Stroke risk prediction using machine learning: a prospective cohort study of 0.5 million Chinese adults.
Ontology highlight
ABSTRACT: Objective
To compare Cox models, machine learning (ML), and ensemble models combining both approaches, for prediction of stroke risk in a prospective study of Chinese adults.Materials and methods
We evaluated models for stroke risk at varying intervals of follow-up (<9 years, 0-3 years, 3-6 years, 6-9 years) in 503 842 adults without prior history of stroke recruited from 10 areas in China in 2004-2008. Inputs included sociodemographic factors, diet, medical history, physical activity, and physical measurements. We compared discrimination and calibration of Cox regression, logistic regression, support vector machines, random survival forests, gradient boosted trees (GBT), and multilayer perceptrons, benchmarking performance against the 2017 Framingham Stroke Risk Profile. We then developed an ensemble approach to identify individuals at high risk of stroke (>10% predicted 9-yr stroke risk) by selectively applying either a GBT or Cox model based on individual-level characteristics.Results
For 9-yr stroke risk prediction, GBT provided the best discrimination (AUROC: 0.833 in men, 0.836 in women) and calibration, with consistent results in each interval of follow-up. The ensemble approach yielded incrementally higher accuracy (men: 76%, women: 80%), specificity (men: 76%, women: 81%), and positive predictive value (men: 26%, women: 24%) compared to any of the single-model approaches.Discussion and conclusion
Among several approaches, an ensemble model combining both GBT and Cox models achieved the best performance for identifying individuals at high risk of stroke in a contemporary study of Chinese adults. The results highlight the potential value of expanding the use of ML in clinical practice.
SUBMITTER: Chun M
PROVIDER: S-EPMC8324240 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BackgroundMany cohort studies have explored the relation between tea consumption and stroke risk; however, the conclusions have been inconsistent. In addition, evidence is lacking in China, where the patterns of tea consumption and main types of tea consumed differ substantially from those in high-income countries.ObjectiveWe aimed to systematically assess the association of tea consumption with the risk of stroke based on a Chinese large-scale cohort study.MethodsA total of 487,377 participants from the China Kadoorie Biobank were included in the present study. Detailed information about tea consumption (including frequency, duration, amount, and tea type) was self-reported at baseline. After ?4.3 million person-years of follow-up, 38,727 incident cases of stroke were recorded, mainly through linkage with mortality and morbidity registries and based on the national health insurance system.ResultsOverall, 128,280 adults (26.3%) reported drinking tea almost daily (41.4% men, 15.9% women), predominantly green tea (86.7%). Tea consumption had an inverse and dose-response relation with the risk of stroke (Ptrend < 0.001). Compared with nonconsumers, those who consumed tea occasionally, weekly, and daily had adjusted HRs and 95% CIs of 0.96 (0.94, 0.99), 0.94 (0.90, 0.98), and 0.92 (0.89, 0.95) respectively, with little difference by stroke type. Among those who consumed tea daily, the HRs for stroke decreased with the increasing duration and amount of tea consumed (all P < 0.001). These inverse associations were significant for green tea but not for other types of tea. Among men, but not women, the inverse relations could be detected, and similar inverse associations could be found for male noncurrent alcohol-consumers and noncurrent smokers as well.ConclusionsAmong Chinese adults, higher consumption of tea, especially green tea, was associated with a lower risk of ischemic and hemorrhagic stroke.
| S-EPMC7223259 | biostudies-literature
Project description:Although the relationship between depression and stroke risk has been investigated, findings in previous reports were conflicting. The aim of this study was to prospectively examine the effect of major depressive episodes (MDE) on stroke incidence and further assess the potential dose-response relationship between number of depression symptoms and subsequent stroke risk in Chinese population.A total of 199?294 men and 288?083 women aged 30 to 79 years without a history of stroke, heart disease, and cancer in the China Kadoorie Biobank cohort were followed from 2004 to 2013. A World Health Organization Composite International Diagnostic Interview-Short Form was used to access MDE according to Diagnostic and Statistical Manual of Mental Disorders-IV criteria. Stroke events were ascertained through death certificates, medical records, and health insurance data.Past year MDE was marginally associated with a 15% increased risk of stroke (adjusted hazard ratio, 1.15; 95% confidence interval, 0.99-1.33) in the fully adjusted model, and the association was steeper and statistically significant in individuals aged <50 years, smokers, drinkers, those with higher education degree, body mass index <24.0 kg/m(2), and no history of diabetes mellitus. Moreover, there was a positive dose-response relationship between the number of depression symptoms and increased stroke risk (Ptrend=0.011). In addition, smoking status significantly interacted with MDE on stroke onset (P for multiplicative interaction=0.025).Findings from this large prospective study suggest that the presence of MDE is a risk factor for stroke, especially in smokers.
