BACKGROUND: Length of stay (LOS) remains a primary focus for hospitals, and patients with prolonged LOS disproportionately affect hospital capacity and costs. We recently showed that long LOS patients are increasingly hospitalized at academic centers, but their effect on the distribution of LOS is unknown. METHODS: Using the Vizient Clinical Data Base/Resource Manager (CDB/RM), which includes over 90% of the academic medical centers in the United States, we examined trends in the distributions of LOS for acute medical/surgical hospitalizations from 2007 to 2016 in 117 hospitals. We excluded patients under 18 years and those with primary psychiatry, obstetric or rehabilitation diagnoses. RESULTS: Two separate trends were evident during this time period. Mean LOS decreased steadily from 2007 to 2010, but then rose steadily from 2011 and reached its maximum in 2016. Median LOS remained consistent at 3 days from 2007 to 2013 but it too rose from 2014 to 2016. As expected from the difference between the mean and median values, LOS at the 99th percentile dropped from 2007 to 2010 but then rose back by 2016. Gini coefficient values, used to measure inequalities in distribution, declined modestly from 2007 to 2010 but then remained unchanged through 2016. Results were similar in analyses adjusted for age, sex, and case-mix index. CONCLUSIONS: The beginning of the study period was characterized by a reduction in mean LOS, driven largely by decreases of the longest hospitalizations and greater uniformity in LOS. The latter portion saw steady increases in LOS that were similar across the entire distribution of hospitalizations. If the nadir in LOS has truly been reached, these trends will complicate the long-term health of academic medical centers and their staff, faculty, and trainees.
BACKGROUND: Length of stay (LOS) remains a primary focus for hospitals, and patients with prolonged LOS disproportionately affect hospital capacity and costs. We recently showed that long LOS patients are increasingly hospitalized at academic centers, but their effect on the distribution of LOS is unknown. METHODS: Using the Vizient Clinical Data Base/Resource Manager (CDB/RM), which includes over 90% of the academic medical centers in the United States, we examined trends in the distributions of LOS for acute medical/surgical hospitalizations from 2007 to 2016 in 117 hospitals. We excluded patients under 18 years and those with primary psychiatry, obstetric or rehabilitation diagnoses. RESULTS: Two separate trends were evident during this time period. Mean LOS decreased steadily from 2007 to 2010, but then rose steadily from 2011 and reached its maximum in 2016. Median LOS remained consistent at 3 days from 2007 to 2013 but it too rose from 2014 to 2016. As expected from the difference between the mean and median values, LOS at the 99th percentile dropped from 2007 to 2010 but then rose back by 2016. Gini coefficient values, used to measure inequalities in distribution, declined modestly from 2007 to 2010 but then remained unchanged through 2016. Results were similar in analyses adjusted for age, sex, and case-mix index. CONCLUSIONS: The beginning of the study period was characterized by a reduction in mean LOS, driven largely by decreases of the longest hospitalizations and greater uniformity in LOS. The latter portion saw steady increases in LOS that were similar across the entire distribution of hospitalizations. If the nadir in LOS has truly been reached, these trends will complicate the long-term health of academic medical centers and their staff, faculty, and trainees.