Derivation Cohort Research Articles

Predictive model of intrahepatic collateral vessels among hepatic veins in patients with liver tumors involving the caval confluence.

Communicating vessels among hepatic veins in patients with tumors invading/compressing hepatic veins at their caval confluence facilitate new surgical solutions. Although their recognition by intraoperative ultrasound has been described, the possibility of preoperative detection still remains uncertain. We aimed to develop a model to predict their presence before surgery. Patients with caval confluence-located tumors who underwent surgery between 2010 and 2022 were collected and analyzed. A total of 136 eligible patients were assigned randomly into the derivation (n= 96) and validation (n= 40) cohorts. Communicating vessels were recorded as visible or not considering the intraoperative ultrasound findings as reference. A predictive model was built and graphically represented as nomogram. Of 136 patients included, 106 (78%) had communicating vessels. Of patients with communicating vessels, a parenchyma-sparing hepatectomy was performed in 98 (92%; 77% with hepatic vein amputation, 75/98). Of 30 patients without communicating vessels, all underwent an R1 vascular parenchyma-sparing hepatectomy. Operative mortality was nil. No liver congestion occurred. In the former, presence of communicating vessels at the intraoperative ultrasound was detected in 75 (78%) patients, whereas in the validation cohort it was detected in 31 (78%). No differences in the distribution of the variables were found in the 2 cohorts. The model included tumor size at the confluence, tumor number, degree of hepatic vein infiltration, and density of hepatic tissue at contrast-computed tomography, as predictors of communicating vessels in derivation cohort (discrimination was area under the receiver operating characteristic curve= 0.82). Identification of communicating vessels in patients with tumors at the caval confluence opens new surgical options, expanding resectability. The present predictive model may play a useful role in facilitating surgical decision-making, developing and implementing surgical strategies, and guiding resectability improvement efforts.

Prediction Trough Concentrations of Valproic Acid Among Chinese Adult Patients with Epilepsy Using Machine Learning Techniques.

This study aimed to establish an optimal model based on machine learning (ML) to predict Valproic acid (VPA) trough concentrations in Chinese adult epilepsy patients. A single-center retrospective study was carried outat the Jinshan Hospital affiliated with Fudan University from January 2022 to December 2023. A total of 102 VPA trough concentrations were split into a derivation cohort and a validation cohort at a ratio of 8:2. Thirteen ML algorithms were developed using 27 variables in the derivation cohort and were filtered by the lowest mean absolute error (MAE) value. In addition, feature selection was applied to optimize the model. Ultimately, the extra tree algorithm was chosen to establish the personalized VPA trough concentration prediction model due to its best performance (MAE = 13.08). The SHapley Additive exPlanations (SHAP) plotswere used to visualize and rank the importance of features, providing insights into how each feature influences the model's predictions. After feature selection, we found that the model with the top 9 variables [including daily dose, last dose, uric acid (UA), platelet (PLT), combination, gender, weight, albumin (ALB), aspartate aminotransferase (AST)] outperformed the model with 27 variables, with MAE of 6.82, RMSE of 9.62, R2 value of 0.720, relative accuracy (±20%) of 61.90%, and absolute accuracy (±20mg/L) of 90.48%. In conclusion, the trough concentration prediction model for VPA in Chinese adult epileptic patients based on the extra tree algorithm demonstrated strong predictive ability which is valuable for clinicians in medication guidance.

Differentiation of Atherosclerotic Carotid Plaque Components With Dual-Energy Computed Tomography.

