Model Development Cohort Research Articles

A Simple Score (MOG-AR) to Identify Individuals at High Risk of Relapse After MOGAD Attack.

To identify predictors for relapse in patients with myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD) and to develop and validate a simple risk score for predicting relapse. In China National Registry of Neuro-Inflammatory Diseases (CNRID), we identified patients with MOGAD from March 2023 and followed up prospectively to September 2023. The primary endpoint was MOGAD relapse, confirmed by an independent panel. Patients were randomly divided into model development (75%) and internal validation (25%) cohorts. Prediction models were constructed and internally validated using Andersen-Gill models. Nomogram and relapse risk score were generated based on the final prediction models. A total of 188 patients (comprising 612 treatment episodes) were included in cohorts. Female (HR: 0.687, 95% CI 0.524-0.899, p = 0.006), onset age 45 years or older (HR: 1.621, 95% CI 1.242-2.116, p < 0.001), immunosuppressive therapy (HR: 0.338, 95% CI 0.239-0.479, p < 0.001), oral corticosteroids >3 months (HR 0.449, 95% CI 0.326-0.620, p < 0.001), and onset phenotype (p < 0.001) were identified as factors associated with MOGAD relapse. A predictive score, termed MOG-AR (Immunosuppressive therapy, oral Corticosteroids, Onset Age, Sex, Attack phenotype), derived in prediction model, demonstrated strong predictive ability for MOGAD relapse. MOG-AR score of 13-16 indicates a higher risk of relapse (HR: 3.285, 95% CI 1.473-7.327, p = 0.004). The risk of MOGAD relapse seems to be predictable. Further validation of MOG-AR score developed from this cohort to determine appropriate treatment and monitoring frequency is warranted. CNRID, NCT05154370, registered December 13, 2021, first enrolled December 15, 2021.

Development and validation of a chronic kidney disease progression model using patient-level simulations

Chronic disease progression models are available for several highly prevalent conditions. For chronic kidney disease (CKD), the scope of existing progression models is limited to the risk of kidney failure and major cardiovascular (CV) events. The aim of this project was to develop a comprehensive CKD progression model (CKD-PM) that simulates the risk of CKD progression and a broad range of complications in patients with CKD. A series of literature reviews informed the selection of risk factors and identified existing risk equations/algorithms for kidney replacement therapy (KRT), CV events, other CKD-related complications, and mortality. Risk equations and transition probabilities were primarily sourced from publications produced by large US and international CKD registries. A patient-level, state-transition model was developed with health states defined by the Kidney Disease Improving Global Outcomes categories. Model validation was performed by comparing predicted outcomes with observed outcomes in the source cohorts used in model development (internal validation) and other cohorts (external validation). The CKD-PM demonstrated satisfactory modeling properties. Accurate prediction of all-cause and CV mortality was achieved without calibration, while prediction of CV events through CKD-specific equations required implementation of a calibration factor to balance time-dependent versus baseline risk. Predicted annual changes in estimated glomerular filtration rate (eGFR) and urine albumin-creatinine ratio were acceptable in comparison to external values. A flexible eGFR threshold for KRT equations enabled accurate prediction of these events. This CKD-PM demonstrated reliable modeling properties. Both internal and external validation revealed robust outcomes.

A prediction model for the walking and balance milestone in Parkinson's disease

BackgroundWalking and balance impairments, represented by freezing of gait and falls, are significant contributors to disability in advanced Parkinson's disease (PD) patients. However, the composite measure of the Walking and Balance Milestone (WBMS) has not been thoroughly investigated. MethodsThis study included 606 early-stage PD patients from the Parkinson's Progression Markers Initiative (PPMI) database, with a disease duration of less than 2 years and no WBMS at baseline. Patients were divided into a model development cohort (70 %) and a validation cohort (30 %) according to the enrollment site. Longitudinal follow-up data over a period of 12 years were analyzed. ResultsAmong all 606 patients, the estimated probability of being WBMS-free at the 5th and 10th year was 88 % and 60 %, respectively. Five clinical variables (Age, Symbol Digit Modalities Test (SDMT), postural instability and gait difficulty (PIGD) score, Movement Disorder Society-Unified Parkinson's Disease Rating Scale Part I (MDS-UPDRS-I) score, and REM Sleep Behavior Disorder (RBD) were used to construct the Cox predictive model. The C-index of the model was 0.75 in the development cohort and 0.76 in the validation cohort. By optimizing the PIGD and MDS-UPDRS-I variables, an easy-to-use model was achieved with comparable predictive performance. ConclusionA predictive model based on five baseline clinical measures (Age, SDMT, PIGD score, MDS-UPDRS-I score, RBD) could effectively estimate the risk of the WBMS in early PD patients. This model is valuable for prognostic counseling and clinical intervention trials for gait and balance impairment.

