"EMMA Study: a Brazilian community-based cohort study of stroke mortality and morbidity"

ABSTRACT CONTEXT AND OBJECTIVE: Stroke has a high burden of disability and mortality. The aim here was to evaluate epidemiology, risk factors and prognosis for stroke in the EMMA Study (Study of Stroke Mortality and Morbidity). DESIGN AND SETTINGS: Prospective community-based cohort carried out in Hospital Universitário, University of São Paulo, 2006-2014. METHODS: Stroke data based on fatal and non-fatal events were assessed, including sociodemographic data, mortality and predictors, which were evaluated by means of logistic regression and survival analyses. RESULTS: Stroke subtype was better defined in the hospital setting than in the local community. In the hospital phase, around 70% were first events and the ischemic subtype. Among cerebrovascular risk factors, the frequency of alcohol intake was higher in hemorrhagic stroke (HS) than in ischemic stroke (IS) cases (35.4% versus 12.3%, P < 0.001). Low education was associated with higher risk of death, particularly after six months among IS cases (odds ratio, OR, 4.31; 95% confidence interval, CI, 1.34-13.91). The risk of death due to hemorrhagic stroke was greater than for ischemic stroke and reached its maximum 10 days after the event (OR: 3.31; 95% CI: 1.55-7.05). Four-year survival analysis on 665 cases of first stroke (82.6% ischemic and 17.4% hemorrhagic) showed an overall survival rate of 48%. At four years, the highest risks of death were in relation to ischemic stroke and illiteracy (hazard ratio, HR: 1.83; 95% CI: 1.26-2.68) and diabetes (HR: 1.45; 95% CI: 1.07-1.97). Major depression presented worse one-year survival (HR: 4.60; 95% CI: 1.36-15.55). CONCLUSION: Over the long term, the EMMA database will provide additional information for planning resources destined for the public healthcare system.


INTRODUCTION
More than half of the global burden relating to cardiovascular disease (CVD) is concentrated in low and middle-income countries like Brazil. 1 Although age-standardized rates of stroke mortality have declined over the last two decades, updated information from the Global Burden of Disease (GBD) study covering 1990-2013 has shown that the absolute numbers of stroke cases have been increasing for both stroke subtypes, to reach around 10 million incident cases of stroke, 6.5 million deaths due to stroke and almost 26 million stroke survivors. 1Moreover, a rise in the absolute number of disability-adjusted life years (DALYs), mainly due to ischemic stroke (IS), which corresponded to 70% of all stroke cases, was observed over the same period.Lower incidences of IS for both sexes were observed in 2013, compared with 1990.However, higher incidence rates of IS among men than among women were still observed in 2013.Over the same period, no statistical differences in hemorrhagic stroke (HS) according to sex were noticed. 2 In a global comparison, the stroke burden remains unequally distributed in developing countries. 14][5][6] Brazilian data show that CVDs have ranked highest among mortality rates since the early 1960s and account for the highest proportion of hospitalizations.Although mortality rates should be interpreted with caution because of improvements in national statistics and the aging of the Brazilian population over recent decades, age-adjusted mortality rates relating to CVDs were seen to decline by around 20% from 2000 to 2011.Nonetheless, an increase in the overall number of CVD deaths was still reported (DATASUS, the data processing system of the Brazilian Ministry of Health).In 2011, CVDs were responsible for 31% of mortality and cerebrovascular diseases for 30% in this country.Similarly to global trends, CVD mortality rates in Brazil have been influenced by race, sex and other socioeconomic status (SES) characteristics.A greater decline in CVD mortality rates has also been observed among women than among men since 1996.Particularly, stroke mortality has declined by 3.6% and 3.3% per year among women and men, respectively. 7 addition, stroke mortality based on death certificate notifications is slightly higher among blacks than among mixed-race and white people. 7Data from the city of São Paulo (1996-2011) also followed the same trend in comparisons of both gender and family income.A greater decline in the stroke mortality rate was seen in relation to coronary heart disease (CHD) rates.The decrease in mortality was greater among women than among men and was inversely related to income, particularly for men. 8 addition to mortality data, the National Health Survey (PNS), which was a Brazilian community-based epidemiological survey with a nationally representative sample, provided estimates relating to around 2,231,000 stroke cases in 2013, of which 568,000 cases presented severe disabilities.The point prevalences were 1.6% and 1.4% for men and women, respectively.The prevalences of post-stroke disability were 29.5% for men and 21.5% for women.Although stroke prevalence rates were especially higher among older individuals without formal education who were urban dwellers than among individuals with high SES, the degree of stroke disability according to SES was not determined. 9milarly to other data published worldwide, most research in developing countries, including in Brazil, has focused on the epidemiology of and therapeutic advances in CHDs rather than cerebrovascular diseases. 5Despite the undoubted importance of evaluating stroke epidemiology from a broad perspective, including prognostic factors and long-term mortality, there is still a lack of consistent data from developing countries. 101][12][13] Most previous stroke surveillance surveys in Brazil have been based on community and population-based studies restricted to one year at most. 14,15[21]

