Risk Prediction Model Research Articles

Development of a risk prediction model for sepsis-related delirium based on multiple machine learning approaches and an online calculator

Background Sepsis-associated delirium (SAD) occurs due to disruptions in neurotransmission linked to inflammatory responses from infections. It poses significant challenges in clinical management and is associated with poor outcomes. Survivors often experience long-term cognitive and behavioral issues that impact their quality of life and place a burden on their families. This study aimed to develop and validate an interpretable machine learning model for early prediction of SAD in critically ill patients. Additionally, we constructed an online risk calculator to facilitate real-time clinical assessment. Methods This study is a retrospective analysis utilizing data from 16,120 patients in the Medical Information Mart for Intensive Care IV database. To manage imbalanced data, we applied the Synthetic Minority Over-sampling Technique (SMOTE) method. Feature selection was conducted using Multivariate Logistic Regression, LASSO regression, and the Boruta algorithm. We developed predictive models using eight machine learning algorithms and selected the best one for validation. The SHapley Additive exPlanations (SHAP) method was used for visualization and interpretation, enhancing the clinical understanding of the model, alongside the creation of an online web calculator. Results We combined three feature selection methods to identify 17 key features for our machine learning prediction model. The Gradient Boosting Machine (GBM) model demonstrated excellent calibration and strong predictive accuracy in the validation cohort. The SHAP feature importance ranking revealed five critical risk factors for predicting outcomes: Glasgow Coma Scale (GCS), ICU stay duration, chloride, sodium, and Sequential Organ Failure Assessment (SOFA). Based on this optimal model, we successfully developed an online web calculator. Conclusion We developed and validated a machine learning model capable of accurately predicting SAD with high clinical applicability. The integration of interpretable machine learning and an online calculator offers a practical tool to support early identification and timely management of SAD in critically ill patients.

In-silico trials of targeted screening for abdominal aortic aneurysms using linked healthcare data: A study protocol

Background The NHS abdominal aortic aneurysm (AAA) screening programme (NAAASP) is both clinically and economically effective. One of the main determinants of this effectiveness is disease prevalence. AAA prevalence is decreasing over time, steadily reducing the efficiency of the current NAAASP screening policy. One alternative to whole population screening is targeted screening of high-risk groups. Whether this would detect a clinically and publicly acceptable proportion of disease, and whether it would improve cost-effectiveness are unknown. The aim of this research is to estimate the clinical outcomes and cost-effectiveness of targeted AAA screening. Methods Rather than conducting an expensive and time-consuming randomized trial to directly test targeted screening, we will undertake in-silico trials of targeted AAA screening. To determine success criteria for in-silico trials, the ethics and issues around the acceptability of targeted screening will first be explored through focus groups and interviews. A qualitative evidence synthesis to identify issues associated with targeted screening will be used to establish themes and topic guides. To perform the in-silico trials, individual men’s outcomes from the NAAASP (2013–2024, ≈ 2,500,000 men, ≈ 1% with AAA) will be linked to primary care data from the Clinical Practice Research Datalink (CPRD) (20% overlap of records). Risk factors for AAA will be identified by developing a risk prediction model and used as targeted screening criteria in in-silico trials, with diagnostic accuracy as the primary outcome. A discrete event simulation model will be adopted to extrapolate the trial findings beyond the observed period. We will estimate the clinical and cost-effectiveness of targeted screening compared with the current whole population screening strategy. Data linkage will be undertaken under Health Research Authority Confidentiality Advisory Group (Section 251) approval. Linked data will be effectively anonymised. Participants in the qualitative substudy will provide informed consent for participation. Discussion We expect this project to have a direct and significant impact on NHS, UK and worldwide AAA screening policies. The study findings will be submitted for publication in peer-reviewed journals and presented at scientific meetings.

