Influence of hospital-level practices on readmission after ischemic stroke

Abstract

Objective: To inform stroke quality improvement initiatives by determining the relationship between hospital-level stroke practices and readmission after accounting for patient-level factors.

Methods: Retrospective cohort study of adult patients hospitalized for ischemic stroke (principal ICD-9-CM codes 433.x1, 434.x1, and 436) in 5 states from 2003 to 2009 from State Inpatient Databases. The primary outcome was any unplanned readmission within 30 days. Multilevel logistic regression was used to estimate the association between hospital-level practice patterns of care (diagnostic testing, procedures, intensive care unit, tissue plasminogen activator, and therapeutic modalities) and readmission after adjustment for patient factors and whether individual patients received a given practice.

Results: Thirty-day unplanned readmission occurred in 15.2% of stroke admissions; the median hospital readmission rate was 13.6% (interquartile range 9.8%-18.2%). Of the 25 hospital practice patterns of care analyzed, 3 practices were associated with readmission: hospitals with higher use of occupational therapy and higher proportion of transfers from other hospitals had lower adjusted readmission rates, whereas hospitals with higher use of hospice had higher predicted readmission rates. Readmission rates in lowest vs highest utilizing quintile were as follows: occupational therapy 16.2% (95% confidence interval [CI] 14.5%-18.0%) vs 12.3% (95% CI 11.3%-13.2%); transfers 13.8% (95% CI 13.2%-14.5%) vs 12.5% (95% CI 11.6%-13.5%); and hospice 13.1% (95% CI 12.3%-14.0%) vs 14.8% (95% CI 13.5%-16.1%).

Conclusions: Hospital practices have a role in stroke readmission that is complex and poorly understood. Further work is needed to identify specific strategies to reduce readmission rates and to ensure that public reporting of readmission rates will not result in adverse unintended consequences.

Figure. Hospital-level readmission rates

Readmission rates of all hospitals in the sample after adjusting for patient-level characteristics. The red dot represents the mean estimated readmission rate for a hospital and the vertical line represents the 95% confidence interval. The horizontal red line represents the mean readmission rate.