Development and Temporal Validation of a Discharge-Based Administrative Data Risk Model for 30-Day Unplanned...

Area of Science:

• Oncology
• Health Services Research
• Medical Informatics

Background:

• Postoperative complications after endometrial cancer staging surgery frequently occur post-discharge.
• Thirty-day unplanned hospital readmission is a critical outcome following gynecologic oncology surgery.
• Existing risk models lack real-time usability at discharge and robust validation.

Purpose of the Study:

• To develop and validate a discharge-time risk prediction model for 30-day unplanned hospital readmission.
• The model utilizes administrative data for patients undergoing endometrial cancer staging surgery.
• To improve risk stratification and transitional care planning at the point of discharge.

Main Methods:

• Retrospective cohort study using the Healthcare Cost and Utilization Project Nationwide Readmissions Database (HCUP-NRD).
• Development and internal validation using multivariable logistic regression with bootstrap resampling (2016-2022).
• Temporal validation using subsequent NRD data (2020-2022) and decision curve analysis.

Main Results:

• The model demonstrated moderate discrimination (AUC 0.719) and good calibration.
• Key predictors included discharge disposition, admission type (non-elective/emergency), and Elixhauser comorbidity index.
• Temporal validation confirmed stable performance (AUC 0.708-0.713), indicating clinical utility.

Conclusions:

• A validated administrative data-based model can identify patients at high risk for 30-day readmission after endometrial cancer surgery.
• The model serves as a potential decision-support tool for discharge planning and transitional care.
• External validation and local recalibration are necessary before widespread clinical implementation.