Validation of a modified Early Warning Score in medical admissions

Artificial intelligence and machine learning: Definition of terms and current concepts in critical care research

With increasing computing power, artificial intelligence (AI) and machine learning (ML) have prospered, which facilitate the analysis of large datasets, especially those found in critical care. It is important to define these terminologies, to inform a standardized approach to critical care research. This manuscript hopes to clarify these terms with examples from medical literature. Three major components that are required for a successful ML implementation: (i) reliable dataset, (ii) ML algorithm, and (iii) unbiased model evaluation, are discussed. A reliable dataset can be structured or unstructured with limited noise, outliers, and missing values. ML, a subset of AI, is typically focused on supervised or unsupervised learning tasks in which the output is based on inputs and derived from iterative pattern recognition algorithms, while AI is the overall ability of a machine to "think" or mimic human behavior; and to analyze data free from human influence. Even with successful implementation, advanced AI and ML algorithms have faced challenges in adoption into practice, mainly due to their lack of interpretability, which hinders trust, buy-in, and engagement from clinicians. Consequently, traditional algorithms, such as linear and logistic regression, that may have reduced predictive power but are highly interpretable, continue to be widely used.

Early Warning Systems for Critical Illness Outside the Intensive Care Unit

Early warning systems (EWSs) are designed and deployed to create a rapid assessment and response for patients with clinical deterioration outside the intensive care unit (ICU). These models incorporate patient-level data such as vital signs and laboratory values to detect or prevent adverse clinical events, such as vital signs and laboratories to allow detection and prevention of adverse clinical events such as cardiac arrest, intensive care transfer, or sepsis. The applicability, development, clinical utility, and general perception of EWS in clinical practice vary widely. Here, we review the field as it has grown from early vital sign-based scoring systems to contemporary multidimensional algorithms and predictive technologies for clinical decompensation outside the ICU.

Wearable wireless continuous vital signs monitoring on the general ward

PURPOSE OF REVIEW

Wearable wireless sensors for continuous vital signs monitoring (CVSM) offer the potential for early identification of patient deterioration, especially in low-intensity care settings like general wards. This study aims to review advances in wearable CVSM - with a focus on the general ward - highlighting the technological characteristics of CVSM systems, user perspectives and impact on patient outcomes by exploring recent evidence.

RECENT FINDINGS

The accuracy of wearable sensors measuring vital signs exhibits variability, especially notable in ambulatory patients within hospital settings, and standard validation protocols are lacking. Usability of CMVS systems is critical for nurses and patients, highlighting the need for easy-to-use wearable sensors, and expansion of the number of measured vital signs. Current software systems lack integration with hospital IT infrastructures and workflow automation. Imperative enhancements involve nurse-friendly, less intrusive alarm strategies, and advanced decision support systems. Despite observed reductions in ICU admissions and Rapid Response Team calls, the impact on patient outcomes lacks robust statistical significance.

SUMMARY

Widespread implementation of CVSM systems on the general ward and potentially outside the hospital seems inevitable. Despite the theoretical benefits of CVSM systems in improving clinical outcomes, and supporting nursing care by optimizing clinical workflow efficiency, the demonstrated effects in clinical practice are mixed. This review highlights the existing challenges related to data quality, usability, implementation, integration, interpretation, and user perspectives, as well as the need for robust evidence to support their impact on patient outcomes, workflow and cost-effectiveness.

Optimal timing for the Modified Early Warning Score for prediction of short-term critical illness in the acute care chain: a prospective observational study

INTRODUCTION

The Modified Early Warning Score (MEWS) is an effective tool to identify patients in the acute care chain who are likely to deteriorate. Although it is increasingly being implemented in the ED, the optimal moment to use the MEWS is unknown. This study aimed to determine at what moment in the acute care chain MEWS has the highest accuracy in predicting critical illness.

METHODS

Adult patients brought by ambulance to the ED at both locations of the Amsterdam UMC, a level 1 trauma centre, were prospectively included between 11 March and 28 October 2021. MEWS was calculated using vital parameters measured prehospital, at ED presentation, 1 hour and 3 hours thereafter, imputing for missing temperature and/or consciousness, as these values were expected not to deviate. Critical illness was defined as requiring intensive care unit admission, myocardial infarction or death within 72 hours after ED presentation. Accuracy in predicting critical illness was assessed using the area under the receiver operating characteristics curve (AUROC).

RESULTS

Of the 790 included patients, critical illness occurred in 90 (11.4%). MEWS based on vital parameters at ED presentation had the highest performance in predicting critical illness with an AUROC of 0.73 (95% CI 0.67 to 0.79) but did not significantly differ compared with other moments. Patients with an increasing MEWS over time are significantly more likely to become critical ill compared with patients with an improving MEWS.

CONCLUSION

The performance of MEWS is moderate in predicting critical illness using vital parameters measured surrounding ED admission. However, an increase of MEWS during ED admission is correlated with the development of critical illness. Therefore, early recognition of deteriorating patients at the ED may be achieved by frequent MEWS calculation. Further studies should investigate the effect of continuous monitoring of these patients at the ED.

Improvement of the rapid response system at an acute rehabilitation hospital in New Mexico

Aim: Enhance the Rapid Response System (RRS) in a free-standing acute rehabilitation hospital (ARH) by improving announcements, crash cart standardization and role assignments. Materials & methods: Pre-intervention (PreIQ) and post-intervention questionnaires (PostIQ), conducted in English and utilizing a Likert scale, were distributed in-person to clinical staff, yielding a 100% response rate. The questionnaire underwent no prior testing. The PreIQ were disseminated in February 2021, and PostIQ in December 2022. Results: PostIQ illustrated the improvement of audibility and improved the clarity of roles. The training positively impacted the RRS in the ARH. Conclusion: This study highlights the value of continuous RRS improvement in ARHs. Interventions led to notable enhancements, emphasizing the need for sustained efforts and future research on broader implementation.

Review of a large trauma registry in Addis Ababa, Ethiopia: insights into prehospital care and provider training for trauma quality improvement

Background

Injury is a major cause of death and disability in Ethiopia. ALERT Hospital, one of only three designated trauma centers in the country, has employed a basic trauma registry since its inception in 2016; however, these data had not been used. In joint efforts with the Federal Ministry of Health, we aimed to understand patient injury characteristics and predictors of mortality, to inform priorities in resource and training investments.

Methods

Data from 12 816 consecutive patients in the first 3 years of the trauma registry were reviewed retrospectively. Modified Early Warning Score was used at triage to indicate injury severity (red=critically injured, green=minor injury). No physiologic data for calculating Injury Severity Scores or in-hospital intervention data were available. Triage groups were compared and multivariate logistic regression conducted to determine predictors of in-emergency department (ED) mortality.

Results

Most patients presented with minor injuries with 64.7% triaged as 'yellow' and 16.4% triaged as 'green', and most (75.9%) referred from another facility. Of those who were critically injured, only 31.0% arrived by ambulance. Most injuries were soft tissue (51.1%) and fractures (23.0%); when stratified by triage category, most critical ('red') patients had sustained head injuries (52.7%). Arrival by ambulance (OR 2.20, p=0.017) and head injury (OR 3.11, p<0.001) were independent predictors of death in the ED.

Conclusion

This study of injured patients presenting to an Ethiopian trauma center is one of the largest to date, highlighting the need for more accessible and streamlined prehospital trauma care. Opportunities for improvement include staff training in initial trauma management and implementation of a more comprehensive trauma registry containing physiologic, intervention, and outcomes data to support a robust quality improvement program. Efforts by the Federal Ministry of Health are ongoing to support these improvements in care.

Level of Evidence

Level 3, observational study.

Thick Liquids and Clinical Outcomes in Hospitalized Patients With Alzheimer Disease and Related Dementias and Dysphagia

Importance

Oropharyngeal dysphagia is common in hospitalized patients with Alzheimer disease and related dementias (ADRD). Although the use of thick liquids in patients with dysphagia has been shown to reduce aspiration on direct visualization, there is no clear evidence that this practice translates into improved clinical outcomes.

Objectives

To determine whether a diet of thick liquids compared with thin liquids is associated with improved outcomes in hospitalized patients with ADRD and dysphagia.

Design, Setting, and Participants

This cohort study included adults aged 65 years and older with ADRD who were admitted to the medicine service across 11 diverse hospitals in New York between January 1, 2017, and September 20, 2022, with clinical suspicion of dysphagia during hospitalization and survival for at least 24 hours after hospital arrival. Patients were grouped according to whether at least 75% of their hospital diet consisted of a thick liquid diet or a thin liquid diet. Propensity score matching was used to balance covariates across the 2 groups for the following covariates: demographics (eg, age, sex), baseline clinical characteristics (eg, Charlson Comorbidity Index), and acute presentation (eg, respiratory diagnosis, illness severity, delirium).

Main Outcomes and Measures

Hospital outcomes included mortality (primary outcome), respiratory complications (eg, pneumonia), intubation, and hospital length of stay (LOS).

Results

Of 8916 patients with ADRD and dysphagia included in the propensity score matched analysis, the mean (SD) age was 85.7 (8.0) years and 4829 were female (54.2%). A total of 4458 patients receiving a thick liquid diet were matched with 4458 patients receiving a thin liquid diet. There was no significant difference in hospital mortality between the thick liquids and thin liquids groups (hazard ratio, 0.92; 95% CI, 0.75-1.14]; P = .46). Compared with patients receiving thin liquids, patients receiving thick liquids were less likely to be intubated (odds ratio [OR], 0.66; 95% CI, 0.54-0.80), but they were more likely to have respiratory complications (OR, 1.73; 95% CI, 1.56-1.91).

Conclusions and Relevance

This cohort study emphasizes the need for prospective studies that evaluate whether thick liquids are associated with improved clinical outcomes in hospitalized patients with ADRD and dysphagia.

Prognostic Accuracy of Screening Tools for Clinical Deterioration in Adults With Suspected Sepsis in Northeastern Thailand: A Cohort Validation Study

Background

We sought to assess the performance of commonly used clinical scoring systems to predict imminent clinical deterioration in patients hospitalized with suspected infection in rural Thailand.

