Introduction

Increasing life expectancy in most countries has resulted in increased proportion of elderly and associated burden of chronic Non-Communicable Diseases (NCDs). NCDs account for about two-thirds of deaths both globally (68%) and in India (60%) [1] leading to significant long term health, economic and social costs. The proven cost-effectiveness of prevention interventions, make NCD prevention and control, a priority in the 21^st century [2].

Stroke is the second leading cause of death and the leading cause of disability [2, 3]. Globally, between 1990 and 2019, there is a 70% increase in stroke incidence, a 43% rise in stroke-related deaths, a 102% increase in stroke prevalence, and 143% surge in Disability-Adjusted Life Years (DALYs) [4]. India, is currently undergoing an epidemiological transition from communicable diseases to NCDs [5]. Stroke is the fourth leading cause of death, fifth leading cause of disability [6], and a major public health concern in India. In addition to being fatal, stroke is associated with short and lifetime disability among survivors, affecting their quality of life and productivity [7]. A high proportion of stroke survivors suffer from permanent impairments resulting in deficient self-care, requiring long-term support from care-givers, adding to individual health and social costs [8].

Epidemiology and care of stroke widely varies across settings in India. Stroke accounted for 7.73% fatalities and 4.26% of DALYs in Karnataka, a state in southern India [9]. An estimated 500,000 stroke cases are prevalent at any given point of time in Karnataka, placing immense burden on the health system [9]. The situation is compounded by lack of regular, valid and reliable stroke related epidemiological and treatment data for evidence-driven programming [10]. Incidence of stroke is largely dependent on incidence of stroke risk factors, which are dynamic and defined based on understanding of epidemiology, from other countries.

Ongoing national surveys and research studies indicate an increasing trend in prevalence of known stroke risk factors like tobacco use, physical inactivity, alcohol use, hypertension and dyslipidemia. However, current knowledge lacks clarity on the combined effect of multiple risk factors and their various combinations. Understanding the individual, combined and synergistic effect of these known stroke risk factors coupled with information on neo-risk factors for stroke, would help to bridge gaps in stroke epidemiology. This enables in-depth understanding of stroke and its risk factors and facilitates evidence-based health systems and services for risk reduction, case management and rehabilitation of stroke.

The NIMHANS-NH-SKAN stroke project, a large population-based cohort study, aims to examine all issues related to stroke covering risk factors, occurrence, treatment and outcomes. This project also intends to develop an India-specific stroke risk predictability calculator for the studied risk factors that shall predict an individual’s risk of stroke over specific time periods. The NIMHANS-NH-SKAN stroke project is implemented by the Department of Epidemiology, Centre for Public Health, NIMHANS, Bengaluru

We describe the methodology of NIMHANS-NH-SKAN Stroke project which aims to:

1. estimate the incidence and incidence rates of first-ever stroke among risk factors (individually and combined) in Karnataka.
2. estimate the strength of association between select risk factors and acute first-ever stroke in Karnataka.
3. estimate short- and long-term survival, disability and quality of life of first-ever stroke patients in Karnataka.
4. identify and estimate strength of association of factors associated with short- and long-term survival, disability and quality of life of first-ever stroke patients in Karnataka.
5. develop a specific risk predictability calculator for development of stroke over specific time periods.

Methodology

This population-based cohort study assesses the effect of various risk factors on stroke occurrence and its outcomes namely survival, disability and quality of life. For the ease of understanding methodology is explained in two parts (Fig 1):

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Overview of the NIMHANS-NH-SKAN stroke project.

https://doi.org/10.1371/journal.pone.0310309.g001

1. Stroke-free period–period from initial recruitment of a study subject until development of stroke or end of follow-up period.
2. Stroke and post-stroke period–period after when a study subject develops stroke until one-year post-stroke or death whichever is earlier.

Follow-up

Stroke-free period.

After the baseline assessment, all survey respondents will be followed up once annually at three time points (year 2, year 3 and year 4) (Fig 2). For those respondents who move out of the study area, attempts will be made to collect data from them, if they are available within a reasonable travel distance (e.g.: within Kolara and close by urban wards within Bengaluru). Efforts will also be made through telephonic calls either to contact them personally or ascertain their stroke status, to the extent possible.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Schematic representation of NIMHANS-NH-SKAN stroke project.

https://doi.org/10.1371/journal.pone.0310309.g002

The study end-point for subjects in the stroke-free period will be when the subject (Fig 2)

1. completes the follow-up period of four years without developing stroke
2. develops stroke during the four-year follow-up period
3. dies/ migrates from the study area/ is loss to follow-up
4. stops participation in the study for any other reason

Stroke and post-stroke period.

