Bangladesh is a country in South Asia and is situated in the Bay of Bengal. It is divided into 8 administrative divisions [27]: Barisal, Chittagong, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur, and Sylhet. These divisions are further divided into districts, with a total of 64 districts in Bangladesh. Each district is then subdivided into subdistricts, and the subdistricts are further divided into unions, which consist of multiple villages. An overview of the administrative geography of Bangladesh is presented in Figure S1 in Multimedia Appendix 1.

The research was carried out as part of the “Enriched Sastho” program, and the operation is currently operational in 7 divisions of Bangladesh. “Enriched Sastho” is a primary and preventive health care program created by Palli Karma Sahayak Foundation, Bangladesh. The program has 1 smartphone app for community health workers (CHWs) to collect data from rural people by visiting different households (Multimedia Appendix 1). During household visits, CHWs collect information such as sociodemographic, economic, environmental, and health-related information from individuals encompassing every age group and sex. Before collecting data, CHWs received 2 weeks of training to ensure the quality of the data collection and to familiarize themselves with the process.

For this investigation, we used data that were collected from July 2018 to June 2021. A total of 104 variables were obtained during data collection, encompassing sociodemographic, economic, environmental, and health-related information. This smartphone app is connected to Amazon Web Service (AWS) and all these data are usually stored in the AWS.

In this study, the presence of stroke was determined by survey questions asked by the CHWs to the individuals. The answer was categorized in binary format. Thus, if a person had a history of stroke, then the “stroke” variable was filled with 1, otherwise 0. In addition, CHWs checked their medical history and medical records to verify the stroke confirmation.

The United International University ethical review board authorized this study ( IREB/2023/008), and all participants gave written consent before participating in this study. Data were deidentified. As this protocol was part of a community-based health care model, no compensation was given to participants for taking part in the survey.

Initially, data were collected from a total of 1,352,724 individuals. A total of 4030 individuals were excluded due to missing birth dates, and 7088 individuals were excluded due to missing gender. Moreover, 17 individuals aged <18 years who had a stroke were excluded due to a lack of evidence of having a stroke. Following these exclusions, the remaining dataset consisted of a total of 1,341,589 individuals.

First, we collected data from AWS as a dump file. Then, we stored them in a MySQL server and extracted the required data from the original database by running MySQL queries. We examined the data for missing values and outliers. All the missing values and outlier-related instances were removed. We used descriptive statistics to assess the population’s age, gender, education, and occupation distribution. Numeric variables were represented by their mean and SD, while nominal variables were represented by their proportion. The prevalence of stroke per 1000 people was examined along with a 95% CI. We analyzed the prevalence of stroke by divisions to understand the stroke prevalence rate in different parts of Bangladesh.

In addition, the association of stroke risk predictors was calculated using multivariate logistic regression and presented in crude odds ratio (OR) and adjusted OR along with 95% CI. The risk predictors were: five socio-demographic variables (age, gender, education level, occupation, and union name) and 5 health-related comorbidities (stroke, diabetes, hypertension, cancer, CVD, and chronic obstructive pulmonary disease). Before using multivariate logistic regression, we checked for normality, multicollinearity, and outliers. The bar chart and radar chart were used to illustrate the major findings. All these analyses were done with the help of Python (Python Software Foundation) libraries like pandas, Matplotlib, NumPy, SciPy, Plotly, stats model, and Seaborn.

An overview of the data sampling procedure is represented in Figure 1. From July 2018 to June 2021, the CHW questioned a total of 1,352,724 people from 242,034 households, encompassing 7 divisions. After removing missing values and outliers, a total of 1,341,589 people remained from the initial total population.

Detailed sociodemographic information is provided in Table 1. In our study, 51.33% (688,649/1,341,589) were male participants. The average (SD) age was 29.23 (19.05) years, with the average (SD) age of male participants being 29.57 (19.28) years and that of female participants being 28.87 (18.80) years. The majority of people (600,160/1,341,589, 44.74%) had an age of <25 years while the smallest proportion (10,865/1,341,589, 0.81%) of people had an age of ≥80 years. Besides, male participants were found in greater numbers than female participants among different age groups.