Methods

Patients with suspected infection were prospectively enrolled within 24 hours of admission to a referral hospital in northeastern Thailand between 2013 and 2017. In patients not requiring intensive medical interventions, multiple enrollment scores were calculated including the National Early Warning Score (NEWS), the Modified Early Warning Score, Between the Flags, and the quick Sequential Organ Failure Assessment score. Scores were tested for predictive accuracy of clinical deterioration, defined as a new requirement of mechanical ventilation, vasoactive medications, intensive care unit admission, and/or death approximately 1 day after enrollment. The association of each score with clinical deterioration was evaluated by means of logistic regression, and discrimination was assessed by generating area under the receiver operating characteristic curve.

Results

Of 4989 enrolled patients, 2680 met criteria for secondary analysis, and 100 of 2680 (4%) experienced clinical deterioration within 1 day after enrollment. NEWS had the highest discrimination for predicting clinical deterioration (area under the receiver operating characteristic curve, 0.78 [95% confidence interval, .74-.83]) compared with the Modified Early Warning Score (0.67 [.63-.73]; P < .001), quick Sequential Organ Failure Assessment (0.65 [.60-.70]; P < .001), and Between the Flags (0.69 [.64-.75]; P < .001). NEWS ≥5 yielded optimal sensitivity and specificity for clinical deterioration prediction.

Conclusions

In patients hospitalized with suspected infection in a resource-limited setting in Southeast Asia, NEWS can identify patients at risk of imminent clinical deterioration with greater accuracy than other clinical scoring systems.

Rethinking Human-AI Collaboration in Complex Medical Decision Making: A Case Study in Sepsis Diagnosis

Today's AI systems for medical decision support often succeed on benchmark datasets in research papers but fail in real-world deployment. This work focuses on the decision making of sepsis, an acute life-threatening systematic infection that requires an early diagnosis with high uncertainty from the clinician. Our aim is to explore the design requirements for AI systems that can support clinical experts in making better decisions for the early diagnosis of sepsis. The study begins with a formative study investigating why clinical experts abandon an existing AI-powered Sepsis predictive module in their electrical health record (EHR) system. We argue that a human-centered AI system needs to support human experts in the intermediate stages of a medical decision-making process (e.g., generating hypotheses or gathering data), instead of focusing only on the final decision. Therefore, we build SepsisLab based on a state-of-the-art AI algorithm and extend it to predict the future projection of sepsis development, visualize the prediction uncertainty, and propose actionable suggestions (i.e., which additional laboratory tests can be collected) to reduce such uncertainty. Through heuristic evaluation with six clinicians using our prototype system, we demonstrate that SepsisLab enables a promising human-AI collaboration paradigm for the future of AI-assisted sepsis diagnosis and other high-stakes medical decision making.

AI-enabled electrocardiography alert intervention and all-cause mortality: a pragmatic randomized clinical trial

The early identification of vulnerable patients has the potential to improve outcomes but poses a substantial challenge in clinical practice. This study evaluated the ability of an artificial intelligence (AI)-enabled electrocardiogram (ECG) to identify hospitalized patients with a high risk of mortality in a multisite randomized controlled trial involving 39 physicians and 15,965 patients. The AI-ECG alert intervention included an AI report and warning messages delivered to the physicians, flagging patients predicted to be at high risk of mortality. The trial met its primary outcome, finding that implementation of the AI-ECG alert was associated with a significant reduction in all-cause mortality within 90 days: 3.6% patients in the intervention group died within 90 days, compared to 4.3% in the control group (4.3%) (hazard ratio (HR) = 0.83, 95% confidence interval (CI) = 0.70-0.99). A prespecified analysis showed that reduction in all-cause mortality associated with the AI-ECG alert was observed primarily in patients with high-risk ECGs (HR = 0.69, 95% CI = 0.53-0.90). In analyses of secondary outcomes, patients in the intervention group with high-risk ECGs received increased levels of intensive care compared to the control group; for the high-risk ECG group of patients, implementation of the AI-ECG alert was associated with a significant reduction in the risk of cardiac death (0.2% in the intervention arm versus 2.4% in the control arm, HR = 0.07, 95% CI = 0.01-0.56). While the precise means by which implementation of the AI-ECG alert led to decreased mortality are to be fully elucidated, these results indicate that such implementation assists in the detection of high-risk patients, prompting timely clinical care and reducing mortality. ClinicalTrials.gov registration: NCT05118035 .

Risk Factors and Scoring Systems to Predict the Mortality Risk of Afebrile Adult Patients with Monomicrobial Gram-Negative Bacteremia: A 10-Year Observational Study in the Emergency Department

BACKGROUND

The mortality rate of afebrile bacteremia has been reported to be as high as 45%. This investigation focused on the risk factors and predictive performance of scoring systems for the clinical outcomes of afebrile patients with monomicrobial gram-negative bacteria (GNB) in the emergency department (ED).

METHODS

We conducted a retrospective analysis of afebrile adult ED patients with monomicrobial GNB bacteremia from January 2012 to December 2021. We dissected the demographics, clinical pictures, and laboratory investigations. We applied five scoring systems and three revised systems to predict the clinical outcomes.

RESULTS

There were 600 patients included (358 males and 242 females), with a mean age of 69.6 ± 15.4 years. The overall mortality rate was 50.17%, reaching 68.52% (74/108) in cirrhotic patients. Escherichia coli was the leading pathogen (42.83%). The non-survivors had higher scores of the original MEDS (p < 0.001), NEWS (p < 0.001), MEWS (p < 0.001), qSOFA (p < 0.001), and REMS (p = 0.030). In univariate logistic regression analyses, several risk factors had a higher odds ratio (OR) for mortality, including liver cirrhosis (OR 2.541, p < 0.001), malignancy (OR 2.259, p < 0.001), septic shock (OR 2.077, p = 0.002), and male gender (OR 0.535, p < 0.001). The MEDS demonstrated that the best predictive power with the maximum area under the curve (AUC) was measured at 0.773 at the cut-off point of 11. The AUCs of the original NEWS, MEWS, qSOFA, and REMS were 0.663, 0.584, 0.572, and 0.553, respectively. We revised the original MEDS, NEWS, and qSOFA by adding red cell distribution width, albumin, and lactate scores and found a better predictive power of the AUC of 0.797, 0.719, and 0.694 on the revised MEDS ≥11, revised qSOFA ≥ 3, and revised NEWS ≥ 6, respectively.

CONCLUSIONS

The original MEDS, revised MEDS, revised qSOFA, and revised NEWS were valuable tools for predicting the mortality risk in afebrile patients with monomicrobial GNB bacteremia. We suggested that clinicians should explore patients with the risk factors mentioned above for possible severe infection, even in the absence of fever and initiate hemodynamic support and early adequate antibiotic therapy in patients with higher scores of the original MEDS (≥11), revised MEDS (≥11), revised NEWS (≥6), and revised qSOFA (≥3).

Risk prediction model based on machine learning for predicting miscarriage among pregnant patients with immune abnormalities

Introduction: It is known that patients with immune-abnormal co-pregnancies are at a higher risk of adverse pregnancy outcomes. Traditional pregnancy risk management systems have poor prediction abilities for adverse pregnancy outcomes in such patients, with many limitations in clinical application. In this study, we will use machine learning to screen high-risk factors for miscarriage and develop a miscarriage risk prediction model for patients with immune-abnormal pregnancies. This model aims to provide an adjunctive tool for the clinical identification of patients at high risk of miscarriage and to allow for active intervention to reduce adverse pregnancy outcomes. Methods: Patients with immune-abnormal pregnancies attending Sichuan Provincial People's Hospital were collected through electronic medical records (EMR). The data were divided into a training set and a test set in an 8:2 ratio. Comparisons were made to evaluate the performance of traditional pregnancy risk assessment tools for clinical applications. This analysis involved assessing the cost-benefit of clinical treatment, evaluating the model's performance, and determining its economic value. Data sampling methods, feature screening, and machine learning algorithms were utilized to develop predictive models. These models were internally validated using 10-fold cross-validation for the training set and externally validated using bootstrapping for the test set. Model performance was assessed by the area under the characteristic curve (AUC). Based on the best parameters, a predictive model for miscarriage risk was developed, and the SHapley additive expansion (SHAP) method was used to assess the best model feature contribution. Results: A total of 565 patients were included in this study on machine learning-based models for predicting the risk of miscarriage in patients with immune-abnormal pregnancies. Twenty-eight risk warning models were developed, and the predictive model constructed using XGBoost demonstrated the best performance with an AUC of 0.9209. The SHAP analysis of the best model highlighted the total number of medications, as well as the use of aspirin and low molecular weight heparin, as significant influencing factors. The implementation of the pregnancy risk scoring rules resulted in accuracy, precision, and F1 scores of 0.3009, 0.1663, and 0.2852, respectively. The economic evaluation showed a saving of ¥7,485,865.7 due to the model. Conclusion: The predictive model developed in this study performed well in estimating the risk of miscarriage in patients with immune-abnormal pregnancies. The findings of the model interpretation identified the total number of medications and the use of other medications during pregnancy as key factors in the early warning model for miscarriage risk. This provides an important basis for early risk assessment and intervention in immune-abnormal pregnancies. The predictive model developed in this study demonstrated better risk prediction performance than the Pregnancy Risk Management System (PRMS) and also demonstrated economic value. Therefore, miscarriage risk prediction in patients with immune-abnormal pregnancies may be the most cost-effective management method.

Comparison of BAP65, DECAF, PEARL, and MEWS Scores in Predicting Respiratory Support Need in Hospitalized Exacerbation of Chronic Obstructive Lung Disease Patients

OBJECTIVE

Prognostic models aid clinical practice with decision-making on treatment and hospitalization in exacerbation of chronic obstructive lung disease (ECOPD). Although there are many studies with prognostic models, diagnostic accuracy is variable within and between models.

SUBJECTS AND METHODS

We compared the prognostic performance of the BAP65 score, DECAF score, PEARL score, and modified early warning score (MEWS) in hospitalized patients with ECOPD, to estimate ventilatory support need.

