Conectar
Frailty affects over 35% of maintenance haemodialysis (MHD) patients globally—2–3 times higher than the general elderly—and is strongly linked to higher mortality, hospitalisation, and functional decline. Despite its clinical impact, frailty is often underdiagnosed in dialysis settings due to inconsistent assessments and limited resources. Existing prediction models vary widely in predictors and methods, requiring systematic review to guide clinical use and improve risk-stratified care.
To systematically identify, describe, and evaluate the existing risk prediction models for frailty in patients undergoing MHD.
Systematic review and Methodological appraisal.
A comprehensive search was conducted across multiple databases—PubMed, Web of Science Core Collection, Embase, Cochrane Library, CINAHL, China Biomedical Literature Database (CBM), Wanfang Database, VIP Database—covering studies up to November 1, 2024.
Two researchers independently conducted literature searches, screening, and data extraction. They used the Prediction Model Risk of Bias Assessment Tool (PROBAST) to evaluate the risk of bias and the applicability of the included models.
Fifteen studies (21 models) were analysed, with sample sizes 141–786 and frailty incidence 11.00%–59.57%. Model AUCs ranged 0.720–0.998 (potential overfitting at extreme values). Key predictors included age, serum albumin, gender, Charlson comorbidity index, and activities of daily living scores. Methodological appraisal using PROBAST revealed moderate applicability but high bias risks: 53% of studies used retrospective designs, 95% lacked external validation, and limitations included small samples, non-standard variable selection, and inadequate handling of missing data.
While models demonstrate initial predictive utility, widespread bias and developmental-stage limitations hinder clinical application. Future research must prioritise TRIPOD-guided model development, emphasising large prospective cohorts, rigorous validation, and transparent reporting to enhance reliability and clinical utility in frailty risk stratification for MHD patients.
To systematically identify and appraise existing risk prediction models for EN aspiration in adult inpatients.
A systematic search was conducted across PubMed, Web of Science Core Collection, Embase, Cochrane Library, CINAHL, China National Knowledge Infrastructure (CNKI), Wanfang Database, China Biomedical Literature Database (CBM) and VIP Database from inception to 1 March 2025.
Systematic review of observational studies.
Two researchers independently performed literature screening and data extraction using the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS). The Prediction Model Risk of Bias Assessment Tool (PROBAST) was employed to evaluate both the risk of bias and the clinical applicability of the included models.
A total of 17 articles, encompassing 29 prediction models, were included. The incidence of aspiration was 9.45%–57.00%. Meta-analysis of high-frequency predictors identified the following significant predictors of aspiration: history of aspiration, depth of endotracheal intubation, impaired consciousness, sedation use, nutritional risk, mechanical ventilation and gastric residual volume (GRV). The area under the curve (AUC) was 0.771–0.992. Internal validation was performed in 12 studies, while both internal and external validation were conducted in 5 studies. All studies demonstrated a high risk of bias, primarily attributed to retrospective design, geographic bias (all from different parts of China), inadequate data analysis, insufficient validation strategies and lack of transparency in the research process.
Current risk prediction models for enteral nutrition-associated aspiration show moderate to high discriminative accuracy but suffer from critical methodological limitations, including retrospective design, geographic bias (all models derived from Chinese cohorts, limiting global generalisability) and inconsistent outcome definitions.
Recognising the high bias of existing models, prospective multicentre data and standardised diagnostics are needed to develop more accurate and clinically applicable predictive models for enteral nutrition malabsorption.
Not applicable.
PROSPERO: CRD420251016435
FreshRSS 