Same-Hospital Readmission Rate An Unreliable Predictor For All-Hospital Readmission Rate
New statistical analysis suggests that a standardized way to track and report hospital readmission rates is needed as a reliable benchmark for surgical quality outcomes
Approximately one in five Medicare patients are rehospitalized within 30 days of discharge. The Centers for Medicare & Medicaid Services (CMS) considers this rate excessive, and began reducing payments to hospitals that have excessive readmission rates in October 2012 under a provision of the Patient Protection and Affordable Care Act. While the Hospital Readmissions Reduction Program penalizes readmission to any hospital, most hospitals are only tracking same-hospital readmissions using administrative data that is recorded for billing purposes. However, according to new research findings presented at the 2013 Clinical Congress of the American College of Surgeons, same-hospital readmission rates are an unreliable surrogate for predicting all-hospital readmissions rates.
“With increasing penalization for readmissions rates, hospitals need complete information to effectively target areas for quality improvement,” said study coauthor Andrew Gonzalez, MD, JD, MPH, a research fellow in vascular surgery at the Center for Healthcare Outcomes and Policy, University of Michigan, Ann Arbor. “Under the current model, hospitals are attempting to solve the readmissions problem without havng all the puzzle pieces–they know about read-missions to their own facilities, but not about readmissions to other facilities.”
For the University of Michigan study, researchers evaluated three years of data on 660,700 Medicare patients undergoing one of three major surgical operations: coronary artery bypass grafts, hip fracture repair, and colectomy. Within this group, 86,200 patients (about 13 percent overall), had at least one readmission within 30 days of their operation. Within that 13 percent, 54,264, or about two-thirds, were readmitted to the same hospital.
Using patient level data, the researchers generated risk-adjusted rates of same-hospital and all-hospital readmissions. They then ranked hospitals from lowest to highest readmission rates and divided them into five groups (quintiles) based upon their ranking. The authors then compared how hospital performance based upon same-hospital readmission rate compared with performance under all-hospital readmissions, and found that 42 percent of hospitals were reclassified into a different quintile of performance. For example, of hospitals performing in the top quintile based on same-hospital readmission, nearly a quarter (24 percent) were reclassified when rankings were based on all-hospital readmission. This pattern was most exaggerated in the median quintile, where 55 percent were reclassified under all-hospital readmissions rankings.
“When you look at reclassification over all of the quintiles of performance, the take home message is that about 42 percent of hospitals reclassify,” Dr. Gonzalez said. “Unless you are a top or bottom performer for readmissions, your same-hospital readmission rate may be very misleading. That’s why using the same-hospital readmission rate is an unreliable predictor for your all-hospital readmission rate, but that rate is exactly what CMS penalizes hospitals for.”
The researchers conclude that in order to decrease readmissions and improve quality of care, hospitals need to have access to real-time data. This access could come in a number of forms, including a surgical quality improvement collaborative where information is quickly and easily exchanged among participants.
“As it currently stands, CMS sends institutions annual hospital-specific reports,” Dr. Gonzalez explained. “These reports include the institution’s all-hospital readmission rate, and, moreover, the provider IDs for all the other hospitals to which an institution’s patients were readmitted. Yet upon receipt, the information is already a year old. The incorporation of real-time data might significantly improve the efficiency of the quality improvement cycle.”
On the Net: