Predicting readmission risk following coronary artery bypass surgery at the time of admission

Area of Science:

• Cardiology
• Health Services Research
• Medical Informatics

Background:

• Reducing hospital readmissions is a national healthcare priority.
• Early identification of patients at high risk for readmission after coronary artery bypass graft surgery (CABG) can optimize discharge planning.
• Predictive models aid in resource allocation and patient care enhancement.

Purpose of the Study:

• To develop and validate predictive models for 30-day inpatient readmission following CABG.
• To assess the utility of data available at admission, discharge, and from the STS Registry for readmission prediction.
• To identify the most effective data points for early risk stratification.

Main Methods:

• Logistic regression models were developed using data from CABG patients (N=1159) between January 2010 and April 2013.
• Models utilized data available at admission, discharge, and from the STS Registry.
• Model performance was evaluated using the c-index and validated on a separate cohort (N=1018) from May 2013 to September 2015.

Main Results:

• The model using admission data demonstrated reasonable discriminative ability (C-index=0.673).
• Models incorporating discharge and STS Registry data showed marginal improvements in predictive accuracy (C-indices=0.700 and 0.714, respectively).
• Internal and external validation confirmed the predictive capability of the admission-based model, with slight enhancements from additional data.

Conclusions:

• Risk prediction models utilizing early admission data are effective for identifying patients at risk of readmission post-CABG.
• Incorporating STS Registry data did not significantly enhance model performance.
• Simplified models based on readily available admission data are sufficient for early identification of high-risk patients, facilitating timely interventions.