Any respondent who develops stroke during the study period will be assessed within the hospital and followed up at four time-points (28^th day, 3^rd month, 6^th month and 12^th month) post-stroke (Fig 2). For those respondents who move out of the study area, efforts similar to stroke free period shall be made to minimize loss to follow up.

The study end-points for post-stroke period subjects will be when the subject

1. completes the follow-up period of 1 year
2. dies/migrates/ lost to follow-up
3. stops participation in the study for any other reason

Ascertainment of exposure and outcome

Stroke-free period.

We plan to study multiple exposures in this study. Naturally existing stroke risk groups namely weight, mental health/ stress (includes depression, anxiety, stress, non-suicidal self-injury, suicidal behavior and workplace stress), non-communicable diseases (diabetes mellitus, hypertension, dyslipidemia and coronary vascular diseases) substance use including tobacco (smoking/smokeless), alcohol and drugs (injecting/sniffing/oral), nutrition, physical inactivity and certain neo-risk factors namely gambling, COVID-19 infection and vaccination, sleep (insomnia and obstructive sleep apnea), exposure to media, risk of cell phone addiction and Restless leg syndrome constitute exposures. The primary outcome assessed is stroke (including its sub-types).

Stroke and post-stroke-period.

We plan to compare/assess the difference in the survival (both long term and short term), disability and quality of life among stroke patients. Type of stroke will be considered apriori exposure along with different exposures assessed for the stroke free period as confounders.

The definitions/ assessments proposed to ascertain different exposures and outcome is shown in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Instruments used for data collection NIMHANS-NH-SKAN stroke project.

https://doi.org/10.1371/journal.pone.0310309.t001

Sample size and power calculations

Stroke-free period.

The study sites constitute two urban wards in Bengaluru and Mulbagal taluka in Kolara district (with 20:80 –urban: rural distribution of population). Each urban ward is expected to have a minimum of 80,000 people. Approximate population within the taluka is 250,000. The proportion of population above 30 years is approximately 58.9% [11]. Thus, about 241,490 subjects (410,000*0.589) are expected to be available for this project. Assuming 20% attrition, we expect ~193,192 subjects to participate in the study.

Assuming 150 new stroke cases per 100,000 population per year and follow-up duration of 3 years at 5% level of significance, sample sizes are calculated for different risk group (Table 2). Since the expected prevalence of obesity is the least among all the risk groups (6.2%), sample size estimation based on obesity as exposure is expected to cover the required sample size for other groups also. With the expected population of ~240,000 subjects, the study results would have a power of 90% to achieve a Hazard Ratio (HR) of 1.4. Even with an attrition rate of 20% (~193,000 subjects) we would be able to achieve the proposed objective with >80% power.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Sample size calculation for different Hazard ratios and power among apriori risk factor groups.

https://doi.org/10.1371/journal.pone.0310309.t002

Stroke and post-stroke period.

With the minimum estimated number of 193,000 respondents available (after accounting for 20% attrition) for stroke-free cohort and an expected incidence of 150 per 100,000 population per year, we expect around 290 cases of stroke occurring in the study areas every year. Thus, a total of 870 cases are expected to occur in the study area during the study period. Efforts shall be made to recruit all these cases into the stroke cohort study.

Discussion

This large-scale community-based prospective-retrospective cohort study is envisaged to assess the effect of known and neo-risk factors in the natural history of stroke beginning from the stroke-free period to development of stroke and its subsequent outcomes (survival, disability and quality of life post-stroke). This study enrolls a sample of ~200,000 individuals naturally stratified into multiple population-based risk groups within Bengaluru and Kolara districts in Karnataka, India. Stratification of the population into six risk/exposure groups of single, dual, or multiple combinations for stroke will be done. At the end of the study, we propose to develop an India-specific stroke risk predictability calculator for the studied risk factors that shall predict an individual’s risk of stroke over five-year periods. Further, data from this study shall also enable understanding the effect of these risk factors, individually and combined, on NCDs other than stroke and their outcomes.

