This thesis examines 5-year clinical and health-economic outcomes of ischaemic stroke, and their relationship to short-term post-stroke disability, as captured by the 3-month modified Rankin Scale (mRS) - the favoured primary outcome measure in acute stroke trials. I use data from the Oxford Vascular Study (recruited 2002-2014), a population-based prospective cohort for which I followed patients in-person and via medical records until 15-May-2017. I demonstrate that 3-month mRS strongly predicts 5-year post-stroke disability and mortality, including in clinically-relevant groups (treatable major strokes, atrial fibrillation-related strokes, and lacunar strokes), reaffirming its use as a trial outcome measure. About one in four patients experience functional recovery between 3-12 months post-stroke, and mortality follow-up beyond 1-year by stroke trials can show translation of early disability gains into lower mortality. Contrary to previously reported apparent sex-differences, I find no evidence of worse outcomes in women after accounting for differences in age and pre-stroke mRS. I find that late recovery between 3-12 months occurs more often in lacunar strokes, supporting the focus of restorative therapies in this group, but highlighting that uncontrolled studies cannot assume that improvements after 3-months are treatment-related. In addition, I demonstrate that like death/disability, outcomes of institutionalization, post-stroke dementia, health/social-care costs, and quality-adjusted life expectancy (QALE) also show meaningful differences with each step up the mRS ladder. Consequently, ordinal analysis of the 3-month mRS (capturing transitions across the scale's range) better predicts long-term outcomes than dichotomous approaches, which also foster high exclusion rates of relevant patient segments from trials owing to their pre-morbid disability. However, the mRS should be weighted in ordinal analyses, as different state transitions carry different implications for long-term outcomes. Using 3-month mRS-stratified data for clinical endpoints, care costs, and QALE, I derive mRS weights that could be used for meaningful ordinal analyses, clinical prognostication, and cost-effectiveness analyses of stroke therapies.