Prevalence and correlates of stroke among older adults in Ghana: Evidence from the Study on Global AGEing and adult health (SAGE)

This study examines the prevalence and correlates of stroke among older adults in Ghana. This cross-sectional study retrieved data from Wave 1 of the World Health Organization (WHO) Survey on Global Ageing and Adult Health (SAGE) conducted between 2007 and 2008. The sample, comprising 4,279 respondents aged 50 years and above, was analysed using descriptive statistics, cross tabulations and Chi-Square tests, and a multivariable binary logistic regression. Respondents ranged in age from 50 to 114 years, with a median age of 62 years. Stroke prevalence was 2.6%, with the correlates being marital status, level of education, employment status, and living with hypertension or diabetes. The results showed that being separated/divorced, having primary and secondary education, being unemployed and living with hypertension and diabetes, significantly increased the odds of stroke prevalence in this population. The results suggest that interventions to reduce stroke prevalence and impact must be developed alongside interventions for hypertension, diabetes and sociodemographic/economic factors such as marital status, level of education, and employment status.


Background
Stroke is the second leading cause of death and the third leading cause of disability worldwide [1]. It is estimated that 15 million people suffer from stroke every year. Out of this number, about six million people die and another five million are left permanently disabled [2]. A total of 44 million disability-adjusted life-years (DALYs) are lost to stroke every year and this is projected to increase to 61 million by 2020 [2]. The prevalence is projected to increase throughout the world because the number of persons aged 60 years and above is expected to more than double by 2050, and more than triple by 2100, increasing from 901 million in 2015 to 2.1billion in 2050 and 3.2 billion in 2100 [3].
females. There are ten regions in the country. Of the ten regions in the country, the most populous is the Ashanti Region, made up of 19.4% of the total population; this is followed by Greater Accra Region (16.3%) with the least populated region being the Upper West Region (2.8%). With regard to the age structure, 38.3% of Ghana's population is less than 15 years; 49.5% is 15-49 years, and; 12.2% is aged 50 years and above [41]. Further, a little over half of the population (50.9%) is living in urban areas [41]. More than 70% are Christians, 17.6% are Muslims, 5.2% are traditional worshippers and 5.3% do not have any religious affiliation. In terms of occupation, a larger proportion engages in agriculture-related activities. Specifically, 41.2% of the economically active population is skilled agricultural, forestry and fishery workers; about 21% is also engaged as service and sales workers, while 15.2% are craft and related trade workers.
Ghana is a lower-middle income country with average per capita income of GH¢5,347 (about US$1,353) [42]. The health system is structured to mainly treat infectious diseases [12]. The health care system in Ghana faces a lack of investment in health system and health workforce generally. Health spending as a percentage of GDP is 5.4% in 2013 [43], higher than the WHO recommendation of 5% [43]. There are three teaching hospitals (Korle Bu in Accra, Komfo Anokye in Kumasi and Tamale Teaching hospital); nine regional hospitals; 96 government hospitals; and 1,106 health centres and clinics [44]. In terms of the distribution of health professionals, there are 4,747 medical and dental practitioners, 1,832 physician assistants, 759, anaesthetics, 24,974 nurses, 1129 pharmacist and 41 health research officers. The doctor-population ratio was 1:11929 and the nurse to population ratio was 1:971 [41,45].