RESULTS

This cross-sectional study consisted of 139 patients. Patients in need of noninvasive or invasive mechanical ventilation support are grouped as ventilatory support groups (n = 54). Comparison between receiver operating characteristic curves revealed that the DECAF score is significantly superior to the PEARL score (p = 0.04) in discriminating patients in need of ventilatory support. DECAF score with a cutoff value of 1 presented the highest sensitivity and BAP65 score with a cutoff value of 2 presented the highest specificity in predicting ventilatory support need. Multivariable analysis revealed that gender played a significant role in COPD exacerbation outcome, and arterial pCO2 and RDW measurements were also predictors of ventilatory support need. Within severity indexes, only the DECAF score was independently associated with the outcome. One-point increase in DECAF score created a 1.43 times higher risk of ventilatory support need. All severity indexes showed a correlation with age, comorbidity index, and dyspnea. BAP65 and DECAF scores also showed a correlation with length of stay.

CONCLUSION

Objective and practical classifications are needed by clinicians to assess prognosis and initiate treatment accordingly. DECAF score is a strong candidate among severity indexes.

Development of an enhanced scoring system to predict ICU readmission or in-hospital death within 24 hours using routine patient data from two NHS Foundation Trusts

RATIONALE

Intensive care units (ICUs) admit the most severely ill patients. Once these patients are discharged from the ICU to a step-down ward, they continue to have their vital signs monitored by nursing staff, with Early Warning Score (EWS) systems being used to identify those at risk of deterioration.

OBJECTIVES

We report the development and validation of an enhanced continuous scoring system for predicting adverse events, which combines vital signs measured routinely on acute care wards (as used by most EWS systems) with a risk score of a future adverse event calculated on discharge from the ICU.

DESIGN

A modified Delphi process identified candidate variables commonly available in electronic records as the basis for a 'static' score of the patient's condition immediately after discharge from the ICU. L1-regularised logistic regression was used to estimate the in-hospital risk of future adverse event. We then constructed a model of physiological normality using vital sign data from the day of hospital discharge. This is combined with the static score and used continuously to quantify and update the patient's risk of deterioration throughout their hospital stay.

SETTING

Data from two National Health Service Foundation Trusts (UK) were used to develop and (externally) validate the model.

PARTICIPANTS

A total of 12 394 vital sign measurements were acquired from 273 patients after ICU discharge for the development set, and 4831 from 136 patients in the validation cohort.

RESULTS

Outcome validation of our model yielded an area under the receiver operating characteristic curve of 0.724 for predicting ICU readmission or in-hospital death within 24 hours. It showed an improved performance with respect to other competitive risk scoring systems, including the National EWS (0.653).

CONCLUSIONS

We showed that a scoring system incorporating data from a patient's stay in the ICU has better performance than commonly used EWS systems based on vital signs alone.

TRIAL REGISTRATION NUMBER

ISRCTN32008295.

Identifying acute illness phenotypes via deep temporal interpolation and clustering network on physiologic signatures

Using clustering analysis for early vital signs, unique patient phenotypes with distinct pathophysiological signatures and clinical outcomes may be revealed and support early clinical decision-making. Phenotyping using early vital signs has proven challenging, as vital signs are typically sampled sporadically. We proposed a novel, deep temporal interpolation and clustering network to simultaneously extract latent representations from irregularly sampled vital signs and derive phenotypes. Four distinct clusters were identified. Phenotype A (18%) had the greatest prevalence of comorbid disease with increased prevalence of prolonged respiratory insufficiency, acute kidney injury, sepsis, and long-term (3-year) mortality. Phenotypes B (33%) and C (31%) had a diffuse pattern of mild organ dysfunction. Phenotype B's favorable short-term clinical outcomes were tempered by the second highest rate of long-term mortality. Phenotype C had favorable clinical outcomes. Phenotype D (17%) exhibited early and persistent hypotension, high incidence of early surgery, and substantial biomarker incidence of inflammation. Despite early and severe illness, phenotype D had the second lowest long-term mortality. After comparing the sequential organ failure assessment scores, the clustering results did not simply provide a recapitulation of previous acuity assessments. This tool may impact triage decisions and have significant implications for clinical decision-support under time constraints and uncertainty.

Clinical decision support tools useful for identifying sepsis risk

PURPOSE

Evaluate the effectiveness of the clinical decision support tools (CDSTs), POC Advisor (POCA), and Modified Early Warning System (MEWS) in identifying sepsis risk and influencing time to treatment for inpatients, comparing their respective alert mechanisms.

METHODS

This study was conducted at two academic university medical center hospitals. Data from adult inpatients in medical-surgical and telemetry units were analyzed from January 1, 2020, to December 31, 2020. Criteria included sepsis-related ICD-10 codes, antibiotic administration, and ordered sepsis labs. Subsequent statistical analyses utilized Fisher's exact test and Wilcoxon Rank Sum test, focusing on mortality differences by age, sex, and race/ethnicity.

RESULTS

Among 744 patients, 143 sepsis events were identified, with 83% already receiving treatment upon CDST alert. Group 1 (POCA alert) showed reduced response time compared with MEWS, while Group 3 (MEWS) experienced longer time to treatment. Group 4 included sepsis events missed by both systems. Mortality differences were not significant among the groups.

CONCLUSION

While CDSTs play a role, nursing assessment and clinical judgment are crucial. This study recognized the potential for alarm fatigue due to a high number of CDST-driven alerts, while emphasizing the importance of a collaborative approach for prompt sepsis treatment and potential reduction in sepsis-related mortality.

A transformer-based diffusion probabilistic model for heart rate and blood pressure forecasting in Intensive Care Unit

BACKGROUND AND OBJECTIVE

Vital sign monitoring in the Intensive Care Unit (ICU) is crucial for enabling prompt interventions for patients. This underscores the need for an accurate predictive system. Therefore, this study proposes a novel deep learning approach for forecasting Heart Rate (HR), Systolic Blood Pressure (SBP), and Diastolic Blood Pressure (DBP) in the ICU.

METHODS

We extracted 24,886 ICU stays from the MIMIC-III database which contains data from over 46 thousand patients, to train and test the model. The model proposed in this study, Transformer-based Diffusion Probabilistic Model for Sparse Time Series Forecasting (TDSTF), merges Transformer and diffusion models to forecast vital signs. The TDSTF model showed state-of-the-art performance in predicting vital signs in the ICU, outperforming other models' ability to predict distributions of vital signs and being more computationally efficient. The code is available at https://github.com/PingChang818/TDSTF.

RESULTS

The results of the study showed that TDSTF achieved a Standardized Average Continuous Ranked Probability Score (SACRPS) of 0.4438 and a Mean Squared Error (MSE) of 0.4168, an improvement of 18.9% and 34.3% over the best baseline model, respectively. The inference speed of TDSTF is more than 17 times faster than the best baseline model.

CONCLUSION

TDSTF is an effective and efficient solution for forecasting vital signs in the ICU, and it shows a significant improvement compared to other models in the field.

A customised down-sampling machine learning approach for sepsis prediction

OBJECTIVE

Sepsis is a life-threatening condition in the ICU and requires treatment in time. Despite the accuracy of existing sepsis prediction models, insufficient focus on reducing alarms could worsen alarm fatigue and desensitisation in ICUs, potentially compromising patient safety. In this retrospective study, we aim to develop an accurate, robust, and readily deployable method in ICUs, only based on the vital signs and laboratory tests.

METHODS

Our method consists of a customised down-sampling process and a specific dynamic sliding window and XGBoost to offer sepsis prediction. The down-sampling process was applied to the retrospective data for training the XGBoost model. During the testing stage, the dynamic sliding window and the trained XGBoost were used to predict sepsis on the retrospective datasets, PhysioNet and FHC.

RESULTS

With the filtered data from PhysioNet, our method achieved 80.74% accuracy (77.90% sensitivity and 84.42% specificity) and 83.95% (84.82% sensitivity and 82.00% specificity) on the test set of PhysioNet-A and PhysioNet-B, respectively. The AUC score was 0.89 for both datasets. On the FHC dataset, our method achieved 92.38% accuracy (88.37% sensitivity and 95.16% specificity) and 0.98 AUC score on the test set of FHC.

CONCLUSION

Our results indicate that the down-sampling process and the dynamic sliding window with XGBoost brought robust and accurate performance to give sepsis prediction under various hospital settings. The localisation and robustness of our method can assist in sepsis diagnosis in different ICU settings.

Effectiveness of Implementing Modified Early Warning System and Rapid Response Team for General Ward Inpatients

This retrospective study assessed the effectiveness and impact of implementing a Modified Early Warning System (MEWS) and Rapid Response Team (RRT) for inpatients admitted to the general ward (GW) of a medical center. This study included all inpatients who stayed in GWs from Jan. 2017 to Feb. 2022. We divided inpatients into GWnon-MEWS and GWMEWS groups according to MEWS and RRT implementation in Aug. 2019. The primary outcome, unexpected deterioration, was defined by unplanned admission to intensive care units. We defined the detection performance and effectiveness of MEWS according to if a warning occurred within 24 h before the unplanned ICU admission. There were 129,039 inpatients included in this study, comprising 58,106 GWnon-MEWS and 71,023 GWMEWS. The numbers of inpatients who underwent an unplanned ICU admission in GWnon-MEWS and GWMEWS were 488 (.84%) and 468 (.66%), respectively, indicating that the implementation significantly reduced unexpected deterioration (p < .0001). Besides, 1,551,525 times MEWS assessments were executed for the GWMEWS. The sensitivity, specificity, positive predicted value, and negative predicted value of the MEWS were 29.9%, 98.7%, 7.09%, and 99.76%, respectively. A total of 1,568 warning signs accurately occurred within the 24 h before an unplanned ICU admission. Among them, 428 (27.3%) met the criteria for automatically calling RRT, and 1,140 signs necessitated the nursing staff to decide if they needed to call RRT. Implementing MEWS and RRT increases nursing staff's monitoring and interventions and reduces unplanned ICU admissions.

Early Physician Gestalt Versus Usual Screening Tools for the Prediction of Sepsis in Critically Ill Emergency Patients

STUDY OBJECTIVE

Compare physician gestalt to existing screening tools for identifying sepsis in the initial minutes of presentation when time-sensitive treatments must be initiated.

