Area Under The Receiver Operating Characteristic Research Articles

Identifying the need for infection-related consultations in intensive care patients using machine learning models

Infection-related consultations on intensive care units (ICU) have a positive impact on quality of care and clinical outcome. However, timing of these consultations is essential and to date they are typically event-triggered and reactive. Here, we investigate a proactive approach to identify patients in need for infection-related consultations by machine learning models using routine electronic health records. Data was retrieved from a mixed ICU at a large academic tertiary care hospital including 9684 admissions. Infection-related consultations were predicted using logistic regression, random forest, gradient boosting machines, and long short-term memory neural networks (LSTM). Overall, 7.8% of admitted patients received an infection-related consultation. Time-sensitive modelling approaches performed better than static approaches. Using LSTM resulted in the prediction of infection-related consultations in the next clinical shift (up to eight hours in advance) with an area under the receiver operating curve (AUROC) of 0.921 and an area under the precision recall curve (AUPRC) of 0.541. The successful prediction of infection-related consultations for ICU patients was done without the use of classical triggers, such as (interim) microbiology reports. Predicting this key event can potentially streamline ICU and consultant workflows and improve care as well as outcome for critically ill patients with (suspected) infections.

Feasibility of a clinical-radiomics combined model to predict the occurrence of stroke-associated pneumonia

PurposeTo explore the predictive value of radiomics in predicting stroke-associated pneumonia (SAP) in acute ischemic stroke (AIS) patients and construct a prediction model based on clinical features and DWI-MRI radiomics features.MethodsUnivariate and multivariate logistic regression analyses were used to identify the independent clinical predictors for SAP. Pearson correlation analysis and the least absolute shrinkage and selection operator with ten-fold cross-validation were used to calculate the radiomics score for each feature and identify the predictive radiomics features for SAP. Multivariate logistic regression was used to combine the predictive radiomics features with the independent clinical predictors. The prediction performance of the SAP models was evaluated using receiver operating characteristics (ROC), calibration curves, decision curve analysis, and subgroup analyses.ResultsTriglycerides, the neutrophil-to-lymphocyte ratio, dysphagia, the National Institutes of Health Stroke Scale (NIHSS) score, and internal carotid artery stenosis were identified as clinically independent risk factors for SAP. The radiomics scores in patients with SAP were generally higher than in patients without SAP (P < 0. 05). There was a linear positive correlation between radiomics scores and NIHSS scores, as well as between radiomics scores and infarct volume. Infarct volume showed moderate performance in predicting the occurrence of SAP, with an AUC of 0.635. When compared with the other models, the combined prediction model achieved the best area under the ROC (AUC) in both training (AUC = 0.859, 95% CI 0.759–0.936) and validation (AUC = 0.830, 95% CI 0.758–0.896) cohorts (P < 0.05). The calibration curves and decision curve analysis further confirmed the clinical value of the nomogram. Subgroup analysis showed that this nomogram had potential generalization ability.ConclusionThe addition of the radiomics features to the clinical model improved the prediction of SAP in AIS patients, which verified its feasibility.

MUCIN MRNA ISOFORM SIGNATURES AS POTENTIAL NOVEL BIOMARKERS TO EVALUATE DISEASE STATUS IN PATIENTS WITH INFLAMMATORY BOWEL DISEASES

Abstract Mucosal barrier dysfunction and aberrant mucin expression are major hallmarks in the pathophysiology of IBD. Mucins are highly polymorphic, and the presence of genetic differences can alter gene expression, resulting in several mRNA isoforms via alternative splicing. While most isoforms encode similar biological functions, others alter protein function, potentially resulting in progression towards disease. Currently, little attention has been given to the importance of mucin mRNA isoforms in IBD. The aim of our study is to investigate the potential of mucin mRNA isoforms as novel biomarkers for the evaluation of IBD activity and subtypes. To obtain this goal, RNA was extracted from colonic and terminal ileal biopsies of IBD patients that underwent an endoscopy at the Antwerp University Hospital (UZA). Additionally, patients without a history of IBD undergoing an endoscopy due to a positive iFOBT which show no endoscopic abnormalities, were included as controls. Library preparation was performed with the PacBio Iso-Seq multiplex protocol adapted for targeted transcriptome sequencing. Targeted capture was accomplished by using a custom-designed pool of probes, developed for the capture of all mucin gene transcripts. Samples were sequenced on the PacBio Sequel platform at the University of Antwerp. The data was analyzed by using the isoseq3-pipeline and additional filtering with SQANTI3 was performed. In total 106 biopsies were sequenced on the PacBio platform. The resulting intestinal mucin transcriptome was merged with the human reference transcriptome. On this combined mucin transcriptome Illumina bulk RNA sequencing data from over 2000 intestinal biopsies (GEO dataset GSE193677) were mapped to determine mucin isoform expression. An external dataset (GEO dataset GSE165512) was used for additional validation. A classification random forest was trained on this data to distinguish inflamed IBD from non-inflamed control patients based on the mucin isoform expression alone. The model performed well on train and test datasets but decreased in the external validation (Table 1). Dividing the samples based on disease phenotype greatly increases performance on the external validation dataset (Table 1). When only training on ileal biopsies, the model proved to be excellent in distinguishing Crohn’s disease patients from controls (Table 1). Classification of inflamed ulcerative colitis from control patients based on only the biopsies from the distal colon was similar to the latter (Table 1). Our machine learning model was able to distinguish Crohn’s disease from control patients and ulcerative colitis from control patients based on mucin mRNA isoform expression. In addition, our data suggests that the mucin isoform expression may vary between different regions within the colon. Table 1 Area under the receiver operator curve (AUCROC) from a random forest classifier trained on mucin isoform expression of inflamed biopsies of IBD and non-inflamed control patients.