OBJECTIVE
Here, we describe this Brazilian initiative, focusing on concepts and the main findings regarding stroke burden, including mortality and prognostic risk factors among stroke survivors living in a low-income area who were enrolled in the EMMA cohort between 2006 and 2014.

Design, ethics and setting
This is an ongoing prospective community-based stroke cohort study conducted at the Hospital Universitário, University of São Paulo (Universidade de São Paulo, USP).
The institutional review board of USP's university hospital approved the main study, and also ancillary studies linked to the EMMA cohort.

Population
We evaluated stroke distribution and mortality within all three settings of the WHO STEPS stroke surveillance approach, in a low-income population living in the Butantan area.This area comprises six districts on the western side of São Paulo, with an estimated population of 424,377 (2009), of whom only 12% are over 60 years of age.Among these six districts, the proportion of households with a family income less than or equal to five monthly minimum wages (2000, National Census data) ranges from 13.1% to 40.8%.This range is narrower than in other districts of this city (6.4-60.3%).Cardiovascular diseases account for 40% of all deaths in Butantan and São Paulo, and stroke mortality represents one quarter of all vascular deaths.The proportion of violent deaths during the last 10 years was slightly lower in the Butantan area (4.8%) than in the city overall (5.9%).
In Butantan, there are 16 primary care facilities, seven with an emergency room.The only hospital in the area is the university hospital (Hospital Universitário) of the University of Sao Paulo (USP), which is a community hospital with 260 beds in which STEP 1 was implemented. 19This hospital supports emergencies from primary care units and paramedic ambulances and it is responsible for 80% of the hospitalizations of people living in this location.The center for neurological referral from this community facility is Hospital das Clinicas, which is a tertiary-care hospital located 8 km away.
The primary care units are affiliated to the University, which also manages both hospitals.

WHO STEPS stroke approach within the EMMA cohort
The data collection for the EMMA study was initially based on the WHO manual Stepwise Approach to Surveillance. 22The methodology for case ascertainment data and management of the entire STEPwise method is described elsewhere. 19 brief, STEP 1 (hospital phase), which is still ongoing, was started at USP's university hospital among patients who had neurological symptoms, fulfilled the initial criteria and agreed to participate in EMMA.The main study enrolled its first participant in April 2006 and its last one in September 2014.In this phase, we described the main objectives of the WHO project and some preliminary data, comprising evaluation of hospitalized events and including sociodemographic data (i.e.name, gender, age, race, income, educational level and occupation), acute stroke information regarding stroke recurrence, date and time of onset of stroke symptoms, hospitalization, history of traditional risk factors associated, medical treatment, neurological functionality (modified Rankin scale) and discharge status. 19r STEP 2 (fatal events in the community), which began in November 2006 and ended in 2007, the WHO methodology for cerebrovascular disease was applied in order to investigate cases that evolved to death through application of questionnaires that had previously been set up by WHO plus additional information relating to local realities. 19Mortality data was obtained from the city of São Paulo's health statistics system (PRO-AIM, "Programa de Aprimoramento das Informações de Mortalidade", i.e. the program for improving mortality information).The objective was to identify the set of characteristics that might make it possible to quantify and qualify deaths that occurred within the university hospital's catchment area.In this approach, information on reported deaths was collected according to health areas determined by the Municipal Health Department of the city of São Paulo, taking into account the respective area covered by each primary care unit.
From this information, a protocol for action towards this disease was recommended.Trained interviewers administered the screening instrument, asking each family member to answer symptom questions and to perform simple physical tasks. 23,24All participants who had been screened positive for events suggestive of stroke in the past were invited to answer an individual questionnaire that asked for information similar to that of the STEP 1 questionnaire.All of these individuals were classified according to their clinical diagnosis and their clinical and tomographic diagnoses of stroke.In relation to this last item, the stroke diagnoses of a subset of the patients were validated by a neurologist. 23,246][27][28][29][30] An extension of the EMMA protocol, which was developed in collaboration with this researcher's time spent within the main study, was also implemented in a municipality of the Amazon region, in northern Brazil. 31

