intensive care unit (ICU) patients. In this study, 300 patients hospitalized in the ICUs (medical and general) of two teaching hospitals affiliated with Tehran University of Medical Sciences were evaluated and followed up for 30 days between June 2023 and April 2024. The samples were selected by convenience sampling based on inclusion criteria. A checklist containing the components of early warning criteria introduced by Braden, SOFA and APACHE II, as well as outcomes of hospitalization in the ward, was used. The early warning criteria were assessed by the researcher at the time of admission, as well as 6 and 24 h after the admission of patients to ICU. The collected data were analysed by SPSS-27 and STATA-14 software using Kruskal-Wallis, Pearson and chi-square tests, as well as the area under the receiver operating characteristic (ROC) curve of early warning system. The mean age of patients was 41 years (31-56), and 63.70% of them were male. Comparison of the area under the ROC curve of early warning tool showed that the early warning score in predicting death at the time of admission, 6 h after admission and 24 h after admission was 0.697, 0.712 and 0.87, respectively. Also, this score in predicting adverse outcomes at the time of admission, 6 h after admission and 24 h after admission was 0.700, 0.784 and 0.936, respectively. In other words, the early warning system had a good predictive power in indicating the occurrence of outcomes at all three times (at admission, 6 hours after admission and 24 h after admission), but the best time to predict death and adverse outcomes was estimated to be 24 h after admission. By evaluating the predictive power of early warning score in identifying the outcomes of patients hospitalized in ICUs, we found that the early warning score at 24 h after admission had a higher predictive power than other times.
In multi-risk traffic, a warning is valuable only if it directs the driver's limited attention to the right interaction counterpart early enough to preserve feasible evasive actions. Existing proximity- or surrogate-metric-based warnings capture direct geometric conflicts but often miss behavior-mediated hazards that become critical through the driver's own evasive maneuvers. We propose a physics-informed framework that makes ego-agent interaction explicit and quantifiable. A neural network decodes time-varying inter-agent coupling and control sensitivity from scene history and map context, yielding an interpretable interaction-strength signal for each surrounding agent. A warning is triggered when a two-sample Kolmogorov-Smirnov test detects an abrupt distributional change in interaction strength, and is anchored to the agent with the strongest intensification. Critical-agent identification reaches 83.07% agreement with a physically interpretable baseline on 3131 naturalistic scenarios from the Waymo Open Motion Dataset. In a driver-in-the-loop virtual reality study (n=31) within a multi-risk scenario, the proposed interaction-aware human-machine interface (HMI) elicits faster attention shifts than an anticipated post-encroachment time (APET)-based HMI (head-turn responses within 0.2s versus a 0.8s lag) and reduces the observed collision events from 12 (no-warning) and 7 (APET-based HMI) to 1. These results indicate that quantifying interaction strength to identify critical counterparts offers a practical route to warning targeting in complex multi-risk driving.