Carotid plaque vulnerability is a strong predictor of recurrent ipsilateral stroke, but differentiation of plaque components using conventional computed tomography (CT) is suboptimal. The aim of our study was to evaluate the ability of dual-energy CT (DECT) to characterize atherosclerotic carotid plaque components based on the effective atomic number and effective electron density using magnetic resonance imaging (MRI) and, where possible, histology as the reference standard. Patients with recent cerebral ischemia and a ≥2-mm carotid plaque underwent computed tomography angiography and MRI. A subgroup underwent carotid endarterectomy. Trained observers delineated plaque components on histology or MRI, independent of computed tomography angiography. DECT was coregistered with MRI and/or histology. Intraplaque hemorrhage (IPH), lipid-rich necrotic core (LRNC), fibrous tissue, and calcifications were delineated on DECT, and ρeff and Zeff values were determined in the derivation cohort (n = 55). Spatial separation of these components was evaluated in a ρeff-Zeff-cluster plot. Ranges that optimally differentiate plaque features were determined. For validation, plaque components were quantified in the validation cohort (n = 29) using these ρeff-Zeff ranges and literature-based Hounsfield unit (HU) ranges and correlated to MRI volumes. Eighty-four participants (68 ± 8 years; 55 male) were evaluated. In the derivation cohort, plaque components were well separated on the cluster plot, resulting in the following ranges: IPH:ρeff < 1.15, Zeff < 7.5, LRNC:ρeff < 1.15, Zeff:7.5-8.75, fibrous tissue:ρeff < 1.15, Zeff > 8.75, and calcifications: ρeff > 1.15, Zeff > 0. In the validation cohort, significant correlations were found between ρeff-Zeff-based and MRI plaque volumes for fibrous tissue (r = 0.69, P < 0.001), LRNC (r = 0.94, P < 0.001), IPH (r = 0.35, P = 0.03), and calcifications (r = 0.70, P < 0.001). Lower correlations were found between HU-based and MRI plaque volumes for fibrous tissue (r = 0.40, P = 0.02), LRNC (r = 0.86, P < 0.001), and calcifications (r = 0.47, P = 0.005), with no correlation for IPH (r = 0.02, P = 0.45). We determined ρeff-Zeff ranges for plaque assessment. ρeff-Zeff-based volumes showed strong-to-very strong correlations with MRI for LRNC, fibrous tissue, and calcifications and a weak correlation for IPH. ρeff-Zeff-based volumes demonstrated superior agreement with MRI for all plaque components compared with HU-based volumes, highlighting the potential of DECT for the identification of patients with vulnerable plaques.

Evidence-Based Application of Natriuretic Peptides in the Evaluation of Chronic Heart Failure With Preserved Ejection Fraction in the Ambulatory Outpatient Setting.

Plasma NT-proBNP (N-terminal pro-B-type natriuretic peptide) is commonly used to diagnose heart failure with preserved ejection fraction (HFpEF), but its diagnostic performance in the ambulatory/outpatient setting is unknown because previous studies lacked objective reference standards. Among patients with chronic dyspnea, diagnosis of HFpEF or noncardiac dyspnea was determined conclusively by exercise catheterization in a derivation cohort (n=414), multicenter validation cohort 1 (n=560), validation cohort 2 (n=207), and a nonobese Japanese validation cohort 3 (n=77). Optimal NT-proBNP cut points for HFpEF rule out (optimizing sensitivity) and rule in (optimizing specificity) were derived and tested, stratified by obesity and atrial fibrillation. Derived cut points were tested in 3 additional validation cohorts (cohorts 4-6) in whom HFpEF was diagnosed by resting catheterization only (n=260), previous hospitalization for heart failure (n=447), or exercise echocardiography (n=517), respectively. Current recommended rule-out NT-proBNP threshold <125 pg/mL had 82% sensitivity (95% CI, 77%-88%) with a body mass index (BMI) <35 kg/m2, decreasing to 67% (95% CI, 58%-77%) with a BMI ≥35 kg/m2. A lower rule-out NT-proBNP threshold <50 pg/mL displayed good sensitivity with a BMI <35 kg/m2 (97% [95% CI, 95%-99%]), with a modest decline in sensitivity with a BMI ≥35 kg/m2 (86% [95% CI, 79%-93%]); diagnostic thresholds were confirmed in validation cohorts 1 and 2 (91% [95% CI, 88%-95%] and 86% [95% CI, 80%-93%] with a BMI <35 kg/m2; 80% [95% CI, 74%-87%] and 84% [95% CI, 74%-93%] with a BMI ≥35 kg/m2). Current consensus age- and BMI-stratified rule-in thresholds demonstrated only 65% specificity (95% CI, 57%-72%). Rule-in NT-proBNP threshold ≥500 pg/mL had 85% specificity (95% CI, 78%-91%) with a BMI <35 kg/m2 (87% [95% CI, 80%-94%] and 90% [95% CI, 81%-99%] in validation cohorts), with 100% specificity at a BMI ≥35 kg/m2 (93% [95% CI, 81%-100%] and 100% in validation cohorts). With a BMI ≥35 kg/m2, lower rule-in thresholds (≥220 pg/mL) provided good specificity (88% [95% CI, 73%-100%]; 93% [95% CI, 81%-100%] and 100% in validation cohorts). Findings were consistent in validation cohorts 3 through 6 (sensitivity of <50 pg/mL, 93%-98%; specificity of ≥500 pg/mL, 82%-89%). NT-proBNP provided no incremental discrimination among patients with history of AF; ≥98% of patients with AF and dyspnea were found to have HFpEF in our cohorts. In patients with chronic unexplained dyspnea, current rule-in and rule-out NT-proBNP diagnostic thresholds lead to unacceptably high error rates, with important interactions by obesity and AF status. In our study, NT-proBNP provided little value in those with AF and dyspnea because the presence of AF is by itself a robust biomarker of HFpEF. Use of separate rule-in and rule-out diagnostic thresholds stratified by BMI reduces miscategorization and can guide more appropriate use of exercise testing for possible HFpEF.

