Areas Under The ROC Curve Research Articles

Predictive value of serum fibroblast growth factor 19 and liver stiffness for intestinal failure associated liver disease-cholestasis

BackgroundIntestinal failure associated liver disease (IFALD)-cholestasis is a common complication of long-term parenteral nutrition (PN) in patients with intestinal failure (IF). The lack of effective early identification indicators often results in poor clinical outcomes. The objective of this study was to evaluate the predictive value of serum FGF19 and liver stiffness in IFALD-cholestasis. MethodsEligible adults diagnosed with IF were identified from Jinling Hospital in China. Diagnostic criteria for IFALD-cholestasis: total bilirubin >1 mg/dL and conjugated bilirubin >0.3 mg/dL for ≥6 months. Fasting blood specimens were prospectively collected and serum FGF19 concentrations were determined using ELISA and liver stiffness was measured by Two-dimensional shear wave elastography. Binary logistic regression analysis identified predictors of IFALD-cholestasis. Receiver operating characteristic (ROC) curves and areas under the ROC curves (AUROC) were used to evaluate the accuracy of serum FGF19 and liver stiffness in identifying IFALD-cholestasis. ResultsOf 203 study patients with IF, 70 (34.5%) were diagnosed with IFALD-cholestasis. The serum FGF19 levels in those with IFALD-cholestasis were significantly decreased compared with those in patients without, and liver stiffness was significantly increased (p<0.001). Multivariate logistic regression analyses suggested that intestinal discontinuity, dependence on PN, liver stiffness >6.5 kPa, and serum FGF19 ≤107 pg/mL were independent risk factors for IFALD-cholestasis. The AUROC for serum FGF19 and liver stiffness, which indicate the occurrence of IFALD-cholestasis, were 0.810 and 0.714, respectively. Serum FGF19 had a superior predictive performance than liver stiffness (p<0.05). ConclusionBoth low circulating serum FGF19 concentration and increased liver stiffness are excellent predictors of IFALD-cholestasis, but serum FGF19 is superior to increased liver stiffness in predicting IFALD-cholestasis.

Applying Machine Learning to Support Early Diagnosis of Light-Chain Amyloidosis: A Combination of Knowledge-Based Approach with Data-Driven Approach

Introduction Immunoglobulin light chain (AL) amyloidosis is a rare disease involving the clonal proliferation of bone-marrow-residing plasma cell and resulting in overproduction of serum immunoglobulin free light chains that affects multiple organs. There are several available effective treatments including autologous stem cell transplant, bortezomib, anti-CD38 antibodies, and immunomodulatory drugs. However, because of the atypical symptoms and signs of this disease, diagnostic delays are still the major challenge resulting in poor prognosis. In recent years, machine learning (ML) models have been used to assist in early diagnosis. Therefore, to address this clinical unmet need, we aim to build ML algorithms from clinical data and assess their performance in differentiating AL amyloidosis from similar conditions. Methods Monocenter medical records data were collected from 49 patients with AL amyloidosis and 198 non-AL amyloidosis patients on a ratio of 1:4 in Peking University People's Hospital between January 1, 2013, and December 31, 2021. The non-AL amyloidosis group were patients with diseases of similar symptoms including autoimmune liver disease, myocarditis, and hypertrophic cardiomyopathy. Variables for model development were selected from 30 demographic characteristics and clinical features from routine clinical examination based on the results of recursive feature elimination and hematologists' knowledge. We proposed a four-step approach to develop and evaluate the diagnostic models. In the first step, all patients were randomly allocated into a training set and a testing set with a ratio of 4:1. Second, we derived five separate ML models including logistic regression, support vector machine (SVM), extreme gradient boosting (XGBoost), light gradient boosting (LightGBM), and CatBoost algorithms to differentiate AL amyloidosis from other diseases with similar symptoms and validated the models using five-fold cross validation methods. Third, parameters of model with the highest areas under the receiver operating characteristic curves (AUROC) were updated in the full training set. Finally, the performances of the selected model were evaluated by AUROC, sensitivity, specificity and F1-score in the testing set. Results Twelve features including alanine aminotransferase, troponin, albumin, aspartate aminotransferase, activated partial thromboplastin time, albumin and globulin (A/G) ratio, direct bilirubin, platelet, fibrinogen, blood urea nitrogen, body weight and age were selected to construct ML models. The AUROC values for AL amyloidosis differential diagnosis were 0.55 with logistic regression, 0.63 with SVM, 0.84 with XGBoost, 0.89 with LightGBM, and 0.88 with CatBoost. The LightGBM model, which achieved the highest AUROC, also achieved the best performance with a sensitivity of 0.92, a specificity of 0.60, a F1-score of 0.73, a negative predictive value of 0.97, a positive predictive value of 0.60, and an accuracy of 0.82. Conclusion Our results show that the LightGBM model has the best performance to identify patients with AL amyloidosis from patients with similar symptoms. This novel ML-based diagnostic model has potential to assist in the earlier diagnosis of AL amyloidosis in clinical settings. Further studies are needed to confirm these findings in different study populations.

