Causes of death GBD 2015 uses a database of 14 236 site-years of vital registration, verbal autopsy (a method of determining cause of death in countries that have no functional vital registration system), and maternal and child death surveillance data, covering the period from 1980 to 2015.9 Trained interviewers administer a structured questionnaire to relatives of a deceased person about their symptoms preceding death. The underlying cause of death is inferred by computer algorithms or physician review of the autopsy interview. We estimated deaths for all neurological disorders apart from headaches, to which no deaths are assigned as the underlying cause. For each neurological cause except dementia, we used the GBD Cause of Death Ensemble model (CODEm) strategy.10,11 CODEm applies mixed effects or spatio temporal Gaussian process regression models to mortality rates or cause fractions in varying com binations with predictive covariates. The ensemble of models with best credentials on out-of-sample predictive validity tests was selected for each cause of death. YLLs were calculated by multiplying the number of deaths at each age by the standard life expectancy at that age.12 Results from CODEm for each disease were scaled to fit all-cause mortality estimates derived from demographic sources by location, age, year, and sex. We decided to use a natural history model for dementia because of a large inconsistency between the data for prevalence and mortality. For instance, in the USA, the rates of death from dementia increased five times between 1990 and 2014, whereas the available prevalence and incidence data showed no significant changes over the same period. Large increases in death rates assigned to dementia have also occurred in some other countries with high-quality vital registration systems. Furthermore, in GBD 2015, the prevalence of dementia varied among 187 countries by a factor of three, whereas dementia death rates varied by more than 20 times.13 The likely explanation was a change in coding practices between countries and within countries over time. To correct for this source of measurement bias, we assessed the most recent data from 23 high-income countries with high-quality vital registration systems and the highest ratio of registered dementia death rates to prevalent cases. This ratio is equivalent to the excess rate of mortality in cases of dementia. We derived a pooled estimate by age and sex using a linear regression of the log of these rates. We added these estimates as data in DisMod-MR 2.1 to derive estimates of cause-specific mortality rates that were consistent with prevalence data and the pooled estimate of excess mortality from the 23 countries that in their most recent year of vital registration were most willing to code a death to dementia as the underlying cause. Data are n or % with 95% uncertainty intervals (UIs). Prevalence is an aggregate of all sequelae for a condition. DALYs=disability-adjusted life-years. *Includes muscular dystrophy, Huntington’s disease, and other less common neurological disorders (appendix p 112). †UIs were not calculated for aggregated estimates of percentage change as this was not built into the Global Burden of Disease Study 2015 computational machinery and would have required a large computational effort competing with demands of completing the Global Burden of Disease Study 2016 Global absolute numbers and age-standardised rates per 100 000 people for neurological disorders and percentage changes between 1990 and 2015 of global deaths. We were unable to report the combined prevalence. Classifying stroke under neurological disorders rather than cardiovascular diseases made neurological disorders the largest cause group of DALYs and the second-largest behind cardiovascular diseases in terms of global deaths. Among the neurological disorders, stroke, migraine, meningitis, Alzheimer’s disease and other dementias, and epilepsy each caused more than 10 million DALYs. The most prevalent neurological disorders were the various types of headache and Alzheimer’s disease and other dementias).Global burden by sex and age; Substantial sex differences in age-standardised death, DALY, and prevalence rates existed globally in 2015. The only neurological disorders with less than 10% difference between males and females were Alzheimer’s disease and other dementias in terms of death and DALY rates, and meningitis and epilepsy in terms of prevalence rate. Apart from multiple sclerosis, Alzheimer’s disease and other dementias, and the three headache types, DALY and prevalence rates were higher in males than females. All these more substantial sex differences were significant apart from those for meningitis, tetanus, and encephalitis death rates and the meningitis and encephalitis DALY rates (appendix pp 153–54). The age pattern varied between the various neurological disorders. The bulk of burden due to communicable neurological disorders occurred in young individuals, particularly the 0–5 years age group. Epilepsy caused the most burden in children and young adults. Headaches peaked at ages 25–49 years, whereas the burden of other neurological disorders increased with age (figure 1).Proportional contribution to the combined burden. Stroke accounted for the largest proportion of total DALYs (47·3%) and deaths (67·3%) among all neurological disorders analysed (figure 2). Migraine, meningitis, and Alzheimer’s disease and other dementias were the second, third, and fourth largest contributors of DALYs.