Human Immunodeficiency Virus (HIV) infection has currently become a manageable chronic disease with the prevalence of antiretroviral therapy (ART). The increase in life expectancy and the rising number of people aged more than 50 years among people living with HIV (PLWH) both indicate an increasing risk of chronic non-communicable diseases, especially cardiovascular disease (CVD). Moreover, the contribution of traditional risk factors to CVD far outweighs that of HIV-related factors, which indicates that interventions for reducing CVD risk factors in PLWH are urgent. We designed a multicentre non-randomized controlled trial (nRCT), which incorporated six acquired immunodeficiency syndrome (AIDS)-designated hospitals in Zhejiang Province, and three rounds of surveys were conducted from January 2023 to January 2024. Assessments were performed at baseline, 3 months and 6 months post-intervention, respectively. For the intervention group, PLWH were informed of their 5/10-year CVD risk, on the basis of the assessment in baseline, verbally or online to warn them of the risk of having CVD; at the same time, personalized risky behaviour reinforcement intervention was implemented face-to-face or online for about 15 min. For the control group, PLWH were not informed of the 5/10-year CVD risk and only received routine health education in the clinic without additional intervention. Intervention effects were assessed using 5/10-year CVD risk, health-related behaviours (e.g., smoking, alcohol consumption, physical activities and dietary nutrition), as well as CVD cognitive level. To minimize baseline differences and reduce confounding due to non-treatment factors (e.g., socio-demographic characteristics), propensity score matching (PSM) was applied to construct comparable intervention and control groups, and a generalized estimating equation (GEE) was used to estimate the effect of the intervention. This trial was retrospectively registered with the Chinese Clinical Trial Registry (ChiCTR2500112178) on 11 November 2025. A total of 948 PLWH were incorporated at baseline, after excluding ineligible participants, including 494 in the intervention group and 454 in the control group. The average age was 42.08 ( ± 11.94) years, with 826 male patients (87.13%). A total of 362 pairs were successfully matched after PSM. After matching, no statistical differences remained between the two groups. After 6 months' intervention, the intervention group showed significant improvements in CVD cognitive level, with a total increase of 2.30 points compared with 0.41 points in the control group (P = 0.028). Additionally, PLWH who almost never drink and have > 8 h of sedentary time per day, increased and decreased 4.54% and 1.46% respectively, compared with the control group (P < 0.05). The 10-year CVD risk decreased by 0.63% in the intervention group and 0.18% in the control group. Moreover, the smoking rate and drinking rate decreased more in the intervention group compared with the control group, and the physical exercise compliance rate increased more as well, though there was no statistical difference compared with the control group (P > 0.05). Risk warning and behavioural intervention effectively improved CVD cognitive level. Although the risk warning and behavioural intervention could not make PLWH completely quit smoking or using alcohol, the intervention led to notable improvements in reducing the frequency of some health risk behaviours.
Cervical injury is a common occupational health concern among pilots. At present, most studies in this area concentrate on post-injury diagnosis or retrospective analysis of high-risk exposure factors. In contrast, research on candidate early warning biomarkers remains limited. To address this gap, we employed New Zealand rabbits as an experimental model to investigate cervical spine injuries induced by horizontal acceleration under conditions relevant to pilots. By integrating approaches of imaging, histopathological evaluation, and multi-omics profiling, we determined the injury thresholds for single impact and cumulative exposure to the cervical spine. Furthermore, the potential biomarkers for candidate early cervical injury warning were identified. The study demonstrated that a single impact of -5Gx could cause acute injury to the trapezius muscle, while a single impact of -9Gx could lead to acute intervertebral disc injury in New Zealand rabbits. Additionally, cumulative exposure to -6Gx over 3 weeks could induce degenerative changes in the cervical intervertebral discs. Furthermore, this research identified four serum-based and five intervertebral disc-derived biomarkers associated with early-stage injury. This study provides theoretical support for the injury thresholds for cervical damage among pilots and proposes candidate biomarkers for pre-injury early warning systems.
Accurate identification of exercise induced fatigue and real time injury early warning is a core requirement for scientific training in competitive sports. Traditional laboratory based biomechanical monitoring is hindered by spatial constraints and limited ecological validity. The integration of flexible sensing textiles and deep learning has emerged as a disruptive solution. This paper reviews recent progress in flexible sensing textiles for athlete monitoring. First, the mechanical response characteristics of Frontier sensing technologies including self-powered triboelectric nanogenerators, piezoresistive or capacitive sensors, and liquid metals are analyzed for capturing microscopic biomechanical signals. Second, deep learning architectures such as CNN, Long Short-Term Memory, and Transformers are discussed for signal denoising, action phase segmentation, and fatigue feature mining. Crucially, the early warning logic based on the fatigue compensation injury causal chain is elaborated, covering real-time high-risk movement monitoring for acute injuries, cumulative load evaluation for overuse injuries, and digital twin driven individualized benchmarking. Finally, future challenges including signal robustness, washability, and multimodal data fusion are addressed. This review establishes a biomechanically informed theoretical framework for smart sports apparel, facilitating a paradigm shift toward closed loop intelligent prediction in injury prevention.
