Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 5 appraisals
Scientific data
Anonymized but Useful Synthetic Tabular Health Data for AI based Fall Risk Assessment
Artificial Intelligence (AI) bears potential for improving health care, but this depends on the availability of open-access, realistic, and useful data. To facilitate AI model development in health care we release SynTabFall, a novel synthetic dataset for fall risk assessment. With a total of 745,380 samples and 44 attributes such as demographics, diseases, mobility and cognition related risk factors, this tabular dataset allows for training fall risk prediction models without access to the original patient data. Models trained on our synthetic dataset can reach predictive performance scores in fall risk assessment which are on par with models trained on real data. To support others in sharing data we also describe a process that was developed over multiple years in one of Germany's largest hospitals in close collaboration between data protection officers, health care staff, informaticians and AI engineers. The proposed data sharing approach combines established methods for anonymization and modern generative AI (genAI) methods for synthesizing tabular data and allows for sharing health care data responsibly without sacrificing its utility. We release the synthetic fall risk dataset along with the software developed for synthetic data generation and evaluation.
26 July 2026
Read appraisal →BMC research notes
From manual entry to machine precision: challenges and evolution of metadata schema development in collaborative research centers
OBJECTIVE: Metadata standardization in collaborative biomedical research must balance interoperability with domain-specific detail. We describe a parent-template approach in which a baseline schema from the nephrology-focused CRC 1453 NephGen was adapted for the tumor-immunology CRC OncoEscape and the perinatal-immunology CRC Pilot. RESULTS: The derivation process produced three structurally compatible yet vocabulary-divergent schemas. Pilot required the highest granularity (324 levels), followed by NephGen (287) and OncoEscape (283). Vocabulary reuse from the NephGen baseline was limited: 134 of 283 OncoEscape levels (47%) and 113 of 324 Pilot levels (35%) were retained unchanged. The main adaptations were not only expanded level lists, such as cell lines and mouse lines, but also new CRC-specific query dimensions, including "Oncogenes" in OncoEscape and "Timeline" in Pilot. In the context of AI-assisted extraction, we use the term instruction set to denote a schema that specifies target fields, expected granularity, example values, and validation resources for each metadata dimension, rather than a simple drop-down form or a free-text prompt template.
8 July 2026
Read appraisal →Journal of medical systems
Security Analysis of a Federated Learning Framework for Medical Image-to-Image Translation
Federated Learning (FL) emerged as a privacy-preserving paradigm for collaborative training of deep learning models across institutions without sharing patient data. This approach has been applied to complex tasks such as medical image-to-image (I2I) translation, including MRI-to-synthetic CT (sCT) generation. However, existing federated I2I frameworks often assume privacy preservation as an inherent property of FL rather than a requirement to be explicitly validated, leaving their robustness to representative adversarial threat scenarios largely unexplored. In this study, we evaluated the vulnerability of a federated MRI-to-sCT translation framework (FedSynthCT-Brain) to three representative attack classes: Deep Leakage from Gradients (DLG), Federated Membership Inference Attack (FedMIA), and data poisoning. The efficacy of corresponding defense mechanisms, such as Secure Aggregation (SecAgg) and Byzantine-robust median aggregation (FedMedian), were assessed. DLG enabled only the recovery of coarse anatomical structures, with no clinically identifiable details (SSIM ≤ 0.16, PSNR ≤ 11 dB) across clients, suggesting limited vulnerability under the evaluated DLG setting. In contrast, FedMIA achieved high membership discrimination, with AUC scores between 0.92 and 0.99, revealing a critical privacy vulnerability. The introduction of SecAgg reduced AUC values to near-random levels (0.23-0.56) across all centers without impacting synthesis quality. Under high-noise poisoning, the standard federated averaging (FedAvg) aggregation rendered the federation inoperative, while FedMedian restored performance close to the no-poisoning baseline in most scenarios, with significant residual degradation in specific center configurations. At low noise levels, the advantage of FedMedian was less consistent, as low-level noise injection may be indistinguishable from natural heterogeneity across centers, potentially enabling stealthy degradation. These findings demonstrate that federated I2I translation frameworks are not inherently secure and require explicit, multi-layered evaluation. As FL is increasingly adopted in clinical workflows, our results underscore the necessity of integrating cryptographic, algorithmic, and infrastructural safeguards for secure deployment.