Cohort study is the most suitable study design to accomplish our stated objectives. However, prolonged latency period of development of stroke in an individual’s life span, multifactorial causation, and indefinite time of onset of risk factors/NCDs are known difficulties to conduct cohort studies on NCDs including stroke [19]. The employed mixed (Prospective-Retrospective) cohort design attempts to address these challenges. The study considers various apriori risk factors i.e., weight, mental health/stress, NCDs, nutrition, physical inactivity and substance use. Recall based retrospective /documentary evidence of the onset of certain risk factor counters the prolonged latency and provides valid information on the time of onset of risk factors. These study subjects are further followed prospectively to have a clear insight of the time period between onset of risk factors and development of stroke. Other major advantage of our mixed cohort study design is ability to study multiple exposures and multiple outcomes namely stroke, survival, disability, quality of life, and other NCDs. Cohort studies, mostly assess single exposures and multiple outcomes [19]. The mixed cohort study design in this scenario allows assessment of multiple exposures as well. Stroke being a rare event [20], a case control design is a better suited design [21]. However, concerns with respect to temporality, recall bias, and availability of funding made us opt for a mixed (Prospective-retrospective) cohort study. Other advantages like less susceptibility to selection bias, possibility to estimate true risk, measures of impact (population attributable risk and attributable risk) made us opt for cohort design compared to a case-control design. Furthermore, our objective to assess multiple exposures and outcomes along with developing a stroke-risk predictability calculator makes cohort study a valuable study design in comparison to other available study designs. Alternatively, stroke registries are an invaluable resource to assess the burden of stroke. Stroke registries in India are predominantly urban-centric and limited to few risk factors and/or outcomes assessed [22, 23]. Difficulties in obtaining informed consent, unwillingness to share data, difficulty in retrieving medical records, and inadequate documentation of deaths are all known issues [24] limiting the utility of stroke registries to accomplish our stated objectives. Further, our study involves follow-up of naturally existing risk factor groups from the stroke-free period to the development of stroke and beyond, looking at stroke outcomes. This life course approach of assessing the effect of risk factors on stroke and its outcomes is unique to our study and is best accomplished by the adapted mixed cohort study design.

Ascertainment of risk factors and confounders, in our study, involves utilization of multiple validated and standardized tools for use within the community coupled with strong quality assessment and monitoring of study processes. Most tools proposed are extensively used routinely in clinical and biomedical research, medical practice, auditing and policy making [10, 25–42]. Age of onset/diagnosis of risk factors is considered to panelize data in this study, instead of the more typical and accurate time of assessment of the risk factor during the study [43]. The Cox model implicitly matches subjects on the time scale used. Since the rates of stroke increase with age, more so after onset of risk factor, it is assumed to probably have a larger effect than time of assessment. Thus, age of onset/diagnosis is considered most appropriate to use as time scale for the model.

One of the challenges in implementing such large-scale community-based cohort studies is ensuring community participation and high response rates [44]. Our study employs a strong community engagement plan to ensure high response rate and community participation. These measures are intended to reduce non-response/ loss-to-follow up–known limitations of a cohort study. The study team is acquainted with the cultural and social ethos of the study sites along with a long working relationship with the respective district administrations within Bengaluru and Kolara is likely to facilitate reduction in non-response.

Uniform and standardized data collection, minimizing known information biases like rounding errors, social desirability, and recall issues related to the exact dates of onset / diagnosis of risk factors assume importance when such large-scale cohort studies are implemented. Appropriate informed consent procedures, conducting interviews at a place convenient to the participant ensuring confidentiality and privacy are likely to reduce information bias. Gender-matched data collectors facilitate minimizing information bias mainly social desirability related to certain sensitive information.

Eligible respondents are those aged ≥30 years. This is based on the recommendation of the EAG since the Indian National Programme for Prevention & Control of Non-Communicable Diseases (NP-NCD) recommends opportunistic screening for risk factors and stroke among individuals aged 30 years and above [45]. Further, including these individuals shall capture dynamics of stroke in the young, an issue in countries of South-east Asian region, especially India [22, 23]. The study also includes pregnant and lactating women since they are known to develop stroke due to various risk factors, including those related to pregnancy [23].

The sample size for both stroke-free and post-stroke period is adequate in terms of the size and power. Although, apriori risk factors are known risk factors for stroke, to our knowledge, there are very minimal studies looking at these risk factors on stroke outcomes namely, survival, disability and quality of life. Not many studies assess the effect of various combinations of stroke risk factors on development of stroke and stroke outcomes. Our study provides an opportunity to understand risk factors of stroke, outcomes of stroke and various combinations of the same. This is a unique strength of our study.

The study employs specific quality assessment procedures along with an IMT, monitoring different processes involved in the study regularly. The entire study methodology is validated and approved by an EAG. In addition, the project team will monitor various activities and processes, including data collection, data storage, training, and documentation focusing on quality. Strict protocols are established for data transfer and management with access-controlled mechanisms. All these efforts likely ensure implementation of study activities to the highest of standards.