Central nervous system (CNS) infections and clinically overlapping neuroinflammatory conditions in the neurocritical care unit (neuro-ICU) are associated with profound mortality and prolonged clinical burdens. Invasive neuromonitoring devices, particularly external ventricular drains (EVDs), significantly increase the risk of infection, while delayed clinical manifestations often hinder early recognition and intervention. This scoping review aimed to systematically map the current landscape of precision nursing device management and artificial intelligence (AI)-integrated early warning systems (EWS) for CNS infections in the neuro-ICU, rather than to definitively evaluate their clinical effectiveness. Guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) framework, a systematic search of four major databases was conducted. Eligibility criteria included studies focusing on adult patients in the neuro-ICU. A rigorous multi-round filtration process was applied. A total of 24 eligible studies, encompassing longitudinal nursing management cohorts, quality improvement protocols, and machine learning predictive models, were meticulously selected for data extraction and narrative synthesis. The systematic optimization of EVD care bundles, specifically the reduction of routine cerebrospinal fluid sampling and the implementation of closed needleless systems, significantly decreased device-related infection rates, occasionally achieving zero infections under optimal interdisciplinary rounding conditions. Furthermore, AI-driven EWS algorithms (e.g., Random Forest, XGBoost {Seattle, WA: University of Washington}) and non-invasive intracranial pressure (ICP) waveform clustering demonstrated exceptional prognostic accuracy. These advanced models successfully predicted ventriculitis up to 24 h prior to positive bacterial cultures. The management of severe complications, such as paroxysmal sympathetic hyperactivity, remains heavily reliant on continuous evidence-based nursing vigilance. The integration of AI-driven predictive modeling with standardized, precision nursing bundles represents a paradigm shift from reactive treatment to proactive prevention in neurocritical care. Translating these technologies into bedside clinical decision support systems will be pivotal in optimizing individualized patient outcomes and redefining neuro-ICU nursing standards.
To construct an early warning model for hypothermia in emergency trauma patients using Random Forest (RF) and compare its predictive performance with Logistic regression. A total of 400 trauma patients were included. Hypothermia was defined as any temperature < 36.0°C within 3 hours after admission. Patients were randomly divided into training and test sets (7:3). Based on the data of the modeling set, the least absolute shrinkage and selection operator (LASSO) regression was used to identify the characteristic variables of hypothermia in patients. Logistic regression model and RF model are constructed based on the identified characteristic variables. Among 400 patients, 92 (23.00%) developed hypothermia. The LASSO regression identified six non-zero coefficient indicators, including RTS score, ISS, SI index, ambient temperature at the time of injury, wet clothing, and shock upon entering the room. Based on these, a Logistic regression model and a RF model were constructed. The RF model achieved an AUC of 0.985 (95% CI: 0.972-0.997) in the training set and 0.956 (95% CI: 0.921-0.990) in the test set; the Logistic model achieved 0.963 (95% CI: 0.939-0.988) and 0.956 (95% CI: 0.923-0.989), respectively. Calibration curves demonstrated good agreement between predicted and observed outcomes. DCA revealed that the RF model provided higher standardized net benefit across a wider range of threshold probabilities. The RF-based early warning model demonstrates high predictive efficacy for hypothermia in emergency trauma patients and outperforms traditional Logistic regression, supporting early identification and preventive intervention.