Relapse following withdrawal of D-penicillamine from combination (D-penicillamine + zinc) therapy in hepatic Wilson disease.

Long-term D-penicillamine (D-pen) therapy in Wilson disease (WD) has numerous adverse effects which advocates its withdrawal, but with an inherent risk of relapse. This prospective observational study was conducted with the objective of evaluating incidence of relapse following withdrawal of D-pen from combination (D-pen + zinc) therapy in maintenance phase of previously symptomatic hepatic WD. Hepatic WD patients <18 years of age and on combination therapy for >2 years with 6 months of biochemical remission were included. Biochemical remission was defined as achievement of (i) aspartate aminotransferase (AST) and alanine aminotransferase (ALT)≤1.5 times upper limit of normal (ULN), (ii) serum albumin >3.5 g/dL, international normalized ratio (INR) <1.5 and (iii) 24-h urinary copper excretion (UCE) <500 mcg/day, nonceruloplasmin-bound-copper (NCC) <15 mcg/dL. After D-pen withdrawal, monthly liver function test (LFT) and INRand 3 monthly UCE and NCCwere done till 1 year or relapse (elevation of AST/ALT/both >2 times ULN or total bilirubin >2 mg/dL), whichever occurred earlier. Forty-five patients enrolled with median combination therapy duration of 36 months. Sixty percent of them had their index presentation as decompensated cirrhosis. Fourteen patients (31.8%) relapsed (cumulative incidence: 4 at 3 months, 11 at 6 months, and 14 at 12 months after D-pen discontinuation). All relapsers had index presentation as decompensated cirrhosis. On Cox-regression, ALT at D-pen withdrawal was an independent predictor of relapse (hazard ratio [HR]: 1.077, 95% confidence interval [CI]: 1.014-1.145, p = 0.017) with area under the receiver operating characteristic (AUROC) of 0.860. ALT ≥40 U/L predicted risk of relapse with 85.7% sensitivity, 70.9% specificity. Incidence of relapse after withdrawal of D-pen from combination therapy is 31.8% in hepatic WD. ALT ≥40 U/L, at the time of D-pen stoppage, predicts future relapse.

Language model enables end-to-end accurate detection of cancer from cell-free DNA.

We present a language model Affordable Cancer Interception and Diagnostics (ACID) that can achieve high classification performance in the diagnosis of cancer exclusively from using raw cfDNA sequencing reads. We formulate ACID as an autoregressive language model. ACID is pretrained with language sentences that are obtained from concatenation of raw sequencing reads and diagnostic labels. We benchmark ACID against three methods. On testing set subjected to whole-genome sequencing, ACID significantly outperforms the best benchmarked method in diagnosis of cancer [Area Under the Receiver Operating Curve (AUROC), 0.924 versus 0.853; P < 0.001] and detection of hepatocellular carcinoma (AUROC, 0.981 versus 0.917; P < 0.001). ACID can achieve high accuracy with just 10000 reads per sample. Meanwhile, ACID achieves the best performance on testing sets that were subjected to bisulfite sequencing compared with benchmarked methods. In summary, we present an affordable, simple yet efficient end-to-end paradigm for cancer detection using raw cfDNA sequencing reads.

