Multilevel Analysis of Readmissions After Radical Cystectomy for Bladder Cancer in the USA: Does the Hospital Make a Difference?
Academic Article
Overview
abstract
BACKGROUND: Hospitals are increasingly being held responsible for their readmissions rates. The contribution of hospital versus patient factors (eg, case mix) to hospital readmissions is unknown. OBJECTIVE: To estimate the relative contribution of hospital and patient factors to readmissions after radical cystectomy (RC) for bladder cancer. DESIGN, SETTING, AND PARTICIPANTS: We identified individuals who underwent RC in 2014 in the Nationwide Readmissions Database (NRD). The NRD is a nationally representative (USA), all-payer database that includes readmissions at index and nonindex hospitals. Survey weights were used to generate national estimates. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: The main outcome was readmission within 30 d after RC. Using a multilevel mixed-effects model, we estimated the statistical association between patient and hospital characteristics and readmission. A hospital-level random-effects term was used to estimate hospital-level readmission rates while holding patient characteristics constant. RESULTS AND LIMITATIONS: We identified a weighted sample of 7095 individuals who underwent RC at 341 hospitals in the USA. The 30-d readmission rate was 29.5% (95% confidence interval [CI] 27.8-31.2%), ranging from 1.4% (95% CI 0.6-2.2%) in the bottom quartile to 73.6% (95% CI 68.4-78.7) in the top. In our multilevel model, female sex and comorbidity score were associated with a higher likelihood of readmission. The hospital random-effects term, encompassing both measured and unmeasured hospital characteristics, contributed minimally to the model for readmission when patient characteristics were held constant at population mean values (pseudo-R2<0.01% for hospital effects). Surgical volume, bed size, hospital ownership, and academic status were not significantly associated with readmission rates when these terms were added to the model. CONCLUSIONS: After adjusting for patient characteristics, hospital-level effects explained little of the large between-hospital variability in readmission rates. These findings underscore the limitations of using 30-d post-discharge readmissions as a hospital quality metric. PATIENT SUMMARY: The chance of being readmitted after radical cystectomy varies substantially between hospitals. Little of this variability can be explained by hospital-level characteristics, while far more can be explained by patient characteristics and random variability.