Hepatocellular carcinoma risk scores for non-viral liver disease: A systematic review and meta-analysis

BackgroundHepatocellular carcinoma (HCC) risk prediction models may provide a more personalised approach to surveillance for HCC among patients with cirrhosis. This systematic review aims to summarise the performance of HCC prediction models in patients with non-viral chronic liver disease. MethodThe study was prospectively registered with PROSPERO (ID: CRD42022370078) and reported in accordance with PRISMA guidelines. MEDLINE and EMBASE databases were searched using a validated search filter for prediction model studies. Two reviewers independently assessed studies for inclusion and risk of bias. Model performance (discrimination and calibration) were identified to assess the risk of HCC at specified time points. A random effects meta-analysis was done on a subset of studies that reported performance of the same model. Results7,854 studies were identified. After review, 14 studies with a total of 94,014 participants were included. 45% of patients had viral hepatitis, 27% alcohol related liver disease (ArLD) and 19% metabolic associated steatotic liver disease (MASLD). Follow-up ranged from 15.1 - 138 months. Only one model was developed using a competing risk approach. Age (7 models) and sex (6 models) were the most frequently included predictors. Model discrimination (AUROC or c-statistic) ranged from 0.61 - 0.947. Only the ‘aMAP’ score (age, male sex, albumin,bilirubin, and platelets) had sufficient external validation for quantitative analysis, with a pooled c-statistic of 0.81 (95% CI 0.80-0.83). Calibration was reported in only 9 of 14 studies. All studies were rated high risk of bias. ConclusionStudies describing risk prediction of HCC in non-viral chronic liver disease are poorly reported, have a high risk of bias and do not account for competing risk events. Patients with ArLD and MASLD are underrepresented in development and validation cohorts. These factors remain barriers to the clinical utility and uptake of HCC risk models in persons with the commonest liver diseases. Impact and implicationsThe recent EASL policy statement emphasises the potential of risk-based surveillance to reduce both HCC-related deaths and surveillance costs. This study addresses the gap in understanding the performance of current HCC risk models in patients with non-viral liver diseases, reflecting the epidemiological landscape of liver disease in Western countries.In our review of these models, we identified several key concerns regarding reporting standards and risk of bias and confirmed that patients with alcohol-related liver disease (ArLD) and metabolic-associated steatotic liver disease (MASLD) are underrepresented in model development and validation cohorts. Additionally, most models fail to account for the significant risk of competing events, leading to potential overestimation of true HCC risk.This study highlights these critical issues that may hinder the implementation of risk models in clinical practice and offers key recommendations for future model development studies.

Novel Nomogram for Prediction of Lymph Node Metastases from Buccal Mucosa Squamous Carcinoma Using Histological Parameters

Objective The aim of the study was to construct a nomogram that is easily reproducible, accurate, and cost-effective in predicting cervical lymph nodal metastasis in buccal mucosa cancer. Methodology Patients who underwent radical resection of a primary tumor of the buccal mucosa with neck dissection were enrolled. Clinical characteristics independently associated with lymph nodal metastasis in multivariate analyses were adopted to build the model. Results Patients who underwent surgery (January 2021–December 2021) were included as the model development cohort (n = 127). Depth of invasion, perineural invasion, lymphovascular invasion, and the worst pattern of invasion were independent predictors of lymph nodal metastasis. The nomogram model based on these four predictors showed good discrimination accuracy in percentage prediction of lymph nodal metastasis. Conclusion This study proposes a simple predictive model for the risk of nodal metastasis in buccal mucosa squamous cell cancer. The study has strength that, it is based on a large sample, proposed model being simple size, and based on parameters empirically supported as well as established in literature, easy to use in routine clinical practice, and cost-effective.

Identification of variables and development of a prediction model for DIBH eligibility in left-sided breast cancer radiotherapy: a prospective cohort study with temporal validation

ObjectiveTo identify variables associated with a patients’ ability to reproducibly hold their breath for deep-inspiration breath-hold (DIBH) radiotherapy (RT) and to develop a predictive model for DIBH eligibility.MethodsThis prospective, single-institution, IRB-approved observational study included women with left-sided breast cancer treated between January 2023 and March 2024. Patients underwent multiple breath-hold sessions over 2–3 consecutive days. DIBH waveform metrics and clinical factors were recorded and analysed. Logistic mixed modelling was used to predict DIBH eligibility, and a temporal validation cohort was used to assess model performance.ResultsIn total, 253 patients were included, with 206 in the model development cohort and 47 in the temporal validation cohort. The final logistic mixed model identified increasing average breath-hold duration (OR, 95% CI: 0.308, 0.104–0.910. p = 0.033) and lower amplitude (OR, 95% CI: 0.737, 0.641–0.848. p < 0.001) as significant predictors of DIBH eligibility. Increasing age was associated with higher odds of being ineligible for DIBH (OR, 95% CI: 1.040, 1.001–1.081. p = 0.044). The model demonstrated good discriminative performance in the validation cohort with an AUC of 80.9% (95% CI: 73.0-88.8).ConclusionThe identification of variables associated with DIBH eligibility and development of a predictive model has the potential to serve as a decision-support tool. Further external validation is required before its integration into routine clinical practice.

Systematic review: risk prediction models for metachronous advanced colorectal neoplasia after polypectomy.

Colorectal cancer (CRC) is the fourth leading cause of cancer death globally. CRC surveillance is a common indication for colonoscopy, representing a considerable burden for endoscopy services. Accurate identification of high-risk patients who would benefit from more intensive surveillance, as well as low-risk patients suitable for less frequent follow-up, could improve the effectiveness of surveillance protocols and resource use. Our aim was to identify and critically appraise published risk models for the occurrence of metachronous advanced colorectal neoplasia (ACN), defined here as CRC or advanced adenomas detected during surveillance colonoscopy. We searched PubMed and EMBASE for primary research studies reporting the development and/or validation of multivariable models that predict metachronous ACN risk. Screening of studies for inclusion, data extraction, and risk of bias assessment were conducted by two researchers independently. We identified nine studies describing nine risk models. Six models were internally validated and two were externally validated. No model underwent both internal and external validation. Good model discrimination (concordance index>0.7) was reported for two models during internal validation and for one model during external validation. Calibration was acceptable when assessed (n=4). Methodological limitations and a high risk of bias were observed for all studies. Several published models predicting metachronous ACN risk showed some promise. However, adherence to methodological standards was limited, and only two models were externally validated. Head-to-head comparisons of existing models using populations independent from model development cohorts should be prioritized to identify models suitable for use in clinical practice.

A novel tool for arrhythmic risk stratification in desmoplakin gene variant carriers.

Pathogenic desmoplakin (DSP) gene variants are associated with the development of a distinct form of arrhythmogenic cardiomyopathy known as DSP cardiomyopathy. Patients harbouring these variants are at high risk for sustained ventricular arrhythmia (VA), but existing tools for individualized arrhythmic risk assessment have proven unreliable in this population. Patients from the multi-national DSP-ERADOS (Desmoplakin SPecific Effort for a RAre Disease Outcome Study) Network patient registry who had pathogenic or likely pathogenic DSP variants and no sustained VA prior to enrolment were followed longitudinally for the development of first sustained VA event. Clinically guided, step-wise Cox regression analysis was used to develop a novel clinical tool predicting the development of incident VA. Model performance was assessed by c-statistic in both the model development cohort (n = 385) and in an external validation cohort (n = 86). In total, 471 DSP patients [mean age 37.8 years, 65.6% women, 38.6% probands, 26% with left ventricular ejection fraction (LVEF) < 50%] were followed for a median of 4.0 (interquartile range: 1.6-7.3) years; 71 experienced first sustained VA events {2.6% [95% confidence interval (CI): 2.0, 3.5] events/year}. Within the development cohort, five readily available clinical parameters were identified as independent predictors of VA and included in a novel DSP risk score: female sex [hazard ratio (HR) 1.9 (95% CI: 1.1-3.4)], history of non-sustained ventricular tachycardia [HR 1.7 (95% CI: 1.1-2.8)], natural logarithm of 24-h premature ventricular contraction burden [HR 1.3 (95% CI: 1.1-1.4)], LVEF < 50% [HR 1.5 (95% CI: .95-2.5)], and presence of moderate to severe right ventricular systolic dysfunction [HR 6.0 (95% CI: 2.9-12.5)]. The model demonstrated good risk discrimination within both the development [c-statistic .782 (95% CI: .77-.80)] and external validation [c-statistic .791 (95% CI: .75-.83)] cohorts. The negative predictive value for DSP patients in the external validation cohort deemed to be at low risk for VA (<5% at 5 years; n = 26) was 100%. The DSP risk score is a novel model that leverages readily available clinical parameters to provide individualized VA risk assessment for DSP patients. This tool may help guide decision-making for primary prevention implantable cardioverter-defibrillator placement in this high-risk population and supports a gene-first risk stratification approach.