Subtle variation in sepsis-III definitions markedly influences predictive performance within and across methods

Early detection of sepsis is key to ensure timely clinical intervention. Since very few end-to-end pipelines are publicly available, fair comparisons between methodologies are difficult if not impossible. Progress is further limited by discrepancies in the reconstruction of sepsis onset time. This retrospective cohort study highlights the variation in performance of predictive models under three subtly different interpretations of sepsis onset from the sepsis-III definition and compares this against inter-model differences. The models are chosen to cover tree-based, deep learning, and survival analysis methods. Using the MIMIC-III database, between 867 and 2178 intensive care unit admissions with sepsis were identified, depending on the onset definition. We show that model performance can be more sensitive to differences in the definition of sepsis onset than to the model itself. Given a fixed sepsis definition, the best performing method had a gain of 1–5% in the area under the receiver operating characteristic (AUROC). However, the choice of onset time can cause a greater effect, with variation of 0–6% in AUROC. We illustrate that misleading conclusions can be drawn if models are compared without consideration of the sepsis definition used which emphasizes the need for a standardized definition for sepsis onset.

The GCS-Pupils (GCS-P) score to assess outcomes after traumatic brain injury: a retrospective study

Introduction The Glasgow Coma Scale (GCS) and pupil response to light are commonly used to assess brain injury severity and predict outcomes. The aim of this study was to investigate whether the GCS combined with pupil response (GCS-P), compared to the GCS alone, could be a better predictor of hospital mortality for patients with traumatic brain injury (TBI). Methods A retrospective cohort study was undertaken at an adult level one trauma centre including patients with isolated TBI of Abbreviated Injury Scale above three. The GCS and pupil response were combined to an arithmetic score (GCS score (range 3–15) minus the number of nonreacting pupils (0, 1, or 2)), or by treating each factor as separate categorical variables. The association of in-hospital mortality with GCS-P as a categorical variable was evaluated using Nagelkerke’s R 2 and compared using areas under the receiver operating characteristic (AUROC) curve. Results There were 392 patients included over the study period of 1 July 2014 and 30 September 2017, with an overall mortality rate of 15.2%. Mortality was highest at GCS-P of 1 (79%), with lowest mortality at a GCS-P 15 (1.6%). Nagelkerke’s R 2 was 0.427 for GCS alone and 0.486 for GCS-P. The AUROC for GCS-P to predict mortality was 0.87 (95%CI: 0.82–0.72), higher than for GCS alone (0.85; 95%CI: 0.80–0.90; p < .001). Discussion GCS-P provided a better predictor of mortality compared to the GCS. As both the GCS and pupillary response are routinely recorded on all patients, combination of these pieces of information into a single score can further simplify assessment of patients with TBI, with some improvement in performance.

Performance of Automated Machine Learning in Predicting Outcomes of Pneumatic Retinopexy

ObjectiveAutomated Machine Learning (AutoML) has emerged as a novel tool for medical professionals lacking coding experience, enabling them to develop predictive models for treatment outcomes. This study evaluated the performance of AutoML tools in developing models predicting the success of pneumatic retinopexy (PR) in treatment of rhegmatogenous retinal detachment (RRD). These models were then compared with custom models created by ML experts. DesignRetrospective multicenter study. Population539 consecutive patients with primary RRD that underwent PR by a vitreoretinal fellow at six training hospitals between 2002 and 2022. MethodsWe used two AutoML platforms, MATLAB Classification Learner and Google Cloud AutoML. Additional models were developed by computer scientists. We included patient demographics and baseline characteristics, including lens and macula status, RRD size, number and location of breaks, presence of vitreous hemorrhage and lattice degeneration, and physicians’ experience. The dataset was split into a training (n = 483) and test set (n = 56). The training set, with a 2:1 success-to-failure ratio, was used to train the MATLAB models. Since Google Cloud AutoML requires a minimum of 1000 samples, the training set was tripled to create a new set with 1449 datapoints. Additionally, balanced datasets with a 1:1 success-to-failure ratio were created using Python. Main Outcome MeasuresSingle-procedure anatomic success rate, as predicted by the ML models. F2-scores and area under the receiver operating curve (AUROC) were used as primary metrics to compare models. ResultsThe best performing AutoML model (F2-score=0.85, AUROC=0.90; MATLAB), showed similar performance to the custom model (0.92, 0.86) when trained on the balanced datasets. However, training the AutoML model with imbalanced data yielded misleadingly high AUROC (0.81) despite low F2-score (0.2) and sensitivity (0.17). ConclusionsWe demonstrated the feasibility of using AutoML as an accessible tool for medical professionals to develop models from real-world data. Such models can ultimately aid in the clinical decision-making, contributing to better patient outcomes. However, outcomes can be misleading or unreliable if used naively. Limitations exist, particularly if datasets contain missing variables or are highly imbalanced. Proper model selection and data preprocessing can improve the reliability of AutoML tools.

