Stroke imposes a substantial burden worldwide. With the rapid economic and lifestyle transition in China, trends of the prevalence of stroke across different geographic regions in China remain largely unknown. Capitalizing on the data in the National Health Services Surveys (NHSS), we assessed the prevalence and risk factors of stroke in China from 2003 to 2018. In this study, data from 2003, 2008, 2013, and 2018 NHSS were collected. Stroke cases were based on participants’ self-report of a previous diagnosis by clinicians. We estimated the trends of stroke prevalence for the overall population and subgroups by age, sex, and socioeconomic factors, then compared across different geographic regions. We applied multivariable logistic regression to assess associations between stroke and risk factors. The number of participants aged 15 years or older were 154,077, 146,231, 230,067, and 212,318 in 2003, 2008, 2013, and 2018, respectively, among whom, 1435, 1996, 3781, and 6069 were stroke patients. The age and sex standardized prevalence per 100,000 individuals was 879 in 2003, 1100 in 2008, 1098 in 2013, and 1613 in 2018. Prevalence per 100,000 individuals in rural areas increased from 669 in 2003 to 1898 in 2018, while urban areas had a stable trend from 1261 in 2003 to 1365 in 2018. Across geographic regions, the central region consistently had the highest prevalence, but the western region has an alarmingly increasing trend from 623/100,000 in 2003 to 1898/100,000 in 2018 (Ptrend<0.001), surpassing the eastern region in 2013. Advanced age, male sex, rural area, central region, hypertension, diabetes, depression, low education and income level, retirement or unemployment, excessive physical activity, and unimproved sanitation facilities were significantly associated with stroke. In conclusion, the increasing prevalence of stroke in China was primarily driven by economically underdeveloped regions. It is important to develop targeted prevention programs in underdeveloped regions. Besides traditional risk factors, more attention should be paid to nontraditional risk factors to improve the prevention of stroke.
- risk factors
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Stroke is a leading cause of disability and mortality worldwide and imposes a severe global burden.1 2 The Global Burden of Disease (GBD) Study showed the cases and deaths due to stroke markedly expanded between 1990 and 2019 across the world.2 With the wide implementation of stroke prevention strategies and good health services, the burden of stroke has decreased in high-income countries. However, reversed trends have been found in low-income and middle-income countries.2 Accompanied by the fast-growing economies and urbanisation, the burden of stroke in China has also changed substantially in the past decades, Recently, stroke has surpassed cancer and coronary heart disease as the top cause of death in China.3 The prevalence of stroke in China in 2013 was more than three times that in the 1990s,4 Thus, monitoring the epidemiological features of stroke has important implications for public health in China. A few epidemiological surveys on the prevalence of stroke and associated risk factors had been completed in China, but most of these studies were either outdated, local or small samples based, or suffered from selection bias.5–8 Prior studies have also noticed regional variations in stroke prevalence before 2013, Most prominent among them is rural–urban difference, to the disadvantage of rural areas.8 9 However, there is no most up-to-date study on the changes of prevalence of stroke across China.
While traditional stroke risk factors such as old age, male, hypertension, diabetes, smoking and cardiac causes, collectively explain the majority of the population attributable risks of stroke,10 there was evidence indicating excessive stroke risk unaccounted by these traditional risk factors.11 Meanwhile, several strategies for preventing these traditional stroke risk factors, such as health screenings for elderly, hypertension and diabetes management, tobacco control, and cardiovascular diseases therapy have been implemented in China to control stroke,12–15 yet stroke burden continue to grow. As a result, a better understanding of the potentially modifiable non-traditional factors is critically important to formulate strategies for stroke prevention.
Using the most updated data from the National Health Services Surveys (NHSS),15–17 a large-scale population-based health status screening project in China, we evaluate the national trends of stroke prevalence and the associated risk factors in China from 2003 to 2018, with a focus on identifying potentially modifiable risk factors.