Neuro - renal syndrome (NRS), a neuroimmunomediated entity that presents with acute nephrotic syndrome (primarily membranous glomerulonephritis), recurrent proteinuria, and even acute renal failure requiring dialysis, has been reported with varying prevalence in patients with immunomediated neuropathies and paraneoplastic syndromes. Pathophysiologically, IgG 3 and IgG 4 antibodies target the peripheral nervous system (PNS) and central nervous system (CNS), leading to predominantly distal weakness, sensory ataxia, and electrophysiological abnormalities. The patient tends to be refractory to intravenous immunoglobulin due to the pathophysiology of IgG 3 or IgG 4 antibodies but responds favorably to plasmapheresis and anti- CD 20 therapy. Additional investigations revealed a definitive diagnosis through positive antibody levels on tests such as the transfected CBA, using commercial or in- house kits. To describe the relationship between the clinical presentation of nodopathies and paranodopathies with immunomediated origins and acute nephrotic syndrome, recurrent proteinuria, and acute renal failure requiring dialysis, and to identify factors that alter prognosis and therapeutic options. Systematic review of NRS. Our postdoctoral project was approved by the local ethics committee (protocol no. 34144620.7.1001.5363.7.1001.5363. 7. 1001. 5363. 7. 1001. 5363) at FMUSP- RP. NRS has been recently described and partially explained, with neuropathies linked to nodal antigens such as neurofascin (NF) 146 and 188, NF paranodal 155, contactin 1 (CNTN- 1), and the contactin 1 (CASPR- 1) complex, representing warning signs consistent with an expanded phenotype and therapeutic refractoriness. We believe that NRS (and its variants) is a comorbidity that contributes to the pathophysiology of immunomediated conditions, with potentially catastrophic evolution if not recognized early. To describe NRS (with its variants) with early renal involvement and failure of first- line immunotherapies, due to the predominance of IgG 3 or IgG 4 antibodies. Identifying the clinical spectrum of NRS in nodopathies and paranodopathies can directly impact management, including early plasmapheresis and anti- CD 20 therapies, aiming to prevent early axonal loss, irreversible comorbidities, and increased mortality in a subacute- onset chronic neuropathy.
The psychological well-being of university students is an important public health concern and a growing implementation challenge for digital health systems. Cross-sectional psychometric screening is limited by temporal lag, selective self-disclosure, and the difficulty of distinguishing transient contextual disruption from clinically meaningful deterioration. Although digital phenotyping and predictive artificial intelligence (AI) have advanced mental health monitoring, real-world deployment in universities remains constrained by intrusive data collection, limited auditability, automation bias, and the risk that routine behavioral variation will be prematurely medicalized. In response to these implementation and governance challenges, this article proposes the Generative Semantic Intermediary Framework (GSIF), a behavior-first framework for explainable mental health early warning in higher education. GSIF is organized around three layers: ecologically feasible multimodal observation, two-stage generative semantic translation, and constrained review prioritization. Large language models (LLMs) are used not as autonomous diagnostic agents but as bounded semantic intermediaries: first translating heterogeneous institutional signals into plain-language descriptions of observable behavioral change, and then mapping these descriptions to cautious, reviewable symptom-related descriptors within established psychopathological frameworks. The framework emphasizes data minimization, role-bounded access, human-in-the-loop (HITL) verification, and explicit escalation thresholds. By making the pathway from routine data to review recommendations more transparent, GSIF offers a testable digital health architecture for earlier, more proportionate, and more governable student support workflows. Future work should evaluate its feasibility, acceptability, reviewer calibration, false-positive burden, and incremental value over existing screening and monitoring approaches.
Predators rarely forage with fixed efficiency; they cooperate, adjust their search effort to prey availability, and impose non-consumptive effects such as fear, while prey often respond through mutualistic associations with non-prey partners. Yet how these adaptive feedbacks collectively shape ecosystem resilience, critical transitions, and collapse in mutualistic communities remains unresolved within a unified theoretical framework. To address this gap, we develop a community-level mathematical model integrating prey-non-prey mutualism, cooperative hunting, prey-dependent predator search efficiency, and fear-mediated reproductive suppression, and investigate how these interacting mechanisms influence ecological stability under environmental variability. Our results reveal that cooperative hunting fundamentally reorganizes the ecosystem's tipping structure. Increasing cooperation drives catastrophic regime shifts through saddle-node bifurcations, promotes oscillatory coexistence, and generates homoclinic transitions absent from earlier mutualistic predator-prey models. Mutualistic support initially enhances resilience and species persistence but, beyond a critical threshold, induces tri-stable dynamics consistent with the paradox of enrichment. Contrary to the conventional view of fear as solely detrimental, moderate fear stabilizes coexistence, whereas its interaction with strong predator cooperation triggers abrupt transitions between alternative ecological states. Furthermore, two-parameter bifurcation analysis uncovers a remarkably rich dynamical landscape containing Bogdanov-Takens, generalized Hopf, cusp, and homoclinic structures, exposing multiple pathways to ecosystem collapse and recovery that remained hidden in previous studies lacking adaptive predator behavior. Under environmental stochasticity, noise-induced switching emerges well before deterministic tipping thresholds are reached, while increasing standard deviation and lag-1 autocorrelation provide reliable early-warning signals of impending collapse. Together, these findings identify adaptive predator behavior as a key determinant of resilience, tipping dynamics, and collapse predictability in mutualistic ecosystems facing environmental change.
