3119 — Within-hospital variation in 30-day readmission rates: implications for measuring quality

Burke RE, Denver VA Medical Center, University of Colorado School of Medicine; Glorioso T, Denver VA Medical Center; Kaboli PJ, CADRE Center Iowa City; University of Iowa School of Medicine; Ho PM, Denver VA Medical Center, University of Colorado School of Medicine;

Objectives:
Medicare penalizes hospitals with high risk-adjusted 30-day readmission rates for certain conditions based on three-year averages. However, these averages may conceal significant within-hospital variability in readmission rates. A fundamental tenet of quality improvement is to reduce variation in care processes and therefore outcomes, delivering high-quality care to each patient in a similar way. Hospitals with high within-hospital variation in readmission rates may therefore be delivering lower-quality care not captured by overall averages. We sought to evaluate within-hospital variability in readmission rates and link this variability to other quality measures, such as mortality rates.

Methods:
We included all Veterans who underwent percutaneous coronary intervention (PCI) at a VA hospital between Oct 1, 2007 and September 30, 2012, excluding low-volume centers. Readmission rates and mortality were measured at 30 days and 1 year, respectively. To assess variation within site over time, we separated the data into six-month periods and generated a generalized linear mixed model using a logit link with a random intercept for PCI site. We controlled for demographics, comorbidities, PCI indication (eg unstable angina), number of diseased vessels, PCI status (ie, elective, urgent, emergent), and discharge medications (eg antiplatelets). Spearman correlations were used to evaluate the relationship between within-hospital variability and outcomes.

Results:
The analysis included 41,277 records at 48 sites. Within-site variability in readmission rates was substantial; most hospitals moved between quintiles of overall readmission rates every six month period. With three exceptions, hospitals did not demonstrate linear trends towards improvement or decline over time, but rather large fluctuation period-to-period. Sites with lower overall mean readmission rates did not have lower variation over time (rs = 0.15, p = NS). There also was no correlation between the size of sites and variation over time (rs = 0.02, p = NS). Within-hospital variability in readmission rates was weakly negatively correlated with 1-year mortality rates (rs = -0.23, p = 0.11).

Implications:
Within-hospital variability in 30-day readmission rates after PCI is substantial despite a relatively uniform population and robust risk adjustment; however, it is not strongly linked to one-year mortality rates.

Impacts:
Evaluation of the relationship of within-hospital variability to other quality measures and in other conditions is needed.