Prediction of sepsis among patients with major trauma using artificial intelligence: a multicenter validated cohort study.

Sepsis remains a significant challenge in patients with major trauma in the ICU. Early detection and treatment are crucial for improving outcomes and reducing mortality rates. Nonetheless, clinical tools for predicting sepsis among patients with major trauma are limited. This study aimed to develop and validate an artificial intelligence (AI) platform for predicting the risk of sepsis among patients with major trauma. This study involved 961 patients, with prospective analysis of data from 244 patients with major trauma at our hospital and retrospective analysis of data from 717 patients extracted from a database in the United States. The patients from our hospital constituted the model development cohort, and the patients from the database constituted the external validation cohort. The patients in the model development cohort were randomly divided into a training cohort and an internal validation cohort at a ratio of 8:2. The machine learning algorithms used to train models included logistic regression (LR), decision tree (DT), extreme gradient boosting machine (eXGBM), neural network (NN), random forest (RF), and light gradient boosting machine (LightGBM). The incidence of sepsis for the model development cohort was 43.44%. Twelve predictors, including gender, abdominal trauma, open trauma, red blood cell count, heart rate, respiratory rate, injury severity score, sequential organ failure assessment score, Glasgow coma scale, smoking, total protein concentrations, and hematocrit, were used as features in the final model. Internal validation showed that the NN model had the highest area under the curve (AUC) of 0.932 (95% CI: 0.917-0.948), followed by the LightGBM and eXGBM models with AUCs of 0.913 (95% CI: 0.883-0.930) and 0.912 (95% CI: 0.880-0.935), respectively. In the external validation cohort, the eXGBM model (AUC: 0.891, 95% CI: 0.866-0.914) had the highest AUC value, followed by the LightGBM model (AUC: 0.886, 95% CI: 0.860-0.906), and the AUC value of the NN model was only 0.787 (95% CI: 0.751-0.829). Considering the predictive performance for both the internal and external validation cohorts, the LightGBM model had the highest score of 82, followed by the eXGBM (81) and NN (76) models. Thus, the LightGBM was emerged as the optimal model, and it was deployed online as an AI application. This study develops and validates an AI application to effectively assess the susceptibility of patients with major trauma to sepsis. The AI application equips healthcare professionals with a valuable tool to promptly identify individuals at high risk of developing sepsis. This will facilitate clinical decision-making and enable early intervention.

Artificial intelligence approaches for phenotyping heart failure in U.S. Veterans Health Administration electronic health record.

