Estimation of Excess Risk of Readmission to Hospital After an Index Inpatient Separation

OBJECTIVES: To develop methods to measure excess risk of readmission following an index admission using linked administrative health data. RESEARCH DESIGN: The cumulative risk of readmission following an index admission was calculated for index and reference subjects using linked hospital morbidity, death, and electoral roll data in cohort, cohort-crossover, and cohort-comparison-crossover designs. SUBJECTS: Index subjects were defined as any man age 20 years or older who separated from an acute hospital in Western Australia in 1990 to 1995 following any form of prostatectomy for a principal diagnosis of benign prostatic hypertrophy. Reference subjects were selected from the general population and the electoral roll (cohort designs). Cases were also used as their own historical controls (cohort-crossover) with and without adjustment for background time difference (cohort-comparison-crossover). MEASURES: The excess risk of readmission following an index admission was estimated by calculating the cumulative risk of readmission in index subjects and subtracting the background risk of admission. The background risk calculation varied according to the study design. RESULTS: The risk of readmission at 30 days increased by 241 to 328% following the procedure. After 1 year of follow-up, the risk of readmission was still increased by 58 to 108%. In general, the absolute differences between index and reference subjects decrease or remain the same with increasing rigor of the methods. CONCLUSIONS: In this example, there was little difference between the cohort-crossover and the cohort-comparison-crossover designs, suggesting that the cohort-crossover method is a justifiable method in the absence of electoral roll controls.