Application value of dual-sequence MRI based nomogram of radiomics and morphologic features in predicting tumor differentiation degree and lymph node metastasis of Oral squamous cell carcinoma

BackgroundOral squamous cell carcinoma is a highly invasive tumor. The degree of histological differentiation and lymph node metastasis are important factors in the treatment and prognosis of patients. There is a lack of non-invasive and accurate preoperative risk prediction model in the existing clinical work.ObjectiveThis study sought to develop and validate a combined model including MRI radiomics and morphological analysis to predict lymph node metastasis and degree of tumor differentiation prior to surgical intervention for oral squamous cell carcinoma (OSCC).MethodsThis study retrospectively included 119 patients which were divided into a training cohort (n=83) and a validation cohort (n=36). To predict lymph node metastasis (LNM) and degree of tumor differentiation, both univariate and multivariate analyses were performed to identify significant features and develop morphological prediction models. Radiomics features were extracted from T2-FS and DWI sequences, followed by feature selection and the establishment of Rad-scores using the LASSO method. Two nomograms was constructed by integrating MRI morphological features with radiomics features. The performance of the models was assessed using the AUC and the Delong test. Calibration curves and DCA were employed to further evaluate the models’ practical applicability.ResultsNine radiomics features were selected to develop the Rad-scores. The morphological features for predicting LNM are depth of invasion and tumor thickness. The morphological features for predicting the degree of tumor differentiation are ADC value and intratumoral necrosis.In the validation cohort, the nomogram for predicting LNM achieved an area under the curve (AUC) of 0.90 (95% CI: 0.84, 0.97), while the nomogram for tumor grade prediction achieved an AUC of 0.87 (95% CI: 0.76, 0.98), demonstrating excellent diagnostic performance. Calibration curve and decision curve further confirmed the accuracy of nomograms prediction.ConclusionNomograms derived from MRI radiomics and morphological characteristics offer a noninvasive and precise method for predicting degree of tumor differentiation and LNM in OSCC preoperatively. The combined model is an accurate risk prediction model with good clinical benefits and prediction accuracy.

Construction and validation of a machine learning model to predict the risk of nasopharyngeal carcinoma using multimodal clinical data: a single-center, retrospective study.

Early detection and treatment of nasopharyngeal carcinoma (NPC) are critical for improving patient prognosis. The aim of this study is to develop and compare multiple machine learning (ML) models using multimodal clinical data to identify a predictive model for NPC risk, increase diagnostic accuracy, and guide personalized treatment strategies. Clinical data were retrospectively collected from 1337 patients suspected of having NPC at the First People's Hospital of Yulin. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) regression. Patients were divided into training and test sets (80:20 ratio), and seven ML models were developed based on the training set. Model performance was assessed using metrics such as the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. The best-performing model was further evaluated through decision curve analysis (DCA), calibration, and learning curves. SHapley Additive exPlanations (SHAP) were used to interpret key clinical features. Seven models were developed using 17 clinical features selected from 53 parameters. The gradient boosting decision tree (GBDT) model demonstrated superior performance (AUC of 0.95 in the training cohort and 0.82 in the validation cohort). Calibration curves and DCA confirmed the model's strong accuracy and clinical benefit. SHAP analysis revealed that age, lymphocyte percentage, serum albumin, sex, and EBV IgM were the five most significant predictors of NPC risk. The GBDT-based ML model, using multimodal clinical data, accurately identifies patients at high risk for NPC, providing a valuable tool for early screening and personalized treatment strategies.

Methodological Considerations for a Risk Model Adopted into the Chronic Disease Prevention Policy of Taiwan. Comment on Chang et al. Developing and Validating Risk Scores for Predicting Major Cardiovascular Events Using Population Surveys Linked with Electronic Health Insurance Records. Int. J. Environ. Res. Public Health 2022, 19, 1319

Chang, H.-Y. et al. (2022) developed a risk prediction model for major adverse cardiovascular events (MACEs), coronary heart disease (CHD), and stroke using nationwide claims data retrieved from the Taiwan National Health Insurance (NHI) records [...]

Lactylation prognostic signature identifies DHCR7 as a modulator of chemoresistance and immunotherapy efficacy in bladder cancer