Accurate prediction of HCC risk after SVR in patients with hepatitis C cirrhosis based on longitudinal data

BackgroundMost existing predictive models of hepatocellular carcinoma (HCC) risk after sustained virologic response (SVR) are built on data collected at baseline and therefore have limited accuracy. The current study aimed to construct an accurate predictive model incorporating longitudinal data using a novel modeling strategy. The predictive performance of the longitudinal model was also compared with a baseline model.MethodsA total of 400 patients with HCV-related cirrhosis who achieved SVR with direct-acting antivirals (DAA) were enrolled in the study. Patients were randomly divided into a training set (70%) and a validation set (30%). Informative features were extracted from the longitudinal variables and then put into the random survival forest (RSF) to develop the longitudinal model. A baseline model including the same variables was built for comparison.ResultsDuring a median follow-up time of approximately 5 years, 25 patients (8.9%) in the training set and 11 patients (9.2%) in the validation set developed HCC. The areas under the receiver-operating characteristics curves (AUROC) for the longitudinal model were 0.9507 (0.8838–0.9997), 0.8767 (0.6972,0.9918), and 0.8307 (0.6941,0.9993) for 1-, 2- and 3-year risk prediction, respectively. The brier scores of the longitudinal model were also relatively low for the 1-, 2- and 3-year risk prediction (0.0283, 0.0561, and 0.0501, respectively). In contrast, the baseline model only achieved mediocre AUROCs of around 0.6 (0.6113, 0.6213, and 0.6480, respectively).ConclusionsOur longitudinal model yielded accurate predictions of HCC risk in patients with HCV-relate cirrhosis, outperforming the baseline model. Our model can provide patients with valuable prognosis information and guide the intensity of surveillance in clinical practice.

Comparisons of APACHE II, CURB-65, Pneumonia severity index, and CHA2DS2-VAS score on new onset cardiovascular events, including atrial fibrillation, for critical illness with pneumonia

Abstract Background pneumonia patients afflicted with cardiac disease were frequented with higher rates of short-term mortality. Furthermore, the most common arrhythmia (25%) in critically ill patients is atrial fibrillation and its increased incidences in hospital mortality and extended length of hospital stays. The early detection of the occurrence of atrial fibrillation and other CV events is important among these severely pneumonia patients. Purpose The purpose of this article is to determine the appropriate risk assessments and to make a comparison of predictive performance among four common practice risk scores: APACHE II, CURB-65, pneumonia severity index and CHA2DS2-VASc. Methods This was a retrospective, observational study from consecutive critical pneumonia patients admitted to medical intensive care unit. Heart failure, acute coronary syndrome, atrial fibrillation, ventricular tachycardia or stroke are defined as a cardiovascular event. Those who had the missing data about APACHE II, CURB-65, PSI and CHA2DS2-VASc scores had already been excluded from this study. The baseline characteristics of patients between the disease groups (CV vs. non-CV; AF vs. non-AF) were compared by using the independent sample t-test for continuous variables and the chi-square test for categorical variables, respectively. The performance of discriminating in-hospital CV and AF events for the four risk scores was evaluated using the ROC curve analysis. The areas under the ROC curve (AUC) of the four risk scores were compared using the Delong’s non-parametric method. Results 252 (22.6%) of the 1,117 patients suffered from CV events during the hospitalization. Compared to patients averting CV events, those who experienced CV events during hospitalization were older, more likely to smoke, experiencing more altered mental states, of poor renal function, of higher prevalence of all comorbidities except liver disease, of a long duration of ICU and yielded higher values of the four risk scores. There were 99 patients who suffered from CV events other than an AF episode. After excluding the 99 patients, 153 (15%) of the remaining 1,018 patients had an AF event during hospitalization. The values of the four risk scores were also significantly greater in the AF group. The performance of discriminating CV events for the four risk scores was fair with the values of AUC ranging from 61.8% of APACHE II to 64.8% of CURB 65 (Fig 1). However, no significant difference of the AUC values between any two risk scores. In terms of discriminating AF episode (Fig 2) for the four risk scores, the performance was obviously poorer with the values of AUC less than 60%, except for the CURB 65. Conclusion CURB 65 is simple and effective tool for predicting new onset cardiovascular events, including atrial fibrillation, in ICU patients with pneumonia.new onset CV eventonw onset AF event

Abstract 15176: Artificial Intelligence Model Detecting Acute Myocardial Infarction Requiring Revascularization