MRI radiomics to monitor therapeutic outcome of sorafenib plus IHA transcatheter NK cell combination therapy in hepatocellular carcinoma

BackgroundHepatocellular carcinoma (HCC) is a common liver malignancy with limited treatment options. Previous studies expressed the potential synergy of sorafenib and NK cell immunotherapy as a promising approach against HCC. MRI is commonly used to assess response of HCC to therapy. However, traditional MRI-based metrics for treatment efficacy are inadequate for capturing complex changes in the tumor microenvironment, especially with immunotherapy. In this study, we investigated potent MRI radiomics analysis to non-invasively assess early responses to combined sorafenib and NK cell therapy in a HCC rat model, aiming to predict multiple treatment outcomes and optimize HCC treatment evaluations.MethodsSprague Dawley (SD) rats underwent tumor implantation with the N1-S1 cell line. Tumor progression and treatment efficacy were assessed using MRI following NK cell immunotherapy and sorafenib administration. Radiomics features were extracted, processed, and selected from both T1w and T2w MRI images. The quantitative models were developed to predict treatment outcomes and their performances were evaluated with area under the receiver operating characteristic (AUROC) curve. Additionally, multivariable linear regression models were constructed to determine the correlation between MRI radiomics and histology, aiming for a noninvasive evaluation of tumor biomarkers. These models were evaluated using root-mean-squared-error (RMSE) and the Spearman correlation coefficient.ResultsA total of 743 radiomics features were extracted from T1w and T2w MRI data separately. Subsequently, a feature selection process was conducted to identify a subset of five features for modeling. For therapeutic prediction, four classification models were developed. Support vector machine (SVM) model, utilizing combined T1w + T2w MRI data, achieved 96% accuracy and an AUROC of 1.00 in differentiating the control and treatment groups. For multi-class treatment outcome prediction, Linear regression model attained 85% accuracy and an AUC of 0.93. Histological analysis showed that combination therapy of NK cell and sorafenib had the lowest tumor cell viability and the highest NK cell activity. Correlation analyses between MRI features and histological biomarkers indicated robust relationships (r = 0.94).ConclusionsOur study underscored the significant potential of texture-based MRI imaging features in the early assessment of multiple HCC treatment outcomes.

Utility of P²MS Index in Predicting Esophageal Varices in Patients with Hepatitis C Associated Chronic Liver Disease in Pakistani Population

Background: Screening for esophageal varices (EVs) in patients with chronic liver disease is critical for identifying those needing primary prophylaxis against variceal bleeding. The gold standard for EV screening is endoscopy, an invasive and costly procedure often avoided by patients. This necessitates the need for non-invasive predictive tools for the presence of EVs and high-risk varices (HRVs). The P²MS index, a relatively new model, has shown promise in this regard. Objective: This study aimed to evaluate the effectiveness of the P²MS index, alongside APRI and FIB-4 scores, in predicting the presence of EVs and HRVs in patients with hepatitis C-related chronic liver disease. Methods: Conducted at the Sindh Institute of Urology and Transplantation's Hepato-Gastroenterology department, this cross-sectional study involved 375 participants aged 18-50 years, recently diagnosed with liver cirrhosis due to hepatitis C. Excluding those with varied medical backgrounds, the study used non-probability consecutive sampling. Diagnostic procedures included esophagogastroduodenoscopy (EGD) and complete blood count. P²MS index, APRI, and FIB-4 scores were calculated for each participant. Data analysis was performed using SPSS version 25.0, focusing on the Area Under the Receiver Operating Curve (AUROC) to assess diagnostic accuracy. Results: In our population, the P²MS index at a cutoff of &lt;10.7 showed a specificity of 93.22% and sensitivity of 87.88% for predicting the presence of EVs, with a PPV and diagnostic accuracy exceeding 90%. For HRVs, a cutoff of &lt;5.4 demonstrated an excellent NPV, alongside good sensitivity, specificity, and diagnostic accuracy. Compared to APRI and FIB-4, the P²MS index had the highest AUROC for predicting the presence of EVs (0.92, 95% CI 0.89-0.95, p&lt;0.001). For HRVs, it was comparable to APRI (AUROC=0.88, 95% CI 0.85-0.92, p&lt;0.01) and superior to FIB-4. Conclusion: The P²MS index is an effective non-invasive tool for screening EVs in patients with hepatitis C-related chronic liver disease, demonstrating high accuracy for predicting the presence of EVs. However, its efficacy in predicting HRVs is comparable to other non-invasive indices and warrants further validation.

Performance of noninvasive seromarkers in predicting liver fibrosis among MAFLD patients with or without viral hepatitis.

