Trends in length of stay, mortality and readmission among patients with community-acquired bacteraemia

In patients hospitalized with severe infection, premature discharge may lead to increased risk of readmission and death. We conducted this population-based cohort study to examine trends in length of stay (LOS) and 30-day mortality and hospital readmission rates after bacteraemia from 1994 through 2013. We used Cox regression to compute hazard ratios (HRs) for 30-day mortality and 30-day postdischarge readmission rates by calendar period and quintiles of LOS, adjusting for age, sex and comorbidity. Among 7618 patients hospitalized with community-acquired bacteraemia during the study period, median LOS decreased from 12 days (quartiles 7-21 days) in 1994-1998 to 9 days (quartiles 6-16 days) in 2009-2013 (25% relative reduction). The 30-day mortality fell from 16.7% to 15.0%, yielding an adjusted 30-day HR of 0.80 (95% confidence interval (CI) 0.68-0.95). Almost one fifth (19.4%) of patients discharged alive were readmitted within 30 days. Concurrently, the adjusted HR of readmission tended to increase (adjusted HR 1.09, 95% CI 0.93-1.28) in 2009-2013 compared with 1994-1998. Compared with the middle quintile of LOS (9-12 days), the risk of readmission was slightly higher for patients discharged within 5 days (adjusted HR 1.12, 95% CI 0.92-1.37), especially for readmission due to infection (adjusted HR 1.38, 95% CI 1.03-1.85). Readmission risk was lowest for 6 to 8 days LOS (adjusted HR 0.80, 95% CI 0.67-0.95) and highest for LOS ≥23 days (adjusted HR 1.30, 95% CI 1.11-1.53). The declining LOS after community-acquired bacteraemia between 1994 and 2013 was not accompanied by increased 30-day mortality but by slightly increased readmission rates.