Background: Rapid myocardial revascularization in patients with acute myocardial infarction (AMI) is essential for increasing the chances and extent of myocardial salvage and better clinical outcomes. There still have been largely missed opportunities in the timely diagnosis of AMI. Research Questions: Can artificial intelligence (AI) be helpful in timely diagnosing AMI? Aims: We aimed to develop an AI algorithm for detecting AMI requiring revascularization using electrocardiograms (ECGs). Methods: We developed an AI model detecting myocardial infarction requiring revascularization from a 12-lead ECG. This retrospective cohort study included 17,907 ECGs from 5,872 coronary angiography-validated AMI patients and 705,482 ECGs from 294,755 non-AMI patients at Chonnam National University Hospital between May 2013 and December 2020. External validation was performed at Hallym University Kangnam Sacred Heart Hospital using 357,072 ECGs from 163,870 patients between January 2006 and December 2020. Results: The prevalence rates of AMI-ECGs were 5.29% (1356/25,627) and 0.87% (3089/357,072) in the internal test and external validation set, respectively. The areas under the receiver operating characteristic curves (AUROC) were 0.958 (95% CI, 0.955 to 0.962), 0.988 (95% CI, 0.986 to 0.991), and 0.930 (95% CI, 0.923 to 0.936) in total, ST-elevation myocardial infarction, and non-ST-elevation myocardial infarction in the external validation, respectively (Figure 1A). Subgroup analysis assessing the efficacy of subsets of 12-lead ECG revealed that the AUROC were 0.926 (95% CI, 0.920 to 0.931), 0.921 (95% CI, 0.915 to 0.926), and 0.846 (95% CI, 0.839 to 0.854) in 6 limb-lead, lead combining I and II, and lead I, respectively (Figure 1B). Conclusion: Our AI model detecting myocardial infarction can be beneficial in identifying patients with acute myocardial infarction requiring revascularization timely.

Association between the non-HDL-cholesterol to HDL- cholesterol ratio and abdominal aortic aneurysm from a Chinese screening program

BackgroundAbdominal aortic aneurysms (AAAs) can result in high mortality upon rupture but are usually undiagnosed because of the absence of symptoms in the early stage. Ultrasound screening is regarded as an impactful way to prevent the AAA-related death but cannot be performed efficiently; therefore, a target population, especially in Asia, for this procedure is lacking. Additionally, although dyslipidaemia and atherosclerosis are associated with AAA. However, it remains undetermined whether the non-high-density lipoprotein-cholesterol to high-density lipoprotein-cholesterol ratio (NHHR) is associated with AAA. Therefore, this study was aimed at examining whether NHHR is associated with AAA.MethodA total of 9559 participants who underwent AAA screening at Guangdong Provincial People’s Hospital and through screening in two communities in Dongguan, from June 2019 to June 2021 joined in this screening program. The diagnosis of AAA was confirmed by the ultrasound examination of the abdominal aorta rather than any known or suspected AAA. Clinical and laboratory data of participants were collected. The participants were separated into a normal group and an AAA group according to the abdominal aortic status. To eliminate confounding factors, a propensity score matching (PSM) approach was utilized. The independent relationship between NHHR and AAA was assessed through the utilization of multivariable logistic regression analysis. In addition, internal consistency was evaluated through subgroup analysis, which controlled for significant risk factors.ResultsOf all the participants, 219 (2.29%) participants were diagnosed with AAA. A significant elevation in NHHR was identified in the AAA group when contrasted with that in the normal group (P < 0.001). As demonstrated by the results of the multivariable logistic regression analysis, AAA was independently associated with NHHR before (odds ratio [OR], 1.440, P < 0.001) and after PSM (OR, 1.515, P < 0.001). Significant extension was observed in the areas under the receiver operating characteristic curves (AUROCs) of NHHR compared to those of single lipid parameters before and after PSM. An accordant association between NHHR and AAA in different subgroups was demonstrated by subgroup analysis.ConclusionIn the Chinese population, there is an independent association between NHHR and AAA. NHHR might be propitious to distinguish individuals with high risk of AAA.

Added value of clinical decision rules for the management of enhancing breast MRI lesions: A systematic comparison of the Kaiser score and the Göttingen score

PurposeWe investigated the added value of two internationally used clinical decision rules in the management of enhancing lesions on breast MRI. MethodsThis retrospective, institutional review board approved study included consecutive patients from two different populations. Patients received breast MRI according to the recommendations of the European Society of Breast Imaging (EUSOBI). Initially, all examinations were assessed by expert readers without using clinical decision rules. All lesions rated as category 4 or 5 according to the Breast Imaging Reporting and Data System were histologically confirmed.These lesions were re-evaluated by an expert reader blinded to the histology. He assigned each lesion a Göttingen score (GS) and a Kaiser score (KS) on different occasions. To provide an estimate on inter-reader agreement, a second fellowship-trained reader assessed a subset of these lesions.Subgroup analyses based on lesion type (mass vs. non-mass), size (>1 cm vs. ≤ 1 cm), menopausal status, and significant background parenchymal enhancement were conducted. The areas under the ROC curves (AUCs) for the GS and KS were compared, and the potential to avoid unnecessary biopsies was determined according to previously established cutoffs (KS > 4, GS > 3) Results527 lesions in 506 patients were included (mean age: 51.8 years, inter-quartile-range: 43.0–61.0 years). 131/527 lesions were malignant (24.9 %; 95 %-confidence-interval: 21.3–28.8). In all subgroups, the AUCs of the KS (median = 0.91) were higher than those of the GS (median = 0.83). Except for “premenopausal patients” (p = 0.057), these differences were statistically significant (p ≤ 0.01). Kappa agreement was higher for the KS (0.922) than for the GS (0.358). ConclusionBoth the KS and the GS provided added value for the management of enhancing lesions on breast MRI. The KS was superior to the GS in terms of avoiding unnecessary biopsies and showed superior inter-reader agreement; therefore, it may be regarded as the clinical decision rule of choice.

