Implementation Experience with a 30-Day Hospital Readmission Risk Score in a Large, Integrated Health System: A...

Area of Science:

• Health Informatics
• Predictive Analytics
• Healthcare Management

Background:

• Healthcare systems face pressure to improve quality outcomes and manage costs.
• Predicting 30-day hospital readmissions is a key focus for healthcare organizations.
• The Cleveland Clinic Health System (CCHS) integrated an internally validated readmission risk score into its electronic medical record (EMR).

Purpose of the Study:

• To evaluate the predictive accuracy of the CCHS readmission risk score.
• To assess performance across different hospital sites, primary discharge diagnoses, medical/surgical specialties, and patient race/ethnicity.
• To determine the score's reliability over time, including during the COVID-19 pandemic.

Main Methods:

• Retrospective cohort study design.
• Inclusion of adult patients discharged from CCHS hospitals between April 2017 and September 2020.
• Data sourced from CCHS EMR and billing databases; exclusion of specific patient groups (e.g., Oncology, Labor/Delivery, hospice, deceased).

Main Results:

• The readmission risk score achieved a c-statistic of 0.6875 across 600,872 discharges from 11 CCHS hospitals.
• Consistent performance observed across hospital sites (c-statistics ranging from 0.6762 to 0.7023), medical/surgical specialties, and racial/ethnic groups (>0.65).
• Variations in performance noted for specific diagnoses (e.g., lower for congenital anomalies, neoplasms) and COVID-19 (c-statistic 0.6387).

Conclusions:

• The CCHS readmission risk score demonstrated robust and consistent predictive accuracy over three years post-implementation.
• The score maintained performance across diverse settings, specialties, and patient demographics, including during the COVID-19 pandemic.
• Ongoing evaluation of clinical decision-making tools is essential for ensuring relevance and optimizing performance.