暂无摘要（点击查看详情）
Cardiac arrhythmias are abnormal heart rhythms arising from disordered electrical dynamics that contribute significantly to global morbidity and mortality. Early prediction from physiological time series remains challenging due to nonlinear, nonstationary, and patient-specific cardiac dynamics. Although machine learning has advanced arrhythmia detection, most methods rely on static classification of electrocardiographic signals and lack online prediction, personalization, and mechanistic interpretability. From a dynamical systems perspective, arrhythmias represent regime transitions preceded by subtle deviations that are difficult to detect online. Here, we introduce CASCADE (Chaotic Attractor Sensitivity for Cardiac Anomaly Detection), an online, personalized anomaly forecasting framework based on Dynamical Systems Machine Learning (DynML). DynML uses ensembles of continuous-time nonlinear dynamical systems as chaotic reservoirs to reconstruct and predict short-term cardiac dynamics, training only a linear readout for efficient online adaptation without retraining. CASCADE identifies arrhythmia as failures of short-term predictability, quantified by statistically significant deviations between predicted and observed dynamics relative to patient-specific baselines. Performance is governed by reservoir complexity, quantified via topological entropy. Reservoirs near critical entropy regimes amplify subtle irregularities, with evidence of improved separability of early arrhythmic signatures at the readout level, particularly for morphologically distinct arrhythmia types. Evaluated on the MIT-BIH Arrhythmia Database and validated externally on the Icentia11k dataset, CASCADE achieves consistently high detection performance across diverse cardiac profiles and recording conditions. By integrating chaotic reservoir computing, entropy-guided design, and online personalization, CASCADE reframes arrhythmia detection as a dynamical regime transition problem, providing a scalable and interpretable framework for online beat-level cardiac monitoring.
To evaluate the predictive value of a thromboinflammatory signature integrating the Systemic Immune-Inflammation Index (SII) and routine coagulation markers for Hospital-Acquired Pneumonia (HAP) in patients with Traumatic Brain Injury (TBI). This retrospective study included two cohorts of patients with imaging-confirmed TBI: a development cohort (n=204) and an external validation cohort (n=80). Candidate predictors included demographic characteristics, Glasgow Coma Scale score, mechanical ventilation (MV), SII, and all routine coagulation markers, including prothrombin time (PT), activated partial thromboplastin time (APTT), international normalized ratio (INR), fibrinogen (FIB), and thrombin time (TT). Multivariable logistic regression was used to identify factors associated with HAP. Among patients who developed HAP, ventilator-associated pneumonia (VAP) was analyzed descriptively as an exploratory subgroup only. In the development cohort, 76 of 204 patients (37.3%) developed HAP. Patients with HAP exhibited significant coagulation abnormalities, including elevated FIB. Multivariable logistic regression with collinearity diagnostics (VIF analysis) identified SII, FIB, and MV as independent predictors. The combined model demonstrated good discrimination (AUC=0.824) and maintained moderate performance in the external validation cohort (AUC=0.675). Admission SII, FIB, and MV are independently associated with HAP in patients with TBI. A combined model based on these variables retained original discrimination and nomogram performance for HAP. VAP-related observations are presented only as exploratory subgroup findings.