Can combined high-resolution vessel wall imaging and multiple post-labeling delay 3D pseudo-continuous arterial spin labeling differentiate moyamoya disease from atherosclerotic moyamoya syndrome?

PurposeTo investigate whether moyamoya disease (MMD) and atherosclerotic moyamoya syndrome (AS-MMS) differ in vascular morphology and perfusion characteristics using T1w-CUBE imaging and multiple post-labeling delay 3D pseudo-continuous arterial spin labeling imaging (MP 3D-PcASL), and to explore the potential of the combined techniques for accurate diagnosis of both diseases. MethodThis prospective study enrolled 51 patients with moyamoya vasculopathy, including 26 with MMD and 25 with AS-MMS. All patients underwent digital subtraction angiography (DSA)/magnetic resonance angiography (MRA), T1w-CUBE imaging, and MP 3D-PCASL examinations. Morphological parameters, including the outer diameter, maximum wall thickness, luminal stenosis morphology, degree of wall enhancement, number of collateral vessels, and perfusion parameters, such as cerebral blood flow (CBF) and arterial transit time (ATT), were measured. After univariate analysis between the two groups, logistic regression models based on the derived parameters of T1w-CUBE imaging, MP 3D-PCASL, and combined imaging were implemented, and receiver operating characteristic (ROC) curves were generated to compare the discriminatory power of the different imaging methods for the diagnosis of MMD. ResultsWith T1w-CUBE imaging, MMD showed a smaller outer diameter (2.76 ± 0.39 vs. 3.07 ± 0.49 mm) and maximum wall thickness (1.27 ± 0.19 vs. 1.49 ± 0.24 mm) than AS-MMS (both P < 0.05). Using MP 3D-pcASL, the resultant CBF (36.64 ± 14.28 vs. 28.77 ± 8.63 mL/100 g/min) was higher in MMD relative to AS-MMS, while an opposite pattern was shown for ATT (1.61 ± 0.09 vs. 1.72 ± 0.13 s; both P < 0.05). Robust diagnostic efficacies for disease differentiation, confirmed by high areas under the ROC curve (AUCs) (>0.808), were separately shown with T1w-CUBE and MP 3D-pcASL derived parameters. However, the combined multivariate logistic regression model showed optimaldiagnostic efficacy(AUC: 0.938; P < 0.05). ConclusionsCombined T1w-CUBE imaging and MP 3D-PCASL provides distinctive morphological and functional features to evaluate vessel walls and cerebral perfusion, and might help distinguish MMD from AS-MMS.

Risk factors predicting pathological degradation after cervical excision in cervical intraepithelial neoplasia grade II P16-positive patients over 25 years old: a cross-sectional study.

Cervical intraepithelial neoplasia (CIN) is a collective term for pre-cancerous lesions associated with cervical invasive carcinoma. Treatment options depend on the development and progression of the disease. Especially for patients with CINII grade who are aged 25 years and older and have fertility requirements, it is a clinical challenge to determine whether to proceed with conservative or excisional treatment. Excisional treatment increases the risk of overtreatment outcomes, such as cervical insufficiency, preterm labor, miscarriage, and premature rupture of membranes, in young women with childbearing potential. P16 immunohistochemical staining has greatly improved the consistency of CINII patient's diagnosis. The aim of this study was to analyze the risk factors predicting pathological degradation after cervical excision in cervical intraepithelial neoplasia grade II P16-positive patients over 25 years old, and to provide information to help optimize clinical treatments patients with CINII disease. Single-factor and logistic regression models were used to analyze the risk factors for pathological downgrading in the CINII/P16-positive (+) group. The predicted probability of pathological downgrading in the CINII/P16(+) group of patients was calculated according to the logistic regression model to generate a new variable multi-indicator association for receiver operating characteristic (ROC) curve plotting to determine the predictive ability. A total of 248 women who met the inclusion and exclusion criteria were included. Statistical analysis showed that the CINII/P16(+) group had a higher pathological downgrading rate compared with the CINIII group after cold knife conization (CKC) (χ2=6.26, P=0.012). Univariate factors showed that the differences were statistically significant when comparing age, number of biopsy-involved quadrants, menopausal status, and involvement of glands, respectively (P<0.05). In contrast, the differences were not statistically significant when comparing cytological findings, type of transformation zone, high-risk human papilloma virus (HR-HPV) testing, abortion status, pregnancy frequency, time from diagnosis to CKC and Ki67 percentage between the two groups. Multifactorial logistic regression showed that the extent of biopsy CINII involvement [odds ratio (OR), 1.589], menopausal status (OR, 4.031), and glandular involvement (OR, 5.549) were all independent risk factors for pathological downgrading in the CINII/P16(+) patient group (P<0.05). The order of significance of the areas under the ROC curve (AUCs) was as follows: combined multiple indicators (AUC 0.716) > gland involvement (AUC 0.625) > biopsy CINII involvement extent (AUC 0.614) > menopausal status (AUC 0.565). A higher rate of pathological downgrading after CKC was found in CINII/P16-positive patients who were aged over 25 years. Overtreatment exists in patients with CINII/P16-positive diagnosis. Independent factors for pathological downgrading were identified by the factors including if the lesion involved the gland, the extent of CINII involvement on biopsy, and menopausal status.