Mortality and prognosis
Vital status was investigated periodically by means of a hotpursuit strategy using telephone contacts and medical registries during the follow-up.Particularly with regard to the main cohort at USP's university hospital in São Paulo, we doubledchecked all the mortality data through collaboration with the municipal statistics system (PRO-AIM), the data analysis system of the state of São Paulo (Fundação Sistema Estadual de Análise de Dados, SEADE) and the Brazilian Ministry of Health offices, every year.The reported causes of death on death certificates were transformed into medical codes in accordance with the Tenth Revision of the International Classification of Diseases (ICD-10). 31Ultimately, mortality was categorized as all-cause, cerebrovascular or cardiovascular.Here, we report data from previous EMMA publications based on all-cause mortality.
In the EMMA study, the mortality analyses included evaluation of case-fatality rates at 10, 28 and 180 days and survival analyses one year after the acute event, with exploration of some prognostic risk factors. 17,20,21Short-term mortality (10 and 28-day case-fatality rates) was compared with other stroke registries in other Brazilian cities located in the northeastern region (João Pessoa and Natal). 26garding prognosis over the course of the follow-up, we implemented an extended evaluation of cognitive impairment, using a specific validated questionnaire (Modified Telephone Interview for Cognitive Status, TICS-M) on a subsample of EMMA survivors, three months after the index event. 27In addition, post-stroke depression (PSD) after the acute phase and its influence on oneyear survival was evaluated among stroke survivors, who answered a questionnaire on depression, the Patient Health Questionnaire (PHQ-9), by means of telephone interviews conducted one to three months after the acute event. 28rthermore, we assessed an experimental open-case series to ascertain the effect of transcranial direct current stimulation (tDCS).This is a novel treatment that may improve clinical outcomes from PSD, which is traditionally refractory to pharmacotherapy, among stroke patients with aphasia during the first year after their stroke. 29

Multicenter evaluation
An extended evaluation based on the original EMMA protocol (19) was also developed in the city of Coari, located in the Brazilian Amazon region.In this municipality, cerebrovascular prevalences were compared between the urban zone and rural riverbank areas of the municipality, between May and October 2011. 30

Stroke ascertainment
We ascertained all consecutive cases of potential acute stroke events in the hospital, including first-ever and recurrent events.
All patients older than 18 years of age were eligible for STEP 1.The WHO definition of stroke was used, i.e. "a focal (or at times global) neurological impairment of sudden onset, and lasting more than 24 hours (or leading to death), of presumed vascular origin." 22 Histories of stroke were based on information from patients, caregivers or hospital records.When it was not possible to obtain information, the item was coded as "incomplete data".Stroke diagnoses were validated by a medical practitioner and supported by non-contrasted computed tomography (CT) scans.We used the codes of chapter I of ICD-10 to categorize stroke according to the following subtypes: ill-defined or unspecified stroke (ICD-10:I64), intracerebral hemorrhage (ICD-10:I61), cerebral infarction (ICD-10:I63), late effects of cerebrovascular diseases (ICD-10:I69) and subarachnoid hemorrhage (ICD-10:I60).All suspected stroke cases were also categorized as previous stroke (recurrent incidence of stroke) or no previous stroke (first-ever incidence of stroke), through access to medical records.
All data collection was performed by trained interviewers and medical researchers in accordance with the instructions in the STEPS stroke manual.Quality control was assured through cross-checking the information, which was done by three medical coordinators of the EMMA study.