Bone metastasis (BM) is common in newly diagnosed prostate cancer (PCa), particularly in patients with advanced disease at presentation. However, the indications for bone scintigraphy remain inconsistent and may lead to unnecessary imaging in low-risk patients. This study aimed to develop and validate a machine learning model for individualized prediction of BM in patients with newly diagnosed PCa. We retrospectively collected data from 327 patients with newly diagnosed PCa from two tertiary hospitals. Patients were randomly assigned to a training set (n = 229) and an internal validation set (n = 98). The Boruta algorithm was used to identify significant predictors. Seven machine learning models, including random forest and logistic regression, were developed and evaluated using five-fold cross-validation. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration, and decision curve analysis (DCA). The best-performing model was interpreted using SHapley Additive exPlanations (SHAP) and deployed as an online prediction tool. Six predictors were identified by the Boruta algorithm: clinical T stage, Gleason score, total prostate-specific antigen (tPSA), alkaline phosphatase (ALP), regional lymph node metastasis, and fibrinogen. Among the seven models, the random forest model achieved the best performance, with an area under the curve (AUC) of 0.902 in the training set and 0.906 in the internal validation set. Calibration curves showed good agreement between predicted and observed outcomes, and decision curve analysis indicated favorable clinical utility. An interactive online prediction tool was developed for individualized risk estimation. We developed and internally validated an interpretable random forest model for predicting BM in newly diagnosed PCa. This model may help identify high-risk patients and guide the use of bone scintigraphy. Prospective multicenter studies with external validation are needed to further confirm its generalizability.
Early warning and severity scores are widely used to support triage and predict clinical deterioration. Their diagnostic performance in low- and middle-income countries (LMICs) however remains uncertain due to differences in patient populations and healthcare resources compared with high-income settings. This study aimed to evaluate and compare the diagnostic accuracy of commonly used scoring systems for predicting intensive care unit (ICU) admission and in-hospital mortality among adults presenting to emergency departments in LMICs. We conducted a systematic review and meta-analysis of PubMed, Science Direct, Web of Science and the Cochrane Library. Observational studies assessing the accuracy of early warning or severity scores for ICU admission or in-hospital mortality of patient admitted to the emergency department were included. Data were pooled using a bivariate random-effects model to estimate sensitivity, specificity and area under the curve (AUC). Study quality was appraised using QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2). Twenty-six studies comprising over 45 000 patients from LMICs were analysed. For in-hospital mortality, quick Sequential Organ Failure Assessment (qSOFA) showed sensitivity of 0.61 and specificity of 0.73 (AUC=0.67); National Early Warning Score (NEWS)-family had higher sensitivity (0.80) but lower specificity (0.56, AUC=0.68); Modified Early Warning Score (MEWS) provided high specificity (0.83) but lower sensitivity (0.53, AUC=0.68). For ICU admission, NEWS demonstrated balanced performance (sensitivity 0.74, specificity 0.77, AUC=0.76), while Rapid Emergency Medicine Score (REMS) achieved the highest overall accuracy (sensitivity 0.79, specificity 0.85, AUC=0.81). Considerable heterogeneity was observed across studies. Early warning scores show moderate accuracy in predicting ICU admission and in-hospital mortality in LMIC emergency departments. qSOFA and MEWS may be most useful when high specificity is required, while NEWS and REMS provide more balanced performance, with REMS showing the best overall accuracy. No single score was optimal across settings, underscoring the need for local validation and adaptation in resource-limited environments.