Heart failure (HF) is a clinical syndrome with no definitive diagnostic tests. HF registries are often based on manual reviews of medical records of hospitalized HF patients identified using International Classification of Diseases (ICD) codes. However, most HF patients are not hospitalized, and manual review of big electronic health record (EHR) data is not practical. The US Department of Veterans Affairs (VA) has the largest integrated healthcare system in the nation, and an estimated 1.5 million patients have ICD codes for HF (HF ICD-code universe) in their VA EHR. The objective of our study was to develop artificial intelligence (AI) models to phenotype HF in these patients. The model development cohort (n=20000: training, 16000; validation 2000; testing, 2000) included 10000 patients with HF and 10000 without HF who were matched by age, sex, race, inpatient/outpatient status, hospital, and encounter date (within 60days). HF status was ascertained by manual chart reviews in VA's External Peer Review Program for HF (EPRP-HF) and non-HF status was ascertained by the absence of ICD codes for HF in VA EHR. Two clinicians annotated 1000 random snippets with HF-related keywords and labelled 436 as HF, which was then used to train and test a natural language processing (NLP) model to classify HF (positive predictive value or PPV, 0.81; sensitivity, 0.77). A machine learning (ML) model using linear support vector machine architecture was trained and tested to classify HF using EPRP-HF as cases (PPV, 0.86; sensitivity, 0.86). From the 'HF ICD-code universe', we randomly selected 200 patients (gold standard cohort) and two clinicians manually adjudicated HF (gold standard HF) in 145 of those patients by chart reviews. We calculated NLP, ML, and NLP+ML scores and used weighted F scores to derive their optimal threshold values for HF classification, which resulted in PPVs of 0.83, 0.77, and 0.85 and sensitivities of 0.86, 0.88, and 0.83, respectively. HF patients classified by the NLP+ML model were characteristically and prognostically similar to those with gold standard HF. All three models performed better than ICD code approaches: one principal hospital discharge diagnosis code for HF (PPV, 0.97; sensitivity, 0.21) or two primary outpatient encounter diagnosis codes for HF (PPV, 0.88; sensitivity, 0.54). These findings suggest that NLP and ML models are efficient AI tools to phenotype HF in big EHR data to create contemporary HF registries for clinical studies of effectiveness, quality improvement, and hypothesis generation.

External Validation of a Risk Score Model for Predicting Major Clinical Events in Adults After Atrial Switch.

A risk model has been proposed to provide a patient individualized estimation of risk for major clinical events (heart failure events, ventricular arrhythmia, all-cause mortality) in patients with transposition of the great arteries and atrial switch surgery. We aimed to externally validate the model. A retrospective, multicentric, longitudinal cohort of 417 patients with transposition of the great arteries (median age, 24 years at baseline [interquartile range, 18-30]; 63% men) independent of the model development and internal validation cohort was studied. The performance of the prediction model in predicting risk at 5 years was assessed, and additional predictors of major clinical events were evaluated separately in our cohort. Twenty-five patients (5.9%) met the major clinical events end point within 5 years. Model validation showed good discrimination between high and low 5-year risk patients (Harrell C index of 0.73 [95% CI, 0.65-0.81]) but tended to overestimate this risk (calibration slope of 0.20 [95% CI, 0.03-0.36]). In our population, the strongest independent predictors of major clinical events were a history of heart failure and at least mild impairment of the subpulmonary left ventricle function. We reported the first external validation of a major clinical events risk model in a large cohort of adults with transposition of the great arteries. The model allows for distinguishing patients at low risk from those at intermediate to high risk. Previous episode of heart failure and subpulmonary left ventricle dysfunction appear to be key markers in the prognosis of patients. Further optimizing risk models are needed to individualize risk predictions in patients with transposition of the great arteries.

A nomogram to predict cognitive function impairment in patients with chronic kidney disease: A national cross-sectional survey

BackgroundCognitive function impairment (CFI) is common in patients with chronic kidney disease (CKD) and significantly impacts treatment adherence and quality of life. This study aims to create a simplified nomogram for early CFI risk detection. MethodsData were obtained from the National Health and Nutrition Examination Survey cycles spanning from 1999 to 2002 and again from 2011 to 2014. Stepwise logistic regression was used to select variables and construct a CFI risk prediction model. Furthermore, C-statistic and Brier Score (BS) assessed model performance. Additionally, Kaplan–Meier survival curves were utilised to assess risk group-death prognosis relationships. ResultsOf the 545 participants in the CKD model development cohort, a total of 146 (26.8 %) had CFI. The final model included the variables of age, race, education, annual family income, body mass index, estimated glomerular filtration rate, serum albumin and uric acid. The model had a C-statistic of 0.808 (95 % confidence interval (CI): 0.769–0.847) and a BS of 0.149. Furthermore, the 5-fold cross-validation internal C-statistic was 0.764 (interquartile range: 0.763–0.807) and BS was 0.154. Upon external validation, the model's C-statistic decreased to 0.752 (95 % CI: 0.654–0.850) and its BS increased to 0.182. The Kaplan–Meier survival curves demonstrated that intermediate-to-high-risk participants had shorter overall survival time than low-risk participants (log-rank test: p = 0.00042). ConclusionsThis study established an effective nomogram for predicting CFI in patients with CKD, which can be used for the early detection of CFI and guide the treatment of patients with CKD.