Data sources and study sample
This study obtained data from the NHSS system, which is a series of national observational cross-sectional studies covering all 31 provinces, autonomous regions and municipalities in mainland China conducted every 5 years since 1993.15–17 The NHSS is representative of national geographical distribution, socialeconomic status and basic characteristics of the population, providing important information about the health status of the Chinese population. In this survey series, we used a multistage stratified cluster sampling procedure, which was described in online supplemental appendix 1. We divided mainland China into three regions: west, central and east, and then sampled counties stratified by urban and rural areas from each region. Covering 0.02% of the total population with a 5% non-response rate, each country required at least 90 counties and 600 households. The 2003 survey selected 95 counties at random, with 28 counties from urban areas and 67 counties from rural areas; the 2008 survey selected the same 94 counties as the 2003 survey, with one country not selected due to administrative division changes; the 2013 survey selected 62 other counties in addition to those surveyed in 2008; while the 2018 survey selected 84 counties from urban areas and 72 counties from rural areas, taking into account China’s urbanisation transition. Then, from each county, five streets (from urban areas) or townships (from rural areas) were selected, and two communities or villages were selected from each street or township. Finally, 60 households were selected at random from each village or community. In addition, we selected 10 standby households at random in each village or community; if we were unable to interview the initially selected households, we could go on to 1 of 10 standby households. The investigation was open to all members of the selected household.
The detailed interview processes have previously been reported.15–17 In brief, local healthcare workers were recruited and trained to conduct interviews in person. Participants aged 15 years or older were questioned after reading a statement explaining the objective of the survey and obtaining consent. Each round used the same stringent quality control programme. All investigators and research staff underwent unified procedure and data collecting training. The interviewers ensured that the questionnaire was completed at the end of each interview, and the questionnaires were checked daily by the supervisors. Five per cent of the total households with completed surveys were randomly selected to be reinterviewed.
Assessment of stroke and related risk factors
Stroke was assessed based on participants’ self-report in the questionnaire according to the International Classification of Diseases 10 at each round of the survey.15–17 We began the questionnaire by asking the respondents whether they had any chronic diseases that had been diagnosed by doctor. If they answered they had stroke, we inquired when they were diagnosed and whether they had been treated within the previous 6 months (online supplemental appendix 2). As proof of the diagnosis, medical records or prescriptions from medical institutions were necessary. These diagnoses were included in the survey data under the supervision by doctors from township or higher-level hospitals, and the investigator then documented them in the questionnaire.
The NHSS questionnaire provided us with data on stroke related factors (online supplemental appendix 2). We assessed demographic (age and sex) and geographical characteristics (residence and region), socioeconomic status (educational level, occupation, gross annual income and marital status), lifestyle (smoking, alcohol consumption and physical activity), health status (hypertension, diabetes and depression) and household environment (sanitation facilities) in each round. At each round of the survey, participants were asked to self-report their history of hypertension and diabetes, and conformation of the diagnosis was required in the form of medical records or prescriptions from medical institutions. Depression was measured using a quality of life questionnaire and self-perceived health. Participants who had smoked a total of at least 100 cigarettes in their lifetime and either continued or ceased smoking during the survey were classified as smokers; that is, both ex-smokers and current smokers are counted as smokers in the analyses. Participants who had an alcoholic drink in the 12 months prior to the survey were considered as alcohol consumption. Physical activity was defined as participating in physical activity (including tai chi, jogging, dancing, swimming, ball sports, aerobics and apparatus exercise) at least once a week in the previous month. Unimproved sanitation facility is defined as not ensuring hygienic separation of human excreta from human contact and open defecation. Improved sanitation facility is defined as likely to ensure hygienic separation of human excreta from human contact. Online supplemental appendix 3 presents a detailed definition of each risk factor.
All data were recorded on a printed questionnaire and double entered into an online system provided by the National Health Commission of the People’s Republic of China. A database was established using Access software.