METHODS

This prospective observational study conducted with consecutive encounter sampling took place in the emergency department (ED) of an academic, urban, safety net hospital between September 2020 and May 2022. The study population included ED patients who were critically ill, excluding traumas, transfers, and self-evident diagnoses. Emergency physician gestalt was measured using a visual analog scale (VAS) from 0 to 100 at 15 and 60 minutes after patient arrival. The primary outcome was an explicit sepsis hospital discharge diagnosis. Clinical data were recorded for up to 3 hours to compare Systemic Inflammatory Response Syndrome (SIRS), Sequential Organ Failure Assessment (SOFA), quick SOFA (qSOFA), Modified Early Warning Score (MEWS), and a logistic regression machine learning model using Least Absolute Shrinkage and Selection Operator (LASSO) for variable selection. The screening tools were compared using receiver operating characteristic analysis and area under the curve calculation (AUC).

RESULTS

A total of 2,484 patient-physician encounters involving 59 attending physicians were analyzed. Two hundred seventy-five patients (11%) received an explicit sepsis discharge diagnosis. When limited to available data at 15 minutes, initial VAS (AUC 0.90; 95% confidence interval [CI] 0.88, 0.92) outperformed all tools including LASSO (0.84; 95% CI 0.82 to 0.87), qSOFA (0.67; 95% CI 0.64 to 0.71), SIRS (0.67; 95% 0.64 to 0.70), SOFA (0.67; 95% CI 0.63 to 0.70), and MEWS (0.66; 95% CI 0.64 to 0.69). Expanding to data available at 60 minutes did not meaningfully change results.

CONCLUSION

Among adults presenting to an ED with an undifferentiated critical illness, physician gestalt in the first 15 minutes of the encounter outperformed other screening methods in identifying sepsis.

Multicenter Development and Prospective Validation of eCARTv5: A Gradient Boosted Machine Learning Early Warning Score

Rationale

Early detection of clinical deterioration using early warning scores may improve outcomes. However, most implemented scores were developed using logistic regression, only underwent retrospective internal validation, and were not tested in important patient subgroups.

Objectives

To develop a gradient boosted machine model (eCARTv5) for identifying clinical deterioration and then validate externally, test prospectively, and evaluate across patient subgroups.

Methods

All adult patients hospitalized on the wards in seven hospitals from 2008- 2022 were used to develop eCARTv5, with demographics, vital signs, clinician documentation, and laboratory values utilized to predict intensive care unit transfer or death in the next 24 hours. The model was externally validated retrospectively in 21 hospitals from 2009-2023 and prospectively in 10 hospitals from February to May 2023. eCARTv5 was compared to the Modified Early Warning Score (MEWS) and the National Early Warning Score (NEWS) using the area under the receiver operating characteristic curve (AUROC).

Measurements and Main Results

The development cohort included 901,491 admissions, the retrospective validation cohort included 1,769,461 admissions, and the prospective validation cohort included 46,330 admissions. In retrospective validation, eCART had the highest AUROC (0.835; 95%CI 0.834, 0.835), followed by NEWS (0.766 (95%CI 0.766, 0.767)), and MEWS (0.704 (95%CI 0.703, 0.704)). eCART's performance remained high (AUROC ≥0.80) across a range of patient demographics, clinical conditions, and during prospective validation.

Conclusions

We developed eCARTv5, which accurately identifies early clinical deterioration in hospitalized ward patients. Our model performed better than the NEWS and MEWS retrospectively, prospectively, and across a range of subgroups.

RAPID-ED: A predictive model for risk assessment of patient's early in-hospital deterioration from emergency department

Introduction

The objective of this multi-center retrospective cohort study was to devise a predictive tool known as RAPID-ED. This model identifies non-traumatic adult patients at significant risk for cardiac arrest within 48 hours post-admission from the emergency department.

Methods

Data from 224,413 patients admitted through the emergency department (2016-2020) was analyzed, incorporating vital signs, lab tests, and administered therapies. A multivariable regression model was devised to anticipate early cardiac arrest. The efficacy of the RAPID-ED model was evaluated against traditional scoring systems like National Early Warning Score (NEWS) and Modified Early Warning Score (MEWS) and its predictive ability was gauged via the area under the receiver operating characteristic curve (AUC) in both hold-out validation set and external validation set.

Results

RAPID-ED outperformed traditional models in predicting cardiac arrest with an AUC of 0.819 in the hold-out validation set and 0.807 in the external validation set. In this critical care update, RAPID-ED offers an innovative approach to assessing patient risk, aiding emergency physicians in post-discharge care decisions from the emergency department. High-risk score patients (≥13) may benefit from early ICU admission for intensive monitoring.

Conclusion

As we progress with advancements in critical care, tools like RAPID-ED will prove instrumental in refining care strategies for critically ill patients, fostering an improved prognosis and potentially mitigating mortality rates.

Prognostic value of early warning scores in patients presenting to the emergency department with exacerbation of COPD

OBJECTIVE

Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is a condition that frequently presents to the emergency department (ED) and its prognosis is not very well understood. Risk tools that can be used rapidly in the ED are needed to predict the prognosis of these patients.

METHODS

This study comprised a retrospective cohort of AECOPD patients presenting to a single center between 2015 and 2022. The prognostic accuracy of several clinical early warning scoring systems, Modified Early Warning Score (MEWS), National Early Warning Score (NEWS), NEWS‑2, Systemic Inflammatory Response Syndrome (SIRS) and the quick Sepsis-related Organ Failure Assessment (qSOFA), were compared. The outcome variable was determined as one-month mortality.

RESULTS

Of the 598 patients, 63 (10.5%) had died within 1 month after presenting to the ED. Patients who died had more often congestive heart failure, altered mental status, and admission to intensive care, and they were older. Although the MEWS, NEWS, NEWS‑2, and qSOFA scores of those who died were higher than those who survived, there was no difference between the SIRS scores of these two groups. The score with the highest positive likelihood ratio for mortality estimation was qSOFA (8.5, 95% confidence interval [CI] 3.7-19.6). The negative likelihood ratios of the scores were similar, the NEWS score had a negative likelihood ratio of 0.4 (95% CI 0.2-0.8) with the highest negative predictive value of 96.0%.

CONCLUSION

In AECOPD patients, most of the early warning scores that are frequently used in the ED were found to have a moderate ability to exclude mortality and a low ability to predict mortality.

Early switch from intravenous to oral antibiotic therapy in patients with cancer who have low-risk neutropenic sepsis: the EASI-SWITCH RCT

Background

Neutropenic sepsis is a common complication of systemic anticancer treatment. There is variation in practice in timing of switch to oral antibiotics after commencement of empirical intravenous antibiotic therapy.

Objectives

To establish the clinical and cost effectiveness of early switch to oral antibiotics in patients with neutropenic sepsis at low risk of infective complications.

Design

A randomised, multicentre, open-label, allocation concealed, non-inferiority trial to establish the clinical and cost effectiveness of early oral switch in comparison to standard care.

Setting

Nineteen UK oncology centres.

Participants

Patients aged 16 years and over receiving systemic anticancer therapy with fever (≥ 38°C), or symptoms and signs of sepsis, and neutropenia (≤ 1.0 × 109/l) within 24 hours of randomisation, with a Multinational Association for Supportive Care in Cancer score of ≥ 21 and receiving intravenous piperacillin/tazobactam or meropenem for < 24 hours were eligible. Patients with acute leukaemia or stem cell transplant were excluded.

Intervention

Early switch to oral ciprofloxacin (750 mg twice daily) and co-amoxiclav (625 mg three times daily) within 12-24 hours of starting intravenous antibiotics to complete 5 days treatment in total. Control was standard care, that is, continuation of intravenous antibiotics for at least 48 hours with ongoing treatment at physician discretion.

Main outcome measures

Treatment failure, a composite measure assessed at day 14 based on the following criteria: fever persistence or recurrence within 72 hours of starting intravenous antibiotics; escalation from protocolised antibiotics; critical care support or death.

Results

The study was closed early due to under-recruitment with 129 patients recruited; hence, a definitive conclusion regarding non-inferiority cannot be made. Sixty-five patients were randomised to the early switch arm and 64 to the standard care arm with subsequent intention-to-treat and per-protocol analyses including 125 (intervention n = 61 and control n = 64) and 113 (intervention n = 53 and control n = 60) patients, respectively. In the intention-to-treat population the treatment failure rates were 14.1% in the control group and 24.6% in the intervention group, difference = 10.5% (95% confidence interval 0.11 to 0.22). In the per-protocol population the treatment failure rates were 13.3% and 17.7% in control and intervention groups, respectively; difference = 3.7% (95% confidence interval 0.04 to 0.148). Treatment failure predominantly consisted of persistence or recurrence of fever and/or physician-directed escalation from protocolised antibiotics with no critical care admissions or deaths. The median length of stay was shorter in the intervention group and adverse events reported were similar in both groups. Patients, particularly those with care-giving responsibilities, expressed a preference for early switch. However, differences in health-related quality of life and health resource use were small and not statistically significant.

Conclusions

Non-inferiority for early oral switch could not be proven due to trial under-recruitment. The findings suggest this may be an acceptable treatment strategy for some patients who can adhere to such a treatment regimen and would prefer a potentially reduced duration of hospitalisation while accepting increased risk of treatment failure resulting in re-admission. Further research should explore tools for patient stratification for low-risk de-escalation or ambulatory pathways including use of biomarkers and/or point-of-care rapid microbiological testing as an adjunct to clinical decision-making tools. This could include application to shorter-duration antimicrobial therapy in line with other antimicrobial stewardship studies.

Trial registration

This trial is registered as ISRCTN84288963.

Funding

This award was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme (NIHR award ref: 13/140/05) and is published in full in Health Technology Assessment; Vol. 28, No. 14. See the NIHR Funding and Awards website for further award information.

The value of C-reactive protein, leucocytes and vital signs in detecting major complications after oncological colorectal surgery

PURPOSE

To assess the association of postoperative C-reactive protein (CRP), leucocytes and vital signs in the first three postoperative days (PODs) with major complications after oncological colorectal resections in a tertiary referral centre for colorectal cancer in The Netherlands.

METHODS

A retrospective cohort study, including 594 consecutive patients who underwent an oncological colorectal resection at Maastricht University Medical Centre between January 2016 and December 2020. Descriptive analyses of patient characteristics were performed. Logistic regression models were used to assess associations of leucocytes, CRP and Modified Early Warning Score (MEWS) at PODs 1-3 with major complications. Receiver operating characteristic curve analyses were used to establish cut-off values for CRP.

