OBJECTIVES: Cost of illness in ischemic stroke has historically been reported as direct total health care cost, not compared to an equally ill, non-stroke, control group. This methodology likely overestimates cost and may affect national burden of illness estimates and cost effectiveness comparisons. The magnitude of the over-estimation is not known. The objective of this study is to estimate the 1-year marginal cost of ischemic stroke in SC Medicare beneficiaries and examine the impact of total versus marginal cost reporting on overall US stroke cost estimates and on the distribution of expenditures of inpatient care and rehabilitation. METHODS: Cost of illness was estimated from a Medicare cost perspective for ischemic stroke in South Carolina. Stroke patients were matched 1:2 to non-stroke beneficiaries using nearest-neighbor propensity score matching. Standard Medicare analytic files for all beneficiaries in SC in 2004 and 2005 were summed for the year following their index ischemic stroke or to death. The total study sample size was 8928.  RESULTS: The use of a marginal costing approach produced lower estimates than the average total costing method. The 2004 marginal costs of stroke were $26.9 million, while the average cost estimate was $81.3 million, a difference of $54.4 million. Average total cost includes expenditures for comorbid conditions which are common in patients with stroke. Cost of stroke for the US in 2012 would be overestimated by $4.89 billion if this difference is generalized to national estimates. Furthermore, the proportion of stroke costs attributable to rehabilitation services would change from 13.7% based on mean cost, to 29.3% using marginal estimates. CONCLUSIONS: Using a marginal costing approach to estimate health care costs for conditions common in patients with a high prevalence of comorbid conditions is essential for accurate estimation of burden of illness, as well as estimating in-hospital and follow-up cost distributions correctly.