6 July 2026
Read appraisal →JMIR medical informatics
Graph Network Feature Space Fusion for Predicting Irregularly Sampled Medical Time-Series Data: Deep Learning Model Development and Validation Study
BACKGROUND: Irregularly sampled data, as a common data structure in the medical field, is frequently observed in emergency clinical datasets. It poses problems such as unequal sampling time intervals and frequencies, making it difficult to align the data without losing information for input into models. Meanwhile, due to information loss in the dataset, it is also difficult for the model to effectively analyze the overall data variation and predict the complex and dynamic conditions of emergency patients in the future. OBJECTIVE: This study aimed to address the impact of issues such as missing values and irregularly sampled time intervals on prediction results. This paper proposes a method that converts time-series data into a graph network structure and develops a prediction model for graph data to better support complex tasks. METHODS: In this paper, feature channels and measurement times in time-series data are constructed as nodes, and feature measurements at time points are constructed as edge weights. The point feature space and edge weight space are extracted by a graph convolutional neural network and a gated convolutional attention mechanism network, respectively, and feature fusion is carried out under the action of the learnable head. RESULTS: The proposed method was tested on 4 datasets and compared with both irregularly sampled time-series predictors and state-of-the-art methods for complete data. In addition, a dedicated analysis was conducted leveraging data from 5 key clinical treatment phases. In single-target classification and single-target regression tasks, the method proposed in this study outperforms other comparative methods in most experiments (accuracy=0.74-0.88, area under the curve=0.70-0.87, and mean absolute error=0.038-0.061). Meanwhile, this study constructs clinical phenotypes during patients' hospitalization as multitarget prediction tasks, and the proposed model also outperforms comparative methods in multitarget classification tasks (mean accuracy=0.75-0.91 and mean area under the curve=0.70-0.75). Ablation experiments verify the effectiveness of fusion between the node feature space and the edge feature space. The interpretability analysis of time points identifies key features that require focused attention for patients across different target tasks and time periods. CONCLUSIONS: The proposed model can effectively analyze medical data with missing values and uneven sampling. It achieves high prediction accuracy in both regression prediction tasks and classification tasks. The graph network-based model structure endows the model with better interpretability, enabling effective exploration of factors that trigger changes.
5 July 2026
Read appraisal →Calcified tissue international
Artificial Intelligence Approaches for Osteoporotic Fracture Risk Prediction Using Administrative Health Data: A Systematic Review
To systematically review Machine Learning (ML) fracture risk prediction models developed solely using administrative data, evaluating their development, performance, risk of bias, and concerns regarding applicability. A systematic search was conducted in PubMed, Embase, IEEE, and Web of Science following PRISMA guidelines (up to November 2025). We included studies developing or validating ML models for osteoporotic fracture risk in adults using administrative data without clinical measurements. Risk of bias and concerns for applicability were assessed using the PROBAST tool. Seven studies were included from 3435 initial records. A range of of ML models were utilized, including Random Forests, XGBoost, LASSO-regularized Logistic Regression and Neural Networks. Discriminative performance was moderate-to-good, with best Area Under the Curve (AUC) values ranging from 0.818 for osteoporotic fractures in general to 0.905 for hip fractures only. While five studies showed low risk of bias, five raised applicability concerns due to using specific subpopulations or features with database characteristics. Only two studies used external validation and none directly evaluated clinical utility of the models. ML models using solely administrative data show some promise as scalable, automated tools for fracture risk prediction without requiring manual clinical input. However, clinical implementation is hindered by limited external validation and a lack of formal utility evaluation of the models. Future research should prioritize more rigorous external validation and recalibration of the existing models, more so than the development of novel models, ensuring they are robust, interpretable, and integrable into future clinical workflows.
28 June 2026
Read appraisal →