Preoperative carcinoembryonic antigen to lymphocyte ratio serves as a prognostic predictor in elderly patients with colorectal cancer: a multicentre retrospective study

ObjectivesThe aim of this study is to derive and validate a reliable indicator for predicting an increased risk of postoperative mortality in elderly patients undergoing curative resection for colorectal cancer...

Prediction Model for POstoperative atriaL fibrillAtion in caRdIac Surgery: The POLARIS Score.

Background: New-onset postoperative atrial fibrillation (POAF) is the most common complication after cardiac surgery, occurring approximately in one-third of the patients. This study considered all-comer patients who underwent cardiac surgery to build a predictive model for POAF. Methods: A total of 3467 (Center 1) consecutive patients were used as a derivation cohort to build the model. The POLARIS score was then derived proportionally from the odds ratios obtained following multivariable logistic regression (MLR). The Brier Score, the area under the receiver operating characteristic curve, and the Hosmer-Lemeshow goodness-of-fit test were used to validate the model. Then, 2272 (Center 2) consecutive patients were used as an external validation cohort. Results: In the overall population (n = 5739), POAF occurred in 32.7% of patients. MLR performed in the derivation cohort showed that age, obesity, chronic renal failure, pulmonary hypertension, minimally invasive surgery, and aortic and mitral valve surgery were predictors of POAF. The derived POLARIS score was used to further stratify the population into four risk clusters: low (1.5-3), intermediate (3.5-5), high (5.5-7), and very high (7.5-9), each progressively showing an increase in POAF incidence. This was confirmed in a correlation analysis (Spearman's rho: 0.636). Conclusions: The POLARIS score is a simple-to-use tool to stratify patients at higher risk of POAF. Precise identification of such patients might be used to implement clinical practice with the introduction of preoperative antiarrhythmic prophylaxis, further reducing the incidence of POAF and, potentially, its clinical sequelae, despite further investigations being warranted to test this model in prospective studies.

Derivation and Validation of D-RISK: An Electronic Health Record-Driven Risk Score to Detect Undiagnosed Dysglycemia in Clinical Practice.

We derive and validate D-RISK, an electronic health record (EHR)-driven risk score to optimize and facilitate screening for undiagnosed dysglycemia (prediabetes + diabetes) in clinical practice. We used retrospective EHR data (derivation sample) and a prospective diabetes screening study (validation sample) to develop D-RISK. Logistic regression with backward selection was used to predict dysglycemia (HbA1c ≥5.7%) using diabetes risk factors consistently captured in structured EHR data. Model coefficients were converted to a points-based risk score. We report discrimination, sensitivity, and specificity and compare D-RISK to the American Diabetes Association (ADA) risk test and the ADA and United States Preventive Services Task Force (USPSTF) screening guidelines. The derivation cohort included 11,387 patients (mean age 48 years; 65% female; 42% Hispanic; 32% non-Hispanic Black; mean BMI 32; 29% with hypertension). D-RISK included age, race, BMI, hypertension, and random glucose. The area under curve (AUC) for the risk score was 0.75 (95% CI 0.74-0.76). In the validation screening study (n = 519), the AUC was 0.71 (95% CI 0.66-0.75) which was better than the ADA and USPSTF diabetes screening guidelines (AUC = 0.52 and AUC = 0.58, respectively; P < 0.001 for both). Discrimination was similar to the ADA risk test (AUC = 0.67) using patient-reported data to supplement EHR data, although D-RISK was more sensitive (75% vs. 61%) at the recommended screening thresholds. Designed for use in EHR, D-RISK performs better than commonly used screening guidelines and risk scores and may help detect undiagnosed cases of dysglycemia in clinical practice.

Development and Validation of a Clinical, CT, Genomic Classifier and BAL Scoring System for Diagnosing IPF.

