Cardiovascular disease (CVD) is a significant global health concern and the leading cause of death in many countries. Early detection and diagnosis of CVD can significantly reduce the risk of complications and mortality. Machine learning methods, particularly classification algorithms, have demonstrated their potential to accurately predict the risk of cardiovascular disease (CVD) by analyzing patient data. This study evaluates seven binary classification algorithms, including Random Forests, Logistic Regression, Naive Bayes, K-Nearest Neighbors (kNN), Support Vector Machines, Gradient Boosting, and Artificial Neural Networks, to understand their effectiveness in predicting CVD. Advanced preprocessing techniques, such as SMOTE–ENN for addressing class imbalance and hyperparameter optimization through Grid Search Cross-Validation, were applied to enhance the reliability and performance of these models. Standard evaluation metrics, including accuracy, precision, recall, F1-score, and Area Under the Receiver Operating Characteristic Curve (ROC-AUC), were used to assess predictive capabilities. The results show that kNN achieved the highest accuracy (99%) and AUC (0.99), surpassing traditional models like Logistic Regression and Gradient Boosting. The study examines the challenges encountered when working with datasets related to cardiovascular diseases, such as class imbalance and feature selection. It demonstrates how addressing these issues enhances the reliability and applicability of predictive models. These findings emphasize the potential of kNN as a reliable tool for early CVD prediction, offering significant improvements over previous studies. This research highlights the value of advanced machine learning techniques in healthcare, addressing key challenges and laying a foundation for future studies aimed at improving predictive models for CVD prevention.

Periprosthetic joint infection (PJI) is a serious complication that can occur after joint arthroplasty, such as hip or knee replacement surgeries. It involves the invasion of the periprosthetic space by pathogens, leading to severe inflammation and often requiring complex medical intervention. PJI is associated with significant morbidity, increased healthcare costs, and a reduced quality of life for patients. This study aims to evaluate the performance of multiple supervised machine learning models in predicting PJI using clinical and demographic data collected from patients who underwent joint arthroplasty. Eight supervised machine learning models—Logistic Regression, Random Forest, XGBoost, Artificial Neural Network (ANN), k-Nearest Neighbors (KNN), AdaBoost, Gaussian Naive Bayes (GNB), and Stochastic Gradient Descent (SGD)—were trained and tested on a dataset of 27,854 patients. Models were evaluated using accuracy, precision, recall, specificity, F1 score, and area under the ROC curve (AUC). Random Forest and XGBoost showed the best overall performance, with high accuracy and balanced metrics across all evaluation criteria. KNN also performed strongly, particularly in minimizing misclassifications. GNB and SGD yielded weaker results, with higher error rates. Random Forest, XGBoost, and KNN are the most promising models for clinical implementation in PJI prediction. Their robust performance may support earlier diagnosis and improved patient outcomes in orthopedic care. This study demonstrates that machine learning models—particularly Random Forest and XGBoost—can accurately predict periprosthetic joint infection (PJI) using structured electronic health record data. By integrating these models into preoperative assessment workflows, clinicians may be able to identify high-risk patients earlier, personalize prophylactic strategies, and reduce infection-related morbidity. The implementation of these predictive tools has the potential to enhance clinical decision-making, improve surgical outcomes, and optimize the use of healthcare resources in orthopedic practice.

The evaluation of large language models (LLMs) increasingly relies on other LLMs acting as automated judges. While this approach offers scalability and efficiency, it raises serious concerns regarding evaluator reliability, positional bias, and ranking stability. This paper presents a scalable framework for diagnosing positional bias and instability in LLM-based evaluation by using controlled pairwise comparisons judged by multiple independent language models. The system supports mirrored comparisons with reversed response order, prompt injection, and surface-level perturbations (e.g., paraphrasing, lexical noise), enabling fine-grained analysis of evaluator consistency and verdict robustness. Over 3600 pairwise comparisons were conducted across five instruction-tuned open-weight models using ten open-ended prompts. The top-performing model (gemma:7b-instruct) achieved a 66.5% win rate. Evaluator agreement was uniformly high, with 100% consistency across judges, yet 48.4% of verdicts reversed under mirrored response order, indicating strong positional bias. Kendall’s Tau analysis further showed that local model rankings varied substantially across prompts, suggesting that semantic context influences evaluator judgment. All evaluation traces were stored in a graph database (Neo4j), enabling structured querying and longitudinal analysis. The proposed framework provides not only a diagnostic lens for benchmarking models but also a blueprint for fairer and more interpretable LLM-based evaluation. These findings underscore the need for structure-aware, perturbation-resilient evaluation pipelines when benchmarking LLMs. The proposed framework offers a reproducible path for diagnosing evaluator bias and ranking instability in open-ended language tasks. Future work will apply this methodology to educational assessment tasks, using rubric-based scoring and graph-based traceability to evaluate student responses in technical domains.