The overall population’s stroke prevalence was determined, as well as subgroups stratified by age, sex, residential area and geographical region. We also assessed socioeconomic factors such as education level, occupation, income and marital status. The age-standardised and sex-standardised prevalence of stroke was standardised for age and sex using the 2010 Chinese census for both the overall population and subgroups. To analyse trends in stroke prevalence across time, we used the one-sided (increasing trend) Cochran-Armitage trend test. The Pearson χ2 test was used to assess between-group differences in stroke prevalence. We also compared the evolution of stroke prevalence between urban and rural areas by sex and geographical subgroups. To estimate the ORs and 95% CIs of all recorded factors potentially associated with stroke, we constructed multiple logistic regression models involving age, sex, residence, region, educational level, occupation, income, marital status, hypertension, diabetes, depression, smoking, alcohol consumption, physical activity and sanitation facilities, separately for each survey (online supplemental appendix 4). Meta-analyses were performed for OR value of risk factors from serial surveys. We assessed heterogeneity using the I2 statistic. Individuals with missing values did not have their values imputed. SAS V.9.4 software was used for all statistical analyses.
We sampled 57 023, 56 456, 93 613 and 94 076 households in 2003, 2008, 2013 and 2018, respectively. A total of 154 077, 146 231, 230 067 and 212 318 participants in 2003, 2008, 2013 and 2018, respectively, were included in the final analysis. Overall, 1435 (0.93%), 1996 (1.36%), 3781 (1.64%) and 6069 (2.86%) people had stroke in 2003, 2008, 2013 and 2018, respectively. The age-standardised and sex-standardised prevalence of stroke per 100 000 people in China was 879 (95% CI 834 to 924) in 2003, 1100 (95% CI 1052 to 1147) in 2008, 1098 (95% CI 1063 to 1133) in 2013 and 1613 (95% CI 1572 to 1655) in 2018, respectively (table 1).
The elderly were more likely to be affected, especially those aged 70 years or older, of whom 8137 per 100 000 had a stroke in 2018. The subgroup aged 50–59 years experienced a rapid increase in stroke prevalence, from 1228 per 100 000 people in 2003 to 2448 in 2018. Comparing residential areas (figure 1, table 2), the prevalence of stroke was significantly higher in urban (1261/100 000) than in rural areas (669/100 000) in 2003, but the difference diminished by 2008 (1159/100 000 in urban vs 1052/100 000 in rural areas). Furthermore, the prevalence of stroke in rural areas (1898/100 000) had surpassed that of urban areas (1365/100 000), by 2018. Similarly, while the prevalence increased in all regions in China, the western region has a more dramatic increase (from 623/100 000 in 2003 to 1439/100 000 people in 2018) than the eastern region (from 918/100 000 in 2003 to 1306/100 000 people in 2018). We also estimated stroke prevalence in different provinces of China (online supplemental appendix 5), among which the Jiangxi province experienced the most rapid increase from 253 cases per 100 000 people in 2003 to 1133 in 2018. While most provinces displayed a significant increasing trend in stroke prevalence (Ptrend<0.05), four provinces had stabilised prevalence, including Liaoning, Hainan, Henan and Xinjiang Uygur Autonomous region.
Table 2 shows the disparity of stroke prevalence between Chinese urban and rural areas by age, sex and geographical regions from 2003 to 2018. Stroke prevalence was higher in urban than in rural areas in 2003 for all geographical regions. However, the prevalence was greater in rural than urban areas in 2018 across all geographical regions. Furthermore, in the western region, the prevalence of stroke increased in both urban and rural areas in the past decades. Similarly, for each age or sex subgroup, the prevalence of stroke significantly increased in rural areas during this period. A different trend was observed in urban areas.
