Risk Prediction Models for Unplanned 30-Day Readmission After Discharge of Patients After Undergoing Coronary...

Unplanned 30-day readmission is a critical quality indicator in clinical health care. Although the number of risk prediction models related to this research is increasing, uncertainties remain regarding their methodological rigor and clinical practical value.

This study aims to systematically evaluate risk prediction models for unplanned 30-day readmission following coronary revascularization, providing evidence to inform clinical practice and guide future model development.

We identified studies that developed or validated risk prediction models for unplanned 30-day readmission following coronary revascularization. Literature searches were conducted from database inception to June 7, 2025. Study screening and data extraction were performed independently by 2 reviewers. Methodological quality of the included studies was assessed using the Prediction model Risk Of Bias Assessment Tool (PROBAST) tool, covering both risk of bias and applicability. Meta-analyses of model discrimination, measured by the area under the receiver operating characteristic curve (AUC), and frequently reported predictors were conducted using R software (version 4.5.1).

A total of 4961 records were initially identified, of which 14 studies meeting the eligibility criteria were included, encompassing 19 prediction models. The reported 30-day unplanned readmission rates ranged from 0.7% to 18.08%. Meta-analysis showed that older age, female sex, diabetes, chronic pulmonary disease, and heart failure were significantly associated with an increased risk of unplanned readmission (P < .05), with AUC values ranging from 0.604 to 0.999. Assessment using the PROBAST tool revealed that the included studies generally had a high risk of bias, primarily attributable to poor reporting quality of the analysis methods.

Existing forecasting models for unintended readmission within 30 days after coronary revascularization show considerable variability in effectiveness and are typically vulnerable to a considerable risk of bias. Although some models show potential predictive value, most are in the early stages of development and lack reliable external validation. Future studies should strictly adhere to Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) reporting guidelines, improve transparency in study design and reporting, and prioritize independent external validation to enhance model stability and generalizability.

This research offers a thorough and current assessment of risk prediction models for unintended readmissions within 30 days after coronary revascularization, identifying the overall quality and limitations of current models. The results provide evidence-based support for clinical risk assessment, promote early screening of high-risk cases, and offer methodological recommendations for the development of future prediction models.