Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 2 appraisals
JMIR formative research
Machine Learning-Based Prediction Model for 30-Day Emergency Department Revisits in a Medically Underserved Tertiary Hospital: Formative Retrospective Cohort Study
BACKGROUND: Emergency department (ED) revisits are critical quality indicators, particularly in medically underserved areas, where traditional prediction tools show limited performance. Machine learning (ML) approaches may offer improved predictive performance for identifying high-risk patients. OBJECTIVE: This formative study aimed to develop and validate an ML-based model for predicting 30-day ED revisits using electronic health records from a tertiary hospital serving a medically underserved area in South Korea and to evaluate its clinical utility through interpretability analysis and risk stratification. METHODS: This retrospective cohort study analyzed 36,230 adult patients visiting the Gangneung Asan Hospital ED in 2023. We developed and compared 3 ML models (extreme gradient boosting [XGBoost], random forest, and ElasticNet) using electronic health records. Model interpretability was ensured through Shapley additive explanations (SHAP) analysis, and clinical utility was evaluated through 5-tier risk stratification. RESULTS: Among 36,230 patients, 798 (2.2%) revisited within 30 days. XGBoost achieved superior performance with an area under the receiver operating characteristic curve of 0.90 (95% CI 0.88-0.92), a sensitivity of 0.94, and a specificity of 0.69. The SHAP analysis identified ED length of stay, oxygen saturation, systolic blood pressure, computed tomography performance, antibiotic use, and liver disease as key predictors. Risk stratification demonstrated a 25-fold difference in the actual revisit rates between the lowest (152/8450, 1.8%) and the highest (686/1500, 45.7%) risk groups. CONCLUSIONS: The XGBoost model demonstrated excellent predictive performance with high interpretability for 30-day ED revisit predictions. The implementation of this model could enable risk-stratified interventions and more efficient resource allocation in medically underserved settings, potentially reducing unnecessary revisits and improving patient outcomes. This formative study establishes feasibility and provides a foundation for future multicenter validation studies in similar medically underserved settings.
30 May 2026
Read appraisal →BMJ open
Developing and validating an electronic health record-embedded AI model for managing multimorbid hospitalisation risk in patients with chronic RESpiratory disease (AiRES): a study protocol
BACKGROUND: Chronic respiratory diseases (CRDs), such as asthma and chronic obstructive pulmonary disease (COPD), are heterogeneous conditions with a high multimorbidity burden. However, existing risk assessment instruments prioritise physiological measures while overlooking systemic comorbidities. We aim to develop and validate an electronic health record (EHR)-embedded artificial intelligence (AI) model-AiRES (AI in patients with RESpiratory disease)-to predict the 30-day, 90-day and 180-day risks of all-cause and index-disease hospitalisations. This model represents a first step towards a clinical decision support tool for personalised multimorbidity management in patients with CRD. METHOD AND ANALYSIS: Patients aged ≥18 years with a validated case definition of asthma and COPD will be identified from Singapore health administrative data (2012-2020). Candidate predictors will include age, sex, ethnicity, housing type, and comorbidities, measured across multiple care settings as visit frequency, grouped at quarterly intervals in Year 1 and annually for Years 2 and 3 over a 3-year lookback window. We will predict 30-day, 90-day, and 180-day risks of (1) all-cause and (2) asthma/COPD-specific hospital admissions using up to five randomly selected index dates per individual. Three machine learning algorithms-logistic regression (LR) with Lasso regularisation, eXtreme Gradient Boosting, and Categorical Boosting-will be trained using 10-fold cross-validation (CV) with an ensemble feature selection strategy. The optimal model, selected based on performance and feature importance, will be benchmarked against two reference models: a full LR and a Zero-Inflated Negative Binomial regression with hospitalisation history as the sole predictor. Discrimination and calibration will be assessed using internal-external cluster-based and temporal CV. Clinical utility will be evaluated using decision curve analysis. ETHICS AND DISSEMINATION: This study obtained ethics approval from the National University of Singapore (NUS-IRB-2024-849). Results will be published in international peer-reviewed journals.
15 May 2026
Read appraisal →