The accuracy of noninvasive seromarkers in predicting liver fibrosis in metabolic dysfunction-associated fatty liver disease (MAFLD) patients with or without viral hepatitis is elusive. The AST to platelet ratio index (APRI), fibrosis-4 index (FIB-4), and NAFLD fibrosis score (NFS) were assessed in 871 MAFLD patients who received elastography in a viral hepatitis-endemic area. The area under the receiver operating characteristic (AUROC) curve increased substantially with increasing fibrotic stage across the three biomarkers. APRI (AUROC range 0.73-0.80) and FIB-4 (AUROC range 0.66-0.82) performed better than NFS (AUROC range 0.63-0.75). When patients were divided into viral and non-viral MAFLD groups, a better AUROC of APRI (range 0.76-0.80) and FIB-4 (range 0.68-0.78) than NFS (range 0.62-70) existed only in viral MALFD but not in non-viral MAFLD. Regarding the NFS, the AUROC was higher in non-viral MAFLD (range 0.69-0.86) and outperformed viral MAFLD at all fibrotic stages. The accuracy in predicting liver fibrosis increased with the advancement of liver disease for the three biomarkers. NFS exerted better diagnostic accuracy in non-viral than in viral MAFLD patients across different fibrotic stages. The best accuracy was 91.1% using the cutoff value of -9.98 for the NFS in predicting liver cirrhosis in non-viral MAFLD patients. The APRI and FIB-4 performed better than the NFS in predicting liver fibrosis in MAFLD as a whole. The suboptimal performance and accuracy of the NFS existed only in viral MAFLD patients. Caution should be taken when assessing the NFS in MAFLD patients with viral hepatitis.

Evaluation of an artificial intelligence-based decision support for the detection of cutaneous melanoma in primary care: a prospective real-life clinical trial.

Use of artificial intelligence (AI), or machine learning, to assess dermoscopic images of skin lesions to detect melanoma has, in several retrospective studies, shown high levels of diagnostic accuracy on par with - or even outperforming - experienced dermatologists. However, the enthusiasm around these algorithms has not yet been matched by prospective clinical trials performed in authentic clinical settings. In several European countries, including Sweden, the initial clinical assessment of suspected skin cancer is principally conducted in the primary healthcare setting by primary care physicians, with or without access to teledermoscopic support from dermatology clinics. To determine the diagnostic performance of an AI-based clinical decision support tool for cutaneous melanoma detection, operated by a smartphone application (app), when used prospectively by primary care physicians to assess skin lesions of concern due to some degree of melanoma suspicion. This prospective multicentre clinical trial was conducted at 36 primary care centres in Sweden. Physicians used the smartphone app on skin lesions of concern by photographing them dermoscopically, which resulted in a dichotomous decision support text regarding evidence for melanoma. Regardless of the app outcome, all lesions underwent standard diagnostic procedures (surgical excision or referral to a dermatologist). After investigations were complete, lesion diagnoses were collected from the patients' medical records and compared with the app's outcome and other lesion data. In total, 253 lesions of concern in 228 patients were included, of which 21 proved to be melanomas, with 11 thin invasive melanomas and 10 melanomas in situ. The app's accuracy in identifying melanomas was reflected in an area under the receiver operating characteristic (AUROC) curve of 0.960 [95% confidence interval (CI) 0.928-0.980], corresponding to a maximum sensitivity and specificity of 95.2% and 84.5%, respectively. For invasive melanomas alone, the AUROC was 0.988 (95% CI 0.965-0.997), corresponding to a maximum sensitivity and specificity of 100% and 92.6%, respectively. The clinical decision support tool evaluated in this investigation showed high diagnostic accuracy when used prospectively in primary care patients, which could add significant clinical value for primary care physicians assessing skin lesions for melanoma.

Evaluating the Efficacy of Type 2 Diabetes Polygenic Risk Scores in an Independent European Population.

Numerous type 2 diabetes (T2D) polygenic risk scores (PGSs) have been developed to predict individuals' predisposition to the disease. An independent assessment and verification of the best-performing PGS are warranted to allow for a rapid application of developed models. To date, only 3% of T2D PGSs have been evaluated. In this study, we assessed all (n = 102) presently published T2D PGSs in an independent cohort of 3718 individuals, which has not been included in the construction or fine-tuning of any T2D PGS so far. We further chose the best-performing PGS, assessed its performance across major population principal component analysis (PCA) clusters, and compared it with newly developed population-specific T2D PGS. Our findings revealed that 88% of the published PGSs were significantly associated with T2D; however, their performance was lower than what had been previously reported. We found a positive association of PGS improvement over the years (p-value = 8.01 × 10-4 with PGS002771 currently showing the best discriminatory power (area under the receiver operating characteristic (AUROC) = 0.669) and PGS003443 exhibiting the strongest association PGS003443 (odds ratio (OR) = 1.899). Further investigation revealed no difference in PGS performance across major population PCA clusters and when compared with newly developed population-specific PGS. Our findings revealed a positive trend in T2D PGS performance, consistently identifying high-T2D-risk individuals in an independent European population.