BackgroundBladder cancer (BLCA), the 10th most common cancer worldwide, presents a worsening prognosis as the disease progresses. Reliable tools for predicting BLCA prognosis and treatment efficacy remain urgently needed.MethodsExpression profiles of lactylation related genes were analyzed utilizing the Cancer Genome Atlas (TCGA) database and BLCA data from the GSE13507 dataset. Two distinct clusters were identified through unsupervised clustering analysis. Lactylation associated gene signatures were established and subsequently validated using training cohort and different validation cohorts. Immune cell infiltration patterns and drug response profiles were systematically evaluated. Parallel analyses of lactylation related genes were conducted at the single-cell resolution. A series of in vivo and in vitro experiments were subsequently performed to validate the findings.ResultsWe examined the mRNA expression profiles of 22 lactylation related genes in BLCA tissues. Through comprehensive analysis, we identified two distinct lactylation clusters that exhibited significantly different clinical outcomes and tumor immune microenvironment characteristics. Building upon these findings, we subsequently stratified patients into two molecular subtypes according to the lactylation clusters and established a robust genetic signature for predicting survival outcomes in BLCA patients. The lactylation risk score showed a strong connection with survival outcomes and correlated with the tumor microenvironment (TME) immunosignature and predicted immunotherapy efficacy. DHCR7 emerged as a pivotal prognostic gene from the nine gene model, prompting subsequent focused analyses. Single-cell analysis confirmed that DHCR7 reached peak expression in tumor epithelial cells, whereas TCGA data and single-cell data demonstrated strong associations between DHCR7 and diverse immune-cell populations. For the first time, we identified that knockdown of DHCR7 enhances the efficacy of both cisplatin chemotherapy and immunotherapy, highlighting DHCR7 as a key player in cisplatin resistance and its influence on immunotherapy effectiveness in BLCA. These findings offer valuable insights into potential combined therapeutic strategies.ConclusionsWe developed a robust lactylation risk prediction model for accurately forecasting BLCA prognosis and identified DHCR7 as a pivotal biomarker involved in cisplatin resistance and influencing immunotherapy efficacy in BLCA.

Validation of the HFA-ICOS Score for Carfilzomib-Induced Cardiotoxicity in Multiple Myeloma: A Real-Life Perspective Study

Background: Despite the inference about the cardiotoxicity induced by Carfilzomib, no validated risk prediction models for adverse cardiovascular events in a real-life population are available. Objectives: The aim of this study was to evaluate the performance of the risk stratification score for Carfilzomib-induced cardiotoxicity of the Heart Failure Association of the European Society of Cardiology and the International Cardio-Oncology Society (HFA-ICOS) in patients with multiple myeloma (MM). Methods: This is a prospective, real-world study including MM patients consecutively enrolled prior to starting Carfilzomib, divided into levels of risk according to the HFA-ICOS proforma. Results: Of 169 patients, 11.8% were classified as ‘low risk’, 38.5% as ‘medium risk’, 45.6% as ‘high risk’ and 4.1% as ‘very high risk’ at baseline. A total of 89 (52.7%) patients experienced one of the following events: 36 (21.3%) had at least one cardiovascular event and 77 (45.6%) had almost one hypertension-related event. No significant differences were observed for the incidence of any cardiovascular events between the different levels of risk (p > 0.05), even considering the HFA-ICOS score as a continuous variable. The integration of the score with the baseline systolic blood pressure and pulse wave velocity enhanced the accuracy of the score (AUC 0.557 vs. 0.736). Conclusions: The HFA-ICOS score did not discriminate between patients at low, medium and high risk, showing a limited discriminatory power in predicting the risk of events in our population. The integration of other parameters in the HFA-ICOS score, such as systolic blood pressure and pulse wave velocity, improved the performance of the score.

Prediction Models for Health Care Workers' Exposure to Type II Workplace Violence: A Systematic Review and Meta-Analysis.

Workplace violence poses a serious threat to safety and well-being of health care workers. The purpose of this systematic review was to assess the accuracy and applicability of predictive models for workplace violence risk among health care workers. Ten databases were searched through May 2025. The Prediction Model Risk of Bias Assessment Tool was used to evaluate model quality. Predictors were classified using the Job Demands-Resources framework. Ten studies reporting 18 models were included. The pooled area under the curve was 0.87 (95% confidence interval [CI], 0.81-0.93). Predictors were categorized into 3 main categories and 7 subcategories. Current workplace violence risk models lack clinical utility; future research must strengthen rigor and validation for practical application.

Development of an early mortality risk prediction model for pediatric patients with secondary hemophagocytic lymphohistiocytosis.