| S-EPMC4998954 | biostudies-other
Project description:BackgroundTea consumption may have favorable effects on risk of fracture. However, little is known about such association in Chinese adults. The aim of this study was to examine the association between tea consumption and risk of hospitalized fracture in Chinese adults.MethodsThe present study included 453,625 participants from the China Kadoorie Biobank (CKB). Tea consumption was self-reported at baseline. Hospitalized fractures were ascertained through linkage with local health insurance claim databases.The resultsDuring a median of 10.1 years of follow-up, we documented 12,130 cases of first-time any fracture hospitalizations, including 1376 cases of hip fracture. Compared with never tea consumers, daily tea consumption was associated with lower risk of any fracture (hazard ratio (HR): 0.88; 95% confidence interval (CI): 0.83, 0.93). Statistically significant reduced risk of hip fracture was shown among daily consumers who most commonly drank green tea (HR: 0.80; 95% CI: 0.65, 0.97) and those who had drunk tea for more than 30 years (HR: 0.68; 95% CI: 0.52, 0.87). Our conclusions: Habitual tea consumption was associated with moderately decreased risk of any fracture hospitalizations. Participants with decades of tea consumption and those who preferred green tea were also associated with lower risk of hip fracture.
| S-EPMC6265708 | biostudies-literature
Project description:Current experimental and epidemiological studies provide inconsistent evidence toward the association between tea consumption and cancer incidence. We investigated whether tea consumption was associated with the incidence of all cancers and six leading types of cancer (lung cancer, stomach cancer, colorectal cancer, liver cancer, female breast cancer and cervix uteri cancer) among 455,981 participants aged 30-79 years in the prospective cohort China Kadoorie Biobank. Tea consumption was assessed at baseline (2004-2008) with an interviewer-administered questionnaire. Cancer cases were identified by linkage to the national health insurance system. Cox proportional hazard regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). In the present population, daily tea consumers were more likely to be current smokers and daily alcohol consumers. 22,652 incident cancers occurred during 10.1 years follow-up (5.04 cases/1000 person-years). When we restricted analyses to non-smokers and non-excessive alcohol consumers to minimize confounding, tea consumption was not associated with all cancers (daily consumers who added tea leaves > 4.0 g/day vs. less-than-weekly consumers: HR, 1.03; 95%CI, 0.93-1.13), lung cancer (HR, 1.08; CI, 0.84-1.40), colorectal cancer (HR, 1.08; CI, 0.81-1.45) and liver cancer (HR, 1.08; CI, 0.75-1.55), yet might be associated with increased risk of stomach cancer (HR, 1.46; CI, 1.07-1.99). In both less-than-daily and daily tea consumers, all cancer risk increased with the amount of tobacco smoked or alcohol consumed. Our findings suggest tea consumption may not provide preventive effect against cancer incidence.
| S-EPMC6602977 | biostudies-literature
Project description:BackgroundLittle prospective evidence exists about risk factors and prognosis of acute pancreatitis in China. We examined the associations of certain metabolic and lifestyle factors with risk of acute pancreatitis in Chinese adults.Methods and findingsThe prospective China Kadoorie Biobank (CKB) recruited 512,891 adults aged 30 to 79 years from 5 urban and 5 rural areas between 25 June 2004 and 15 July 2008. During 9.2 years of follow-up (to 1 January 2015), 1,079 cases of acute pancreatitis were recorded. Cox regression was used to estimate adjusted hazard ratios (HRs) for acute pancreatitis associated with various metabolic and lifestyle factors among all or male (for smoking and alcohol drinking) participants. Overall, the mean waist circumference (WC) was 82.1 cm (SD 9.8) cm in men and 79.0 cm (SD 9.5) cm in women, 6% had diabetes, and 6% had gallbladder disease at baseline. WC was positively associated with risk of acute pancreatitis, with an adjusted HR of 1.35 (95% CI 1.27-1.43; p < 0.001) per 1-SD-higher WC. Individuals with diabetes or gallbladder disease had HRs of 1.34 (1.07-1.69; p = 0.01) and 2.42 (2.03-2.88; p < 0.001), respectively. Physical activity was inversely associated with risk of acute pancreatitis, with each 4 metabolic equivalent of task (MET) hours per day (MET-h/day) higher physical activity associated with an adjusted HR of 0.95 (0.91-0.99; p = 0.03). Compared with those without any metabolic risk factors (i.e., obesity, diabetes, gallbladder disease, and physical inactivity), the HRs of acute pancreatitis for those with 1, 2, or ?3 risk factors were 1.61 (1.47-1.76), 2.36 (2.01-2.78), and 3.41 (2.46-4.72), respectively (p < 0.001). Among men, heavy alcohol drinkers (?420 g/week) had an HR of 1.52 (1.11-2.09; p = 0.04, compared with abstainers), and current regular smokers had an HR of 1.45 (1.28-1.64; p = 0.02, compared with never smokers). Following a diagnosis of acute pancreatitis, there were higher risks of pancreatic cancer (HR = 8.26 [3.42-19.98]; p < 0.001; 13 pancreatic cancer cases) and death (1.53 [1.17-2.01]; p = 0.002; 89 deaths). Other diseases of the pancreas had similar risk factor profiles and prognosis to acute pancreatitis. The main study limitations are ascertainment of pancreatitis using hospital records and residual confounding.ConclusionsIn this relatively lean Chinese population, several modifiable metabolic and lifestyle factors were associated with higher risks of acute pancreatitis, and individuals with acute pancreatitis had higher risks of pancreatic cancer and death.