Sampling design
Ghana SAGE Wave 1 used a stratified, multistage cluster design that was based on the design for the World Health Survey [46] and presented a nationally representative sample. The primary sampling units were stratified by administrative region (Ashanti, Brong Ahafo, Central, Eastern, Greater Accra, Northern, Upper East, Upper West, Volta, and Western) and type of locality (urban/rural). Based on this, a total of 20 strata were developed [46,47]. From each of the strata, a total of 10-15 Enumeration Areas (EAs) were selected according to the population size. Household listings were done for each selected EA. Twenty households with persons aged 50 years and above, and four households with persons aged 18-49 years were then selected for interview [46]. All persons aged 50 + in 'older' households (households with at least one person aged 50-plus years) were invited to participate, whereas only one person was randomly selected in the 'younger' households (households with no person aged 50-plus years). Further, for those who were incapable of completing an interview for reasons of health or cognition, a proxy questionnaire was completed [46]. Standardized training in all aspects of the interview was provided to all interviewers. The questionnaires were translated into respective local languages, following a translation protocol, and modified to take into account the local context where needed [20]. The interview response rate was 86% [46]. had a stroke?" Hence, the prevalence of stroke in this study was determined as the proportion of Ghanaians aged 50 years and above who had ever been told by a health professional that they have a stroke [46]. Self-reported diagnosis has been used to ascertain stroke prevalence in similar population-based studies in Africa [48], Asia [49,50], Oceania [51,52], Europe [53,54], and North America [55][56][57]. These studies showed that self-reported diagnosis is a valid approach for estimating stroke prevalence and has high sensitivity and specificity values [51-53, 56, 57]. During the data collection, respondents were provided with the definition of stroke. Stroke was described as a sudden and severe attack to the brain, which can cause permanent or temporary paralysis (inability to move, usually one side of the body) and loss of speech [58].
Physical activity. Physical activity was measured as the number of days respondents spent doing moderate-intensity activities including sports, fitness, or recreational leisure activities. This was re-categorized into three: physically inactive, partially active (those engaged in physical activities less than 3 times a week), and fully active (those who engaged in physical activities 3 or more times a week. Smoking and alcohol consumption. Two questions were used to categorize smoking status. These included whether the respondent had ever smoked and if they currently smoked. These categories were re-grouped into non-smokers, previous smokers and current smokers. Those who responded 'No' to the two questions were referred to as 'non-smokers' while those who responded 'Yes' to whether they ever smoked and 'No' to whether they currently smoked were categorized as 'previous-smokers'. Those who responded 'yes' to both questions were categorized as 'current smokers'. Alcohol consumption was represented by three categories: nondrinkers (those who had 'never' consumed alcohol), occasional drinkers (those who had taken alcohol 'once in a while'), and; regular drinkers (those who take alcohol 'all the time').
Comorbidities. The comorbidities examined in this study were diabetes and hypertension. Respondents who had been diagnosed with hypertension by a health professional, and/or had the average of three blood pressure (BP) measurements to be systolic �140mm Hg, and/ or diastolic BP of � 90mm Hg, or were on antihypertensive medications, were regarded as living with hypertension. Diabetes prevalence was measured on a self-reported diagnosis by a health professional or use of insulin or other blood sugar lowering medications.
Other independent variables. The other independent variables included in this study were sex, age, place of residence, marital status, level of education, wealth status, religion, employment status and ethnicity.

Data analysis
Descriptive statistics such as frequency distributions and median were used to describe the socioeconomic and demographic characteristics of the respondents, lifestyle factors, and comorbid conditions. At the bivariate level, cross tabulations and Chi-Square tests were used to determine the variation in stroke prevalence by socioeconomic and demographic characteristics, lifestyle factors and comorbid conditions. Further, a multivariable binary logistic regression was used to examine correlates of stroke prevalence in Ghana and the alpha level for statistical significance was set at 0.05. At the multivariable analyses, the reference categories were theoretically selected based on what existing literature has shown. For instance, research shows that stroke prevalence is relatively higher among males compared to females; hence, we made the female gender the reference category in the multivariable analysis. Further, since the WHO SAGE used a stratified, multistage cluster sampling, appropriate sampling weights were applied before the analysis was done to adjust for the survey design. We also tested for multicollinearity using variance inflation factor (VIF) and tolerance; the outcomes indicate that there was no high intercorrelations among the variables (S1 Table).

Socioeconomic and demographic characteristics of respondents
The socioeconomic and demographic characteristics of the respondents are shown in Table 1. The respondents' ages ranged from 50 to 114 years with a median age of 62 years. More than one-third (39.3%) were between 50-59 years while 38.0% were 60-69 years and 9.9% was 80 years and above. More than half (52.3%) was female and close to 60.0% lived in the rural areas. The highest proportion was currently married (56.8%), more than one-fourth (27.9%) was widowed and the least proportion (1.2%) was never married.
Further, more than half (55.0%) had no education and about one-fifth (19.9%) had tertiary education. With regard to wealth status, almost equal proportions were in the middle, richer and richest categories (20.0%, 20.2% and 20.1% respectively). More than six out of ten older adults (68.7%) were Christians, about 16.0% were Muslims and 5.0% had no religion. In terms of the employment status, close to 70% was employed. The largest proportion (48.8%) was Akan, followed by Ga Adangbe (10.4%), and the least proportion (1.0%) was Grusi.

Lifestyle factors and comorbidities
The results showed that stroke prevalence was 2.6% (Table 2). While almost three out of four (74.2%) had never smoked, 13.3% and 12.5% were previous and current smokers, respectively. More than 40% had never consumed alcohol, while 28.5% were occasional consumers and 30.1% were regular consumers of alcohol. While a large proportion of the sample was not physically active (86.5%), 5.2% were partially active while 8.3% were fully active. With regard to the BMI status, more than half of the respondents (51.8%) had normal BMI; close to onefifth (19.6%) were underweight and 28.6% were overweight or obese. More than half of the respondents (57.8%) were living with hypertension, and 4.2% were living with diabetes.