Deep learning on pre-procedural computed tomography and clinical data predicts outcome following stroke thrombectomy

BackgroundDeep learning using clinical and imaging data may improve pre-treatment prognostication in ischemic stroke patients undergoing endovascular thrombectomy (EVT).MethodsDeep learning models were trained and tested on baseline clinical and imaging...

Predicting Risk of Colorectal Cancer After Adenoma Removal in a Large Community-Based Setting.

Colonoscopy surveillance guidelines categorize individuals as high or low risk for future colorectal cancer (CRC) based primarily on their prior polyp characteristics, but this approach is imprecise, and consideration of other risk factors may improve postpolypectomy risk stratification. Among patients who underwent a baseline colonoscopy with removal of a conventional adenoma in 2004-2016, we compared the performance for postpolypectomy CRC risk prediction (through 2020) of a comprehensive model featuring patient age, diabetes diagnosis, and baseline colonoscopy indication and prior polyp findings (i.e., adenoma with advanced histology, polyp size ≥10 mm, and sessile serrated adenoma or traditional serrated adenoma) with a polyp model featuring only polyp findings. Models were developed using Cox regression. Performance was assessed using area under the receiver operating characteristic curve (AUC) and calibration by the Hosmer-Lemeshow goodness-of-fit test. Among 95,001 patients randomly divided 70:30 into model development (n = 66,500) and internal validation cohorts (n = 28,501), 495 CRC were subsequently diagnosed; 354 in the development cohort and 141 in the validation cohort. Models demonstrated adequate calibration, and the comprehensive model demonstrated superior predictive performance to the polyp model in the development cohort (AUC 0.71, 95% confidence interval [CI] 0.68-0.74 vs AUC 0.61, 95% CI 0.58-0.64, respectively) and validation cohort (AUC 0.70, 95% CI 0.65-0.75 vs AUC 0.62, 95% CI 0.57-0.67, respectively). A comprehensive CRC risk prediction model featuring patient age, diabetes diagnosis, and baseline colonoscopy indication and polyp findings was more accurate at predicting postpolypectomy CRC diagnosis than a model based on polyp findings alone.

Abstract TP144: Deep Learning on Pre-Procedural CT and Clinical Data Predicts Outcome Following Stroke Thrombectomy

Background: Deep learning using clinical and imaging data may improve pre-treatment prognostication in ischemic stroke patients undergoing endovascular thrombectomy (EVT). Methods: Deep learning models were trained and tested on pre-treatment clinical and imaging (CT head and CT angiography) data to predict 3-month functional outcomes in ischemic stroke patients who underwent EVT. Logistic regression models were constructed to compare their performance with the deep learning models. An external validation dataset was used to validate the models. The MR PREDICTS prognostic tool was tested on the external validation set, and its performance was compared with the deep learning and logistic regression models. Results: There were a total of 975 patients (550 men; mean±SD age 67.5±15.1 years), including 778 patients from the model development cohort, and 197 patients in the external validation cohort. The deep learning model trained on CT head and clinical data, and the logistic regression model (clinical data alone) demonstrated the strongest discriminative abilities for 3-month functional outcome and were comparable (AUC 0.811 versus 0.817, Q=0.82). Both models exhibited superior prognostic performance than the other deep learning (CT head alone, CT angiography alone) and MR PREDICTS models (all Q&lt;0.05). Conclusions: The discriminative performance of deep learning for predicting functional outcome following EVT was comparable to logistic regression. Future studies should focus on whether incorporating procedural and post-procedural data significantly improves model performance.