To establish a visual prediction model for the risk of early mortality in pediatric patients with secondary hemophagocytic lymphohistiocytosis (sHLH). A retrospective analysis was conducted on 128 pediatric sHLH cases diagnosed at the Affiliated Hospital of Zunyi Medical University between January 2012 and April 2023. Logistic regression analysis was employed to identify prognostic factors for pediatric patients with sHLH. A nomogram prediction model was constructed using R software. The receiver operating characteristic (ROC) curve, calibration curve, decision clinical curve (DCA), and clinical impact curve (CIC) were plotted to evaluate the discriminative power and clinical application value of the prediction model. Among the 128 pediatric patients with sHLH included in the study, 90 (70.31%) were associated with infections, and 37 (28.9%) died within 30days after admission. In the fully adjusted model, central nervous system involvement (CNSI) (OR = 9.496, 95% CI = 2.965-30.410), prognostic nutrition index (PNI) (OR = 0.931, 95% CI = 0.872-0.994), and activated partial thromboplastin time (APTT) (OR = 1.029, 95% CI = 1.007-1.052) were identified as independent risk factors affecting early mortality risk in pediatric patients with sHLH, while the use of blood purification combined with HLH-94/2004 treatment (OR = 0.097, 95% CI = 0.024-0.395) was an independent protective factor. Based on these independent prognostic factors, a nomogram prediction model was constructed. The ROC curve (AUC = 0.880) and calibration curve (χ2 = 10.12, P = 0.257) demonstrated good discriminability and fit. The DCA curve and CIC curve indicated that the model had good clinical applicability. The nomogram model has good discriminability and accuracy in predicting the risk of early death in pediatric patients with sHLH.

From dosimetry to deep learning: personalized risk prediction models for radiation pneumonitis

Abstract Radiation pneumonitis (RP) is a common complication following radiotherapy for thoracic tumors, significantly impacting treatment efficacy and patient quality of life. Identifying and predicting risk factors for RP has become a key research focus. This study aims to summarize current knowledge by analyzing previously published studies and large clinical trials. A systematic literature search was conducted in Embase, PubMed, and Web of Science for publications up to November 1, 2024, using keywords such as “radiation pneumonitis”, “risk factors”, “machine learning”, etc. Inclusion criteria prioritized clinical relevance and methodological rigor. Identified RP-related factors include radiation dose parameters (e.g., V20, mean lung dose [MLD]), clinical characteristics (e.g., age, interstitial lung disease), inflammatory markers (e.g., IL-6, neutrophil-to-lymphocyte ratio [NLR]), and features from imaging and multi-omics analyses. In addition, traditional dosimetric indicators remain central, while recent advances integrate radiomics and artificial intelligence (AI)-driven models to improve predictive accuracy. Despite progress, challenges such as limited sample sizes, lack of standardization, and insufficient multi-center validation persist. Future efforts should prioritize data integration, model optimization, and clinical translation to better predict RP risk and guide individualized interventions.

Machine learning-based risk prediction models for bronchopulmonary dysplasia in preterm infants: a high-altitude cohort study.

Bronchopulmonary dysplasia (BPD) is a significant cause of morbidity in preterm infants, yet its development and severity at high altitudes (>1500 m) remain poorly understood. This study aimed to identify altitude-specific risk factors and develop robust, interpretable predictive models for BPD in this unique population. In this retrospective matched cohort study, 378 preterm infants (<32 weeks gestation, <1500 g birth weight) admitted to a high-altitude (1500 m) NICU(Neonatal Intensive Care Unit) between 2019 and 2023 were analysed. The cohort included 189 BPD cases (91 mild, 61 moderate, 37 severe) and 189 matched controls. Maternal, perinatal and postnatal data were collected. Machine learning models (XGBoost, logistic regression, random forest) were developed and rigorously evaluated using comprehensive performance metrics to predict BPD occurrence and severity. SHAP (SHapley Additive exPlanations) analysis was employed to interpret the best-performing model. Key risk factors for BPD development included maternal hypertension (OR 2.31, 95% CI 1.56 to 3.42), initial oxygen requirement >30% (OR 3.15, 95% CI 2.13 to 4.65) and lack of exclusive breast milk feeding (OR 1.89, 95% CI 1.28 to 2.79). Severe BPD was independently associated with prolonged invasive ventilation (>7 days) (OR 4.12, 95% CI 2.78 to 6.11), elevated C reactive protein (>10 mg/L) (OR 2.87, 95% CI 1.93 to 4.26) and patent ductus arteriosus (OR 2.53, 95% CI 1.71 to 3.74). Machine learning models demonstrated strong predictive performance; the optimal XGBoost model achieved an area under the curve of 0.89 (95% CI 0.85 to 0.93), an F1 score of 0.82, a Matthews Correlation Coefficient of 0.73 and a balanced accuracy of 0.85. SHAP analysis identified initial FiO2 >30%, mechanical ventilation and maternal hypertension as the top three most influential predictors for the XGBoost model. This study provides the first comprehensive analysis of BPD risk factors at a specific high altitude and validates effective, interpretable machine learning models for its prediction. These findings highlight the critical importance of altitude-specific adjustments in risk assessment and emphasise the potential for model-guided early interventions to improve outcomes for this vulnerable population.