The utility of incorporating a usual interstitial pneumonia (UIP) genomic classifier (GC) and bronchoalveolar lavage (BAL) cell count analysis alongside traditional clinical-imaging assessment in aiding in the multidisciplinary diagnosis of IPF in patients with a non-definite HRCT UIP pattern is uncertain.We reviewed consecutive adult patients presenting with fibrotic interstitial lung disease (fILD) and non-definite HRCT UIP pattern who underwent BAL and GC. The initial fILD diagnoses were re-evaluated after bronchoscopy and a final multidisciplinary consensus diagnosis was provided. We created a clinical score by analyzing fILD clinical characteristics, GC, and BAL results from 139 National Jewish Health (NJH) patients and validated it at the University of Arizona, n=52. A multivariable model was developed and assessed using receiver operating characteristic curves.43/139 (31%) and 29/52 (56%) of patients in the derivation and validation cohort, respectively, were diagnosed with IPF after bronchoscopy, and 85/139 (61%) and 32/52 (61%) had a change in treatment, respectively. Compared to non-IPF, IPF patients had a similar progression-free survival (HR, 1.50; 95%CI, 0.76, 2.95). The final model, assigned a score to eight predictors: age, sex, HRCT probable UIP pattern, exposures, connective tissue disease signs/symptoms, Velcro crackles, CG results, and BAL lymphocyte and monocyte count. The final score demonstrated an AUC of 0.90 (95% CI 0.85 to 0.95) in the derivation cohort and 0.91 (95% CI 0.83 to 0.99) in the validation cohort.The clinical-HRCT-BAL-GC IPF score may accurately estimate the post-test probability of IPF in patients with a non-definite HRCT UIP pattern.

Imaging-based method to quantify left ventricular diastolic pressures.

To establish an imaging-based method to quantify left ventricular (LV) diastolic pressures. In 115 patients suspected of coronary artery disease, LV pressure was measured by micromanometers and images by echocardiography. LV filling pressure was measured as LV pre-atrial contraction pressure (pre-A PLV). Based on previous observations we hypothesized that pre-A PLV approximates the sum of minimum PLV and maximum transmitral pressure difference. Parameters used for pressure estimates included LV volumes and strain, left atrial strain, mitral flow velocities, systolic arterial cuff pressure and body mass index. Minimum PLV was estimated by predictors identified in a derivative cohort (n=81). Mitral pressure difference was calculated by a simplified Navier-Stokes equation. Accuracy of estimates of minimum PLV, pre-A PLV and end-diastolic PLV were investigated in a testing cohort (n=19). Patient-specific LV diastolic pressure curves were constructed by adjusting a reference curve according to pressure estimates at key diastolic events.There was good agreement between estimated and measured pre-A PLV: Bias 0.0, limits of agreement <3.1 mmHg (±1.96SD). Estimated minimum PLV and end-diastolic PLV also showed good agreement with measured pressures. Furthermore, there was good agreement between measured and estimated LV diastolic pressure curves, quantified as mean LV diastolic pressure: Bias 0.2, limits of agreement <3.2 mmHg. The proposed non-invasive method provided estimates of minimum PLV, pre-A PLV and end-diastolic PLV, each reflecting different features of diastolic function. Additionally, it provided an estimate of the LV diastolic pressure curve. Validation in larger populations with different phenotypes is necessary to determine the validity of the method in clinical practice.

Updating the Scottish national cardiovascular risk score: ASSIGN version 2.0

BackgroundThe Assessing cardiovascular risk using Scottish Intercollegiate Guidelines Network (ASSIGN) risk score, developed in 2006, is used in Scotland for estimating the 10-year risk of first atherosclerotic cardiovascular disease (ASCVD)....

Clinical sub-phenotypes of acute kidney injury in children and their association with prognosis

To investigate the clinical sub-phenotype (SP) of pediatric acute kidney injury (AKI) and their association with clinical outcomes. General status and initial values of laboratory markers within 24 hours after admission to the pediatric intensive care unit (PICU) were recorded for children with AKI in the derivation cohort (n=650) and the validation cohort (n=177). In the derivation cohort, a least absolute shrinkage and selection operator (LASSO) regression analysis was used to identify death-related indicators, and a two-step cluster analysis was employed to obtain the clinical SP of AKI. A logistic regression analysis was used to develop a parsimonious classifier model with simplified metrics, and the area under the curve (AUC) was used to assess the value of this model. This model was then applied to the validation cohort and the combined derivation and validation cohort. The association between SPs and clinical outcomes was analyzed with all children with AKI as subjects. In the derivation cohort, two clinical SPs of AKI (SP1 and SP2) were identified by the two-step cluster analysis using the 20 variables screened by LASSO regression, namely SPd1 group (n=536) and SPd2 group (n=114). The simplified classifier model containing eight variables (P<0.05) had an AUC of 0.965 in identifying the two clinical SPs of AKI (P<0.001). The validation cohort was clustered into SPv1 group (n=156) and SPv2 group (n=21), and the combined derivation and validation cohort was clustered into SP1 group (n=694) and SP2 group (n=133). After adjustment for confounding factors, compared with the SP1 group, the SP2 group had significantly higher incidence rates of multiple organ dysfunction syndrome and death during the PICU stay (P<0.001), and SP2 was significantly associated with the risk of death within 28 days after admission to the PICU (P<0.001). This study establishes a parsimonious classifier model and identifies two clinical SPs of AKI with different clinical features and outcomes.The SP2 group has more severe disease and worse clinical prognosis.