Safety control for automated vehicles under occluded visibility has emerged as a key research topic. Nevertheless, prior work largely focuses on risk-aware control based on onboard sensing, with limited discussion of cloud-supported proactive safety decision-making and control (CPSC) for occluded scenarios. To address this gap, we analyze the key characteristics of the vehicle-road-cloud integrated control system (VRCICS) and develop a generic and scalable CPSC architecture across different SAE Levels of Driving Automation. Using the lead-vehicle cut-out with occluded hazard exposure (LVCO) as a representative scenario, we propose a cloud-based TTC interaction-matrix risk assessment method that leverages predicted spatiotemporal trajectory interactions to dynamically identify potential collision points and risk targets. A risk-transfer-aware proactive safety decision algorithm is then developed to enable timely risk warnings and adaptive safe-speed planning. Simulation and ablation studies verify risk-transfer identification and adaptive safety control in potential multi-vehicle rear-end crash chains, and highlight the role of vehicle-side safety functions in the hierarchical architecture. Field tests show that conventional AEB collides at 50 km/h (headway level 1) and 60 km/h (headway level 1/2), whereas CPSC avoids collisions with early warning and proactive deceleration (mean≤4m/s2). Moreover, a sensitivity analysis of roadside perception refinement indicates that the false-trigger rate decreases by 72.7% and the collision-avoidance success rate increases by 18.5%. Additionally, Joint experiments across multiple OEMs confirm the system's engineering feasibility, scalability, and robust performance across diverse vehicle platforms. This study provides a practical guidance and insights for VRCICS-based assisted driving applications.
Sepsis remains one of the leading causes of mortality in adult critical care, representing a dysregulated host response to infection. Despite advances in therapeutics and supportive care, it continues to account for over 31 million deaths annually. Its unpredictable trajectory and variable physiologic presentations make early recognition and timely intervention crucial to improving outcomes. Early warning systems were initially developed to standardize the recognition of physiologic instability and guide early recognition. The Modified Early Warning Score, National Early Warning Score 2, and Quick Sequential Organ Failure Assessment remain the most frequently used bedside tools for adult deterioration and sepsis.
Clinical deterioration in hospitalized patients is often preventable, but traditional early warning scores based on structured data are limited by delayed or inconsistent documentation. We developed and evaluated a machine learning pipeline that predicts clinical deterioration using real-time pager messages exchanged between clinicians. We conducted a retrospective study of adult non-ICU hospitalizations at Vanderbilt University Medical Center between January 2018 and June 2021. Using the content and frequency of messages, we trained long short-term memory models to predict rapid response activation, unplanned ICU transfer, or cardiac arrest within the next 6, 12, or 24 hours. Model performance was compared with the Epic's Deterioration Index and with a logistic regression ensemble combining predictions from both models. There were 1 519 445 pages associated with 111 346 hospitalizations among 74 912 patients. Deterioration events were observed in 5114 (4.6%) hospitalizations. The model achieved moderate discriminative ability, with areas under the receiver operating characteristic curve (AUROC) of 0.684 (95% CI, 0.66-0.71), 0.724 (95% CI, 0.71-0.74), and 0.669 (95% CI, 0.65-0.68) for predicting deterioration within 6, 12, and 24 h, respectively. Predictions showed moderate correlation with EDI scores (Pearson's R, 0.297-0.323) suggesting complementary signals. The ensemble model consistently outperformed either approach alone, achieving an AUROC of 0.797 (95% CI, 0.78-0.82) for 12-h prediction. Clinical pages represent an underutilized data source capturing clinicians' intuition and observations before they appear in formal documentation. Machine learning on pages can augment early warning systems by providing real-time information and intuition without increasing workload.