RESULTS

A total of 364 (61.3%) patients have recovered without any postoperative complications, 134 (22.6%) patients have encountered minor complications and 96 (16.2%) developed major complications. CRP levels reached their peak on POD 2, with a mean value of 155 mg/L. This peak was significantly higher in patients with more advanced stages of disease and patients undergoing open procedures, regardless of complications. A cut-off value of 170 mg/L was established for CRP on POD 2 and 152 mg/L on POD 3. Leucocytes and MEWS also demonstrated a peak on POD 2 for patients with major complications.

CONCLUSIONS

Statistically significant associations were found for CRP, Δ CRP, Δ leucocytes and MEWS with major complications on POD 2. Patients with CRP levels ≥ 170 mg/L on POD 2 should be carefully evaluated, as this may indicate an increased risk of developing major complications.

Real-Time Machine Learning Alerts to Prevent Escalation of Care: A Nonrandomized Clustered Pragmatic Clinical Trial

OBJECTIVES

Machine learning algorithms can outperform older methods in predicting clinical deterioration, but rigorous prospective data on their real-world efficacy are limited. We hypothesized that real-time machine learning generated alerts sent directly to front-line providers would reduce escalations.

DESIGN

Single-center prospective pragmatic nonrandomized clustered clinical trial.

SETTING

Academic tertiary care medical center.

PATIENTS

Adult patients admitted to four medical-surgical units. Assignment to intervention or control arms was determined by initial unit admission.

INTERVENTIONS

Real-time alerts stratified according to predicted likelihood of deterioration sent either to the primary team or directly to the rapid response team (RRT). Clinical care and interventions were at the providers' discretion. For the control units, alerts were generated but not sent, and standard RRT activation criteria were used.

MEASUREMENTS AND MAIN RESULTS

The primary outcome was the rate of escalation per 1000 patient bed days. Secondary outcomes included the frequency of orders for fluids, medications, and diagnostic tests, and combined in-hospital and 30-day mortality. Propensity score modeling with stabilized inverse probability of treatment weight (IPTW) was used to account for differences between groups. Data from 2740 patients enrolled between July 2019 and March 2020 were analyzed (1488 intervention, 1252 control). Average age was 66.3 years and 1428 participants (52%) were female. The rate of escalation was 12.3 vs. 11.3 per 1000 patient bed days (difference, 1.0; 95% CI, -2.8 to 4.7) and IPTW adjusted incidence rate ratio 1.43 (95% CI, 1.16-1.78; p < 0.001). Patients in the intervention group were more likely to receive cardiovascular medication orders (16.1% vs. 11.3%; 4.7%; 95% CI, 2.1-7.4%) and IPTW adjusted relative risk (RR) (1.74; 95% CI, 1.39-2.18; p < 0.001). Combined in-hospital and 30-day-mortality was lower in the intervention group (7% vs. 9.3%; -2.4%; 95% CI, -4.5% to -0.2%) and IPTW adjusted RR (0.76; 95% CI, 0.58-0.99; p = 0.045).

CONCLUSIONS

Real-time machine learning alerts do not reduce the rate of escalation but may reduce mortality.

Real-time machine learning-assisted sepsis alert enhances the timeliness of antibiotic administration and diagnostic accuracy in emergency department patients with sepsis: a cluster-randomized trial

Machine learning (ML) has been applied in sepsis recognition across different healthcare settings with outstanding diagnostic accuracy. However, the advantage of ML-assisted sepsis alert in expediting clinical decisions leading to enhanced quality for emergency department (ED) patients remains unclear. A cluster-randomized trial was conducted in a tertiary-care hospital. Adult patient data were subjected to an ML model for sepsis alert. Patient visits were assigned into one of two groups. In the intervention cluster, staff received alerts on a display screen if patients met the ML threshold for sepsis diagnosis, while patients in the control cluster followed the regular alert process. The study compared triage-to-antibiotic (TTA) time, length of stay, and mortality rate between the two groups. Additionally, the diagnostic performance of the ML model was assessed. A total of 256 (intervention) and 318 (control) sepsis patients were analyzed. The proportions of patients who received antibiotics within 1 and 3 h were higher in the intervention group than in the control group (in 1 h; 68.4 vs. 60.1%, respectively; P = 0.04, in 3 h; 94.5 vs. 89.0%, respectively; P = 0.02). The median TTA times were marginally shorter in the intervention group (46 vs. 50 min). The area under the receiver operating characteristic curve (AUROC) of ML in early sepsis identification was significantly higher than qSOFA, SIRS, and MEWS. The ML-assisted sepsis alert system may help sepsis ED patients receive antibiotics more rapidly than with the conventional, human-dedicated alert process. The diagnostic performance of ML in prompt sepsis detection was superior to that of the rule-based system.Trial registration Thai Clinical Trials Registry TCTR20230120001. Registered 16 January 2023-Retrospectively registered, https://www.thaiclinicaltrials.org/show/TCTR20230120001 .

Designing a Virtual Hospital-at-Home Intervention for Patients with Infectious Diseases: A Data-Driven Approach

BACKGROUND

Virtual hospital-at-home care might be an alternative to standard hospital care for patients with infectious diseases. In this study, we explore the potential for virtual hospital-at-home care and a potential design for this population.

METHODS

This was a retrospective cohort study of internal medicine patients suspected of infectious diseases, admitted between 1 January and 31 December 2019. We collected information on delivered care during emergency department visits, the first 24 h, between 24 and 72 h, and after 72 h of admission. Care components that could be delivered at home were combined into care packages, and the potential number of eligible patients per package was described. The most feasible package was described in detail.

RESULTS

763 patients were included, mostly referred for general internal medicine (35%), and the most common diagnosis was lower respiratory tract infection (27%). The most frequently administered care components were laboratory tests, non-oral medication, and intercollegiate consultation. With a combination of telemonitoring, video consultation, non-oral medication administration, laboratory tests, oxygen therapy, and radiological diagnostics, 48% of patients were eligible for hospital-at-home care, with 35% already eligible directly after emergency department visits.

CONCLUSION

While the potential for virtual hospital-at-home care is high, it depends greatly on which care can be arranged.

Causes, Diagnostic Testing, and Treatments Related to Clinical Deterioration Events among High-Risk Ward Patients

OBJECTIVE

Timely intervention for clinically deteriorating ward patients requires that care teams accurately diagnose and treat their underlying medical conditions. However, the most common diagnoses leading to deterioration and the relevant therapies provided are poorly characterized. Therefore, we aimed to determine the diagnoses responsible for clinical deterioration, the relevant diagnostic tests ordered, and the treatments administered among high-risk ward patients using manual chart review.

DESIGN

Multicenter retrospective observational study.

SETTING

Inpatient medical-surgical wards at four health systems from 2006-2020 PATIENTS: Randomly selected patients (1,000 from each health system) with clinical deterioration, defined by reaching the 95th percentile of a validated early warning score, electronic Cardiac Arrest Risk Triage (eCART), were included.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Clinical deterioration was confirmed by a trained reviewer or marked as a false alarm if no deterioration occurred for each patient. For true deterioration events, the condition causing deterioration, relevant diagnostic tests ordered, and treatments provided were collected. Of the 4,000 included patients, 2,484 (62%) had clinical deterioration confirmed by chart review. Sepsis was the most common cause of deterioration (41%; n=1,021), followed by arrhythmia (19%; n=473), while liver failure had the highest in-hospital mortality (41%). The most common diagnostic tests ordered were complete blood counts (47% of events), followed by chest x-rays (42%), and cultures (40%), while the most common medication orders were antimicrobials (46%), followed by fluid boluses (34%), and antiarrhythmics (19%).

CONCLUSIONS

We found that sepsis was the most common cause of deterioration, while liver failure had the highest mortality. Complete blood counts and chest x-rays were the most common diagnostic tests ordered, and antimicrobials and fluid boluses were the most common medication interventions. These results provide important insights for clinical decision-making at the bedside, training of rapid response teams, and the development of institutional treatment pathways for clinical deterioration.

KEY POINTS

Question: What are the most common diagnoses, diagnostic test orders, and treatments for ward patients experiencing clinical deterioration? Findings: In manual chart review of 2,484 encounters with deterioration across four health systems, we found that sepsis was the most common cause of clinical deterioration, followed by arrythmias, while liver failure had the highest mortality. Complete blood counts and chest x-rays were the most common diagnostic test orders, while antimicrobials and fluid boluses were the most common treatments. Meaning: Our results provide new insights into clinical deterioration events, which can inform institutional treatment pathways, rapid response team training, and patient care.

Temporal-Spatial Correlation Attention Network for Clinical Data Analysis in Intensive Care Unit

Recent advancements in medical information technology have enabled electronic health records (EHRs) to store comprehensive clinical data which has ushered healthcare into the era of "big data". However, medical data are rather complicated, making problem-solving in healthcare be limited in scope and comprehensiveness. The rapid development of deep learning in recent years has opened up opportunities for leveraging big data in healthcare. In this article we introduce a temporal-spatial correlation attention network (TSCAN) to address various clinical characteristic prediction problems, including mortality prediction, length of stay prediction, physiologic decline detection, and phenotype classification. Leveraging the attention mechanism model's design, our approach efficiently identifies relevant items in clinical data and temporally correlated nodes based on specific tasks, resulting in improved prediction accuracy. Additionally, our method identifies crucial clinical indicators associated with significant outcomes, which can inform and enhance treatment options. Our experiments utilize data from the publicly accessible Medical Information Mart for Intensive Care (MIMIC-IV) database. Finally, our approach demonstrates notable performance improvements of 2.0% (metric) compared to other SOTA prediction methods. Specifically, we achieved an impressive 90.7% mortality rate prediction accuracy and 45.1% accuracy in length of stay prediction.

Measurement accuracy of a microwave doppler sensor beneath the mattress as a continuous respiratory rate monitor: a method comparison study

PURPOSE

Non-contact continuous respiratory rate monitoring is preferred for early detection of patient deterioration. However, this technique is under development; a gold standard respiratory monitor has not been established. Therefore, this prospective observational method comparison study aimed to compare the measurement accuracy of a non-contact continuous respiratory rate monitor, a microwave Doppler sensor positioned beneath the mattress, with that of other monitors.