RESULTS
Previous publications from the EMMA study are summarized in Table 1.

Main findings from STEPS within the EMMA study
The first published data from EMMA study reported on 682 stroke cases out of 1,023 cases of cerebrovascular disease (66.6%).The participants were over 18 years of age and their acute stroke event was confirmed through medical diagnosis and CT within the first 24-48 hours, upon hospital admission.All of them were attended at the university hospital's emergency care sector and were enrolled in EMMA (STEP 1) between April 2006 and May 2009. 19ring the surveillance of fatal events within the community (STEP 2), 256 deaths due to stroke were identified over a 12-month follow-up period.The primary cause of death (causa mortis) according to stroke subtype showed that 30.5% of the deaths were due to IS, 26.6% were due to HS and 43% had an unspecified form of stroke as the primary cause. 19e initial screening of non-fatal stroke cases in the community (STEP 3) included 4,446 individuals living in the reference area of USP's university hospital, near to the Jardim São Jorge primary Design/population: cross-sectional evaluation using questionnaire with six questions concerning limb and facial weakness, articulation, sensory disturbances, impaired vision and past diagnosis of stroke was completed door-to-door in a well-defined area of 15,000 people.Questionnaires were considered positive when a participant answered two or more questions about stroke symptoms or when the presence of stroke was confirmed by a physician, or when at least three questions had positive findings, even if not confirmed by a doctor.Exclusion criteria: people who did not belong to the reference area and were not able to give responses to survey.Statistics: prevalence rates (95% CI).
A total of 243 people initially screened positive for stroke.Age-adjusted prevalence rate for men was 4.6% (95% CI: 3.5-5.7).For women, the prevalence rate was 6.5% (95% CI: 5.5-7.5);when considering only one question, the rate was 4.8% (95% CI: 3.9-5.7).Most commonly reported symptoms were limb weakness and sensory disorders.To determine the CBV prevalence in a town in the Brazilian Amazon region, comparing urban versus rural population in the same municipality.
Design/population: cross-sectional evaluation of CBV to calculate prevalence rates among 6,216 residents ≥ 35 y old in the town of Coari, Amazonas, using a screening questionnaire, the Stroke Symptom Questionnaire.CBV prevalence rates (PRs) from the door-to-door survey were calculated according to the location of the home.Exclusion criteria: people who did not belong to the reference area and were not able to give responses to survey.

Statistics: prevalence rates (95% CI).
There were 4,897 respondents in the urban area and 1,028 in the rural area.The crude prevalence of stroke was 6.3% in the rural area and 3.7% in the urban area, regardless of age and sex.Among stroke cases, people in the rural area were those with less access to medical care in comparison with the urban area (32.1% versus 52.5%; P = 0.01), and there was a positive association between the rural area and no medical care (PR: 1.33; 95% CI: 1.03-1.71),independently of age, sex, education and functional impairment.Paulo.Among these individuals, 618 (13.7%) were not found, 204 (4.5%) refused to participate in the study and 13 (0.29%) were incapable of answering the questions.In total, 3,661 individuals (81.4%) answered a familial screening questionnaire, and 582 of them were identified as having screened neurologically positive, based on additional information relating to their treatment, disability and neurological recovery after stroke.A total of 577 subjects answered the final questionnaire, of whom 243 were screened positive for stroke, based on a questionnaire, and were validated by a board-certified neurologist. 23,24The age-adjusted prevalence rate for men was 4.6% (95% CI: 3.5-5.7).For women, the prevalence rate was 6.5% (95% CI: 5.5-7.5)and, when considering only one question, the rate 4.8% (95% CI: 3.9-5.7).The most commonly reported symptoms were limb weakness and sensory disturbances.Hypertension and heart disease were very frequent conditions associated with previous stroke. 23 all settings, most of the participants were white and married, and had low education (1-7 years).We observed that most of the subjects in STEP 1 and 2 were older (mean ages: 66 and 74 years, respectively) than those who participated in STEP 3 (50.7% of subjects were of ages ranging from 45 to 64 years).Regarding the respondents' gender, we observed that in STEP 3, more females participated in the study (59.4% in STEP 3, 49.2% in STEP 2 and 45.3% in STEP 1).