Table 4 represents the distribution of patients with stroke having other noncommunicable diseases (NCDs). From Table 4, it was observed that, among patients with stroke, hypertension was the most prevalent among all NCDs. A total of 67.55% (95% CI 65.13‐69.97) of patients with stroke had hypertension. The second most prevalent was diabetes; 25.21% (95% CI 22.96‐27.45) of patients with stroke had diabetes. On the other hand, cancer was the least prevalent among patients with stroke, affecting only 0.63% (95% CI 0.22‐1.03). In addition, it is evident that female patients with stroke had a higher proportion of hypertension and diabetes compared to male patients with stroke. Conversely, male patients with stroke exhibited a higher proportion of cancer, CVD, and chronic obstructive pulmonary disease in comparison to female patients with stroke.

Table 5 represents the ORs of risk factors of stroke. From Table 5, it can be seen that individuals aged 65‐79 years were at a higher risk of having a stroke than other age groups (crude OR 9.883, 95% CI 8.834‐11.057, P<.05; adjusted OR 9.728 [adjusted for sex], 95% CI 8.694‐10.884, P<.05). In addition, male participants were at greater risk of having a stroke than female participants were (crude OR 1.565, 95% CI 1.406‐1.741, P<.05; adjusted OR 1.469 [adjusted for age], 95% CI 1.320‐1.636, P<.05). In addition, it can be recognized that, among hypertension, diabetes, and CVD, people who had CVD were at greater risk of having a stroke (crude OR 109.888, 95% CI 92.069‐131.156, P<.05; adjusted OR 40.232 [adjusted for age and sex], 95% CI 33.42‐48.437, P<.05). Besides, “Khulna” divisions (southwest of Bangladesh) people were at a greater risk of having a stroke (crude OR 1.881, 95% CI 1.671‐2.117, P<.05; adjusted OR 1.769 [adjusted for age and sex], 95% CI 1.571‐1.992, P<.05) than other division’s people.

The main findings of our study were (1) the overall prevalence of stroke was 1.07 per 1000 people on analyzing 1.3 million individuals’ data, (2) the prevalence of stroke was the highest in the “Khulna” division (southwest of Bangladesh), and (3) hypertension was the major risk factor in patients with stroke. To the best of our knowledge, this study was the largest population-based investigation on the prevalence of strokes in rural Bangladesh.

From Table 6, it is evident that our findings have more similarity with those obtained by Saha et al [17] who conducted the study on 94,965 individuals and reported a prevalence of 1.9 per 1000 in the general population. We reported around a 14-fold higher sample size than that reported by Saha et al [17] and found a prevalence of 1.07 per 1000 in the general population. Other studies have calculated the prevalence on a much lower sample size as shown in Table 6. Moreover, it is noteworthy to mention that most of the studies have been performed in 1 specific region (Zaman et al [16] sampled data from only 1 rural area of “Chittagong” division; Saha et al [17] conducted a study from the data that were collected from only 1 rural area of “Rajshahi” division), whereas our study covered 7 divisions of Bangladesh.

Furthermore, this study was focused on rural settings. Mohammad et al [15] and Mondal et al [18] examined data from urban and rural settings and reported higher prevalence. However, these studies did not compare the prevalence between urban and rural settings. Therefore, we could not draw any conclusion from the fact that stroke prevalence may be higher in an urban setting than a rural setting. In addition, Mohammad et al [15] considered individuals aged 40 years and performed the study on 15,627 individuals. In our study of individuals aged ≥40 years, the prevalence was 3.42 per 1000 people, which is slightly higher than that observed in the study by Mohammad et al. Zaman et al [16] conducted a study on people aged ≥30 years on a sample size of 1709 and reported a prevalence of 9.4 per 1000. In our study, the prevalence was 2.23 per 1000 people for patients aged ≥30 years on a population of over 4,00,000. Here, it is noteworthy to mention that our sampled population had a higher proportion of younger people (44.74% of total individuals were aged <25 years) than other studies, which also could play a role in general prevalence calculation. Taken together, it is fair to assume that the lower prevalence reported in our study than in previous literature could be due to the following reasons: (1) the large sample size, (2) the vast region covered, (3) the rural setting, and (4) the proportion of the younger population.