Maternal and fetal factors associated with stillbirth in singleton pregnancies in 13 hospitals across six states in India: A prospective cohort study.

This study aimed to investigate the incidence of and risk factors for stillbirth in an Indian population. We conducted a secondary data analysis of a hospital-based cohort from the Maternal and Perinatal Health Research collaboration, India (MaatHRI), including pregnant women who gave birth between October 2018-September 2023. Data from 9823 singleton pregnancies recruited from 13 hospitals across six Indian states were included. Univariable and multivariable Poisson regression analysis were performed to examine the relationship between stillbirth and potential risk factors. Model prediction was assessed using the area under the receiver-operating characteristic (AUROC) curve. There were 216 stillbirths (48 antepartum and 168 intrapartum) in the study population, representing an overall stillbirth rate of 22.0 per 1000 total births (95% confidence interval [CI]: 19.2-25.1). Modifiable risk factors for stillbirth were: receiving less than four antenatal check-ups (adjusted relative risk [aRR]: 1.75, 95% CI: 1.25-2.47), not taking any iron and folic acid supplementation during pregnancy (aRR: 7.23, 95% CI: 2.12-45.33) and having severe anemia in the third trimester (aRR: 3.37, 95% CI: 1.97-6.11). Having pregnancy/fetal complications such as hypertensive disorders of pregnancy (aRR: 1.59, 95% CI: 1.03-2.36), preterm birth (aRR: 4.41, 95% CI: 3.21-6.08) and birth weight below the 10th percentile for gestational age (aRR: 1.35, 95% CI: 1.02-1.79) were also associated with an increased risk of stillbirth. Identified risk factors explained 78.2% (95% CI: 75.0%-81.4%) of the risk of stillbirth in the population. Addressing potentially modifiable antenatal factors could reduce the risk of stillbirths in India.

Validation of Three Models for Prediction of Blood Transfusion during Cesarean Delivery Admission.

Prediction of blood transfusion during delivery admission allows for clinical preparedness and risk mitigation. Although prediction models have been developed and adopted into practice, their external validation is limited. We aimed to evaluate the performance of three blood transfusion prediction models in a U.S. cohort of individuals undergoing cesarean delivery. This was a secondary analysis of a multicenter randomized trial of tranexamic acid for prevention of hemorrhage at time of cesarean delivery. Three models were considered: a categorical risk tool (California Maternal Quality Care Collaborative [CMQCC]) and two regression models (Ahmadzia et al and Albright et al). The primary outcome was intrapartum or postpartum red blood cell transfusion. The CMQCC algorithm was applied to the cohort with frequency of risk category (low, medium, high) and associated transfusion rates reported. For the regression models, the area under the receiver-operating curve (AUC) was calculated and a calibration curve plotted to evaluate each model's capacity to predict receipt of transfusion. The regression model outputs were statistically compared. Of 10,785 analyzed individuals, 3.9% received a red blood cell transfusion during delivery admission. The CMQCC risk tool categorized 1,970 (18.3%) individuals as low risk, 5,259 (48.8%) as medium risk, and 3,556 (33.0%) as high risk with corresponding transfusion rates of 2.1% (95% confidence interval [CI]: 1.5-2.9%), 2.2% (95% CI: 1.8-2.6%), and 7.5% (95% CI: 6.6-8.4%), respectively. The AUC for prediction of blood transfusion using the Ahmadzia and Albright models was 0.78 (95% CI: 0.76-0.81) and 0.79 (95% CI: 0.77-0.82), respectively (p = 0.38 for difference). Calibration curves demonstrated overall agreement between the predicted probability and observed likelihood of blood transfusion. Three models were externally validated for prediction of blood transfusion during cesarean delivery admission in this U.S. Overall, performance was moderate; model selection should be based on ease of application until a specific model with superior predictive ability is developed. · A total of 3.9% of individuals received a blood transfusion during cesarean delivery admission.. · Three models used in clinical practice are externally valid for blood transfusion prediction.. · Institutional model selection should be based on ease of application until further research identifies the optimal approach..

A comparison of the original major depression inventory with a modified version: A Danish validation study.

