Objectives To examine the effect of an emergency department (ED)‐based transitional care nurse (TCN) on hospital use. Design Prospective observational cohort. Setting Three U.S. (NY, IL, NJ) EDs from January 1, 2013, to June 30, 2015. Participants Individuals aged 65 and older in the ED (N = 57,287). Intervention The intervention was first TCN contact. Controls never saw a TCN during the study period. Measurements We examined sociodemographic and clinical characteristics associated with TCN use and outcomes. The primary outcome was inpatient admission during the index ED visit (admission on Day 0). Secondary outcomes included cumulative 30‐day admission (any admission on Days 0–30) and 72‐hour ED revisits. Results A TCN saw 5,930 (10%) individuals, 42% of whom were admitted. After accounting for observed selection bias using entropy balance, results showed that when compared to controls, TCN contact was associated with lower risk of admission (site 1: −9.9% risk of inpatient admission, 95% confidence interval (CI) = −12.3% to −7.5%; site 2: −16.5%, 95% CI = −18.7% to −14.2%; site 3: −4.7%, 95% CI = −7.5% to −2.0%). Participants with TCN contact had greater risk of a 72‐hour ED revisit at two sites (site 1: 1.5%, 95% CI = 0.7–2.3%; site 2: 1.4%, 95% CI = 0.7–2.1%). Risk of any admission within 30 days of the index ED visit also remained lower for TCN patients at both these sites (site 1: −7.8%, 95% CI = −10.3% to −5.3%; site 2: −13.8%, 95% CI = −16.1% to −11.6%). Conclusion Targeted evaluation by geriatric ED transitions of care staff may be an effective delivery innovation to reduce risk of inpatient admission.