Derivation and validation of a simple prognostic risk score to predict short-term mortality in acute cardiogenic pulmonary edema: SABIHA Score.

In the context of acute cardiogenic pulmonary edema (ACPE), a frequently encountered medical emergency associated with high early mortality rates, there is a need to predict short-term outcomes for risk stratification.Our aim was to derive and validate a model, a simple clinical scoring system using baseline vital signs, clinical and presenting characteristics, and readily available laboratory tests, that allows accurate prediction of short-term mortality in individuals experiencing ACPE. This retrospective cohort study included 1088 patients with ACPE from six health centers. Subjects were randomly allocated into derivation and validation cohorts at a 4:3 ratio, facilitating comprehensive examination and validation of prognostic model. Independent predictors of mortality (p<0.05) from the multivariable model were included in the risk score. The discriminant ability of the score was tested by ROC analysis. In the derivation cohort (n=623), age, blood urea nitrogen, heart rate, intubation, anemia, and systolic blood pressure were identified as independent predictors of mortality in multivariable analysis. These variables were used to develop a risk score ranging from 0 to 6 by scoring 0 and 1. The SABIHA score provided a good calibration with a concordance index of 0.879 (95% CI: 0.821-0.937). While the probability of short-term mortality was 80.0% in the high risk group, this rate was only 3.3% in the low risk group. This score also performed well on the validation set. A simple clinical score consisting of routinely obtained variables can be used in risk stratification to predict short-term outcomes in patients with ACPE.

Machine learning validation of a simple prediction model for the correlation between advanced age and clinical outcomes in patients with aneurysmal subarachnoid hemorrhage.

Adverse effects of advanced age and poor initial neurological status on outcomes of patients with aneurysmal subarachnoid hemorrhage (SAH) have been documented. While a predictive model of the non-linear correlation between advanced age and clinical outcome has been reported, no previous model has been validated. Therefore, we created a prediction model of the non-linear correlation between advanced age and clinical outcome by machine learning and validated it using a separate cohort. Data of aneurysmal SAH patients treated by surgical clipping or endovascular coiling between 2003 and 2019 were obtained from the Japanese Stroke Databank (derivation cohort, n = 9,657) and "Predict for Outcome Study of Aneurysmal Subarachnoid Hemorrhage" (validation cohort, n = 5,085). Generalized additive models (GAMs) for poor outcome (modified Rankin Scale score ≥ 3 at discharge) were fitted with age transformation using spline curves for each World Federation of Neurological Societies grade. The discrimination property and calibration plot of unadjusted and adjusted models were assessed using the validation cohort. The derivation and validation cohorts included 3,610 and 3,251 patients, respectively. Regarding discrimination, areas under the receiver operating characteristic curve for the derivation and validation cohorts were 0.835 and 0.827, respectively, in the unadjusted model and 0.844 and 0.836, respectively, in the adjusted model. An unbiased correlation was confirmed between predicted and observed probabilities of poor outcomes. GAM could help visualize the correlation between age and clinical outcomes. Our prediction model can quantitatively aid in treatment decision-making and can be effective for most diseases and treatment settings. Trial Registration: UMIN Clinical Trials Registry (Date 2/22/2022/ ID, UMIN000046282 number, R000052809 URL, https//www.umin.ac.jp/ctr/index.htm) and the Japan Registry of Clinical Trials (Date 3/28/2022 /No. jRCT1060210092 URL, https//jrct.niph.go.jp/).

Prognostic model for esophagogastric variceal rebleeding after endoscopic treatment in liver cirrhosis: A Chinese multicenter study.