The Major Depression Inventory (MDI) is a patient-reported outcome measure used by general practitioners to assist with diagnosing and evaluation of the severity of a patient's depression. However, recent studies have questioned the structural validity of the MDI. We proposed a modified version (mMDI) of the MDI with fewer response categories and four rephrased items and aimed to compare the psychometric properties of the changes in a joint cohort of patients from general practice and mental health associations. We used Rasch analysis, confirmatory factor analysis, and the area under the receiver operating curve (AUROC) to assess the validity and reliability of the two versions. Equipercentile linking was used to compute cut-off points for the mMDI. For both versions, local dependence was found between the three item pairs (loss of interest, lack of energy), (lack of self-confidence, feelings of guilt), and (concentration problems, feeling restless/slowed down). The mMDI displayed lower measurement error in the upper end of the scale and better item level fit for three of the four reformulated items compared to the MDI. For the MDI, 5.3% of the respondents gave improbable responses; the corresponding number was 3.4% for the mMDI. The mMDI displayed better fit to a one-factor model compared to the MDI. When comparing the correlation of the scales with the WHO-5 instrument, the corresponding AUROC estimates for the mMDI and MDI were found to be 0.93 (0.92; 0.96) and 0.91 (0.87; 0.94), respectively. The cut-off points for mild, moderate, and severe depression in the mMDI were found to be 17, 20, and 23, respectively. The proposed changes of the MDI are psychometrically sound upgrades of the original.

Enhancing Health Equity by Predicting Missed Appointments in Health Care: Machine Learning Study.

The phenomenon of patients missing booked appointments without canceling them-known as Did Not Show (DNS), Did Not Attend (DNA), or Failed To Attend (FTA)-has a detrimental effect on patients' health and results in massive health care resource wastage. Our objective was to develop machine learning (ML) models and evaluate their performance in predicting the likelihood of DNS for hospital outpatient appointments at the MidCentral District Health Board (MDHB) in New Zealand. We sourced 5 years of MDHB outpatient records (a total of 1,080,566 outpatient visits) to build the ML prediction models. We developed 3 ML models using logistic regression, random forest, and Extreme Gradient Boosting (XGBoost). Subsequently, 10-fold cross-validation and hyperparameter tuning were deployed to minimize model bias and boost the algorithms' prediction strength. All models were evaluated against accuracy, sensitivity, specificity, and area under the receiver operating characteristic (AUROC) curve metrics. Based on 5 years of MDHB data, the best prediction classifier was XGBoost, with an area under the curve (AUC) of 0.92, sensitivity of 0.83, and specificity of 0.85. The patients' DNS history, age, ethnicity, and appointment lead time significantly contributed to DNS prediction. An ML system trained on a large data set can produce useful levels of DNS prediction. This research is one of the very first published studies that use ML technologies to assist with DNS management in New Zealand. It is a proof of concept and could be used to benchmark DNS predictions for the MDHB and other district health boards. We encourage conducting additional qualitative research to investigate the root cause of DNS issues and potential solutions. Addressing DNS using better strategies potentially can result in better utilization of health care resources and improve health equity.

Charlson Comorbidity Index and Frailty as Predictors of Resolution Following Middle Meningeal Artery Embolization for Chronic Subdural Hematoma

Research on variables associated with chronic subdural hematoma (cSDH) resolution following middle meningeal artery embolization (MMAE) is limited. This study investigated the clinical utility of age-adjusted Charlson Comorbidity Index (ACCI) and modified 5-item Frailty Index (mFI-5) for predicting cSDH resolution following MMAE. We identified patients who underwent MMAE at our institution between January 2018, and December 2022, with at least 20 days of follow-up and one radiographic follow-up study. Patient demographics, characteristics, and outcomes were collected. Complete resolution was defined as absence of subdural collections on CT-scan at last follow-up. Non-age adjusted CCI (CCI), ACCI, and mFI-5 scores were calculated. Univariate and multivariable logistic regression analyzed the relationship between cSDH resolution and variables. A receiver operating characteristic (ROC) curve established the utility of ACCI and mFI-5 in predicting hematoma resolution. The study included 85 MMAE procedures. In univariate analysis, patients without resolution were older, had higher CCI, higher ACCI, higher mFI-5, and were more likely to have diabetes mellitus. In multivarible analysis, CCI (OR: 0.66, 95% CI: 0.48, 0.91) was independently associated with resolution controlling for age and antithrombotic resumption. The area under the ROC (AUROC) curve was 0.75 (95% CI: 0.65 - 0.85) for ACCI and 0.64 (95% CI: 0.52 - 0.76) for mFI-5. The optimal cutoffs for predicting resolution were ACCI ≥ 5 (sensitivity = 0.63, specificity = 0.77), and mFI-5 > 0 (sensitivity = 0.84, specificity = 0.43). ACCI and mFI-5 moderately predict MMAE resolution and may aid in medical decision-making.