Comparison of ANN and LR models for predicting Carbapenem-resistant Klebsiella pneumoniae isolates from a southern province of China's RNSS data

Carbapenem-resistant Klebsiella pneumoniae (CRKP) is a serious threat to public health due to its limited treatment options and high mortality rate. This study aims to identify the risk factors of carbapenem resistance in patients with K. pneumoniae isolates and develop CRKP prediction models using logistic regression (LR) and artificial neural network (ANN) methods. We retrospectively analysed the data of 49,774 patients with Klebsiella pneumoniae isolates from a regional nosocomial infection surveillance system (RNSS) between 2018 and 2021. We performed logistic regression analyses to determine the independent predictors for CRKP. We then built and evaluated LR and ANN models based on these predictors using calibration curves, ROC curves, and decision curve analysis (DCA). We also applied the Synthetic Minority Over-Sampling Technique (SMOTE) to balance the data of CRKP and non-CRKP groups. The LR model showed good discrimination and calibration in both training and validation sets, with areas under the ROC curve (AUROC) of 0.824 and 0.825, respectively. The DCA indicated that the LR model had clinical usefulness for decision making. The ANN model outperformed the LR model both in the training set and validation set. The SMOTE technique improved the performance of both models for CRKP detection in training set, but not in the validation set. We developed and validated LR and ANN models for predicting CRKP based on RNSS data. Both models were feasible and reliable for CRKP inference and could potentially assist clinicians in selecting appropriate empirical antibiotics and reducing unnecessary medical resource utilization.

A python system for regional landslide susceptibility assessment by integrating machine learning models and its application

Landslide susceptibility assessment is considered the first step in landslide risk assessment, but current studies mostly rely on GIS platforms or other software for data preprocessing. The modeling process is relatively complicated and multi-models cannot be integrated. With regard to this issue, this study develops a Python system for automatic assessment of regional landslide susceptibility. The Python system implements landslide susceptibility assessment through three modules: geographic data processing, machine learning modeling and result evaluation analysis. For geographic data processing, ten landslide influencing factors can be used to construct an evaluation factor dataset and reclassify the thematic maps based on the frequency ratio method. Four built-in machine learning models (logistic regression (LR), multi-layer perceptron (MLP), support vector machine (SVM) and extreme gradient boosting (XGBoost)) are integrated into the system to complete susceptibility modeling and calculation. Additionally, receiver operating characteristic (ROC) curves can be automatically generated to evaluate the accuracy. The system was then applied into Lantian County in Shaanxi Province as a demonstration example. The results show that the areas under the ROC curve (AUC) of the four models are 0.838 (LR)、0.882 (SVM)、0.809 (MLP) and 0.812 (XGBoost), respectively, indicating that the SVM model was the most suitable model for landslide susceptibility assessment in Lantian County in the Loess Plateau of China. The system has now been made open source on Github, which can effectively improve the efficiency of regional landslide susceptibility assessment, especially provide tools for data processing and modeling for non-professionals.

Machine Learning Models for ASCVD Risk Prediction in an Asian Population - How to Validate the Model is Important.

Atherosclerotic cardiovascular disease (ASCVD) is prevalent worldwide including Taiwan, however widely accepted tools to assess the risk of ASCVD are lacking in Taiwan. Machine learning models are potentially useful for risk evaluation. In this study we used two cohorts to test the feasibility of machine learning with transfer learning for developing an ASCVD risk prediction model in Taiwan. Two multi-center observational registry cohorts, T-SPARCLE and T-PPARCLE were used in this study. The variables selected were based on European, U.S. and Asian guidelines. Both registries recorded the ASCVD outcomes of the patients. Ten-fold validation and temporal validation methods were used to evaluate the performance of the binary classification analysis [prediction of major adverse cardiovascular (CV) events in one year]. Time-to-event analyses were also performed. In the binary classification analysis, eXtreme Gradient Boosting (XGBoost) and random forest had the best performance, with areas under the receiver operating characteristic curve (AUC-ROC) of 0.72 (0.68-0.76) and 0.73 (0.69-0.77), respectively, although it was not significantly better than other models. Temporal validation was also performed, and the data showed significant differences in the distribution of various features and event rate. The AUC-ROC of XGBoost dropped to 0.66 (0.59-0.73), while that of random forest dropped to 0.69 (0.62-0.76) in the temporal validation method, and the performance also became numerically worse than that of the logistic regression model. In the time-to-event analysis, most models had a concordance index of around 0.70. Machine learning models with appropriate transfer learning may be a useful tool for the development of CV risk prediction models and may help improve patient care in the future.