Rebleeding after recovery from esophagogastric variceal bleeding (EGVB) is a severe complication that is associated with high rates of both incidence and mortality. Despite its clinical importance, recognized prognostic models that can effectively predict esophagogastric variceal rebleeding in patients with liver cirrhosis are lacking. To construct and externally validate a reliable prognostic model for predicting the occurrence of esophagogastric variceal rebleeding. This study included 477 EGVB patients across 2 cohorts: The derivation cohort (n = 322) and the validation cohort (n = 155). The primary outcome was rebleeding events within 1 year. The least absolute shrinkage and selection operator was applied for predictor selection, and multivariate Cox regression analysis was used to construct the prognostic model. Internal validation was performed with bootstrap resampling. We assessed the discrimination, calibration and accuracy of the model, and performed patient risk stratification. Six predictors, including albumin and aspartate aminotransferase concentrations, white blood cell count, and the presence of ascites, portal vein thrombosis, and bleeding signs, were selected for the rebleeding event prediction following endoscopic treatment (REPET) model. In predicting rebleeding within 1 year, the REPET model exhibited a concordance index of 0.775 and a Brier score of 0.143 in the derivation cohort, alongside 0.862 and 0.127 in the validation cohort. Furthermore, the REPET model revealed a significant difference in rebleeding rates (P < 0.01) between low-risk patients and intermediate- to high-risk patients in both cohorts. We constructed and validated a new prognostic model for variceal rebleeding with excellent predictive performance, which will improve the clinical management of rebleeding in EGVB patients.

Establishment and validation of predictive model of ARDS in critically ill patients

BackgroundAcute respiratory distress syndrome (ARDS) is a prevalent complication among critically ill patients, constituting around 10% of intensive care unit (ICU) admissions and mortality rates ranging from 35 to 46%. Hence, early recognition and prediction of ARDS are crucial for the timely administration of targeted treatment. However, ARDS is frequently underdiagnosed or delayed, and its heterogeneity diminishes the clinical utility of ARDS biomarkers. This study aimed to observe the incidence of ARDS among high-risk patients and develop and validate an ARDS prediction model using machine learning (ML) techniques based on clinical parameters.MethodsThis prospective cohort study in China was conducted on critically ill patients to derivate and validate the prediction model. The derivation cohort, consisting of 400 patients admitted to the ICU of the Peking University Third Hospital(PUTH) between December 2020 and August 2023, was separated for training and internal validation, and an external data set of 160 patients at the FU YANG People's Hospital from August 2022 to August 2023 was employed for external validation. Least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression were used to screen predictor variables. Multiple ML classification models were integrated to analyze and identify the best models. Several evaluation indexes were used to compare the model performance, including the area under the receiver-operating-characteristic curve (AUC) and decision curve analysis (DCA). SHapley Additive ex Planations (SHAP) is used to interpret ML models.Results400 critically ill patients were included in the analysis, with 117 developing ARDS during follow-up. The final model included gender, Lung Injury Prediction Score (LIPS), Hepatic Disease, Shock, and combined Lung Contusion. Based on the AUC and DCA in the validation group, the logistic model demonstrated excellent performance, achieving an AUC of 0.836 (95% CI: 0.762–0.910). For external validation, comprising 160 patients, 44 of whom developed ARDS, the AUC was 0.799 (95% CI: 0.723–0.875), significantly outperforming the LIPS score alone.ConclusionCombining the LIPS score with other clinical parameters in a logistic regression model provides a more accurate, clinically applicable, and user-friendly ARDS prediction tool than the LIPS score alone.

Electrocardiogram-based machine learning for risk stratification of patients with suspected acute coronary syndrome.

The importance of risk stratification in patients with chest pain extends beyond diagnosis and immediate treatment. This study sought to evaluate the prognostic value of electrocardiogram feature-based machine learning models to risk-stratify all-cause mortality in those with chest pain. This was a prospective observational cohort study of consecutive, non-traumatic patients with chest pain. All-cause death was ascertained from multiple sources, including the CDC National Death Index registry. Six machine learning models were trained for survival analysis using 73 morphological electrocardiogram features (80% training with 10-fold cross-validation and 20% testing), followed by a variational Bayesian Gaussian mixture model to define distinct risk groups. The resulting classification performance was compared against the HEART score. The derivation cohort included 4015 patients (age 59 ± 16 years, 47% women). The mortality rate was 20.3% after a median follow-up period of 3.05 years (interquartile range 1.75-5.32). Extra Survival Trees outperformed other forecasting models, and the derived risk groups successfully classified patients into low-, moderate-, and high-risk groups (log-rank test statistic = 121.14, P < .001). This model outperformed the HEART score, reducing the rate of missed events by >90% with a negative predictive value and sensitivity of 93.4% and 85.9%, compared to 89.0% and 75.0%, respectively. In an independent external testing cohort (N = 3095, age 59 ± 15 years, 44% women, 30-day mortality 3.5%), patients in the moderate [odds ratio 3.62 (1.35-9.74)] and high [odds ratio 6.12 (2.38-15.75)] risk groups had significantly higher odds of mortality compared to those in the low-risk group. The externally validated machine learning-based model, exclusively utilizing features from the 12-lead electrocardiogram, outperformed the HEART score in stratifying the mortality risk of patients with acute chest pain. This may have the potential to impact the precision of care delivery and the allocation of resources to those at highest risk of adverse events.