Short and long-term mortality during extended follow-up
Regarding the mortality rates from the hospital registry, we evaluated case-fatality rates from ten days to one year among all consecutive patients with first-ever stroke who sought acute care at the USP university hospital's emergency service and were enrolled in the EMMA study between April 2006 and December 2008.Among 430 first-ever stroke events, 365 (84.9%) were IS and 65 (15.1%) were HS. 25 Among cerebrovascular risk factors, the frequency of alcohol intake was higher in hemorrhagic stroke (HS) than in ischemic stroke (IS) cases (35 with increased stroke mortality. 25r early case-fatality rates were also compared with other stroke registries located in general hospitals in the northeastern region of Brazil (cities of João Pessoa and Natal). 26Out of 962 firstever events recorded in three centers, the proportions of ischemicto-hemorrhagic cases were maintained at 5:1, as we previously observed in our single-center analysis at USP's university hospital, where the EMMA study was set up. 25 Additional long-term mortality data, including survival analyses from April 2006 to December 2010, were used to evaluate 665 first-ever stroke cases, of which 545 (82.6%) were IS and 116 (17.4%) were HS during the four-year follow-up.We found an overall survival rate of 48% (mean survival of 40 months).Again, we confirmed that lack of formal education and diabetes were independent predictors of poor survival, particularly among IS subjects during long term-follow-up 15 (Table 1).

Cognitive impairment
Cognitive as the comparison. 28We found that cognitive impairment was present in 22.9% of the individuals, post-stroke.The test-retest reliability and intraclass correlation from TICS-M were found to be good, with coefficients ranging from 0.87 to 0.97 across the evaluations.Principal-component analysis extracted three meaningful domains: working memory, recall memory and orientation.The best cutoff point for screening for cognitive impairment was 14 out of 15 (91.5% sensitivity and 71.4% specificity), based on MMSE as the comparison.The area under the curve was 0.89 and, in the end, we concluded that the Brazilian version of the TICS-M was a reliable, stable and homogeneous instrument for screening for cognitive impairment among stroke patients. 28

Post-stroke depression
In a subsample of 191 EMMA participants who reported their depressive status using PHQ-9, one to three months after the acute event, we found that 164 (85.9%) had suffered IS and 27 (14.1%),HS.Among these, the overall incidence of major depression disorder was 25.1% during the one-year follow-up, regardless of stroke subtype.The peak rate of major depression subsequent to the acute event was more than one month afterwards.
We observed that there was a lower survival rate among individuals who developed post-stroke major depression disorder than among those who did not develop this condition, after one year of follow-up (85.4% versus 96.5%; log rank P = 0.006).After multiple analysis, we found that there continued to be higher risk of all-cause mortality among those who developed major depression disorder than among participants without major depression disorder (HR: 4.60; 95% CI: 1.36-15.55;P = 0.01), thus suggesting that incident major depression disorder is a potential marker for poor prognosis, one year after stroke. 29CS in post-stroke depression (PSD) and aphasia The sample comprised four females (mean age: 48 years) with aphasia after stroke who developed the onset of post-stroke depression after the index event (mean time elapsed: six months).