METHODS

The respiratory rate of intensive care unit patients was simultaneously measured using a microwave Doppler sensor, capnography, thoracic impedance pneumography, and a piezoelectric sensor beneath the mattress. Bias and 95% limits of agreement between the respiratory rate measured using capnography (standard reference) and that measured using the other three methods were calculated using Bland-Altman analysis for repeated measures. Clarke error grid (CEG) analysis evaluated the sensor's ability to assist in correct clinical decision-making.

RESULTS

Eighteen participants were included, and 2,307 data points were analyzed. The bias values (95% limits of agreement) of the microwave Doppler sensor, thoracic impedance pneumography, and piezoelectric sensor were 0.2 (- 4.8 to 5.2), 1.5 (- 4.4 to 7.4), and 0.4 (- 4.0 to 4.8) breaths per minute, respectively. Clinical decisions evaluated using CEG analyses were correct 98.1% of the time for the microwave Doppler sensor, which was similar to the performance of the other devices.

CONCLUSION

The microwave Doppler sensor had a small bias but relatively low precision, similar to other devices. In CEG analyses, the risk of each monitor leading to inadequate clinical decision-making was low.

TRIAL REGISTRATION NUMBER

UMIN000038900, February 1, 2020.

Machine learning for risk stratification in the emergency department (MARS-ED) study protocol for a randomized controlled pilot trial on the implementation of a prediction model based on machine learning technology predicting 31-day mortality in the emergency department

BACKGROUND

Many prediction models have been developed to help identify emergency department (ED) patients at high risk of poor outcome. However, these models often underperform in clinical practice and their actual clinical impact has hardly ever been evaluated. We aim to perform a clinical trial to investigate the clinical impact of a prediction model based on machine learning (ML) technology.

METHODS

The study is a prospective, randomized, open-label, non-inferiority pilot clinical trial. We will investigate the clinical impact of a prediction model based on ML technology, the RISKINDEX, which has been developed to predict the risk of 31-day mortality based on the results of laboratory tests and demographic characteristics. In previous studies, the RISKINDEX was shown to outperform internal medicine specialists and to have high discriminatory performance. Adults patients (18 years or older) will be recruited in the ED. All participants will be randomly assigned to the control group or the intervention group in a 1:1 ratio. Participants in the control group will receive care as usual in which the study team asks the attending physicians questions about their clinical intuition. Participants in the intervention group will also receive care as usual, but in addition to asking the clinical impression questions, the study team presents the RISKINDEX to the attending physician in order to assess the extent to which clinical treatment is influenced by the results.

DISCUSSION

This pilot clinical trial investigates the clinical impact and implementation of an ML based prediction model in the ED. By assessing the clinical impact and prognostic accuracy of the RISKINDEX, this study aims to contribute valuable insights to optimize patient care and inform future research in the field of ML based clinical prediction models.

TRIAL REGISTRATION

ClinicalTrials.gov NCT05497830. Machine Learning for Risk Stratification in the Emergency Department (MARS-ED). Registered on August 11, 2022. URL: https://clinicaltrials.gov/study/NCT05497830 .

Systematic review and longitudinal analysis of implementing Artificial Intelligence to predict clinical deterioration in adult hospitals: what is known and what remains uncertain

OBJECTIVE

To identify factors influencing implementation of machine learning algorithms (MLAs) that predict clinical deterioration in hospitalized adult patients and relate these to a validated implementation framework.

MATERIALS AND METHODS

A systematic review of studies of implemented or trialed real-time clinical deterioration prediction MLAs was undertaken, which identified: how MLA implementation was measured; impact of MLAs on clinical processes and patient outcomes; and barriers, enablers and uncertainties within the implementation process. Review findings were then mapped to the SALIENT end-to-end implementation framework to identify the implementation stages at which these factors applied.

RESULTS

Thirty-seven articles relating to 14 groups of MLAs were identified, each trialing or implementing a bespoke algorithm. One hundred and seven distinct implementation evaluation metrics were identified. Four groups reported decreased hospital mortality, 1 significantly. We identified 24 barriers, 40 enablers, and 14 uncertainties and mapped these to the 5 stages of the SALIENT implementation framework.

DISCUSSION

Algorithm performance across implementation stages decreased between in silico and trial stages. Silent plus pilot trial inclusion was associated with decreased mortality, as was the use of logistic regression algorithms that used less than 39 variables. Mitigation of alert fatigue via alert suppression and threshold configuration was commonly employed across groups.

CONCLUSIONS

: There is evidence that real-world implementation of clinical deterioration prediction MLAs may improve clinical outcomes. Various factors identified as influencing success or failure of implementation can be mapped to different stages of implementation, thereby providing useful and practical guidance for implementers.

Three Years of Continuous Vital Signs Monitoring on the General Surgical Ward: Is It Sustainable? A Qualitative Study

Continuous monitoring of vital signs using a wireless wearable device was implemented in 2018 at a surgical care unit of an academic hospital. This study aimed at gaining insight into nurses' and patients' perspectives regarding the use and innovation of a continuous vital signs monitoring system, three years after its introduction. This qualitative study was performed in a surgical, non-intensive care unit of an academic hospital in 2021. Key-user nurses (nurses with additional training and expertise with the device) and patients were selected for semi-structured interviews, and nurses from the ward were selected for a focus group interview using a topic list. Transcripts of the audio tapes were deductively analysed using four dimensions for adoptions of information and communication technologies (ICT) devices in healthcare. The device provided feelings of safety for nurses and patients. Nurses and patients had a few issues with the device, including the size and the battery life. Nurses gained knowledge and skills in using the system for measurement and interpretations. They perceived the system as a tool to improve the recognition of clinical decline. The use of the system could be further developed regarding the technical device's characteristics, nurses' interpretation of the data and the of type of alarms, the information needs of patients, and clarification of the definition and standardization of continuous monitoring. Three years after the introduction, wireless continuous vital signs monitoring is the new standard of care according to the end-users at the general surgical ward.

Digital remote monitoring for screening and early detection of urinary tract infections

Urinary Tract Infections (UTIs) are one of the most prevalent bacterial infections in older adults and a significant contributor to unplanned hospital admissions in People Living with Dementia (PLWD), with early detection being crucial due to the predicament of reporting symptoms and limited help-seeking behaviour. The most common diagnostic tool is urine sample analysis, which can be time-consuming and is only employed where UTI clinical suspicion exists. In this method development and proof-of-concept study, participants living with dementia were monitored via low-cost devices in the home that passively measure activity, sleep, and nocturnal physiology. Using 27828 person-days of remote monitoring data (from 117 participants), we engineered features representing symptoms used for diagnosing a UTI. We then evaluate explainable machine learning techniques in passively calculating UTI risk and perform stratification on scores to support clinical translation and allow control over the balance between alert rate and sensitivity and specificity. The proposed UTI algorithm achieves a sensitivity of 65.3% (95% Confidence Interval (CI) = 64.3-66.2) and specificity of 70.9% (68.6-73.1) when predicting UTIs on unseen participants and after risk stratification, a sensitivity of 74.7% (67.9-81.5) and specificity of 87.9% (85.0-90.9). In addition, feature importance methods reveal that the largest contributions to the predictions were bathroom visit statistics, night-time respiratory rate, and the number of previous UTI events, aligning with the literature. Our machine learning method alerts clinicians of UTI risk in subjects, enabling earlier detection and enhanced screening when considering treatment.

The relationship between the strength of religious faith and spirituality in relation to post-traumatic growth among nurses caring for COVID-19 patients in eastern Poland: a cross-sectional study

Introduction

The COVID-19 pandemic had forced intensive care unit (ICU) nurses to adapt to extreme conditions in a short period of time. This resulted in them experiencing extremely stressful situations. The aim of this study was to assess the relationship between post-traumatic growth (PTG) and religiosity and spirituality (R/S) among nurses caring for COVID-19 patients in intensive care during the pandemic.

Materials and methods

120 nurses working in Lublin, eastern Poland, participated in the cross-sectional study. The questionnaire was made up of three standardised tools: The Post-Traumatic Growth Inventory, The Santa Clara Strength of Religious Faith Questionnaire, The Spiritual Attitude and Involvement List.

Results

In terms of spirituality, the study group of nurses achieved the highest score in the Connectedness with Nature subscale (4.37 ± 1.07), while the strength of religious beliefs had a positive correlation with post-traumatic growth only in the Spiritual changes subscale (r = 0.422, p < 0.001). The following dimensions of spirituality were significantly correlated with post-traumatic growth in the multi-factor model that included religiosity and spirituality: Transcendent experiences, Spiritual activities, Meaningfulness, Acceptance, and Trust. We saw that increase in the assessment of the Transcendent experiences, Meaningfulness and Trust subscales significantly mirrors increase in post-traumatic growth, while increase in the assessment of the Spiritual activities and Acceptance subscales significantly mirrors decrease in post-traumatic growth. The above variables explained up to 44% of the dependent variable.

Conclusion

Both religiosity and spirituality were significantly associated with post-traumatic growth in the group of ICU nurses, but spirituality appears to have played a larger role. Our findings support the value and significance of the development of spiritual and religious identity as a means of enhancing positive psychological changes in the face of traumatic events.

Good Performance of Revised Scoring Systems in Predicting Clinical Outcomes of Aeromonas Bacteremia in the Emergency Department: A Retrospective Observational Study

BACKGROUND

Aeromonas species, Gram-negative, non-sporulating, facultative, and anaerobic bacilli, widely distributed in aquatic environments, derive various infections, including bacteremia. Most of these infections were opportunistic and found in patients with predisposing conditions. Among the infections, bacteremia remains with notable mortality, reported from 15% to 45%. However, predicting systems for assessing the mortality risk of this disease have yet to be investigated. We aimed to validate the performance of specific predictive scoring systems to assess the clinical outcomes of Aeromonas bacteremia and applied the revised systems to predict mortality risk.

METHODS

A retrospective observational study reviewed patients with bacteremia caused by Aeromonas spp. based on at least one positive blood culture sample collected in the emergency department from January 2012 to December 2020. The outcome was in-hospital mortality. We used seven predictive scoring systems to predict the clinical outcome. According to the effectiveness in predicting mortality, we revised three of the seven predictive scoring systems by specific characteristics to refine their risk-predicting performances.

