Predicting 30-day readmissions with preadmission electronic health record data

Area of Science:

• Health Informatics
• Predictive Modeling
• Healthcare Management

Background:

• Readmission prevention is crucial and should commence early during a patient's hospital stay.
• Early identification of high-risk patients facilitates targeted interventions within the hospital.

Purpose of the Study:

• To develop and validate a predictive model for 30-day all-cause readmissions.
• Utilize electronic health record (EHR) data available prior to hospital admission for prediction.

Main Methods:

• Retrospective cohort study of adult internal medicine admissions.
• Developed the Preadmission Readmission Detection Model (PREADM) using EHR and administrative data.
• Employed decision trees, neural networks, and logistic regression for model development and validation.

Main Results:

• The PREADM model incorporates 11 variables, including chronic conditions and prior healthcare utilization.
• Achieved a c-statistic of 0.70 in the derivation set and 0.69 in the validation set.
• Model performance was not significantly improved by including length of stay.

Conclusions:

• The PREADM model enables early identification of high-risk patients for readmission by health plans.
• Demonstrates discriminatory power comparable or superior to existing models, particularly those using post-discharge data.
• Facilitates proactive readmission prevention strategies through pre-admission risk assessment.