Quality evaluation of the usefulness of an emergency department fall risk assessment tool

IntroductionFalls that occur within a hospital setting are difficult to predict, however, are preventable adverse events with the potential to negatively impact patient care. Falls have the potential to cause serious or fatal injuries and may increase patient morbidity. Many hospitals utilize fall “predictor tools” to categorize a patient's fall risk, however, these tools are primarily studied within in-patient units. The emergency department (ED) presents a unique environment with a distinct patient population and demographic. The Memorial Emergency Department Fall Risk Assessment Tool (MEDFRAT) has shown to be effective with predicting a patient's fall risk in the ED. This IRB-approved study aims to assess the predictive validity of the MEDFRAT by evaluating the sensitivity and specificity for predicting a patient's fall risk in an emergency department at a level 1 trauma center. MethodsA retrospective cohort analysis was conducted using an electronic medical record (EMR) for patients who met study inclusion criteria at a level 1 trauma center ED. Extracted data includes MEDFRAT components, demographic information, and data from the Moving Safely Risk Assessment (MSRA) Tool, our institution's current fall assessment tool. A receiver operating characteristic (ROC) curve was constructed to determine the best cutoff for identifying any fall risk. Sensitivity, specificity, accuracy, positive likelihood ratio (LR+) and negative LR (LR-), with 95% CIs were then calculated for the cutoff value determined from the ROC curve. To compare overall tool performance, the areas under the ROC curves (AUC) were determined and compared with a z-test. ResultsThe MEDFRAT had a significantly higher sensitivity compared to the MSRA (83.1% vs. 66.1%, p = 0.002), while the MSRA had a significantly higher specificity (84.5% vs. 69.0%, p = 0.012). For identifying any level of fall risk, ROC curve analysis showed that the cutoff providing the best trade-off between sensitivity and specificity for the MEDFRAT was a score of ≥1. Additionally, area under the curve was determined for the MEDFRAT and MSRA (0.817 vs. 0.737). ConclusionThis study confirms the validity of the MEDFRAT as an acceptable tool to predict in-hospital falls in a level 1 trauma center ED. Accurate identification of patients at a high risk of falling is critical for decreasing healthcare costs and improving health outcomes and patient safety.

Development and validation of artificial intelligence models to predict urinary tract infections and secondary bloodstream infections in adult patients

BackgroundTraditional culture methods are time-consuming, making it difficult to utilize the results in the early stage of urinary tract infection (UTI) management, and automated urinalyses alone show insufficient performance for diagnosing UTIs. Several models have been proposed to predict urine culture positivity based on urinalysis. However, most of them have not been externally validated or consisted solely of urinalysis data obtained using one specific commercial analyzer. MethodsA total of 259,187 patients were enrolled to develop artificial intelligence (AI) models. AI models were developed and validated for the diagnosis of UTI and urinary tract related-bloodstream infection (UT-BSI). The predictive performance of conventional urinalysis and AI algorithms were assessed by the areas under the receiver operating characteristic curve (AUROC). We also visualized feature importance rankings as Shapley additive explanation bar plots. ResultsIn the two cohorts, the positive rates of urine culture tests were 25.2% and 30.4%, and the proportions of cases classified as UT-BSI were 1.8% and 1.6%. As a result of predicting UTI from the automated urinalysis, the AUROC were 0.745 (0.743–0.746) and 0.740 (0.737–0.743), and most AI algorithms presented excellent discriminant performance (AUROC > 0.9). In the external validation dataset, the XGBoost model achieved the best values in predicting both UTI (AUROC 0.967 [0.966–0.968]) and UT-BSI (AUROC 0.955 [0.951–0.959]). A reduced model using ten parameters was also derived. ConclusionsWe found that AI models can improve the early prediction of urine culture positivity and UT-BSI by combining automated urinalysis with other clinical information. Clinical utilization of the model can reduce the risk of delayed antimicrobial therapy in patients with nonspecific symptoms of UTI and classify patients with UT-BSI who require further treatment and close monitoring.

Passive leg raising test induced changes in plethysmographic variability index to assess fluid responsiveness in critically ill mechanically ventilated patients with acute circulatory failure

BackgroundPassive leg raising (PLR) reliably predicts fluid responsiveness but requires a real-time cardiac index (CI) measurement or the presence of an invasive arterial line to achieve this effect. The plethysmographic variability index (PVI), an automatic measurement of the respiratory variation of the perfusion index, is non-invasive and continuously displayed on the pulse oximeter device. We tested whether PLR-induced changes in PVI (ΔPVIPLR) could accurately predict fluid responsiveness in mechanically ventilated patients with acute circulatory failure. MethodsThis was a secondary analysis of an observational prospective study. We included 29 mechanically ventilated patients with acute circulatory failure in this study. We measured PVI (Radical-7 device; Masimo Corp., Irvine, CA) and CI (Echocardiography) before and during a PLR test and before and after volume expansion of 500 mL of crystalloid solution. A volume expansion-induced increase in CI of >15% defined fluid responsiveness. To investigate whether ΔPVIPLR can predict fluid responsiveness, we determined areas under the receiver operating characteristic curves (AUROCs) and gray zones for ΔPVIPLR. ResultsOf the 29 patients, 27 (93.1%) received norepinephrine. The median tidal volume was 7.0 [IQR: 6.6–7.6] mL/kg ideal body weight. Nineteen patients (65.5%) were classified as fluid responders (increase in CI > 15% after volume expansion). Relative ΔPVIPLR accurately predicted fluid responsiveness with an AUROC of 0.89 (95%CI: 0.72–0.98, p < 0.001). A decrease in PVI ≤ −24.1% induced by PLR detected fluid responsiveness with a sensitivity of 95% (95%CI: 74–100%) and a specificity of 80% (95%CI: 44–97%). Gray zone was acceptable, including 13.8% of patients. The correlations between the relative ΔPVIPLR and changes in CI induced by PLR and by volume expansion were significant (r = −0.58, p < 0.001, and r = −0.65, p < 0.001; respectively). ConclusionsIn sedated and mechanically ventilated ICU patients with acute circulatory failure, PLR-induced changes in PVI accurately predict fluid responsiveness with an acceptable gray zone. Trial registrationClinicalTrials.govNCT03225378.