Predicting the risk of intensive care unit admission in patients with COVID-19 presenting in the emergency room: Development and evaluation of the CROSS score.

Existing risk evaluation tools underperform in predicting intensive care unit (ICU) admission for patients with the Coronavirus Disease 2019 (COVID-19). This study aimed to develop and evaluate an accurate and calculator-free clinical tool for predicting ICU admission at emergency room (ER) presentation. Data from patients with COVID-19 in a nationwide German cohort (March 2020-January 2023) were analyzed. Candidate predictors were selected based on literature and clinical expertise. A risk score, predicting ICU admission within seven days of ER presentation, was developed using elastic net logistic regression on a northern German cohort (derivation cohort), evaluated on a southern German cohort (evaluation cohort) and externally validated on a Colombian cohort. Performance was evaluated through discrimination, calibration, and clinical utility against existing tools. ICU admission rates within seven days were 30.8% (derivation cohort, n=1295, median age 60, 38.1% female), 28.1% (evaluation cohort, n=1123, median age 58, 36.9% female), and 30.3% (Colombian cohort, n=780, median age 57, 38.8% female). The 11-point CROSS score, based on Confusion, Respiratory rate, Oxygen Saturation (with or without concurrent supplemental oxygen), and oxygen Supplementation, demonstrated good discrimination (area under the curve (AUC): 0.77 in the evaluation cohort; 0.69 in the Colombian cohort), good calibration, and superior clinical utility compared to existing tools. Mortality-predicting tools performed poorly in predicting ICU admission risk for patients with COVID-19. The calculator-free CROSS score effectively predicts ICU admission for patients with COVID-19 in the ER. Further studies are needed to assess its generalizability in other settings. Mortality-predicting tools are not recommended for ICU admission prediction.

Machine Learning for Predicting Primary Graft Dysfunction After Lung Transplantation: An Interpretable Model Study.

Primary graft dysfunction (PGD) develops within 72 h after lung transplantation (Lung Tx) and greatly influences patients' prognosis. This study aimed to establish an accurate machine learning (ML) model for predicting grade 3 PGD (PGD3) after Lung Tx. This retrospective study incorporated 802 patients receiving Lung Tx between July 2018 and October 2023 (640 in the derivation cohort and 162 in the external validation cohort), and 640 patients were randomly assigned to training and internal validation cohorts in a 7:3 ratio. Independent risk factors for PGD3 were determined by integrating the univariate logistic regression and least absolute shrinkage and selection operator regression analyses. Subsequently, 9 ML models were used to construct prediction models for PGD3 based on selected variables. Their prediction performances were further evaluated. Besides, model stratification performance was assessed with 3 posttransplant metrics. Finally, the SHapley Additive exPlanations algorithm was used to understand the predictive importance of selected variables. We identified 9 independent clinical risk factors as selected variables. Among 9 ML models, the random forest (RF) model displayed optimal performance (area under the curve [AUC] = 0.9415, sensitivity [Se] = 0.8972, specificity [Sp] = 0.8795 in the training cohort; AUC = 0.7975, Se = 0.7520, Sp = 0.7313 in the internal validation cohort; and AUC = 0.8214, Se = 0.8235, Sp = 0.6667 in the external validation cohort). Further assessments on calibration and clinical usefulness indicated the promising applicability of the RF model in PGD3 prediction. Meanwhile, the RF model also performed best in terms of risk stratification for postoperative support (extracorporeal membrane oxygenation time: P < 0.001, mechanical ventilation time: P = 0.006, intensive care unit time: P < 0.001). The RF model had the optimal performance in PGD3 prediction and postoperative risk stratification for patients after Lung Tx.

Up-front resection for hepatocellular carcinoma: Assessing futility in the preoperative setting.

