| Currently, chronic disease is a major public health problem with the characteristics of long incubation period, hidden incidence and difficult healing or curing. Chronic disease is a kind of disease that can be prevented and controled effectively. Related research showed that more than about half of the reasons for significant decline of the mortality rate of coronary heart disease, diabetes and other diseases were attributed to the decrease of the risk factors. Particularly, decreased smoking rate and cholesterol levels played important role in the decline of the mortality rate of chronic diseases. It is important to carry out the research on the methods of measuring risk factors and risk levels for developing targeted interventions and effective prevention and treatment of chronic diseases. Health examination is widely recognized as the foundation and important content of chronic disease risk assessment studies. Health examination data, accumulating a large number of population health information, is the main source of information for population health assessment and common chronic disease risk assessment, which also provides an important platform for clinical epidemiology and chronic disease prevention research. Currently, there is no unified structure and standard for health examination data collection. Because of the different medical data storage structure, data items, and range in different medical institutions, examination results of different medical institutions can not be cross-agency summary, which resulting in a large waste of population health medical examination information and bring some difficulties in the smooth development of chronic disease risk assessment work. Currently, because there are no standardized data collection and data standards for chronic disease risk assessment based on the data of health examination, correlation analysis between health information and health outcomes and health risk assessment cann’t be done timely, and therefore can not achieve individualized healthy lifestyle guidance and intervention. In order to use the health examination platform as soon as possible, unify medical examination database standards, extract important health-related data, mine meaningful clinical information, it is necessary to carry out the research on health examination data collection standards, risk factors measurement, risk assessment methods and other related research. Objectives:1. To understand and identify the main risk factors for high blood pressure, coronary heart disease, stroke, diabetes, cancer, lung cancer and breast cancer seven common chronic diseases according to the evidence-based medical literature search results, referring to clinical expert advice, combining with the actual situation of our country. It can provide a theoretical basis for measuring risk factors and risk levels of common chronic disease.2. Based on health examination data, to develop and design standard common chronic disease risk assessment questionnaire and several special post-disease risk assessment questionnaire, which can provide a reference standard and practical basis for chronic disease risk assessment.3. To develop common chronic disease risk level assessment tools according to some statistical modeling methods such as Harvard Cancer Index, Synthetic analysis method and Joint model. Methods:1. To understand and identify the main risk factors for high blood pressure, coronary heart disease, stroke, diabetes, cancer, lung cancer and breast cancer seven common chronic diseases according to the evidence-based medical literature search results, referring to clinical expert advice, combining with the actual situation of our country.2. Based on the concept of health management, according to domestic and international information standards and industry business service specifications, refering to health management paper inquiry questionnaire and electronic health records from large domestic medical examination institutions and other relevant research results, to study and design common chronic risk assessment questionnaire and several special post-disease risk assessment questionnaire. To establish common chronic disease risk assessment basic data set based on the standard assessment questionnaire.3. Refering to the relevant literature, the mainstream score method was used to construct hypertension, coronary heart disease, stroke, diabetes, stomach cancer, lung cancer and breast cancer risk assessment model. Reliability and validity was made for each disease risk assessment models.4. Based on evidence based medicine, Harvard cancer index was used to build hypertension risk index. The prediction performance and other features of the hypertension risk index were assessed through studying the correlation between the risk level of hypertension and the number of hypertension patients.5. Based on cross-sectional health examination data, synthetic analysis methods was used to build hypertension risk assessment model. According to the area of under operating characteristic curve(ROC) curve, sensitivity, specificity and other main evaluation index, the prediction performance and accuracy of logistic regression analysis model and synthetic analysis model were evaluated.6. With the construction of hypertension risk prediction model as an example, to study and discuss three statistical modeling methods of traditional logistic regression analysis, Cox regression model and Joint model in constructing disease risk assessment model. Prediction performance and accuracy of the three types of models were evaluated through the area of under operating characteristic curve(ROC) curve, sensitivity, specificity