Background: Patients vary greatly in their use of care after hospitalization for stroke. We classified stroke patients according to their care trajectories and associated costs in the 12-month period after hospital discharge. Methods: We followed a cohort of 3,811 veterans for one year after hospitalization with ischemic stroke in Veterans Health Administration facilities in 2007. Three discharge outcomes -- nursing home care, home care, and mortality -- were modeled jointly with Latent Class Growth Analysis. VA and Medicare costs were obtained for use of institutional care (inpatient acute, rehabilitation facility, and nursing home) and home care (home health, other home care, and outpatient rehabilitation). Covariates included patient age, NIHSS stroke severity and FIM scores measured at hospital discharge. Results: Members of the cohort had one of five care trajectories: 49% had a Rapid Recovery with little or no use of care in the 12 months after discharge, 15% had a Gradual Recovery with initially high nursing or home care use that tapered off over time, 9% had consistent use of Long-Term Home Care (HC), 13% had consistent use of Long-Term Nursing Home Care (NH), and 14% had an Unstable trajectory with multiple transitions between long-term and acute care. Patients with Long-Term NH and Unstable trajectories had the highest average total costs (greater than $60,000 per person) and patients with the Rapid Recovery trajectory had the lowest cost (less than $11,000 per person). Medicare accounted for 23% of total costs. In a multinomial regression model, the likelihood of a Long-Term NH, Long-Term HC or an Unstable Trajectory was greatest for persons with more severe strokes (higher NIHSS score), more disability (lower FIM score), and age 65 or older. About half of the veterans received rehabilitation services. Most rehabilitation was delivered in the NH. There was no clear association between use of rehabilitation and subsequent care trajectory. Conclusions: Care trajectories were explained partly by veteran health and functional status. However, we need a better understanding of system factors shaping care trajectories, particularly access to and use of rehabilitation services.