We sought to develop a predictive model to preoperatively identify patients with hepatocellular carcinoma (HCC) at risk of undergoing futile upfront liver resection (LR). Patients undergoing curative-intent LR for HCC were identified from a large multi-institutional database. Futile LR was defined by death or disease recurrence within six months postoperatively. Backward logistic regression was performed to identify factors associated with futility. Additionally, binary criteria were established for surgical candidacy, aiming to keep the likelihood of futility below 20%. Among 1633 patients with HCC, 264 (16.2%) underwent futile upfront LR. Tumor burden score (TBS) (coefficient: 0.083, 95%CI: 0.067-0.099), alpha-fetoprotein (AFP) (coefficient: 0.254, 95%CI: 0.195-0.310), and albumin-bilirubin (ALBI) grade 2/3 (coefficient: 0.566, 95%CI: 0.420-0.718) were independently associated with an increased risk of futile LR. The model demonstrated strong discrimination and calibration in both derivation and validation cohorts. Low, intermediate, and high-risk groups were determined based on the risk model, each with an escalating likelihood of futility, worse histological features, and worse survival outcomes. Six distinct conditions based on AFP-adjusted-to-TBS criteria were established, all with a futility likelihood of less than 20%. Patients fulfilling these criteria had significantly better long-term recurrence-free and overall survival. The futility risk model was made available online for wide clinical applicability: (https://altaf-pawlik-hcc-futilityofsurgery-calculator.streamlit.app/). A preoperative risk model and AFP-adjusted-to-TBS criteria were developed and validated to predict the likelihood of futile LR among patients with HCC. This pragmatic clinical tool may assist clinicians in preoperative decision-making, helping them avoid futile surgery unlikely to offer long-term benefits.

Derivation and validation of a prediction model for inadequate bowel preparation in Chinese outpatients

The quality of bowel preparation is an important factor in the success of colonoscopy. However, multiple influencing factors that function together can lead to inadequate bowel preparation. The main objective of this study was to explore the specific factors that affect the quality of bowel preparation, with the goal of deriving and validating a predictive model for inadequate bowel preparation in Chinese outpatients. A prospective observational study. We conducted a prospective study in a tertiary hospital in Zhejiang Province that included elective colonoscopy outpatients treated between December 15, 2022 and August 12, 2023. Clinical data related to the patient characteristics and patient bowel preparation were collected to analyze the risk factors of inadequate bowel preparation in outpatients. The quality of bowel preparation was assessed by using the Boston bowel preparation scale (BBPS). Inadequate bowel preparation was defined as a total BBPS score of < 6 or any segment score of < 2. The predictive model was constructed based on multivariate logistic regression, and the discrimination and calibration of the prediction model were evaluated via internal and external validation; additionally, a clinical decision curve was drawn to evaluate the clinical utility of the model. A total of 1314 patients were included from December 15, 2022 through May 31, 2023 (derivation cohort, n = 1035) and from June 1 through August 12, 2023 (external validation cohort, n = 279). Inadequate bowel preparation occurred in 260 of 1035 patients in the derivation cohort (25.1%). Multivariate analysis identified that male sex (OR = 1.690, 95% CI: 1.242–2.300), diabetes (OR = 1.769, 95% CI: 1.059–2.954), constipation (OR = 2.375, 95% CI: 1.560–3.617), history of colorectal surgery (OR = 2.915, 95% CI: 1.455–5.840), a high fiber diet used at 24 h before examination (OR = 2.662, 95% CI: 1.636–4.334) and the time interval from the end of the bowel preparation to the start of the colonoscopy (PC) >5 h (OR = 2.471, 95% CI: 1.814–3.366) were independent risk factors. We derived a model to identify patients with inadequate cleansing by using data from patients in the derivation cohort and tested it in the validation cohort. The area under the curve (AUC) was 0.704 (95% CI: 0.667–0.741), with a calibration value of p = 0.632 in the derivation cohort. Bootstrap cross-validation showed a good model calibration condition. For the validation cohort, the AUC was 0.704 (95% CI: 0.628–0.779), and the calibration value was p = 0.376. We identified the influencing factors of outpatient bowel cleansing that are associated with patient clinical characteristics and bowel preparation-related behaviors. A model was constructed and validated to identify patients who were at high risk of inadequate bowel preparation by using six simple variables, which included male sex, diabetes, constipation, history of colorectal surgery, a high fiber diet used at 24 h prior to examination, and PC > 5 h. The clinical decision curve showed that the constructed prediction model has some clinical utility based on results from the derivation cohort and validation cohort.