Univariate analysis
Variation in stroke prevalence by sociodemographic/socioeconomic characteristics. Table 3 shows the variation in stroke prevalence by the socioeconomic and demographic characteristics of respondents. The table shows that stroke prevalence significantly varied by age, place of residence, marital status, level of education, wealth status, employment status, and ethnicity. Specifically, stroke prevalence was higher among the older age groups. The prevalence of stroke among those who were 50-59 years, 60-69 years and 70-79 years were 1.5%, 2.8%, and 4.2%, respectively. Stroke prevalence was higher among urban dwellers compared to their rural counterparts (3.8% and 1.7%, respectively).
In addition, stroke prevalence was highest among respondents who were separated/ divorced and lowest among those who were currently married (4.6% and 2.1% respectively). With regard to education, stroke prevalence was higher among those with higher level of education. Stroke prevalence was highest among those in the richest wealth quintile (3.7%) and lowest among those in the poorest wealth quintile (1.4%). Concerning variation in stroke prevalence by employment status, the results showed that stroke prevalence was significantly higher among those who were unemployed (6.1%) compared to those who were employed (1.0%). Among the ethnic groups, stroke prevalence was highest among the Ewe (3.9%) and none of the respondents who were Grusi reported any stroke prevalence.

Variation in stroke prevalence by lifestyle factors and comorbidities
None of the lifestyle factors was significantly associated with stroke (Table 4). Nevertheless, stroke prevalence was highest among those who were previous smokers, occasional drinkers, partially or fully active, and those who were overweight/obese. Further, hypertension and diabetes were significantly associated with stroke. While 3.8% of those living with hypertension had stroke, 10.4% of those living with diabetes had stroke.

Correlates of stroke
This section reports the results of the multivariable analysis. Table 5 shows that being separated/divorced, having primary and secondary education, being unemployed and living with hypertension and diabetes, were the correlates of stroke in Ghana. The results showed that the odds of stroke prevalence were 1.5 times higher among those who were separated/divorced compared to those who were currently married. Those with primary and secondary education had higher odds of stroke prevalence compared to those with no education (1.1 times and 2.4 times, respectively). Further, those who were unemployed were 2.8 times more likely to be living with stroke compared to those who were employed. Respondents living with hypertension were 2.0 times more likely to be living with stroke; those living with diabetes were likewise 3.0 times more likely to be living with stroke, compared to their respective counterparts.

Discussion
This study examined the prevalence and correlates of stroke among Ghanaians who were 50 years and above. The findings showed that stroke prevalence was 2.6% and its correlates were being separated/divorced, having primary and secondary education, being unemployed and living with hypertension or diabetes. This study showed that the stroke prevalence rate reported in this study is lower than the ones reported in similar population in China (3.1%) [49], South Africa (4.0%) [48] and Singapore (4.1%) [50].

Sociodemographic/socioeconomic factors and stroke prevalence
The findings showed that marital status, level of education and employment status were correlates of stroke. Our findings showed that the odds of stroke prevalence were higher among those who were separated/divorced compared to those who were married. Plausible explanation for this is that marriage is a protective factor for stroke through spousal support, early  health seeking, encouragement of healthy behaviours and adherence to medication [59]. Also, this association may be a reverse causality. For instance, a qualitative study on lived experience of stroke in Ghana showed that living with stroke may result in marital separation [60], thus the higher odds of stroke prevalence among those who were separated. The odds of stroke were higher among those with higher levels of education. An explanation for this is that those with low levels of education may have experienced higher mortality since higher education has a positive association with longer life expectancy [35]. Further, the results showed that those who were unemployed had higher odds of living with stroke. One plausible explanation for this is that a large proportion of those with stroke probably may have stopped working as a result of the illness. This is likely because among those who were unemployed, the mean age at which they stopped working was 41.5 years. Further, about one-third (33.6%) of these people stopped work because of health issues or disabilities. In addition, when those who were unemployed were asked whether they were looking for work, 85.5% responded in the negative. This may suggest that they were not actively looking for work probably because of the stroke. Studies have particularly shown that stroke usually comes with physical, cognitive and behavioural disabilities that limit the sufferer's ability to be independent or engage in productive activity [61].
On the other hand, sex, age, wealth status, religion and ethnicity were not statistically related to stroke prevalence. This is not surprising because there have been contradictory findings with respect to how these sociodemographic and socioeconomic factors are associated with stroke [6,32,[35][36][37]62]. An interesting finding that is worth explaining was the non-statistical association between age and stroke prevalence. This may be because this study only considered people who were 50 years and above, and perhaps, the risk of developing stroke is the same for all age groups after 50 years in Ghana. A review study in Ghana has shown that the risk of stroke is high in the fifth decade of life [34].

