Examining Post-Hospital Care Coordination by its Outcomes: How Should We Measure Hospital Readmissions, ED visits, and Timely E&M Service Use
Abstract: Health policymakers considering hospital quality have indicated the need for measures of post-hospital care coordination, and specifically of care transitions from hospital to other settings. The goal of care co-ordination measures is to extend expectations that hospitals will link their patients with appropriate post-hospital care. CMS currently reports on hospital performance up to 30 days following discharge, signaling the position that hospital responsibility does not end at time of discharge but need to include further co-ordination efforts. Without such, patients discharged from some hospitals may experience excessive rates of emergency department utilization, or readmission to the inpatient hospital setting. Some direct measures of post-hospital care-coordination have been proposed, dealing with the issues of communication between patient and physician, management of setting transitions, and medication reconciliation. These measures, however, focus on activities, not demonstrated results for patients, and largely have not been implemented, in part due to data collection burdens. Recognizing the limitations of such direct measures of processes, we have developed a set of measures based results related to post-hospital care coordination– negative events including an ED visit or hospital readmission, and positive events such as an E&M service within 30 days of discharge. There exists a 30-day readmission measure, developed for CMS, which determines hospital risk standardized readmission rates for AMI and heart failure cases by use of a hierarchical logistic model on pooled discharge data across a three-year time period. Many of the specifications for this CMS 30-day readmission measure were employed again for our measures of readmissions, 30-day ED visits, and E&M services. We test which design features and specifications of the original approach continue to perform well, and which could be improved by modification. For this investigation we used a 20% sample of AMI and heart failure hospital discharges among Medicare elders for years 2003-2007. Alternative measure specifications tested include a) 3-year rolling average similar to the current CMS approach, b) weighted average of 3 individual years, and c) longitudinal hierarchical logistic model with a marker for time in the model. We present all results of these three approaches, including some diagnostics and reliability statistics, and discuss the strengths and weaknesses of the three approaches.