We further found that male participants exhibited a considerably higher prevalence of stroke (1.55 times higher than that in female participants), similar to the findings of the previous study [17]. Moreover, we found that the stroke prevalence rate increased with age. This finding is also in line with the previous studies [12,17,18]. It is worth mentioning that we divided the age groups similarly to a previous study [17], and the age group ranges were different in previous studies [15,18].

Our division-based analysis showed that the highest stroke prevalence rate (1.77 per 1000 people) was observed in the “Khulna” division (southwest Bangladesh). This could be due to the high salt intake of the people since the “Khulna” division is one of the coastal areas of Bangladesh [28]. Islam et al [29] reported that 70% of groundwater in the “Khulna” district has a very high salinity hazard. In addition, several studies have associated a greater risk of developing hypertension with drinking high-salinity water in Bangladesh [30,31]. Furthermore, it is reported in other studies that the “Khulna” Division lacks primary and preventive health care facilities [32]. Due to this, the people of “Khulna” lack education on the disadvantages of drinking high-salinity water, smoking, and so on. Moreover, due to the lack of primary health care facilities, individuals may not seek regular medical consultations for ongoing monitoring of vital health indicators such as blood pressure and glucose levels. As a consequence, they may remain unaware of underlying conditions such as hypertension or diabetes, which are major risk factors of stroke. Taken together, we assumed that high salinity in water and lack of primary health care could be the main reasons for higher stroke prevalence in the “Khulna” division.

We found that hypertension (970/1436, 67.55%) and diabetes (362/1436, 25.21%) were the first and second most common risk factors of stroke than other NCDs. These findings are in line with the previous studies conducted in Bangladesh [17-19,22,23].

This research has several strengths. First, this study is the largest sample size representation of the rural population of Bangladesh. In addition, we collected data from 50 unions representing rural settings all over Bangladesh. This study demonstrated the prevalence of stroke in 7 divisions of Bangladesh, which could be an accurate prevalence estimate of stroke. Second, we found that the stroke prevalence is higher in the “Khulna” division than in other divisions. It is a major finding in the context of Bangladesh, as the “Khulna” division has a high salt intake. We found that hypertension, diabetes, and other CVDs are leading risk factors for stroke in rural Bangladesh, which are known risk factors of stroke. Therefore, managing these diseases may help in managing stroke in rural Bangladesh.

However, this research has some limitations. Since all the data for this study were collected in a survey setting, the assessment of stroke with neurologists was not performed. However, to mitigate this challenge, the CHWs checked their medical records and medication history to make sure the information was accurate. However, we acknowledge the fact that the neurological assessment could lead to a more accurate prevalence estimation. Furthermore, we did not collect information regarding smoking, dietary habits, or family history of stroke. Future studies should include these variables to undertake better risk factor association studies.

This study presented a comprehensive analysis of stroke prevalence in rural Bangladesh covering 7 divisions. One of our major findings was that the total prevalence rate was 1.07 per 1000 individuals in the general population with increasing prevalence in male participants and older populations. Although we reported lower stroke prevalence than in the contemporary studies, we believe this is due to the large sample size of our study. However, one major finding is that the southeast region of Bangladesh (“Khulna” division) has a higher prevalence than other divisions do. As the water of the “Khulna” division has a high salinity hazard along with a lack of primary health care facilities, our findings could help the policymakers to take necessary steps to mitigate these issues.