Extended EMMA initiative
In the area studied in Coari, 6,216 residents over 35 years of age were interviewed using a screening questionnaire, the Stroke Symptom Questionnaire.From this door-to-door surveillance, cerebrovascular prevalence rates (PR) were calculated according to the location of the home. 31The total numbers of respondents were 4,897 in the urban area and 1,028 in the rural area.
The crude prevalence rate (PR) of stroke was 6.3% in the rural area and 3.7% in the urban area, regardless of age and sex.As expected, lower levels of medical care were observed in the rural area than in the urban area (32.1 versus 52.5%, P = 0.01).
There was a positive association between living in the rural area and no medical care for stroke (PR: 1.33; 95% CI: 1.03-1.71),regardless of SES. 31

DISCUSSION
Since 2006, unified data provided through the EMMA study have demonstrated trends regarding stroke surveillance in three spheres of investigation (hospital, official mortality data and community sources). 19,23The data have also shown potential risk factors and disability and mortality statistics based on case-fatality and survival rates in this low-income population over the course of four-year follow-up. 17,20,21,25,26 first view, the demographic characteristics among the EMMA participants were similar in the three STEPS, except for age and sex.In the community (STEP 3), we found younger survivors and more females than in other settings. 19As expected, information on stroke subtype was better defined in the hospital setting (STEP 1) than in the community (STEP 3). 19Among the cases included in the hospital phase, about 70% were confirmed as first-ever stroke and ischemic subtype (ratio of hemorrhagic-to-ischemic cases of 1:6) during the period 2006-2009. 25We noticed that there was a slight increase in the proportions of hemorrhagic-to-ischemic cases.
to 1:4, by adding one year of follow-up (2006-2010). 17In addition to aging, regular alcohol consumption was closely associated with intracerebral hemorrhage.
Comparisons across epidemiological studies on stroke worldwide are difficult because of divergences of methodology, especially regarding the study sample (hospital or community or populationbased data), the criteria used for ascertaining cases and the stroke subtype enrolled.Nevertheless, the EMMA study, which used a stroke cohort based on a low-income community on the western side of the city of São Paulo, had results that were in accordance with those from most population-based studies. 32In addition, the extension of the EMMA study to a community in the Brazilian Amazon region confirmed that the prevalence of stroke in rural areas is higher than in urban areas. 31systematic review conducted on 56 population-based studies, including Brazilian data, 14 which reported stroke incidence and case-fatality from 1970 to 2008 in low to middle-income countries, found that the proportion of ischemic stroke ranged from 54% to 85% and that the proportion of intracerebral hemorrhage ranged from 14% to 27% over the period from 2000 to 2008. 32As expected, hemorrhagic cases were more commonly detected in low-income countries. 32Although we found a slight increase in the proportion of incident cases of HS, compared with IS, over a four-year period, our rates were similar to those reported in developed countries. 32Regarding mortality, our casefatality rate over a one-year period was 25% (ratio of ischemic to hemorrhagic stroke cases of 1:4).As expected, the risk of death due to HS was greater than the risk due to IS and this gradient reached its maximum at 10 days (OR: 3.31; 95% CI: 1.55-7.05).
Low education was the main sociodemographic factor associated with higher risk of death, particularly among those with ischemic stroke.The influence of lack of education on mortality was markedly higher at 10 days. 25r case rates for IS (85%) and HS (15%) were similar to those reported in other Brazilian studies. 14,33However, they differed from those reported in other countries in Latin America, such as Chile (72% for IS and 28% for HS) 33 and were much more divergent from data from southern African (Mozambique), from where the highest rate of HS (46%) versus IS (56%) was reported. 34comparison of our findings with those from two other Brazilian population-based studies showed that our study had a lower one-year overall case fatality rate than the rate reported in Matão (22.7% versus 30.9%) 14 33 Overall, life expectancy within the first four years after stroke was about 50% in the EMMA study.Our cumulative survival rate for hemorrhagic stroke was 44%.18 The main determinants of poor survival up to four years were hemorrhagic stroke and lack of education for ischemic cases.Moreover, we found that diabetes was an independent predictor of all-cause mortality in a long-term follow-up on our study data.18 Regarding functionality, there was a slight decrease in the hospital phase of EMMA, from 40% at 28 days to 34% at six months after the acute event, particularly among individuals of low education level with IS.21 Reinforcing these findings, in EMMA STEP 3 (community level), the referral rate for rehabilitation services was roughly 25% for all participants within the community who were identified as presenting a previous history of stroke.23 Overall, our statistics revealed similar proportions for incident cases of ischemic and hemorrhagic stroke cases in comparison with developed countries. On thther hand, we continue to be behind developed countries in terms of decreasing the mortality rates.The case-fatality rates remain high and the survival rates remain poor in our setting.There is also a high degree of dysfunctionality among stroke survivors, particularly those with low SES, which is much more similar to the stroke pattern in developing countries.1 These findings reaffirm the trends recently reported by GBD in 2013, regarding the significant burden of stroke concentrated in developing countries.
The impact on mortality rates, DALYs and years lived with disability (YLDs) that comes mainly from hemorrhagic stroke was shown to be significantly higher in developing than in developed countries.During the period from 1990 to 2013, the proportional contributions of deaths due to HS and IS increased by 1.8% and 2.2% in developing countries, respectively.Meanwhile, in developed countries, these rates decreased by 0.73% for HS and by 1.45% for IS. 1 The EMMA study has some strengths.Although our cohort was not a population-based study, we based our data on a community area with low SES, located in a developing country.This may have contributed towards filling the gap in the information on stroke epidemiology.We developed an extended evaluation that included epidemiology, mortality and predictors associated with poor prognosis among individuals who survived the acute phase of the cerebrovascular event.We implemented a standard protocol in order to follow up our participants by means of telephone contacts and thus update the following data: vital status; functional disability; non-fatal outcomes such as hospitalization; recurrence of stroke or other CVD outcomes (heart failure and myocardial infarction); progression of depression; and cognition.All of these data will be available in the near future for prospective analysis.
The information acquired over the course of the follow-up was all double-checked by the medical researcher responsible, based on the patient's medical records and additional examinations such as CT, echocardiography and electrocardiogram for the main study.
The mortality information was all confirmed through official death certificates provided by the local health statistics departments at all centers involved in the stroke surveillance.Thus, mortality specified as due to cerebrovascular or cardiovascular causes will be available for survival analyses.
Other than the EMMA cohort, only a few Brazilian studies have reported on the big picture of stroke epidemiology, including long-term follow-up with its admixture of outcomes. 16,18,35[38] Our main limitations related mainly to the initial data collection, which lacked acute neurological evaluation for quantifying stroke severity in most cases.We implemented the NIH Stroke Scale (NIHSS) from the outset of the study, but less than 10% of the cohort presented trustworthy data.In addition, we implemented some protocols to investigate post-stroke depression and cognitive impairment/dementia after 2010.