Early prediction of pediatric asthma in the Canadian Healthy Infant Longitudinal Development (CHILD) birth cohort using machine learning

BackgroundEarly identification of children at risk of asthma can have significant clinical implications for effective intervention and treatment. This study aims to disentangle the relative timing and importance of early markers of asthma.MethodsUsing the CHILD Cohort Study, 132 variables measured in 1754 multi-ethnic children were included in the analysis for asthma prediction. Data up to 4 years of age was used in multiple machine learning models to predict physician-diagnosed asthma at age 5 years. Both predictive performance and variable importance was assessed in these models.ResultsEarly-life data (≤1 year) has limited predictive ability for physician-diagnosed asthma at age 5 years (area under the precision-recall curve (AUPRC) < 0.35). The earliest reliable prediction of asthma is achieved at age 3 years, (area under the receiver-operator curve (AUROC) > 0.90) and (AUPRC > 0.80). Maternal asthma, antibiotic exposure, and lower respiratory tract infections remained highly predictive throughout childhood. Wheezing status and atopy are the most important predictors of early childhood asthma from among the factors included in this study.ConclusionsChildhood asthma is predictable from non-biological measurements from the age of 3 years, primarily using parental asthma and patient history of wheezing, atopy, antibiotic exposure, and lower respiratory tract infections.ImpactMachine learning models can predict physician-diagnosed asthma in early childhood (AUROC > 0.90 and AUPRC > 0.80) using ≥3 years of non-biological and non-genetic information, whereas prediction with the same patient information available before 1 year of age is challenging.Wheezing, atopy, antibiotic exposure, lower respiratory tract infections, and the child’s mother having asthma were the strongest early markers of 5-year asthma diagnosis, suggesting an opportunity for earlier diagnosis and intervention and focused assessment of patients at risk for asthma, with an evolving risk stratification over time.

Prediction of the acceptance of telemedicine among rheumatic patients: a machine learning-powered secondary analysis of German survey data

Telemedicine (TM) has augmented healthcare by enabling remote consultations, diagnosis, treatment, and monitoring of patients, thereby improving healthcare access and patient outcomes. However, successful adoption of TM depends on user acceptance, which is influenced by technical, socioeconomic, and health-related factors. Leveraging machine learning (ML) to accurately predict these adoption factors can greatly contribute to the effective utilization of TM in healthcare. The objective of the study was to compare 12 ML algorithms for predicting willingness to use TM (TM try) among patients with rheumatic and musculoskeletal diseases (RMDs) and identify key contributing features. We conducted a secondary analysis of RMD patient data from a German nationwide cross-sectional survey. Twelve ML algorithms, including logistic regression, random forest, extreme gradient boosting (XGBoost), and neural network (deep learning) were tested on a subset of the dataset, with the inclusion of only RMD patients who answered “yes” or “no” to TM try. Nested cross-validation was used for each model. The best-performing model was selected based on area under the receiver operator characteristic (AUROC). For the best-performing model, a multinomial/multiclass ML approach was undertaken with the consideration of the three following classes: “yes”, “no”, “do not know/not answered”. Both one-vs-one and one-vs-rest strategies were considered. The feature importance was investigated using Shapley additive explanation (SHAP). A total of 438 RMD patients were included, with 26.5% of them willing to try TM, 40.6% not willing, and 32.9% undecided (missing answer or “do not know answer”). This dataset was used to train and test ML models. The mean accuracy of the 12 ML models ranged from 0.69 to 0.83, while the mean AUROC ranged from 0.79 to 0.90. The XGBoost model produced better results compared with the other models, with a sensitivity of 70%, specificity of 91% and positive predictive value of 84%. The most important predictors of TM try were the possibility that TM services were offered by a rheumatologist, prior TM knowledge, age, self-reported health status, Internet access at home and type of RMD diseases. For instance, for the yes vs. no classification, not wishing that TM services were offered by a rheumatologist, self-reporting a bad health status and being aged 60–69 years directed the model toward not wanting to try TM. By contrast, having Internet access at home and wishing that TM services were offered by a rheumatologist directed toward TM try. Our findings have significant implications for primary care, in particular for healthcare professionals aiming to implement TM effectively in their clinical routine. By understanding the key factors influencing patients' acceptance of TM, such as their expressed desire for TM services provided by a rheumatologist, self-reported health status, availability of home Internet access, and age, healthcare professionals can tailor their strategies to maximize the adoption and utilization of TM, ultimately improving healthcare outcomes for RMD patients. Our findings are of high interest for both clinical and medical teaching practice to fit changing health needs caused by the growing number of complex and chronically ill patients.