ObjectiveTo assess differences in acute ischaemic stroke (AIS) in-hospital mortality between referral stroke hospitals and provide evidence on the association of those differences with the overtime adoption of effective reperfusion therapies.DesignRetrospective, longitudinal observational study using administrative data for virtually all hospital admissions from 2003 to 2015.SettingThirty-seven referral stroke hospitals in the Spanish National Health System.ParticipantsPatients aged 18 years and older with a hospital episode with an admission diagnosis of AIS in any referral stroke hospital (196 099 admissions).Main endpoints(1) Hospital variation in 30-day in-hospital mortality measured in terms of the intraclass correlation coefficient (ICC); and (2) the difference in mortality between the hospital of treatment and the trend of utilisation of reperfusion therapies (including intravenous fibrinolysis and endovascular mechanical thrombectomy) in terms of median OR (MOR).ResultsAdjusted 30-day AIS in-hospital mortality decreased over the study period. Adjusted in-hospital mortality after AIS rates varied from 6.66% to 16.01% between hospitals. Beyond differences in patient characteristics, the relative contribution of the hospital of treatment was higher in the case of patients undergoing reperfusion therapies (ICC=0.031 (95% Bayesian credible interval (BCI)=0.017 to 0.057)) than in the case of those who did not (ICC=0.016 (95% BCI=0.010 to 0.026)). Using the MOR, the difference in risk of death was as high as 46% between the hospital with the highest risk and the hospital with the lowest risk of patients undergoing reperfusion therapy (MOR 1.46 (95% BCI 1.32 to 1.68)); in patients not undergoing any reperfusion therapy, the risk was 31% higher (MOR 1.31 (95% BCI 1.24 to 1.41)).ConclusionsIn the referral stroke hospitals of the Spanish National Health System, the overall adjusted in-hospital mortality decreased between 2003 and 2015. However, between-hospital variations in mortality persisted.