Table 3 summarises the risk factors for stroke in 2003, 2008, 2013 and 2018, respectively. We found that advanced age, as well as males, was associated with an increased prevalence of stroke. People in the rural area and central regions had higher stroke prevalence than those in urban areas and western regions. Retired or unemployed people tended to be associated with higher stroke prevalence compared with employed people. We also observed the prevalence of stroke was also associated with hypertension and diabetes. Meanwhile, a higher prevalence of stroke was found in people with severe depression, with the highest OR values. Interestingly, we found unimproved sanitation facilities were consistently associated with a high risk of stroke. The meta-analysis of serial surveys also showed that advanced age, male sex (OR 1.41, 95% CI 1.34 to 1.47), rural area (OR 1.16, 95% CI 1.10 to 1.21), central region (OR 1.37, 95% CI 1.31 to 1.43), retirement (OR 1.78, 95% CI 1.67 to 1.90) or unemployment (OR 1.74, 95% CI 1.65 to 1.82), hypertension (OR 1.79, 95% CI 1.71 to 1.87), diabetes (OR 1.79, 95% CI 1.71 to 1.87), depression (moderate depression: OR 2.71, 95% CI 2.59 to 2.83; severe depression: OR 4.01, 95% CI 3.53 to 4.49) and unimproved sanitation facilities (OR 1.38 95% CI 1.32 to 1.44) were associated with an increased prevalence of stroke. Furthermore, people with high income and having a college education were protective against stroke. Cigarette smoking was not proven to be a risk factor (OR 1.01, 95% CI 0.97 to 1.06), but alcohol consumption was found to be a protective factor against stroke. Excessive physical activity was also associated with a higher risk of stroke in 2003 (OR 1.22, 95% CI 1.04 to 1.43) and 2013 (OR 1.27, 95% CI 1.16 to 140). Results of the analyses stratified by urban and rural areas are shown in online supplemental appendices 6 and 7, respectively. Similar to the overall population, advanced age, male sex, central region, retirement or unemployment, hypertension, diabetes, depression and unimproved sanitation facilities were risk factors for stroke in both urban and rural areas. Risk factors did not differ between urban and rural areas.
This study involves serially collected participants representative of all regions in mainland China with the largest sample size to date, enabling accurate estimation of the trend of stroke prevalence over time and across the country. We observed distinct trends in the stroke prevalence between underdeveloped (rural, western) and developed (urban, eastern) regions in China, with the curves resembling the shape of scissors (figure 1). Furthermore, we identified advanced age, male sex, rural area, central region, hypertension, diabetes, depression, low education and income level, retirement or unemployment, excessive physical activity, and unimproved sanitation facilities as risk factors for stroke.
Our results highlight a marked increase in the age-adjusted prevalence of stroke in 2018 (1613 cases per 100 000 people) compared with that in 2003 (879 cases per 100 000 people), consistent with previous studies.9 18 19 According to the National Epidemiological Survey of Stroke in China, the age-standardised stroke prevalence had reached 1115 cases per 100 000 people in 2013.9 The China National Stroke Screening and Prevention Project, done between 2014 and 2015, showed higher stroke prevalence (2450 cases per 100 000 people) in adults aged 40 years or older.19 The age-standardised prevalence of stroke reported in the GBD Study increased by 13.2% from 1990 to 2019 in China, reaching 1469 cases per 100 000 people in 2019.18 Such an increasing trend in stroke prevalence is comparable to the pattern in other low-income and middle-income countries, whereas it is decreasing prevalence in high-income countries.2 All these results point to a high and growing prevalence of stroke in China. Our findings could be partly explained by changes in the demographic structure, such as rapid population ageing.20 In addition, the prevalence of stroke depends on the stroke incidence, mortality and length of survival after stroke. First, according to updated GBD Study statistics, over 4 million new patients who had a stroke were diagnosed in China in 2019 and the age-standardised incidence rate of stroke was 201 cases per 100 000 people.2 Although the current stroke incidence rate was a little lower than that in 2013 when reported as 247 cases per 100 000 people. It was still significantly higher than in previous comparable surveys, suggesting a substantial increase in stroke incidence over the past three decades.9 Moreover, in comparison with results in 1990, the age-standardised stroke mortality rate fell by 39.8%, reaching 127 cases per 100 000 people in 2019.18 Improvements in emergency services, stroke prevention and treatment decreased the stroke mortality rate and increased length of survival after stroke. Finally, another possible explanation for the heightened prevalence of stroke could be the improvement in access to diagnosis, such as the use of better diagnostic methods. Together, these factors could have led to the rapid increase in stroke prevalence in China.