CONCLUSIONS
Data provided by the EMMA cohort study have depicted stroke surveillance in three spheres of investigation (hospital, official mortality data and community sources).The foremost findings of high rates of post-stroke disability and mortality and poor longterm survival have mainly been influenced by low education levels so far, up to the four-year follow-up.

STEP 3 (
non-fatal events in the community) started in February 2008 and finished in May 2008.The community area was previously delimited through a public family healthcare program at one primary care unit within the university hospital's catchment area (at Jardim São Jorge).A potential total number of 4,725 subjects older than 35 years of age was estimated for this first part.
status three months after acute stroke was evaluated during the follow-up as an additional action relating to poststroke disabilities.In this context, we previously adapted and validated the Brazilian version of the TICS-M for cognitive impairment among post-stroke patients, in a subset of EMMA participants, using the Mini-Mental State Examination (MMSE)

Table 1 .
Executive summary of previous publications from the Study of Stroke Mortality and Morbidity in Adults (EMMA Study), 2006-2014

Table 1 .
Continuationcare unit, which is located in the western area of the city of São and a lower 180-day lethality rate than the rate in Joinville (overall stroke rate: 19.5% versus 25%; ischemic stroke rate: 17.6% versus 19%; and hemorrhagic stroke rate: 29.1% versus 49%).