Predicting acute pancreatitis severity with enhanced computed tomography scans using convolutional neural networks

This study aimed to evaluate acute pancreatitis (AP) severity using convolutional neural network (CNN) models with enhanced computed tomography (CT) scans. Three-dimensional DenseNet CNN models were developed and trained using the enhanced CT scans labeled with two severity assessment methods: the computed tomography severity index (CTSI) and Atlanta classification. Each labeling method was used independently for model training and validation. Model performance was evaluated using confusion matrices, areas under the receiver operating characteristic curve (AUC-ROC), accuracy, precision, recall, F1 score, and respective macro-average metrics. A total of 1,798 enhanced CT scans met the inclusion criteria were included in this study. The dataset was randomly divided into a training dataset (n = 1618) and a test dataset (n = 180) with a ratio of 9:1. The DenseNet model demonstrated promising predictions for both CTSI and Atlanta classification-labeled CT scans, with accuracy greater than 0.7 and AUC-ROC greater than 0.8. Specifically, when trained with CT scans labeled using CTSI, the DenseNet model achieved good performance, with a macro-average F1 score of 0.835 and a macro-average AUC-ROC of 0.980. The findings of this study affirm the feasibility of employing CNN models to predict the severity of AP using enhanced CT scans.

Development and validation of a nomogram for predicting pulmonary complications after video-assisted thoracoscopic surgery in elderly patients with lung cancer.

Postoperative pulmonary complications (PPCs) significantly increase the morbidity and mortality in elderly patients with lung cancer. Considering the adverse effects of PPCs, we aimed to derive and validate a nomogram to predict pulmonary complications after video-assisted thoracoscopic surgery in elderly patients with lung cancer and to assist surgeons in optimizing patient-centered treatment plans. The study enrolled 854 eligible elderly patients with lung cancer who underwent sub-lobectomy or lobectomy. A clinical prediction model for the probability of PPCs was developed using univariate and multivariate analyses. Furthermore, data from one center were used to derive the model, and data from another were used for external validation. The model's discriminatory capability, predictive accuracy, and clinical usefulness were assessed using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis, respectively. Among the eligible elderly patients with lung cancer, 214 (25.06%) developed pulmonary complications after video-assisted thoracoscopic surgery. Age, chronic obstructive pulmonary disease, surgical procedure, operative time, forced expiratory volume in one second, and the carbon monoxide diffusing capacity of the lung were independent predictors of PPCs and were included in the final model. The areas under the ROC curves (AUC) of the training and validation sets were 0.844 and 0.796, respectively. Ten-fold cross-validation was used to evaluate the generalizability of the predictive model, with an average AUC value of 0.839. The calibration curve showed good consistency between the observed and predicted probabilities. The proposed nomogram showed good net benefit with a relatively wide range of threshold probabilities. A nomogram for elderly patients with lung cancer can be derived using preoperative and intraoperative variables. Our model can also be accessed using the online web server https://pulmonary-disease-predictor.shinyapps.io/dynnomapp/. Combining both may help surgeons as a clinically easy-to-use tool for minimizing the prevalence of pulmonary complications after lung resection in elderly patients.

A novel artificial intelligence model for fetal facial profile marker measurement during the first trimester

BackgroundTo study the validity of an artificial intelligence (AI) model for measuring fetal facial profile markers, and to evaluate the clinical value of the AI model for identifying fetal abnormalities during the first trimester.MethodsThis retrospective study used two-dimensional mid-sagittal fetal profile images taken during singleton pregnancies at 11–13+ 6 weeks of gestation. We measured the facial profile markers, including inferior facial angle (IFA), maxilla-nasion-mandible (MNM) angle, facial-maxillary angle (FMA), frontal space (FS) distance, and profile line (PL) distance using AI and manual measurements. Semantic segmentation and landmark localization were used to develop an AI model to measure the selected markers and evaluate the diagnostic value for fetal abnormalities. The consistency between AI and manual measurements was compared using intraclass correlation coefficients (ICC). The diagnostic value of facial markers measured using the AI model during fetal abnormality screening was evaluated using receiver operating characteristic (ROC) curves.ResultsA total of 2372 normal fetuses and 37 with abnormalities were observed, including 18 with trisomy 21, 7 with trisomy 18, and 12 with CLP. Among them, 1872 normal fetuses were used for AI model training and validation, and the remaining 500 normal fetuses and all fetuses with abnormalities were used for clinical testing. The ICCs (95%CI) of the IFA, MNM angle, FMA, FS distance, and PL distance between the AI and manual measurement for the 500 normal fetuses were 0.812 (0.780–0.840), 0.760 (0.720–0.795), 0.766 (0.727-0.800), 0.807 (0.775–0.836), and 0.798 (0.764–0.828), respectively. IFA clinically significantly identified trisomy 21 and trisomy 18, with areas under the ROC curve (AUC) of 0.686 (95%CI, 0.585–0.788) and 0.729 (95%CI, 0.621–0.837), respectively. FMA effectively predicted trisomy 18, with an AUC of 0.904 (95%CI, 0.842–0.966). MNM angle and FS distance exhibited good predictive value in CLP, with AUCs of 0.738 (95%CI, 0.573–0.902) and 0.677 (95%CI, 0.494–0.859), respectively.ConclusionsThe consistency of fetal facial profile marker measurements between the AI and manual measurement was good during the first trimester. The AI model is a convenient and effective tool for the early screen for fetal trisomy 21, trisomy 18, and CLP, which can be generalized to first-trimester scanning (FTS).