Digital image analysis and machine learning-assisted prediction of neoadjuvant chemotherapy response in triple-negative breast cancer

BackgroundPathological complete response (pCR) is associated with favorable prognosis in patients with triple-negative breast cancer (TNBC). However, only 30–40% of TNBC patients treated with neoadjuvant chemotherapy (NAC) show pCR, while the remaining 60–70% show residual disease (RD). The role of the tumor microenvironment in NAC response in patients with TNBC remains unclear. In this study, we developed a machine learning-based two-step pipeline to distinguish between various histological components in hematoxylin and eosin (H&E)-stained whole slide images (WSIs) of TNBC tissue biopsies and to identify histological features that can predict NAC response.MethodsH&E-stained WSIs of treatment-naïve biopsies from 85 patients (51 with pCR and 34 with RD) of the model development cohort and 79 patients (41 with pCR and 38 with RD) of the validation cohort were separated through a stratified eightfold cross-validation strategy for the first step and leave-one-out cross-validation strategy for the second step. A tile-level histology label prediction pipeline and four machine-learning classifiers were used to analyze 468,043 tiles of WSIs. The best-trained classifier used 55 texture features from each tile to produce a probability profile during testing. The predicted histology classes were used to generate a histology classification map of the spatial distributions of different tissue regions. A patient-level NAC response prediction pipeline was trained with features derived from paired histology classification maps. The top graph-based features capturing the relevant spatial information across the different histological classes were provided to the radial basis function kernel support vector machine (rbfSVM) classifier for NAC treatment response prediction.ResultsThe tile-level prediction pipeline achieved 86.72% accuracy for histology class classification, while the patient-level pipeline achieved 83.53% NAC response (pCR vs. RD) prediction accuracy of the model development cohort. The model was validated with an independent cohort with tile histology validation accuracy of 83.59% and NAC prediction accuracy of 81.01%. The histological class pairs with the strongest NAC response predictive ability were tumor and tumor tumor-infiltrating lymphocytes for pCR and microvessel density and polyploid giant cancer cells for RD.ConclusionOur machine learning pipeline can robustly identify clinically relevant histological classes that predict NAC response in TNBC patients and may help guide patient selection for NAC treatment.

Development and External Validation of a Prediction Model for Colorectal Cancer Among Patients Awaiting Surveillance Colonoscopy Following Polypectomy

Demand for surveillance colonoscopy can sometimes exceed capacity, such as during and following the coronavirus disease 2019 pandemic, yet no tools exist to prioritize the patients most likely to be diagnosed with colorectal cancer (CRC) among those awaiting surveillance colonoscopy. We developed a multivariable prediction model for CRC at surveillance comparing performance to a model that assigned patients as low or high risk based solely on polyp characteristics (guideline-based model). Logistic regression was used for model development among patients receiving surveillance colonoscopy in 2014-2019. Candidate predictors included index colonoscopy indication, findings, and endoscopist adenoma detection rate, and patient and clinical characteristics at surveillance. Patients were randomly divided into model development (n= 36,994) and internal validation cohorts (n= 15,854). External validation was performed on 30,015 patients receiving surveillance colonoscopy in 2020-2022, and the multivariable model was then updated and retested. One hundred fourteen, 43, and 71 CRCs were detected at surveillance in the 3 cohorts, respectively. Polyp size ≥10 mm, adenoma detection rate <32.5% or missing, patient age, and ever smoked tobacco were significant CRC predictors; this multivariable model outperformed the guideline-based model (internal validation cohort area under the receiver-operating characteristic curve: 0.73, 95% confidence interval (CI): 0.66-0.81 vs 0.52, 95% CI: 0.45-0.60). Performance declined at external validation but recovered with model updating (operating characteristic curve: 0.72 95% CI: 0.66-0.77). When surveillance colonoscopy demand exceeds capacity, a prediction model featuring common clinical predictors may help prioritize patients at highest risk for CRC among those awaiting surveillance. Also, regular model updates can address model performance drift.

Machine learning models to predict the warfarin discharge dosage using clinical information of inpatients from South Korea