Construction and verification of a risk prediction model for patients with kinesophobia after cerebral hemorrhage surgery

ObjectiveTo establish a risk prediction model of kinesophobia in patients after cerebral hemorrhage surgery and verify its effect.MethodsA total of 218 patients after cerebral hemorrhage surgery were selected, and the differences in clinical data between kinesophobia patients and non-kinesophobia patients were analyzed. Using 20 indexes as independent variables, the characteristic variables were screened by LASSO regression, and then multivariate Logistic regression analysis was carried out. Based on the results, the nomogram prediction model was constructed, and the model was verified from the aspects of clinical applicability, discrimination, and calibration.ResultsSignificant differences were found in age, electronic health literacy score, depression score, NIHSS score, VAS pain score, intraoperative blood loss, and anxiety score between patients with phobia and non-phobia (P < 0.05). 12 characteristic variables were selected by LASSO regression. Multivariate Logistic regression analysis showed that age, NIHSS score, VAS pain score and depression score were independent risk factors for the occurrence of kinesophobia after cerebral hemorrhage surgery (OR > 1 and P < 0.05), and electronic health literacy score was an independent protective factor (OR < 1 and P < 0.05). Based on age, NIHSS score, VAS pain score, e-health literacy score, and depression score, a nomogram prediction model was constructed. The DCA curve shows that the model has the highest clinical net benefit when the threshold probability is between 0.14 and 0.99, indicating good clinical applicability. The area under the ROC curve (AUC) is 0.836(95% CI: 0.782–0.890), which indicates good discrimination. Spiegelhalter’s z test and the calibration curve show that the calibration degree is good, and the C statistic after Bootstrap self-sampling internal verification is 0.820 (95% CI: 0.772–0.877), indicating that the prediction is robust.ConclusionThe nomogram prediction model of the risk of kinesophobia after cerebral hemorrhage based on multivariate regression analysis has a good prediction effect, which can provide reference for the clinical prevention of kinesophobia after cerebral hemorrhage.

Machine learning-based risk prediction model for central nervous system involvement in diffuse large B-cell lymphoma.

Accurate prediction of CNS relapse in DLBCL remains challenging despite existing models like IPI and CNS-IPI. This study aimed to develop a machine learning (ML)-based prognostic model. A retrospective cohort of 664 R-CHOP-treated DLBCL patients was analyzed; 44 (6.6%) experienced CNS relapse at a median of 9.3 months. ML models, including Random Survival Forests (RSF) and Gradient Boosting Machines (GBM), were developed and validated using the entire cohort (n = 664), irrespective of CNS relapse. RSF demonstrated high discriminative ability (C-index: 0.91) and low prediction error (Integrated Brier Score [IBS]: 0.057), while GBM yielded comparable performance (C-index: 0.88, IBS: 0.042), both outperforming traditional scores such as IPI and CNS-IPI. Key predictors included extranodal site number, high-risk organ involvement, and ECOG performance status, although ECOG lost significance in Fine and Gray competing risks analysis, likely due to early mortality. ML-based models offer enhanced predictive accuracy and support personalized CNS risk assessment in DLBCL.

Liver Retransplantation: Identifying Factors and Developing a Risk Prediction Model to Predict Futility in the Modern Era.