Lower caliceal stones (LCS) account for 25%-36% of all renal stones. Flexible ureteroscopic lithotripsy (FURL) is first-line treatment for LCS, but its stone-free rate (SFR) is significantly affected by renal anatomy and stone characteristics. Currently, there is a lack of specific, convenient scoring systems for predicting SFR of LCS after FURL. To assess the factors influencing the SFR of LCS after FURL, and establish a practical, convenient, and accurate scoring system. A single-center retrospective cohort study. A total of 149 patients who underwent FURL between September 2020 and December 2022 were divided into two groups based on the results of NCCT of the kidney performed 1 month after surgery: Group A (stone clearance group) with 112 patients and Group B (residual stone group) with 37 patients. Variables with p values less than 0.1 from univariate analysis were incorporated into binary unconditional multivariate logistic regression analysis. The degree of correlation was assessed based on the odds ratio of independent risk factors. The optimal cutoff value for the early warning scoring system was determined using the Youden index. According to binary unconditional multivariate logistic regression analysis, stone length (p = 0.003), stone CT value (p = 0.043), infundibulopelvic angle (IPA; p = 0.031), infundibular width (IW; p = 0.043), calyceal pelvic height (CPH; p = 0.006), and renal pelvis morphology (p = 0.036) were identified as independent risk factors. Each factor was assigned one point based on OR values, resulting in a total score of six points. The area under the ROC curve was 0.859, and the cutoff value was 1.5. Internal validation was performed via bootstrap resampling (1000 iterations). The optimism-corrected AUC was 0.842 (95% CI: 0.797-0.913). When the total score was less than 2, the overall SFR was 96.20%; however, when the total score was ⩾2, this rate significantly decreased to 49.32%. Stone length, stone CT value, renal pelvis shape, IPA, IW, and CPH were identified as independent factors influencing the SFR of LCS after FURL. Each factor was assigned a score of 1, resulting in a prediction scoring system with a total score of six points. A total score exceeding two was associated with a significantly reduced SFR. FURL was strongly recommended for scores of 0-1, used with caution for scores of 2-3, and not routinely recommended for scores 4-6. An innovative predictive scoring system for assessing stone-free rate of lower caliceal stones after flexible ureteroscopic lithotripsy At present, most prediction score models of stone-free rate are targeted at renal stones rather than lower calyceal stones. Additionally, the morphology of renal pelvis and renal calyces has been less extensively analyzed. In the current work, we retrospective analyzed abundant stone-related imaging and renal anatomy information to identify the factors influencing the stone-free rate of lower calyceal stones, and established a practical, convenient, and accurate scoring system. Excitingly, the AUC was as high as 0.859.
Global biodiversity and ecosystem services are threatened by invasive alien plants. Ragweed (Ambrosia artemisiifolia L.), a globally problematic weed, has rapidly spread across Xinjiang's Ili Prefecture-particularly in Xinyuan County-and presents serious obstacles to local agriculture, animal husbandry, and human health. However, in the context of climate change, the fine-scale distribution dynamics and key driving mechanisms of this species remain poorly understood. Based on field survey data from Xinyuan County, a Kuenm-optimized MaxEnt model was combined with four shared socioeconomic pathway (SSP) scenarios from CMIP6 to systematically simulate the spatiotemporal evolution of the potential habitat of ragweed at present and from the 2030s to the 2090s. (1) The optimized model exhibited exceptional predictive accuracy (AUC = 0.991; Partial ROC ratio = 1.959, p < 0.001; empirical omission rate = 0.0267), with precipitation seasonality (BIO15) and isothermality (BIO3) as the primary environmental factors constraining ragweed distribution. (2) Current high-risk zones predominantly exhibit distributions along river valley alluvial plains and road networks. Overlay analysis revealed that mountain steppe grasslands face the most severe stress, followed by low-middle mountain meadow grasslands, whereas mountain meadow steppe is least affected. (3) In the future, the total area of highly suitable ragweed habitats will decrease, with a significant decrease in the area of core suitable zones (mountain steppe). Under most future scenarios and periods, the distribution centroid is projected to exhibit predominantly westward and northwestward passive displacement (maximum displacement: 0.829 km under the SSP585 scenario in the 2050s), driven by the contraction of suitable habitat in low-elevation river valleys rather than by active colonization of high-elevation zones. High-elevation alpine grassland ecosystems (mountain meadow steppe and mountain desert steppe) consistently demonstrated strong resilience against ragweed invasion across all projected periods and scenarios. This study reveals the response pattern of ragweed, characterized by "contraction in low-elevation core areas, passive westward-northwestward displacement of the distribution centroid, and maintenance of the alpine barrier." A zoned control strategy focused on implementing physical eradication and replacement restoration in river valley core areas and establishing early warning systems in the transitional zones between mountain grasslands and meadows is recommended to safeguard regional ecological security.