As warfarin has a narrow therapeutic window and obvious response variability among individuals, it is difficult to rapidly determine personalized warfarin dosage. Adverse drug events(ADE) resulting from warfarin overdose can be critical, so that typically physicians adjust the warfarin dosage through the INR monitoring twice a week when starting warfarin. Our study aimed to develop machine learning (ML) models that predicts the discharge dosage of warfarin as the initial warfarin dosage using clinical data derived from electronic medical records within 2 days of hospitalization. During this retrospective study, adult patients who were prescribed warfarin at Asan Medical Center (AMC) between January 1, 2018, and October 31, 2020, were recruited as a model development cohort (n = 3168). Additionally, we created an external validation dataset (n = 891) from a Medical Information Mart for Intensive Care III (MIMIC-III). Variables for a model prediction were selected based on the clinical rationale that turned out to be associated with warfarin dosage, such as bleeding. The discharge dosage of warfarin was used the study outcome, because we assumed that patients achieved target INR at discharge. In this study, four ML models that predicted the warfarin discharge dosage were developed. We evaluated the model performance using the mean absolute error (MAE) and prediction accuracy. Finally, we compared the accuracy of the predictions of our models and the predictions of physicians for 40 data point to verify a clinical relevance of the models. The MAEs obtained using the internal validation set were as follows: XGBoost, 0.9; artificial neural network, 0.9; random forest, 1.0; linear regression, 1.0; and physicians, 1.3. As a result, our models had better prediction accuracy than the physicians, who have difficulty determining the warfarin discharge dosage using clinical information obtained within 2 days of hospitalization. We not only conducted the internal validation but also external validation. In conclusion, our ML model could help physicians predict the warfarin discharge dosage as the initial warfarin dosage from Korean population. However, conducting a successfully external validation in a further work is required for the application of the models.

Assessing survival post-kidney transplantation in Australia: A multivariable prediction model.

Kidney transplantation remains the preferred standard of care for patients with kidney failure. Most patients do not access this treatment and wide variations exist in which patients access transplantation. We sought to develop a model to estimate post-kidney transplant survival to inform more accurate comparisons of access to kidney transplantation. Development and validation of prediction models using demographic and clinical data from the Australia and New Zealand Dialysis and Transplant Registry. Adult deceased donor kidney only transplant recipients between 2000 and 2020 were included. Cox proportional hazards regression methods were used with a primary outcome of patient survival. Models were evaluated using Harrell's C-statistic for discrimination, and calibration plots, predicted survival probabilities and Akaike Information Criterion for goodness-of-fit. The model development and validation cohorts included 11 302 participants. Most participants were male (62.8%) and Caucasian (79.2%). Glomerulonephritis was the most common cause of kidney disease (45.6%). The final model included recipient, donor, and transplant related variables. The model had good discrimination (C-statistic, 0.72; 95% confidence interval (CI) 0.70-0.74 in the development cohort, 0.70; 95% CI 0.67-0.73 in the validation cohort and 0.72; 95% CI 0.69-0.75 in the temporal cohort) and was well calibrated. We developed a statistical model that predicts post-kidney transplant survival in Australian kidney failure patients. This model will aid in assessing the suitability of kidney transplantation for patients with kidney failure. Survival estimates can be used to make more informed comparisons of access to transplantation between units to better measure equity of access to organ transplantation.

Development and validation of CT‑based radiomics model of PET-negative residual CT masses: apotential biomarker forpredicting relapse‑free survival innon-Hodgkin lymphoma patients showing complete metabolic response.

PET-negative residual CT masses (PnRCMs) are usually dismissed as nonviable post-treatment lesions in non-Hodgkin lymphoma (NHL) patients showing complete metabolic response (CMR). We aimed to develop and validate computed tomography (CT)-based radiomics model of PET-negative residual CT mass (PnRCM) for predicting relapse-free survival (RFS) in NHL patients showing CMR. A total of 224 patients who showed CMR after completing first-line chemotherapy for PET-avid NHL were recruited for model development. Patients with PnRCM were selected in accordance with the Lugano classification. Three-dimensional segmentation was done by two readers. Radiomic scores (RS) were constructed using features extracted using the Least-absolute shrinkage and selection operator analysis among radiomics features of PnRCMs showing more than substantial interobserver agreement (> 0.6). Cox regression analysis was performed with clinical and radiologic features. The performance of the model was evaluated using area under the curve (AUC). For validation, 153 patients from an outside hospital were recruited and analyzed in the same way. In the model development cohort, 68 (30.4%) patients had PnRCM. Kaplan-Meier analysis showed that patients with PnRCM had significantly (p = 0.005) shorter RFS than those without PnRCM. In Kaplan-Meier analysis, the high RS group showed significantly (p = 0.038) shorter RFS than the low-scoring group. Multivariate Cox regression analysis showed that high IPI score [hazard ratio (HR) 2.46; p = 0.02], treatment without rituximab (HR 3.821; p = 0.019) were factors associated with shorter RFS. In estimating RFS, combined model in both development and validation cohort showed AUC values of 0.81. The combined model that incorporated both clinical parameters and CT-based RS showed good performance in predicting relapse in NHL patients with PnRCM.