Specialty-Specific Readmission Risk Models Outperform General Models in Estimating Hepatopancreatobiliary Surgery...

Area of Science:

• Hepatopancreatobiliary (HPB) Surgery
• Medical Informatics
• Health Services Research

Background:

• Readmissions pose significant financial and patient-related burdens in hepatopancreatobiliary (HPB) surgery.
• Existing readmission prediction tools are not specifically designed or validated for the unique HPB patient population.

Purpose of the Study:

• To develop and validate procedure-specific readmission risk models for pancreatectomy and hepatectomy patients.
• To compare the performance of these novel specialty-specific models against existing general readmission risk tools.

Main Methods:

• Developed subspecialty readmission risk assessments (SRRA) for pancreatectomy and hepatectomy using 2014-15 ACS NSQIP data.
• Validated these SRRA models and six other general models in institutional cohorts (2013-2017) of HPB surgery patients.
• Compared model performance using area under the receiver operating characteristic curves (AUC).

Main Results:

• Specialty-specific readmission models demonstrated superior performance compared to general models.
• The derived pancreatectomy SRRA achieved an AUC of 0.66, and hepatectomy SRRA achieved an AUC of 0.74.
• Modified models like Readmission After Pancreatectomy (AUC=0.76) and Readmission Risk Score for hepatectomy (AUC=0.78) also outperformed general tools (e.g., LACE Index, ARIA).

Conclusions:

• Readmission risk models tailored to HPB surgery subspecialties using 30-day data are more effective than general prediction tools.
• Hospitals and surgical practices should implement specialty-specific readmission reduction strategies for HPB patients.