| S-EPMC6070164 | biostudies-literature
Project description:A few prospective studies have suggested that tea, alcohol, and fruit consumption may reduce the risk of kidney stones. However, little is known whether such associations and their combined effect persist in Chinese adults, for whom the popular tea and alcohol drinks are different from those investigated in the aforementioned studies. The present study included 502,621 participants from the China Kadoorie Biobank (CKB). Information about tea, alcohol, and fruit consumption was self-reported at baseline. The first documented cases of kidney stones during follow-up were collected through linkage with the national health insurance system. Cox regression was used to calculate the hazard ratio (HR) and 95% confidence interval (CI). During a median of 11.1 years of follow-up, we collected 12,407 cases of kidney stones. After multivariable adjustment, tea, alcohol, and fruit consumption were found to be negatively associated with kidney stone risk, but the linear trend was only found in tea and fruit consumption. Compared with non-tea consumers, the HR (95% CI) for participants who drank ≥7 cups of tea per day was 0.73 (0.65-0.83). Compared with non-alcohol consumers, the HR (95% CI) was 0.79 (0.72-0.87) for participants who drank pure alcohol of 30.0-59.9 g per day but had no further decrease with a higher intake of alcohol. Compared with less-than-weekly consumers, the HR (95% CI) for daily fruit consumers was 0.81 (0.75-0.87). Even for those who did not drink alcohol excessively, increasing tea and fruit consumption could also independently reduce the stone risk. Among Chinese adults, tea, alcohol, and fruit consumption was associated with a lower risk of kidney stones.
| S-EPMC8065818 | biostudies-literature
Project description:BackgroundThe effect of the overall diet quality on cardiometabolic diseases has been well studied in the Western population. However, evidence is still in need regarding dietary patterns depicting unique Chinese dietary habits and their associations with cardiometabolic diseases.MethodsA prospective cohort recruited around 0.5 million Chinese residents aged 30-79 years from 10 diverse survey sites during 2004-08. Dietary patterns were obtained using factor analysis based on the habitual consumption of 12 food groups collected at baseline. Among 477,465 eligible participants free of prior heart disease, stroke and cancer, linkages to multiple registries and health insurance database recorded 137,715 cardiovascular diseases (CVD) and 17,412 diabetes cases (among 451,846 non-diabetic participants) until 31 December 2017. Adjusted hazard ratios (HRs) were estimated to compare the risks of cardiometabolic diseases across quintiles of dietary pattern scores using the Cox regression.ResultsTwo dietary patterns were derived: the traditional northern pattern, characterised by wheat, other staples, egg and dairy products; and the modern pattern, featured with fresh fruit, meat, poultry, fish, dairy products and soybean. Adherence to either dietary pattern was associated with lower risks of major cardiometabolic diseases in a dose-response relationship way. After multivariate adjustment, participants adhering to the traditional northern pattern the most had an 8% (95%CI: 5-11%) lower risk of CVD in comparison with those adhering the least. Corresponding risk reductions were 12% (11-32%) for haemorrhagic stroke (HS), 14% (8-19%) for ischaemic stroke (IS), and 15% (6-24%) for diabetes, respectively. When comparing extreme quintiles of the modern pattern, the adjusted HR of HS was 0.67 (95%CI: 0.59-0.77). Corresponding HRs were 0.89 (0.86-0.92) for CVD, 0.88 (0.77-0.99) for MCE, 0.85 (0.80-0.89) for IS, and 0.89 (0.81, 0.97) for diabetes.ConclusionAmong Chinese adults, both traditional northern and modern dietary patterns were associated with lower risks of cardiovascular disease and diabetes beyond other risk factors.