Deep venous thrombosis is a critical medical condition that occurs when a blood clot forms in a deep vein, usually in the legs, and can lead to life-threatening complications such as pulmonary embolism if not detected early. Hospitalized patients, especially those with immobility or post-surgical recovery, are at higher risk of developing deep venous thrombosis, making early prediction and intervention vital for preventing severe outcomes. In this study, we evaluated the following eight machine learning models to predict deep venous thrombosis risk: logistic regression, random forest, XGBoost, artificial neural networks, k-nearest neighbors, gradient boosting, CatBoost, and LightGBM. These models were rigorously tested using key metrics, including accuracy, precision, recall, F1-score, specificity, and receiver operating characteristic curve, to determine their effectiveness in clinical prediction. Logistic regression emerged as the top-performing model, delivering high accuracy and an outstanding receiver operating characteristic curve score, which reflects its strong ability to distinguish between patients with and without deep venous thrombosis. Most importantly, the model’s high recall underscores its ability to identify nearly all true deep venous thrombosis cases, significantly reducing the risk of false negatives—a critical concern in clinical settings, where delayed or missed diagnoses can result in life-threatening complications. Although models such as random forest and eXtreme Gradient Boosting also demonstrated competitive performances, logistic regression proved the most reliable across all metrics. These results suggest that machine learning models, particularly logistic regression, have great potential for early deep venous thrombosis detection, enabling timely clinical interventions and improved patient outcomes.

Periprosthetic joint infections (PJIs) represent one of the most problematic complications following total joint replacement, with a significant impact on the patient’s quality of life and healthcare costs. The early and accurate diagnosis of a PJI remains the key factor in the management of such cases. However, with traditional diagnostic measures and risk assessment tools, the early identification of a PJI may not always be adequate. Artificial intelligence (AI) algorithms have been integrated in most technological domains, with recent integration into healthcare, providing promising applications due to their capability of analyzing vast and complex datasets. With the development and implementation of AI algorithms, the assessment of risk factors and the prediction of certain complications have become more efficient. This review aims to not only provide an overview of the current use of AI in predicting PJIs, the exploration of the types of algorithms used, and the performance metrics reported, but also the limitations and challenges that come with implementing such tools in clinical practice.

Background and objectives: Large language models (LLMs) show promise in automating open-ended evaluation tasks, yet their reliability in rubric-based assessment remains uncertain. Variability in scoring, feedback, and rubric adherence raises concerns about transparency and pedagogical validity in educational contexts. This study introduces CourseEvalAI, a framework designed to enhance consistency and fidelity in rubric-guided evaluation by fine-tuning a general-purpose LLM with authentic university-level instructional content. Methods: The framework employs supervised fine-tuning with Low-Rank Adaptation (LoRA) on rubric-annotated answers and explanations drawn from undergraduate computer science exams. Responses generated by both the base and fine-tuned models were independently evaluated by two human raters and two LLM judges, applying dual-layer rubrics for answers (technical or argumentative) and explanations. Inter-rater reliability was reported as intraclass correlation coefficient (ICC(2,1)), Krippendorff’s α, and quadratic-weighted Cohen’s κ (QWK), and statistical analyses included Welch’s t tests with Holm–Bonferroni correction, Hedges’ g with bootstrap confidence intervals, and Levene’s tests. All responses, scores, feedback, and metadata were stored in a Neo4j graph database for structured exploration. Results: The fine-tuned model consistently outperformed the base version across all rubric dimensions, achieving higher scores for both answers and explanations. After multiple-testing correction, only the Generative Pre-trained Transformer (GPT-4)—judged Technical Answer contrast remains statistically significant; other contrasts show positive trends without passing the adjusted threshold, and no additional significance is claimed for explanation-level results. Variance in scoring decreased, inter-model agreement increased, and evaluator feedback for fine-tuned outputs contained fewer vague or critical remarks, indicating stronger rubric alignment and greater pedagogical coherence. Inter-rater reliability analyses indicated moderate human–human agreement and weaker alignment of LLM judges to the human mean. Originality: CourseEvalAI integrates rubric-guided fine-tuning, dual-layer evaluation, and graph-based storage into a unified framework. This combination provides a replicable and interpretable methodology that enhances the consistency, transparency, and pedagogical value of LLM-based evaluators in higher education and beyond.