Hospital Phenotypes of Observation Care Use Among Medicare Beneficiaries Visiting the Emergency Department

Objective: Background: To characterize hospital phenotypes with respect to trends in observation care use and examine whether patterns differ for people with AD/ADRD. Design: Retrospective cohort study. Subjects: Traditional Medicare beneficiaries aged 68 + who visited an emergency department from 2012 to 2019. Main Measures: For each beneficiary, we determined visit disposition (e.g., discharge, inpatient admission, or observation stay). We calculated hospital-level slopes for each disposition using linear regression and categorized each hospital as having an increase, decrease, or no change in each disposition category. Cross-tabulations of these trends determined hospital phenotypes (e.g., increasing observation stays along with decreasing admissions). We compared phenotypes by hospital characteristics and repeated these analyses stratified by AD/ADRD diagnosis. Key Results: Sample included 22,780,334 ED visits among 5,162,037 beneficiaries at 4835 hospitals. Nationally, the percentage of visits ending in observation increased over time, but there was a substantial decrease in observation stays among 17.5% of hospitals, and 41.4% saw no change. There were 68.8% of hospitals with identifiable phenotypes of observation use trends, the most common of which (N = 788 hospitals, 16.3%) was an increase in the rate of observation stays and a decrease in the rate of admissions. Large, teaching, and urban hospitals saw a disproportionate increase in observation use, while small, non-teaching rural and critical access hospitals saw no meaningful change. A greater share of hospitals saw an increase in observation for beneficiaries with AD/ADRD compared to those without AD/ADRD (49.0% vs. 40.0%). Conclusions: While there has been a large national increase in observation use for Medicare beneficiaries, trends among individual hospitals varied substantially, with differential patterns by hospital characteristics and beneficiary AD/ADRD status.