and other indicators. Results:1. Evidence-based medicine literature search results showed that family history, blood lipids abnormalities, obesity, smoking, passive smoking, heavy drinking, lack of vegetable intake, inadequate fruit intake, lack of physical exercise, negative emotions and events, chronic stress and sleep problems were the common risk factors for high blood pressure, coronary heart disease, stroke, diabetes, stomach cancer, lung cancer and breast cancer.Different chronic disease has its specific risk factors, such as family history of premature for cardiovascular disease, history of gestational diabetes for diabetes, helicobacter pylori infection and A-type blood for stomach cancer, special occupational exposure history for lung cancer and menarche age, menopause age and frequency of abortion for breast cancer.2. Standardized electronic document architecture designed in this study based on chronic disease risk assessment questionnaire includes two sections, the document header and body of document. The document header contained examination table identifies, health examination institution, subjects signs and demographic characteristics. The document body part included questionnaire(including medical history, past history, psychological factors and special post-disease risk assessment items) and physical examination. Past medical history in the inquiry questionnaire not only covered the content of medical history, family history, exposure history(environmental exposure and occupational exposure) and lifestyle(diet, smoking, alcohol consumption, physical exercise, physical activity and sleep), but also included a women’s history-related content such as menstruation and obstetrical history.Special post-disease risk assessment items were designed for assessing the disease risk of cardiovascular diseases(hypertension, coronary heart disease and stroke), metabolic diseases(diabetes) and cancer(stomach cancer, lung cancer and breast cancer). Mental and psychological factors include temperaments, work and life stress, negative events and emotions, etc.3. According to pre-experiment analysis and expert guidance, common chronic disease risk assessment questionnaire(it is called chronic risk assessment basis questionnaire in this study) was designed. In addition to basic information and routine physical examination of an individual, the basis questionnaire consisted of 113 indicators(items), all which came from Chinese people healthy standardized indicators library. The basis questionnaire contained a total of 210 standard data elements, in which 25 data element were document header data elements and other 185 data elements were document body data elements. Among the 185 document body data elements, 167 data elements were common data elements, and the other 18 were female-specific data elements.4. Based on the standardized chronic disease risk assessment basis questionnaire, referring to the relevant literature, several special post-disease risk assessment questionnaires were designed to assess specifically hypertension, coronary heart disease, stroke, diabetes, stomach cancer, lung cancer and breast cancer. Each questionnaire and every indicators were normalized. Different special post-disease risk assessment questionnaire contained different data elements. Hypertension special assessment questionnaire was composed of 10 items and it contained a total of 31 data elements, in which 17 data elements were core data elements. Coronary artery disease special assessment questionnaire contained a total of 15 items, included 48 data elements, in which 30 data elements were core data elements. Stroke special assessment questionnaire contained 15 items, with a total of 48 data elements, in which 26 data elements were core data elements. Diabetes special risk assessment questionnaire with a total of 13 items contained 20 data elements, in which 16 data elements were core data elements. Stomach cancer risk assessment questionnaire,with 12 indicators, contained 27 data elements and in which 13 were core data elements. Lung cancer risk assessment questionnaire,with 8 indicators, contained 19 data elements and in which 10 were core data elements. Breast cancer special risk assessment questionnaire with a total of 14 items contained 31 data elements, in which 26 data elements were core data elements. Among the above special disease risk assessment data elements, in addition to five additional data elements are added to assess breast cancer risk, the remaining data elements were come from chronic disease risk assessment basis questionnaire.5. Based on special post-disease risk assessment questionnaires, seven kinds of diseases models were constructed and each model was tested in reliability and validity. Reliability test results showed that the Cronbach alpha coefficients of hypertension, coronary heart disease, stroke, diabetes, stomach cancer, lung cancer and breast cancer disease risk assessment models were 0.632, 0.667, 0.688, 0.689, 0.657, 0.635 and 0.618, respectively. The above internal consistency coefficients were all higher than 0.6, which showed that the consistency of each disease risk assessment model is better.Split-half reliability test results showed that Spearman-Brown coefficients of hypertension, coronary heart disease, stroke, diabetes, stomach cancer, lung cancer and breast cancer disease risk assessment models were 0.677, 0.672, 0.566, 0.629, 0.634, 0.643 and 0.534, respectively. In addition to the Spearman-Brown coefficients of stroke and breast cancer, other disease models split-half reliability coefficients were larger than 0.6, indicating a strong correlation between