RESULTS

We enrolled 165 patients with bacteremia caused by Aeromonas spp., including 121 males (73.3%) and 44 females (26.7%), with a mean age of 66.1 ± 14.9 years and an average length of hospital stay of 12.4 ± 10.9 days. The overall mortality rate was 32.7% (54/165). The non-survivors had significantly higher scores in MEDS (6.7 ± 4.2 vs. 12.2 ± 3.3, p < 0.001), NEWS (4.0 ± 2.8 vs. 5.3 ± 3.0, p = 0.008), and qSOFA (0.3 ± 0.6 vs. 0.6 ± 0.7, p = 0.007). Regarding mortality risk prediction, the MEDS demonstrated the best predictive power with AUC of ROC measured up to 0.834, followed by NEWS (0.626) and qSOFA (0.608). We revised the MEDS, NEWS, and qSOFA by hemoglobin and lactate. We found that the revised scores had better powerful performance, including 0.859, 0.767, and 0.691 of the AUC of ROC, if the revised MEDS ≥10, revised NEWS ≥8, and revised qSOFA ≥2, respectively.

CONCLUSIONS

MEDS, NEWS, and qSOFA were good tools for predicting outcomes in patients with Aeromonas spp. bacteremia. The revised MEDS, NEWS, and qSOFA demonstrated more powerful predicting performance than the original scoring systems. We suggested that patients with higher scores in revised MEDS (≥10), revised NEWS (≥8), and revised qSOFA (≥2) received early goal-directed therapy and appropriate broad-spectrum antibiotic treatment as early as possible to reduce mortality.

Clinical assessment and short-term mortality prediction of older adults with altered mental status using RASS and 4AT tools

BACKGROUND

Altered mental status (AMS) in older adults is a common reason for admission to emergency departments (EDs) and usually results from delirium, stupor, or coma. It is important to proficiently identify underlying factors and anticipate clinical outcomes for those patients.

AIM

The primary objective of this study was to reveal and compare the clinical outcomes and etiologic factors of older patients with delirium, stupor, and coma. The secondary objective was to identify the 30-day mortality risk for those patients.

METHOD

The study was conducted as prospective and observational research. We included patients aged 65 years and older who presented with new-onset neurological and cognitive symptoms or worsening in baseline mental status. Patients who presented no change in their baseline mental status within 48 h and those who needed urgent interventions were excluded. Selected patients were assessed using RASS and 4AT tools and classified into three groups: stupor/coma, delirium, and no stupor/coma or delirium (no-SCD). Appropriate statistical tests were applied to compare these 3 groups. The 30-day mortality risks were identified by Cox survival analysis and Kaplan-Meier curve.

RESULTS

A total of 236 patients were eligible for the study. Based on their RASS and 4AT test scores: 56 (23.7%), 94 (40.6%), and 86 (36.4%) patients formed the stupor/coma, delirium and no-SCD groups, respectively. There was no statistical difference in the three groups for gender, mean age, and medical comorbidities. Neurological (34.7%), infectious (19.4%), and respiratory (19.0%) diseases were the leading factors for AMS. Post-hoc tests showed that CCI scores of the delirium (6, IQR = 3) and stupor/coma (7, IQR = 3) groups were not significantly different. The 30-day mortality rates of stupor/coma, delirium, and no-SCD groups were 42.%, 15.9%, and 12.8%, respectively (p < 0.005). The hazard ratio of the stupor/coma group was 2.79 (CI: 95%, 1.36-5.47, p = 0.005).

CONCLUSION

AMS remains a significant clinical challenge in EDs. Using the RASS and 4AT tests provides benefits and advantages for emergency medicine physicians. Neurological, infectious, and respiratory diseases can lead to life-threatening mental deterioration. Our study revealed that long-term mortality predictor CCI scores were quite similar among patients with delirium, stupor, or coma. However, the short-term mortality was significantly increased in the stupor/coma patients and they had 2.8 times higher 30-day mortality risk than others.

Timing of prophylactic antibiotics administration and suspected systemic infection after percutaneous biliary intervention

BACKGROUND/PURPOSE

Prophylactic antibiotics administration before percutaneous biliary intervention (PBI) is currently recommended, but the underlying evidence is mostly extrapolated from prophylactic antibiotics before surgery. The aim of this study was to evaluate the impact of prophylactic antibiotics administration timing on the incidence of suspected systemic infection after PBI.

METHODS

The incidence of suspected systemic infection after PBI was compared in patients who received prophylactic antibiotics at four different time intervals between antibiotics administration and skin puncture for PBI. Suspected post-intervention systemic infection was assessed according to predetermined clinical criteria.

RESULTS

There were 98 (21.6%) suspected systemic infections after 454 PBIs in 404 patients. There were significant differences among the four groups in the incidence of suspected systemic infection after the intervention (p = .020). Fever was the most common sign of suspected systemic infection. Administration of prophylactic antibiotics more than an hour before PBI was identified as an independent risk factor of suspected systemic infection after adjusting for other relevant factors (adjusted odds ratio = 10.54; 95% confidence interval, 1.40-78.86).

CONCLUSIONS

The incidence of suspected systemic infection after the PBI was significantly lower when prophylactic antibiotics were administered within an hour before the intervention.

Improved interpretable machine learning emergency department triage tool addressing class imbalance

Objective

The Score for Emergency Risk Prediction (SERP) is a novel mortality risk prediction score which leverages machine learning in supporting triage decisions. In its derivation study, SERP-2d, SERP-7d and SERP-30d demonstrated good predictive performance for 2-day, 7-day and 30-day mortality. However, the dataset used had significant class imbalance. This study aimed to determine if addressing class imbalance can improve SERP's performance, ultimately improving triage accuracy.

Methods

The Singapore General Hospital (SGH) emergency department (ED) dataset was used, which contains 1,833,908 ED records between 2008 and 2020. Records between 2008 and 2017 were randomly split into a training set (80%) and validation set (20%). The 2019 and 2020 records were used as test sets. To address class imbalance, we used random oversampling and random undersampling in the AutoScore-Imbalance framework to develop SERP+-2d, SERP+-7d, and SERP+-30d scores. The performance of SERP+, SERP, and the commonly used triage risk scores was compared.

Results

The developed SERP+ scores had five to six variables. The AUC of SERP+ scores (0.874 to 0.905) was higher than that of the corresponding SERP scores (0.859 to 0.894) on both test sets. This superior performance was statistically significant for SERP+-7d (2019: Z = -5.843, p < 0.001, 2020: Z = -4.548, p < 0.001) and SERP+-30d (2019: Z = -3.063, p = 0.002, 2020: Z = -3.256, p = 0.001). SERP+ outperformed SERP marginally on sensitivity, specificity, balanced accuracy, and positive predictive value measures. Negative predictive value was the same for SERP+ and SERP. Additionally, SERP+ showed better performance compared to the commonly used triage risk scores.

Conclusions

Accounting for class imbalance during training improved score performance for SERP+. Better stratification of even a small number of patients can be meaningful in the context of the ED triage. Our findings reiterate the potential of machine learning-based scores like SERP+ in supporting accurate, data-driven triage decisions at the ED.

Estimated plasma volume status in COVID-19 patients and its relation to comorbidities and clinical outcomes

Blood plasma is a large reservoir of circulating mediators of inflammation and its expansion has been associated with unfavorable outcomes in patients with inflammatory and cardiovascular diseases. The aim of this study was to determine clinical and prognostic value of estimated plasma volume status (ePVS) in hospitalized patients with COVID-19. We retrospectively investigated 5871 consecutive COVID-19 patient hospitalized in our tertiary-level institution in period 3/2020-6/2021. ePVS was determined using the Strauss-derived Duarte formula and was correlated with clinical characteristics and unwanted outcomes. Median ePVS was 4.77 dl/g with interquartile range 4.11-5.74. Higher ePVS was significantly associated with older age, female sex, higher comorbidity burden, worse functional status, less severe COVID-19 clinical presentation with lower severity and longer duration of symptoms, but more pronounced inflammatory profile with higher C-reactive protein, interleukin-6 and D-dimer levels (P < 0.05 for all analyses). In the multivariate regression analysis U shaped relationship of ePVS with mortality was revealed, present independently of age, sex, COVID-19 severity and comorbidity burden. In addition, higher ePVS was independently associated with higher tendency for mechanical ventilation, intensive care unit treatment, venous thromboembolism, major bleeding and bacteriemia and lower ePVS was independently associated with tendency for arterial thrombotic events. Higher ePVS, indicative of plasma volume expansion and inflammatory cytokine accumulation, may predispose respiratory deterioration and venous thromboembolism, despite less severe initial clinical presentation. Lower ePVS, indicative of hemoconcentration, may predispose arterial thrombotic events. Both may be associated with higher mortality in hospitalized COVID-19 patients.

Application of national early warning score in assessing postoperative illness severity in elderly patients with gastrointestinal illnesses

BACKGROUND

Population aging is a social problem that is being faced in most countries.

OBJECTIVE

To apply the National Early Warning Score (NEWS) for an early warning on the vital signs and consciousness of elderly patients who are hospitalized in the gastrointestinal surgical department and to provide a reference for early detection of changes in illness severity in elderly patients by studying the correlation between NEWS value and changes in illness severity.

METHODS

We enrolled 528 elderly patients who were hospitalized in the gastrointestinal surgical department of a tertiary grade A hospital in Guizhou Province between June 2020 and May 2021, to analyze how NEWS max value correlates with illness severity and obtain the optimal NEWS cutoff value for both potentially critically ill and critically ill elderly patients using the receiver operating characteristic (ROC) curve.

RESULTS

There were statistically significant differences in NEWS values between elderly patients with various illness severities (P< 0.05). NEWS values correlated positively with illness severity (r= 0.605, P< 0.001). Based on the ROC curve, early warning trigger values for NEWS to identify potentially critically ill, critically ill and terminally ill elderly patients were 6, 7 and 8, respectively. The area under the curve (AUC) for potentially critically ill, critically ill and terminally ill elderly patients was 0.907, 0.921 and 0.939, respectively. NEWS performed better in detecting patient illness severity than Modified Early Warning Score (MEWS) in AUC, sensitivity, specificity, and Youden's index, with statistically significant differences (P< 0.05).