There are largely geographical and regional variations in stroke burden in China.4 Epidemiological studies have reported a north-south gradient, with the highest stroke prevalence in northern China and the lowest in the southern region.8 9 However, less research has focused on the eastern, central and western regions. We found that the age-standardised and sex-standardised prevalence of stroke in the eastern was consistently higher than that in the western region up to 2008, but became lower than in the western region in 2018. Historically, China has had a higher stroke burden in urban areas than in rural areas.21 A large-scale Chinese population survey in 1986 indicated that stroke prevalence was much higher in cities than in rural areas.22 In our study, stroke prevalence was also significantly higher in urban than in rural participants in 2003, but this difference became smaller over time, and the trend has reversed in 2018. The ‘scissors phenomenon’ describes vividly the disparity of stroke prevalence between underdeveloped and developed regions in China from 2003 to 2018. This phenomenon was also supported by several epidemiological studies.4 9 23 Recent studies have reported a higher stroke incidence in rural China,9 and that the stroke mortality in rural regions has surpassed that of urban areas.9 23 Thus, the current burden of stroke in China appears to be more serious in the central and western regions, as well as rural areas. The regional variations in stroke prevalence may be attributed to the rapid transformations in socioeconomic conditions, and lifestyle over the past decades in China, especially in underdeveloped regions. As a result of urbanisation and economic boom, the main risk factors for stroke, such as hypertension and diabetes, are becoming more prevalent in rural areas than in urban areas.23–25 Urban and eastern regions performed more effectively in preventing and controlling these risk factors than rural and western regions.26 In addition, coupled with the advancement of diagnostic tools, the adoption of CT and MRI ensured the accuracy of stroke diagnosis in underdeveloped regions.27 All these changes could be linked to the dramatic rise in stroke prevalence in rural areas and western regions.
An increase in stroke prevalence likely reflects the change in lifestyle and socioeconomic status. The significance of socioeconomic risk factors as predictors of stroke burden has already been discussed.28 Overall, people with lower socioeconomic level tended to have a higher prevalence of stroke. Our results confirmed previous observations that people with higher income and the highest level of education have a lower risk of stroke. However, it is unclear what explains the association between socioeconomic status and stroke prevalence. Traditional stroke risk factors such as older age, male, hypertension and diabetes may help to explain why people with lower socioeconomic status have a higher rate of stroke, as a larger burden of stroke risk factors was found in people with lower socioeconomic status.29 30 Moreover, we identified occupational status as the most reliable risk factor for stroke; a higher prevalence was seen in unemployed or retired individuals. Of the few studies that have investigated the relationship between occupational status and stroke, one reported that unemployed/retired Japanese women could be at risk for stroke,31 and another from Finland also suggested that occupation status is one of the most common health indicators.32 As far as those who were retired, they generally lost their jobs because of old age or poor health, including possible stroke or other diseases. For those who were unemployed, financial stress, depression and social stigma could have triggered unhealthy behaviours and poor mental health.33 Our findings suggest that, of all the socioeconomic factors, occupational status is the strongest risk factor for stroke among the Chinese population. As such, encouraging individuals to work or further study may help minimise the risk of sustaining a stroke. Occupational status is certainly also influenced by education, income and marital status; their independent impacts cannot be separated altogether. We should thus consider all these factors together during analysis, instead of just focusing on one.
It is generally known that modifiable lifestyles, such as smoking, alcohol consumption, diet and physical activity, have been consistently linked to stroke risk. However, the link between some of these factors and stroke is yet unclear. Several studies have shown drinking is associated with a higher stroke burden, due to increased blood pressure caused by alcohol,34 35 whereas other studies have reported a null or inverse association36 37 and still others reported a J-shaped relationship.38 In our study, drinking was consistently associated with low stroke prevalence. A plausible explanation is that alcohol raises high-density lipoprotein cholesterol levels while reducing platelet aggregation and fibrinolytic activity.39 Although alcohol consumption may be beneficial in terms of stroke prevention, high intake is linked to an increased risk of alcohol-related cancers and injuries.36 Therefore, estimating the health influence of drinking is essential. Physical activity is considered beneficial for stroke prevention by reducing hypertension and diabetes. However, high-intensity physical activity was shown to be associated with an increased risk of stroke in this study. These mirrors findings from a Japanese public health centre-based prospective study.40 Several studies have suggested that high-intensity physical