Prognostic Nutritional Index Combined with Triglyceride-Glucose Index to Contrast a Nomogram for Predicting Contrast-Induced Kidney Injury in Type 2 Diabetes Mellitus Patients with Acute Coronary Syndrome After Percutaneous Coronary Intervention.

Our objective was to develop and validate a nomogram model aiming at predicting the risk of contrast-induced acute kidney injury (CI-AKI) following percutaneous coronary intervention (PCI) in patients suffering fromtype 2 diabetes mellitus (T2DM) and also diagnosed with acute coronary syndrome (ACS). The study gathered data from 722 T2DM patients with ACS who received PCI treatment at the Affiliated Hospital of Xuzhou Medical University between February 2019 and December 2022, serving as the training set. Considering the validation set, the study included 217 patients who received PCI at the East Affiliated Hospital of Xuzhou Medical University. The patients were classified into CI-AKI and non-CI-AKI groups. The study employed univariate and multivariate logistic analysis for identifying independent risk factors for CI-AKI, followed by developing a predictive nomogram model for CI-AKI risk using R software. The predictive performance and clinical utility of the nomogram were assessed through internal and external validation, utilizing the areas under the receiver operating characteristic curve (AUC-ROC), the Hosmer-Lemeshow test and calibration correction curve, and decision curve analysis (DCA). The nomogram comprised four variables: age, estimated glomerular filtration rate (eGFR), triglyceride-glucose (TyG) index, and prognostic nutritional index (PNI). The AUC-ROC were 0.785 (95% confidence interval (CI) 0.729-0.841) and 0.802 (95% CI 0.699-0.905) for the training and validation cohorts, respectively, indicating a high discriminative ability of the nomogram. The calibration assessment and decision curve analysis have substantiated the strong concordance and clinical usefulness of the aforementioned. The nomogram exhibits favorable discrimination and accuracy, enabling it to visually and individually identify pre-procedure high-risk patients, and possesses a predictive capacity regarding CI-AKI incidence after PCI in patients diagnosed with both T2DM and ACS.

Serum Cholinesterase, C-reactive Protein, Interleukin 6, and Procalcitonin Levels as Predictors of Mortality in Patients in the Intensive Care Unit.

The prognostic utility of inflammatory markers in survival has been suggested in patients with cancer; however, evidence on their prognostic value in severely ill patients is very limited. We aimed to explore the prognostic value of cholinesterase (ChE), C-reactive protein (CRP), interleukin-6 (IL-6), and procalcitonin (PCT) in predicting mortality in patients from the intensive care unit (ICU). Serum levels of ChE, CRP, IL-6 and PCT were measured in ICU patients from December 13th, 2019 to June 28th, 2022. We assessed the predictive power of ChE, CRP, IL-6, and PCT using the receiver operating characteristic (ROC) curves. Furthermore, we evaluated their diagnostic accuracy by comparing the areas under the ROC curve (AUCs) along with their corresponding 95% confidence intervals (CIs). The cut-off values were determined to dichotomise these biomarkers, which were then included in multivariable logistic regression models to examine their relationship with ICU mortality. Among 253 ICU patients included in the study, 66 (26%) died during the ICU stay. The AUCs to predict ICU mortality were 0.643 (95% CI, 0.566-0.719), 0.648 (95% CI, 0.633-0.735), 0.643 (95% CI, 0.563-0.723) and 0.735 (95% CI, 0.664-0.807) for ChE, CRP, IL-6 and PCT, respectively. After adjusting for age, sex and disease severity, lower ChE level (<3.668 × 103 U L-1) and higher levels of CRP (>10.546 mg dL-1), IL-6 (>986.245 pg mL-1) and PCT (>0.505 μg L-1) were associated with higher mortality risk, with odd ratios of 2.70 (95% CI, 1.32-5.54), 4.99 (95% CI, 2.41-10.38), 3.24 (95% CI, 1.54-6.78) and 3.67 (95% CI, 1.45-9.95), respectively. ChE, CRP, IL-6 and PCT were independent ICU mortality risk factors in severely ill patients. Elevated PCT levels exhibited better predictive value than the other three biomarkers that were evaluated.