reports

Building a predictive model for 30-day inpatient readmission using PROC PHREG

Presented at NESUG 2012. Baltimore, MD, USA: North East SAS User Group.

Department of Health Policy and Management, Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD, USA

CONTEXT: Hospitals assess the readmission risk of their inpatients, and use this assessment to prioritize patients at risk of readmission who will benefit from transition management to outpatient care with the goal of becoming less likely to be readmitted within a 30 day time period.

PURPOSE: We are building a predictive 30-day readmission model
by using a large claims database¹ and applying a tool² that provides several claims-based markers for a time-to-readmission analysis. Cox proportional hazard regression, and in particular the SAS/STAT® PHREG procedure enables us to model time to readmission while adjusting for covariates and accounting for time censoring and person attrition. The approach also enables us to assess the impact of recurrent readmissions and time-varying covariates.

IMPACT: This paper will give the reader insight into applying PROC PHREG to model the likelihood of 30-day readmission.