Liver retransplantation (rLT) results have traditionally been inferior compared with those of primary liver transplantation. Understanding the risks and anticipated outcomes is essential for patient counseling and obtaining informed consent. Using the Scientific Registry of Transplant Recipients database, we analyzed a large cohort of adult rLT cases in the contemporary era to identify variables associated with posttransplant outcomes (with a focus on 1- and 5-y survival). Model predictions were made with random survival forests, a machine learning approach integrated into survival analysis. The difference in the out-of-bag C-index between a model with and without the variable was used to define variable importance. A prospective holdout cohort was used to validate the model predictions. Of the 3774 patients studied, the overall adjusted 1- and 5-y patient survival rates increased from 76.3% and 63.0%, respectively, for those transplanted in 2010, to 81.1% and 67.4% in 2019, then decreased to 78.0% and 63.9%, respectively, in 2022. The most important predictors of posttransplant mortality include recipient characteristics (being on life support before transplant, number of previous liver transplants, age, body mass index, and Karnofsky score) and donor organ characteristics (cold ischemia time and donor age). In a prospective validation cohort stratified into risk tertiles, the high-risk group had significantly lower 1-y survival (63.7%) compared with medium-risk (83.2%) and low-risk (88.7%, P < 0.001) groups. We developed a user-friendly online application using recipient and donor characteristics to predict 1- and 5-y survival. The study model could be used as an additional tool to predict 1- and 5-y patient survival to help counsel prospective rLT candidates and guide donor selection in this technically challenging recipient group.

Development and validation of personalized risk prediction models for patients with IgA nephropathy: anationwide multicenter cohort study.

Effective prediction of immunoglobulin A nephropathy (IgAN) progression is crucial for early intervention and management. We aimed to develop and validate distinct IgAN prediction models for clinical and researchapplications. We analyzed data from the Japanese Nationwide Retrospective Cohort Study in IgAN (n = 1174) gatheredover 10years. The models were developed and tested using data from general physicians in primary care, specialists in tertiary care hospitals, and researchers at academic research institutes. Three tailored prediction models (Primary Care, Tertiary Care, and Research Institute Models) were created to address the unique needs of different clinical environments. The primary outcome was a composite renal event defined as a 1.5-fold increase in serum creatinine level or progression to kidney failure. The predictive performance was assessed using C-statistics. In the derivation cohort, the primary care model included predictors such as estimated glomerular filtration rate < 45mL/min/1.73 m2, proteinuria ≥ 0.5g/day, and non-use of corticosteroids, achieving a C-statistic of 0.796 (95% confidence interval [CI] 0.686-0.895). The tertiary care model showed a C-statistic of 0.807 (95% CI 0.713-0.886), using predictors such as glomerular number and histological severity. The research institute model, incorporating 38 variables, demonstrated a C-statistic of 0.802 (95% CI 0.686-0.906). The prediction models for primary and tertiary care settings provided effective tools for forecasting renal outcomes in IgAN patients and are competitive with more complex machine learning-based models used in research. These models can help guide clinical decisions in various healthcare settings.

Blood Biomarker-Based Machine Learning Model for Predicting Cognitive Impairment in Stroke Patients.

Blood Biomarker-Based Machine Learning Model for Predicting Cognitive Impairment in Stroke Patients.

Risk factors analysis and prediction model construction of hospital-acquired pneumonia after traumatic brain injury

BackgroundHospital-acquired pneumonia (HBP) is a common and serious infections disease that affects the patients with traumatic brain injury (TBI). Severe pneumonia can lead to high mortality and morbidity in TBI patients. Therefore, it is important to investigate the risk factors and develop a prediction model for HBP following TBI.MethodsThe clinical data of 285 patients with TBI, admitted to Shanxi Bethune Hospital and Shanxi Provincial People’s Hospital, were collected. Patients were divided into two groups based on the presence or absence of pneumonia. Risk factors for HBP were identified, a predictive model was constructed, and its performance was validated.ResultsSignificant differences were observed between the pneumonia and non-pneumonia groups regarding several factors, including age, history of diabetes, smoking history, white blood cell count, platelet count, albumin levels, Glasgow Coma Scale (GCS) score upon admission, thoracic trauma, craniocerebral surgery, and the need for tracheal intubation post-admission (p &amp;lt; 0.05). Among these, age, smoking history, thoracic trauma, white blood cell count, albumin levels, and admission GCS score were identified as independent risk factors for HBP following TBI. The predictive model based on these six factors demonstrated high accuracy.ConclusionAge, smoking history, thoracic trauma, white blood cell count, albumin levels, and admission GCS score are independent risk factors for HBP after TBI. The predictive model developed based on these factors shows strong predictive accuracy and clinical utility.