Lifestyle factors and stroke prevalence
The results showed that none of the lifestyle factors was a correlate of stroke. An explanation for this finding may be because this study used a cross-sectional data. Even though a study in the Ashanti Region of Ghana showed that current levels of smoking are higher among older adults [63], it is possible that the effect of smoking on stroke prevalence does not differ significantly beyond age 50 in Ghana. Also, the non-statistically significant association between alcohol consumption and stroke prevalence found in this study contradicts what other studies have shown [15,16,23,25,64]. The findings from this study may be due to inappropriate measurement of alcohol consumption. Perhaps, inclusion of length, types and quantity of alcohol consumption in the analysis may have produced different results. Further, the analysis shown here indicated that there was no statistically significant relationship between physical activity and stroke prevalence. In contrast, research has shown that the risk of stroke is lower for partially and fully active individuals compared to low active individuals [26]. An explanation for the finding from this study may be time-sequence based, or possibly a result of inadequate measurement. Therefore, in order to fully capture the true effects of the lifestyle factors on stroke prevalence, it will be important to consider re-conceptualizing these measures in the WHO SAGE survey. For instance, using the occupations of participants or other established measures of physical activity can be used to capture the association between physical activity and stroke prevalence. The associations between lifestyle factors and stroke prevalence observed in this study also indicate that perhaps there are other factors beyond the 'traditional risk factors' that are worth considering when looking at factors associated with stroke prevalence in Ghana.

Comorbidities and stroke prevalence
This study showed that hypertension and diabetes were correlates of stroke. This is not surprising because studies have shown that hypertension and diabetes are the two leading risk factors of stroke [14,17,18]. This indicates the need for proper management of hypertension and diabetes in reducing the burden of stroke in Ghana. This study recommends that tackling hypertension and diabetes should be a major priority by the Ghana's Ministry of Health. This may be achieved through routine screening of individuals in the country in order to enhance early detection of hypertension and diabetes in the population, before the age of 50.

Limitations
This study has three key limitations. First, there may be possible under-estimation of stroke prevalence in this study. Measurement of stroke prevalence was based on self-report and it is likely that some of the respondents did not report living with a stroke most especially if this was not visible to the research staff. Second, some of the findings may be time-sequencing challenged with respect to determining the factors associated with stroke. For instance, it is possible that some of the socioeconomic and demographic characteristics of the stroke survivors may have changed after diagnosis. Hence, these characteristics may not determine stroke prevalence but may rather be consequences of stroke. For instance, it is possible that those who have been diagnosed of stroke stopped working and that was why those who were unemployed had higher odds of stroke. It would have been good to know the employment status of the stroke survivors before and after the stroke diagnosis; this would have revealed the true effect of employment status on stroke prevalence. Nonetheless, the findings from this study suggested that living with stroke affect the employment status of an individual and this is worth paying attention to.
Further, the findings suffered from the problematic nature of measurements for some of the lifestyle factors. For instance, the measure of physical activity in this study was based on the number of days respondents spent doing recreational activities such as sports, fitness, or recreational leisure activities. Based on this measure, 86.5% of Ghanaian adults were not physically active. This may be erroneous because recreational activities were used as proxies for physical activity and this may have led to under-estimating the physically active in this study. In addition, using recreational leisure as a proxy for physical activity also ignores other forms of physical activity such as that associated with blue collar work and street training; this may have biased some of our estimates. We recommend that the findings should be interpreted with caution.
Lastly, the measures for smoking and alcohol consumption may have contributed to the non-statistically significant associations between smoking, alcohol consumption and stroke prevalence. The questions on smoking and alcohol consumption focused on whether an individual had ever smoked/consumed alcohol and/or currently smoking/consuming alcohol. However, there were no question on the frequency of smoking/consumption of alcohol or types of alcohol consumed; these may be better measures of smoking or alcohol consumption to capture the effect of smoking/alcohol on stroke prevalence.
Despite these limitations, this is the first study in Ghana that used nationally representative sample to examine the prevalence and correlates of stroke. As a result, the findings from this study can be generalized to Ghanaians who are 50 years and above.

Conclusion
This study examined the prevalence and correlates of stroke in Ghana, among people aged 50 years and above. The results suggest that interventions to reduce stroke prevalence and impact must be developed alongside interventions for hypertension, diabetes and sociodemographic/ economic factors such as marital status, level of education, and employment status. The findings also raised concerns about re-conceptualizing the measures of lifestyle factors (i.e. smoking, alcohol consumption and physical activity) used in the WHO SAGE Survey in a way that reflects the Ghanaian setting. Finally, this study suggests that examining the correlates of stroke should go beyond the traditional risk factors to exploring other plausible non-traditional risk factors that may be associated with stroke.
Supporting information S1