3099 — High Facility Variation in 30-Day Readmissions in Veterans >= 65 Years of Age
Pugh JA, Pugh MJ, and Finley EP, STXVHCS, UTHSCSA; Amuan M, Bedford COE; Noel PH, South Texas Veterans Healthcare System, University of Texas Health Science Center;
Thirty-day readmission rates and their preventability is a hotly debated topic for Medicare and the VA. We sought evidence of facility-level variation after adjusting for numerous patient-level predictors of readmission, as well as some facility-level characteristics.
We identified Veterans aged 65 by October 1, 2006 who received VA care at least once each year (FY04-FY06). We identified individuals who had a readmission within 30 days of the discharge from an index admission in FY06. We controlled for individual demographic (e.g., age, sex, race, copay status) and clinical characteristics such as: mental health diagnoses, chronic medical conditions, the number of prescribed medications, incident or chronic high-risk medication exposure, prior year emergency, and hospital or geriatric care. Facility-level variables included: teaching hospital, presence of geriatric education program, and percentage of patients cared for by facility that are >= 65 years. We used GLIMMIX to identify facility-level variables associated with 30-day readmissions, and to determine if additional facility-level variance remained a significant predictor of readmissions after controlling for individual-level characteristics.
The mean readmission rate was 17.3%, with a range among the 128 VA facilities from 5.4% to 25.5%. After controlling for demographic and clinical characteristics of patients, teaching status (OR: 1.24, 1.05-1.48) predicted readmissions. Examination of random effects identified significant remaining variation among facilities, with 30 facilities having significantly worse readmission rates than the group as a whole, and 21 facilities having significantly better rates. Significant patient-level characteristics in the model included non-modifiable (male, black, >85, comorbid mental and physical illness, copay status) and potentially modifiable (high-risk medication exposure, polypharmacy, prior utilization) factors.
High-facility variation in 30-day readmission rates are not explained by patient-level characteristics. Examination in more depth of facility practices with regard to patient transitions, care coordination, and level of PACT implementation will be necessary to both explain and reduce this variation.
High-facility level variation in readmission rates is an opportunity to study both best practices at the 21 high-performing facilities and opportunities for improvement at the low-performing facilities.