Development of cardiovascular and all-cause mortality risk prediction models for maintenance hemodialysis patients based on metabolomics

IntroductionThe outcome of maintenance hemodialysis (MHD) remains poor, with cardiovascular death accounting for more than half of all-cause death cases. In this study, cardiovascular mortality and all-cause mortality prediction models were developed to investigate the predictive role of metabolites in MHD patients.MethodsClinical and metabolomics data of 135 hemodialysis patients from a single center were collected with a 6-year follow-up. Univariate Cox regression and random forest were respectively applied to preliminarily screen clinical and metabolomics characteristics, followed by multivariate Cox regression for identifying features predicting cardiovascular or all-cause mortality. Multivariate Cox proportional regression risk models were constructed using clinical, metabolomics, and combined features. Subgroup survival differences were compared via risk score stratification.ResultsThe combined model showed significant superiority in predicting cardiovascular mortality (3-year AUC = 0.901, 5-year AUC = 0.876), surpassing the clinical-only model (0.868/0.826) and metabolomics-only model (0.659/0.641). For all-cause mortality, the combined model demonstrated modest improvement (0.859/0.834) but still outperformed the metabolomics model (0.534/0.653). Thirty 5-fold cross-validations confirmed stable performance. High-risk groups had significantly higher cumulative mortality than low-risk groups (p < 0.0001).ConclusionThe metabolomics-alone model showed limited predictive performance, but its synergistic integration with clinical indicators further improved the predictive performance of mortality risk models, particularly for cardiovascular mortality.

Retraining the veterans health administration’s REACH VET suicide risk prediction model for patients involved in the legal system

Although patients with criminal legal system involvement have among the highest rates of suicide, the model that identifies patients at high risk of suicide at the United States Veterans Health Administration (VHA) does not include predictors specific to criminal legal system involvement. We explored whether the model’s predictive ability would be improved (1) by retraining the model for legal-involved veterans and (2) by adding additional predictors associated with legal-involvement. For a combined outcome of suicide attempt or suicide death, the retrained models showed a positive predictive value (PPV) of 0.124 and false negative rate (FNR) of 0.527. Adding additional predictors associated with being legal-involved did not improve predictive accuracy. Retraining the VHA suicide risk prediction model for legal-involved patients improves the model’s predictive ability for this group of high-risk patients, more so than adding predictors associated with being legal-involved. A similar approach for other high-risk patients is worth exploring.

Construction and validation of a predictive model for the efficacy of valproic acid monotherapy in epilepsy based on Lasso-logistic regression.

Valproic acid (VPA) is a broad-spectrum antiepileptic drug; but its therapeutic efficacy varies significantly among individuals. The objective of this study is to identify the specific biomarkers that can predict the efficacy of VPA. The GSE143272 dataset from the Gene Expression Omnibus (GEO) was utilized to identify Differentially Expressed Genes (DEGs) between responders and non-responders to VPA. Weighted Gene Co-expression Network Analysis (WGCNA) was employed to identify genes related to the non-responder phenotype. Intersection genes were selected to obtain the core genes affecting VPA tolerance. Lasso regression was applied to determine the core genes that influence the VPA effect. Lasso regression was applied to screen these core genes, using their expression values as independent variables and VPA response as the dependent variable in constructing a univariate logistic regression model. Peripheral blood samples from epileptic patients treated solely with VPA were collected according to nano-discharge standard. The expression levels of target genes were determined by qPCR to validate the accuracy of the model. 86 genes were closely related to the response phenotype through WGCNA. 13 intersection genes were obtained by intersection with 97 DEGs, which mainly involve mRNA splicing function and transport pathway. Ultimately, 3 genes-NELL2, SNORD3A and mir-1974 were included in the final model. The Area Under Curve (AUC) for this predictive model was found to be 0.70 (95 % CI: 0.70). qPCR analysis revealed a significant difference in the relative expression of the SNORD3A gene between the responder and non-responder groups. Epilepsy patients are at an increased risk of developing drug resistance when undergoing VPA monotherapy. The risk prediction model based on Lasso-Logistic regression demonstrates strong predictive capabilities. The SNORD3A gene may serve as a valuable biomarker for predicting the likelihood of VPA resistance.