Title: Alternative Method to Determine High and Low Performing Facilities
Authors: Allen Hass, James Graham, Yong-Fang Kuo
Introduction: The Risk-Standardized Rate (RSR) is a statistic CMS uses to evaluate facilities. Since the standard error for the RSR has no closed form solution, a computationally intensive bootstrap is needed for confidence intervals. We calculated the 95% CI of the adjusted prediction rates at the mean based on the random intercept from the hierarchical logistic model, and compared its performance to the CMS method.
Methods: Using Medicare records from 2012 to 2015, 3 cohorts were constructed. The first evaluated potentially preventable readmission in inpatient rehabilitation facilities (IRFs). The second evaluated community discharge (CD) in IRFs. The third evaluated CD in skilled nursing home facilities.
Results: All Cohorts showed very strong agreement (? = 0.94, 0.89 and 0.97 for cohorts 1, 2 and 3 respectively).
Conclusion: As agreement between the methods was seen in each cohort, we believe that the methods will show agreement in most cases. Further study by simulation is necessary to verify whether this outcome is generalizable.
|