Strengths and limitations

This study has several strengths and certain limitations that need mention. The large sample size of ~200,000 individuals is a strength of the study. This serves to estimate the effect of individual and multiple combinations of risk factors to be assessed on stroke. Five among the six risk factors being assessed in this study are known risk factors. One of the risk factors (mental health/ stress cohort) is unique and a strength of the study. There are other neo-risk factors namely of gambling, COVID-19 infection and vaccination, sleep (Insomnia and obstructive sleep apnea), exposure to media, risk of cell phone addiction and restless leg syndrome being assessed in our study; which are unique & strength of the study. The study setting includes both urban (both slum and non-slum) and rural settings in Bengaluru and Kolara, which reflects the diversity and heterogeneity of the Indian population. Using a mixed cohort design (retrospective-prospective), allows for understanding the age of onset/diagnosis and for quantification of risk factors from onset/diagnosis. This enables assessment of time and dose response relationship between the risk factors, stroke and stroke outcomes. Thus, enabling comprehensive understanding of natural history of stroke across the lifespan, from the stroke-free period to outcomes of stroke. Development of stroke risk predictability calculator is a strength of this study. There are no such calculators available in India. Despite the strengths of this study, there are certain limitations that needs mention. Although data collectors would have been trained to collect information on certain sensitive issues, social desirability bias in reporting certain sensitive information like substance use cannot be completely ruled out. The adapted study design is known to increase the period of observation and negate the long latency associated with development of non-communicable outcomes. While ascertaining exposures, it is ideal to have time of onset as the beginning of exposure. However, date of diagnosis was the best alternative available given the elusive nature of onset of NCDs. This might pose a challenge of lead time bias associated with development of stroke as well as its consequent survival. This will be assessed and adjusted for in the analysis stage [46]. Adaptation of positive behaviors or change in risk behaviors are known challenges while conducting cohort studies [19]. This may reduce the incidence and risk of stroke among the study groups. However, sample size is calculated on the lowest possible incidence of stroke based on incidences reported in literature. This is likely to counter the reduced incidence and risk of stroke to a certain extent due to adoption of positive behavior. Intermixing of risk factors might complicate ascertaining measures of effect and impact. However, we propose to calculate rate ratio as the measure of effect. This negates the contamination of exposure ascertainment due to inter-mixing/changing of exposure status over the study period. The project is limited to specific areas, namely one taluka of Kolara district and two selected urban wards of Bengaluru city. Due to this limited coverage, the project may not fully statistically represent the diverse population and stroke burden across the entire state or country limiting the generalizability of results. But the choice of the study setting has its significance. The residents of the study sites come from various socio-demographic backgrounds (urban-rural-linguistic-slum/non-slum backgrounds). This ensures representation of various socio demographic/ economic and cultural strata representing diversity of the state or country. However, this might need to be further explored to ensure generalizability.

Conclusion

This large population-based stroke cohort assessing stroke and its outcomes, first endeavor of its kind, would help address the need to have sufficient epidemiological data regarding the natural history of stroke, its management and outcomes for evidence-based stroke care in India. Being one of the largest population-based cohort studies in India the NIMHANS-NH-SKAN project has the potential to offer valuable information about stroke epidemiology, risk factors, and care in India with major implications in policy making and health programming both at community and national level including introduction of stroke risk predictability calculator in Indian National Programme for Prevention & Control of Non-Communicable Diseases (NP-NCD) in India in specific and similar countries in general.

Supporting information

Acknowledgments

The authors extend their sincere gratitude to the Expert Advisory Group, whose invaluable insights and expertise significantly contributed to the development of this scientific paper. Special thanks to Dr. Pratima Murthy, Dr. Jeyaraj Durai Pandian, Dr. Sreekumaran Nair, Dr. Prashant Mathur, Dr. Srinivasa G A, Dr. Girish Baburao Kulkarni, Mr. Sundar Ramaswamy, and Dr. G Gururaj for their guidance and support. Additionally, the Independent Monitoring Team comprising of Dr. Akshaya KM, Dr. Usha S, Dr. Ramesh Holla, Dr. K Vidusha, Dr. Aravind Karinagannanavar, Dr. Malatesh Undi, Dr. Sharankumar Holyachi, and Dr. Anwith HS played a crucial role in ensuring the robustness and quality of the research. Their dedication and expertise are sincerely acknowledged. We would also like to thank the State and district health administration of Karnataka and Kolara respectively, health authorities of BBMP for providing administrative approvals for the conduct of the study. We would also like to thank our assistant research officers in our team Ms. Ashwini A and Shraddha S Pada.