CONCLUSION

An early warning on the vital signs and consciousness of hospitalized elderly patients using NEWS can facilitate advanced detection of changes in illness severity of elderly patients by medical staff and enable timely treatment, thus significantly lowering the risks of illness deterioration.

Use of Evidence-Based Vital Signs in Pediatric Early Warning Score to Predict Clinical Deterioration on Acute Care Units

The pediatric early warning score (PEWS) is a tool used to predict clinical deterioration. Referenced vital sign parameters are based on expert opinion but heart rate and respiratory rate percentiles in hospitalized children have been published. This retrospective case-control study of unplanned intensive care unit (ICU) transfers compares evidence-based vital signs (EBVS) effect on PEWS sensitivity and specificity, determines the impact of age categories on PEWS deterioration prediction, and evaluates whether EBVS PEWS is associated with need for invasive ICU supports. EBVS PEWS improved sensitivity (43%-71% vs 30%-63%) for unplanned transfers with slightly decreased specificity (88%-98% vs 93%-99%). Logistic regression analysis and odds ratios (ORs) demonstrated EBVS PEWS was associated with increased risk for ICU-specific supports (OR = 1.16, 95% confidence interval [CI] = 1.0-1.34, P = .0498). Evidence-based vital signs can improve PEWS sensitivity to identify unplanned ICU transfers and identify patients requiring ICU-specific interventions.

Application of Machine Learning Techniques to Development of Emergency Medical Rapid Triage Prediction Models in Acute Care

Given the critical and complex features of medical emergencies, it is essential to develop models that enable prompt and suitable clinical decision-making based on considerable information. Emergency nurses are responsible for categorizing and prioritizing injuries and illnesses on the frontlines of the emergency room. This study aims to create an Emergency Medical Rapid Triage and Prediction Assistance model using electronic medical records and machine learning techniques. Patient information was retrieved from the emergency department of a large regional teaching hospital in Taiwan, and five supervised learning techniques were used to construct classification models for predicting critical outcomes. Of these models, the model using logistic regression had superior prediction performance, with an F1 score of 0.861 and an area under the receiver operating characteristic curve of 0.855. The Emergency Medical Rapid Triage and Prediction Assistance model demonstrated superior performance in predicting intensive care and hospitalization outcomes compared with the Taiwan Triage and Acuity Scale and three clinical early warning tools. The proposed model has the potential to assist emergency nurses in executing challenging triage assessments and emergency teams in treating critically ill patients promptly, leading to improved clinical care and efficient utilization of medical resources.

Development and clinical empirical validation of the chronic critical illness prognosis prediction model

BACKGROUND

The evolution of critical care medicine and nursing has aided and enabled the rescue of a large number of patients from numerous life-threatening diseases. However, in many cases, patient health may not be quickly restored, and the long-term prognosis may not be optimistic.

OBJECTIVES

In this study, we aimed to develop and validate a prediction model for accurate, precise, and objective identification of the severity of chronic critical illness (CCI) in patients.

METHODS

We used a retrospective case-control and prospective cohort study with no interventions. Patients diagnosed with CCI admitted to the ICU of a large metropolitan public hospital were selected. In the case-control study, 344 patients (case: 172; control:172) were enrolled to develop the prognosis prediction model of chronic critical illness (PPCCI Model); 88 patients (case:44; control: 44) in a prospective cohort study, served as the validation cohort. The discrimination of the model was measured using the area under the curve (AUC) of the receiver operating characteristic curve (ROC).

RESULTS

Age, prolonged mechanical ventilation (PMV), sepsis or other severe infections, Glasgow Coma Scale (GCS), mean artery pressure (MAP), heart rate (HR), respiratory rate (RR), oxygenation index (OI), and active bleeding were the nine predictors included in the model. In both cohorts, the PPCCI model outperformed the Acute Physiology And Chronic Health Evaluation II (APACHE II), Modified Early Warning Score (MEWS), and Sequential Organ Failure Assessment (SOFA) in identifying deceased patients with CCI (development cohort: AUC, 0.934; 95%CI, 0.908-0.960; validation cohort: AUC, 0.965; 95% CI, 0.931-0.999).

CONCLUSION

The PPCCI model can provide ICU medical staff with a standardized measurement tool for assessing the condition of patients with CCI, enabling them to allocate ward monitoring resources rationally and communicate with family members.

Explainable Artificial Intelligence Warning Model Using an Ensemble Approach for In-Hospital Cardiac Arrest Prediction: Retrospective Cohort Study

BACKGROUND

Cardiac arrest (CA) is the leading cause of death in critically ill patients. Clinical research has shown that early identification of CA reduces mortality. Algorithms capable of predicting CA with high sensitivity have been developed using multivariate time series data. However, these algorithms suffer from a high rate of false alarms, and their results are not clinically interpretable.

OBJECTIVE

We propose an ensemble approach using multiresolution statistical features and cosine similarity-based features for the timely prediction of CA. Furthermore, this approach provides clinically interpretable results that can be adopted by clinicians.

METHODS

Patients were retrospectively analyzed using data from the Medical Information Mart for Intensive Care-IV database and the eICU Collaborative Research Database. Based on the multivariate vital signs of a 24-hour time window for adults diagnosed with heart failure, we extracted multiresolution statistical and cosine similarity-based features. These features were used to construct and develop gradient boosting decision trees. Therefore, we adopted cost-sensitive learning as a solution. Then, 10-fold cross-validation was performed to check the consistency of the model performance, and the Shapley additive explanation algorithm was used to capture the overall interpretability of the proposed model. Next, external validation using the eICU Collaborative Research Database was performed to check the generalization ability.

RESULTS

The proposed method yielded an overall area under the receiver operating characteristic curve (AUROC) of 0.86 and area under the precision-recall curve (AUPRC) of 0.58. In terms of the timely prediction of CA, the proposed model achieved an AUROC above 0.80 for predicting CA events up to 6 hours in advance. The proposed method simultaneously improved precision and sensitivity to increase the AUPRC, which reduced the number of false alarms while maintaining high sensitivity. This result indicates that the predictive performance of the proposed model is superior to the performances of the models reported in previous studies. Next, we demonstrated the effect of feature importance on the clinical interpretability of the proposed method and inferred the effect between the non-CA and CA groups. Finally, external validation was performed using the eICU Collaborative Research Database, and an AUROC of 0.74 and AUPRC of 0.44 were obtained in a general intensive care unit population.

CONCLUSIONS

The proposed framework can provide clinicians with more accurate CA prediction results and reduce false alarm rates through internal and external validation. In addition, clinically interpretable prediction results can facilitate clinician understanding. Furthermore, the similarity of vital sign changes can provide insights into temporal pattern changes in CA prediction in patients with heart failure-related diagnoses. Therefore, our system is sufficiently feasible for routine clinical use. In addition, regarding the proposed CA prediction system, a clinically mature application has been developed and verified in the future digital health field.

Head-to-head comparison of 19 prediction models for short-term outcome in medical patients in the emergency department: a retrospective study

INTRODUCTION

Prediction models for identifying emergency department (ED) patients at high risk of poor outcome are often not externally validated. We aimed to perform a head-to-head comparison of the discriminatory performance of several prediction models in a large cohort of ED patients.

METHODS

In this retrospective study, we selected prediction models that aim to predict poor outcome and we included adult medical ED patients. Primary outcome was 31-day mortality, secondary outcomes were 1-day mortality, 7-day mortality, and a composite endpoint of 31-day mortality and admission to intensive care unit (ICU).The discriminatory performance of the prediction models was assessed using an area under the receiver operating characteristic curve (AUC). Finally, the prediction models with the highest performance to predict 31-day mortality were selected to further examine calibration and appropriate clinical cut-off points.

RESULTS

We included 19 prediction models and applied these to 2185 ED patients. Thirty-one-day mortality was 10.6% (231 patients), 1-day mortality was 1.4%, 7-day mortality was 4.4%, and 331 patients (15.1%) met the composite endpoint. The RISE UP and COPE score showed similar and very good discriminatory performance for 31-day mortality (AUC 0.86), 1-day mortality (AUC 0.87), 7-day mortality (AUC 0.86) and for the composite endpoint (AUC 0.81). Both scores were well calibrated. Almost no patients with RISE UP and COPE scores below 5% had an adverse outcome, while those with scores above 20% were at high risk of adverse outcome. Some of the other prediction models (i.e. APACHE II, NEWS, WPSS, MEWS, EWS and SOFA) showed significantly higher discriminatory performance for 1-day and 7-day mortality than for 31-day mortality.

CONCLUSIONS

Head-to-head validation of 19 prediction models in medical ED patients showed that the RISE UP and COPE score outperformed other models regarding 31-day mortality.

Utility of Routine Blood Cultures for Upper Extremity Abscesses

BACKGROUND

Patients with skin and soft tissue infections (SSTIs) are often admitted by the emergency department for intravenous antibiotic therapy and surgical drainage of abscesses if necessary. As part of the initial diagnostic workup, blood cultures are routinely drawn at our institution in patients with SSTIs. This study seeks to identify the utility of performing blood cultures in patients with upper extremity abscesses as it relates to the number of incision and drainage (I&D) procedures performed, patient readmission rates, and length of hospital stay.

METHODS

A retrospective chart review of 314 patients aged 18 to 89 years who underwent 1 or more I&D procedures of upper extremity abscesses were included in the study. Patient demographic data, comorbidities, laboratory values, wound and blood culture results, number of I&D procedures performed, length of stay, and readmission rates were evaluated.

RESULTS

Increasing age and white blood count were associated with an increased number of I&Ds performed. Obtaining blood cultures, whether positive or negative, was associated with increased length of stay. There was no association between obtaining blood cultures and number of procedures performed on multivariable analysis. Positive blood cultures were associated with increased readmission rates.

CONCLUSIONS

Routinely obtaining blood cultures in patients with upper extremity abscesses may not be beneficial. Obtaining blood cultures is not associated with an increased number of I&D procedures or readmission rates. Furthermore, obtaining blood cultures, regardless of positivity, is associated with increased lengths of hospital stay.