activity may cause haemorrhagic stroke by triggering a sudden and short-lasting increase in blood pressure.40–42 Moreover, greater physical activity might enhance the effect of the increased risk of stroke due to longer exposure to polluted air in developing countries.43 Thus, high-intensity physical activity might not be suitable for the prevention of stroke in China. Depression is highly prevalent not only in China but also worldwide, imposing a huge burden on public health. Several prospective studies have confirmed that in developed countries, depression is related with a considerably increased risk of stroke.44 45 In general, the development of poststroke depression is well recognised, but the function of depression as a risk factor for stroke is less well studied in China. Depression was found to be a significant risk factor for stroke in our study. This suggests that as in developed countries, depression may have a significant influence in stroke prevalence in China. Interestingly, we found people living with unimproved sanitation facilities have a higher risk of stroke, particularly in rural areas. In low-income and middle-income countries, unimproved sanitation facilities are more commonly used in rural areas compared with urban areas. It is likely that poor sanitation is a surrogate for the low socioeconomic status, which is often related to poor access to care. Many diseases have been linked to inadequate sanitation such as malnutrition, diarrhoea, intestinal nematode infections and trachoma.46 47 Nonetheless, no prior research has looked into the association between poor sanitation and stroke. The mechanism through which unimproved sanitation facilities contribute to stroke risk is unknown. Infection and alteration of gut microbiota caused by poor sanitation might increase stroke risk through pathways such as platelet hyperreactivity and immunomodulation.48 49 It is recognised that the gut microbiota-brain axis affects the brain’s pathophysiology.50 A prospective clinical study also showed intestinal microbiota-dependent metabolism of phosphatidylcholine was associated with an increased risk of stroke.51 Furthermore, a study from India indicated over a third of stroke occurred in toilets because squatting could increase blood pressure and thus trigger stroke.52 Most people living with unimproved sanitation facilities perform their ritual in the squatting posture, which might explain why unimproved sanitation was associated with significantly higher stroke risk. Overall, we speculate that improving socioeconomic status and sanitation may help lower stroke risk in rural areas of China, as well as other developing countries.
There are some limitations in our study. First, due to the cross-sectional nature of this study’s methodology, we cannot infer causality from the findings. Also, the self-reported questionnaire may cause recall bias. Second, several well-documented contributing factors that may affect stroke prevalence, such as hyperlipidaemia, dietary factors and obesity, were not included in the questionnaire, making it impossible to analyse their relationship to stroke prevalence. Third, people with stroke risk factors like hypertension and diabetes were likely to have better access to medical care due to their current morbidity and hence more likely to be diagnosed with stroke and this association may be due to a care-seeking bias. Furthermore, risk factors such as lifestyle and health status may have changed after suffering stroke, which may introduce bias. This bias may explain some results such as alcohol consumption which is found to be a protective factor, and high-intensity physical activity might not be suitable for the prevention of stroke. Finally, our study only analysed stroke prevalence, with no data on the incidence and mortality. In the future, we plan to prospective follow-up participants, investigate stroke incidence and collect information about mortality.
In summary, our study presents updated estimates of the prevalence and risk factors of stroke from 2003 to 2018 in China. In the past decade, the scissors phenomenon of stroke prevalence occurred in China. These novel findings indicate that the stroke prevalence may continue to increase in rural areas, and in western and central regions without interventions. Therefore, it is important to develop targeted programmes for stroke prevention in these regions. In addition to traditional risk factors of stroke, more attention should be given to nontraditional risk factors in the public health policies for stroke prevention.
Patient consent for publication
The NHSS was approved by the institutional review board of the Chinese National Bureau of Statistics.
We thank the Centre for Health Statistics Information, National Health Commission of the People’s Republic of China, which provided outstanding support in the data collection and analysis of this study.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
D-ST, C-CL and C-LW contributed equally.
Contributors WW, D-ST and C-CL conceived the idea for the study and managed the project. WW, XL, S-BX, D-JP and J-PH designed the study. D-ST, C-CL, C-LW, CQ, M-HW, W-HL, JL, H-WZ, R-GZ, S-KW, X-XZ and LW collected the data, performed the statistical analyses and wrote the statistical analysis plan. D-ST, C-CL and C-LW wrote the manuscript. WW had full access to all of data in the study and took full responsibility for the integrity of data and the accuracy of the analyses. WW is responsible for the overall content as guarantor. All authors read and approved the final manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.