| S-EPMC8418004 | biostudies-literature
Project description:AIMS/HYPOTHESIS:Season of birth as a surrogate for potential environmental exposure during fetal development and early postnatal life has shown an inconsistent association with adult type 2 diabetes in white populations living in high-latitude regions. The present study aimed to examine the association between birth seasonality and risk of adult type 2 diabetes in Chinese individuals living across wide regions of low latitude and lower to middle latitude. METHODS:Participants from the China Kadoorie Biobank were enrolled during 2004-2008 and followed up until 31 December 2013. After excluding participants with cancer, heart disease, stroke and diabetes at baseline, the present study included 189,153 men and 272,058 women aged 30-79 years. We used multivariable Cox proportional hazards model to estimate the HR and 95% CI. RESULTS:During a median follow-up of 7.2 years (3.3 million person-years), we documented 8784 incident cases of type 2 diabetes. In the whole cohort, compared with summer-born participants, the adjusted HRs (95% CIs) were 1.09 (1.02, 1.16), 1.08 (1.02, 1.15) and 1.09 (1.02, 1.15) for those who were born in Spring, Autumn and Winter, respectively. The association was consistent in both men and women and across subgroups defined by residence and lifestyle factors later in life. CONCLUSIONS/INTERPRETATION:In this large prospective study, participants born in summer had a lower risk of adult type 2 diabetes compared with other seasons of birth, suggesting exposures in early life with some degree of seasonal variation might influence the risk of adult diabetes.
| S-EPMC6521727 | biostudies-literature
Project description:Previous research found tobacco smoking and solid fuel use for cooking to increase the risk of chronic liver disease mortality, but previous cohort studies have not investigated their independent and joint associations with liver cancer incidence in contemporary China. The China Kadoorie Biobank (CKB) study recruited 0.5 million adults aged 30 to 79 years from 10 areas across China during 2004 to 2008. Participants reported detailed smoking and fuel use information at baseline. After an 11.1-year median follow-up via electronic record linkage, we recorded 2997 liver cancer cases. Overall, 29.4% participants were current smokers. Among those who cooked at least once per month, 48.8% always used solid fuels (ie, coal or wood) for cooking. Tobacco smoking and solid fuel use for cooking were independently associated with increased risks of liver cancer, with hazard ratios (95% confidence intervals [CIs]) of 1.28 (1.15-1.42) and 1.25 (1.03-1.52), respectively. The more cigarettes consumed each day, the earlier the age of starting smoking or the longer duration of solid fuels exposure, the higher the risk (Ptrend < .001, =.001, =.018, respectively). Compared with never smokers who had always used clean fuels (ie, gas or electricity), ever-smokers who had always used solid fuels for cooking had a 67% (95% CIs: 1.29-2.17) higher risk. Among Chinese adults, tobacco smoking and solid fuel use for cooking were independently associated with higher risk of liver cancer incidence. Stronger association was observed with higher number of daily cigarette consumption, the earlier age of starting smoking and longer duration of solid fuel use.
| S-EPMC7612779 | biostudies-literature
Project description:Timely stroke diagnosis and intervention are necessary considering its high prevalence. Previous studies have mainly focused on stroke prediction with balanced data. Thus, this study aimed to develop machine learning models for predicting stroke with imbalanced data in an elderly population in China. Data were obtained from a prospective cohort that included 1131 participants (56 stroke patients and 1075 non-stroke participants) in 2012 and 2014, respectively. Data balancing techniques including random over-sampling (ROS), random under-sampling (RUS), and synthetic minority over-sampling technique (SMOTE) were used to process the imbalanced data in this study. Machine learning methods such as regularized logistic regression (RLR), support vector machine (SVM), and random forest (RF) were used to predict stroke with demographic, lifestyle, and clinical variables. Accuracy, sensitivity, specificity, and areas under the receiver operating characteristic curves (AUCs) were used for performance comparison. The top five variables for stroke prediction were selected for each machine learning method based on the SMOTE-balanced data set. The total prevalence of stroke was high in 2014 (4.95%), with men experiencing much higher prevalence than women (6.76% vs. 3.25%). The three machine learning methods performed poorly in the imbalanced data set with extremely low sensitivity (approximately 0.00) and AUC (approximately 0.50). After using data balancing techniques, the sensitivity and AUC considerably improved with moderate accuracy and specificity, and the maximum values for sensitivity and AUC reached 0.78 (95% CI, 0.73-0.83) for RF and 0.72 (95% CI, 0.71-0.73) for RLR. Using AUCs for RLR, SVM, and RF in the imbalanced data set as references, a significant improvement was observed in the AUCs of all three machine learning methods (p < 0.05) in the balanced data sets. Considering RLR in each data set as a reference, only RF in the imbalanced data set and SVM in the ROS-balanced data set were superior to RLR in terms of AUC. Sex, hypertension, and uric acid were common predictors in all three machine learning methods. Blood glucose level was included in both RLR and RF. Drinking, age and high-sensitivity C-reactive protein level, and low-density lipoprotein cholesterol level were also included in RLR, SVM, and RF, respectively. Our study suggests that machine learning methods with data balancing techniques are effective tools for stroke prediction with imbalanced data.
| S-EPMC7142983 | biostudies-literature