OBJECTIVES: To compare the characteristics of populations admitted to hospital-at-home services with the population admitted to hospital and assess the association of these services with healthcare costs and mortality. DESIGN: In a retrospective observational cohort study of linked patient level data, we used propensity score matching in combination with regression analysis. PARTICIPANTS: Patients aged 65 years and older admitted to hospital-at-home or hospital. INTERVENTIONS: Three geriatrician-led admission avoidance hospital-at-home services in Scotland. OUTCOME MEASURES: Healthcare costs and mortality. RESULTS: Patients in hospital-at-home were older and more socioeconomically disadvantaged, had higher rates of previous hospitalisation and there was a greater proportion of women and people with several chronic conditions compared with the population admitted to hospital. The cost of providing hospital-at-home varied between the three sites from £628 to £2928 per admission. Hospital-at-home was associated with 18% lower costs during the follow-up period in site 1 (ratio of means 0.82; 95% CI: 0.76 to 0.89). Limiting the analysis to costs during the 6 months following index discharge, patients in the hospital-at-home cohorts had 27% higher costs (ratio of means 1.27; 95% CI: 1.14 to 1.41) in site 1, 9% (ratio of means 1.09; 95% CI: 0.95 to 1.24) in site 2 and 70% in site 3 (ratio of means 1.70; 95% CI: 1.40 to 2.07) compared with patients in the control cohorts. Admission to hospital-at-home was associated with an increased risk of death during the follow-up period in all three sites (1.09, 95% CI: 1.00 to 1.19 site 1; 1.29, 95% CI: 1.15 to 1.44 site 2; 1.27, 95% CI: 1.06 to 1.54 site 3). CONCLUSIONS: Our findings indicate that in these three cohorts, the populations admitted to hospital-at-home and hospital differ. We cannot rule out the risk of residual confounding, as our analysis relied on an administrative data set and we lacked data on disease severity and type of hospitalised care received in the control cohorts.