PURPOSE: To analyse mortality for spontaneous intracerebral haemorrhage (ICH), myasthenia gravis (MG) and Guillain-Barré syndrome (GBS) from 1996 to 2009 in UK intensive care units (ICUs). METHODS: We used the Intensive Care National Audit & Research Centre (ICNARC) database. We identified specialised neurosciences critical care units (NCCUs) (n = 16), general ICUs with full neurological support (n = 48) and general ICUs with limited neurological support (n = 138) and undertook descriptive analyses for each condition. Poisson regression was used to identify trends in admission rates, median regression to identify trends in lengths of stay (LOS), and logistic regression (Wald test) to analyse interaction between unit type and time period; odds ratios were calculated for hospital mortality associated with unit types. RESULTS: For ICH (n = 10,313 cases), overall ICU mortality was 42.4 %, and acute hospital mortality 62.1 %. In NCCU, LOS was longer, but mortality lower, and over time, mortality from ICH decreased faster. For MG (n = 1,064 cases) and GBS (n = 1,906 cases), overall mortality was relatively high (MG: 8.7 % ICU mortality and 22 % acute hospital mortality; GBS: 7.7 and 16.7 %, respectively); overall mortality did not decrease over time. CONCLUSIONS: This first large-scale analysis of outcomes in acute neurological disease in the UK demonstrates real-life mortality higher than published series. NCCU care is associated with increased survival in conditions requiring highly specialised intensive care techniques, but high-quality step-down care is pivotal in others. Strategies that truly improve outcomes must integrate emergency department management, ICU admission criteria, NCCU treatment, high-quality step-down care and neurorehabilitation.