items. Retest reliability showed that the basis questionnaire had high test-retest reliability, with a a retest reliability coefficient of 0.650(P <0.05). The construct validity test results showed that the Kaiser Meyer Olkin measure of sampling adequacy of seven kinds of disease risk assessment model were all about 0.60. In each questionnaire, contribution rate of common factors can explain more than 50% of the total variance. Apart from a few items(such as blood type item of gastric questionnaire, heavy drinking item of breast cancer questionnaire), the vast majority of the items reflect the expected measurement dimensions.The above test results showed that each disease risk assessment questionnaire has good reliability and validity.6. Based on the literature and the results of previous studies, refering to Harvard Cancer index, age, gender, family history of hypertension, overweight or obese, TG abnormalities, smoking and drinking were used to develop the hypertension risk index.The results showed that the risk score values of age, gender, family history of hypertension, overweight or obese, TG abnormalities, smoking and drinking were 5 points, 10 points, 10 points, 10 points, 10 points, 10 points and 10 points, respectively. The population hypertension average risk score value was 18 points. 17,032 individuals were assessed by the hypertension risk index designed in this study. The assessment showed that, according to the hypertension risk index, 3770 were assessed as “much lower risk”, 3860 were assessed as "low" risk, 5363 were assessed as “high risk”, and 4841 individual were assessed as " much higher risk ". In order to evaluate the performance of the hypertension risk index, the incidence of hypertension in different risk levels was analyzed. The results showed that 802 people suffered from high blood pressure during the 7-year follow-up, Among 802 hypertensions, 29(3.62%), 74(9.23%), 204(25.44%) and 495(61.72%) individuals were assessed as “much lower”, “lower”, “higher” and “much higher” risk, respectively. Hypertension incidence in different risk levels was different(2c=557.650,P<001.0). With the increase of risk level, the number of high blood pressure patients increase.7.Based on cross-sectional examination of data sets, synthetic analysis methods was used to build hypertension predictive models. Depending on the two different univariate regression coefficient values of family history of hypertension, two different synthetic analysis model were built. One was logit P =-7.664 + 1.320 × age(≥45 years) + 2.823 × sex + 2.159 ×family history of hypertension + 2.190 ×overweight or obese + 1.734 × triglyceride abnormalities + 1.211 × smoking + 1.973 × heavy drinking(No.1 synthetic analysis model with 2.280 of historyfamilyu()b), and another was logit P =-8.303 + 1.320 × age(≥45 years old) + 2.823 × sex + 4.669 × family history + 2.237 × overweight or obese + 1.734 × triglyceride abnormalities + 1.211 × smoking + 1.973 × heavy drinking(No.2 synthetic analysis model with 4.790 of historyfamilyu()b). Based on the same data set, prediction performance of three models were made. With the same numbers of variables, the range of the difference of AUROC between No.1 synthetic analysis model and logistic regression model was from 0.009 to 0.043, and the range of the difference of AUROC between No.2 synthetic analysis model and logistic regression model was from 0.008 to 0.020. The areas under the ROC curve of two synthetic analysis models were very close to the traditional logistic regression model. No.2 synthetic analysis model was closer to the traditional logistic regression. The area under the ROC curve of three models(containing seven variable) were 0.772,0.793 and 0.815, respectively. There was no difference between the sizes of the area under the ROC curve analysis of three models(P> 0.05). Joint model was used to study the influence of triglyceride measurements at different time points on the risk of hypertension. We assessed the bivariate relations between the following covariates and the occurrence of hypertension: age, gender, family history of hypertension, overweight or obesity, smoking and alcohol consumption. The estimated hazard ratio of the longitudinal measurement of triglyceride was 2.436. Apart from smoke, other covariates had impacts on the incidence of hypertension(P <0.05).8. Based on the same longitudinal cohort data set, logistic regression model, Cox regression model and Joint model were built for hypertension risk assessment. Univariate logistic regression analysis and univariate Cox regression analysis showed the similar results that 14 variables(such as age, gender, family history of hypertension and abnormal triglycerides, etc.) were the the main risk factors for hypertension. With 14 variables as independent variables, further multivariate logistic regression and multivariate Cox regression analysis were made. Multivariate logistic analysis results showed that age(larger than 45 years), sex(male), family history of hypertension, overweight, abnormal triglycerides, smoking and heavy drinking were the main risk factors of hypertension, with the final equation of logit P =-4.743 + 1.229 × age(≥45 years old) + 0.444 × sex + 1.759 × + 0.692 × family history of hypertension are overweight or obese + 0.218 × smoking + 0.459 × heavy drinking + 0.389 × triglyceride exception. Multivariate logistic analysis results showed that age(larger than 45 years), sex(male), family history of hypertension, overweight, abnormal triglycerides, smoking and heavy drinking were the main risk factors of hypertension, with the hypertension prognostic index(PI) of PI = 1.145 × age(≥45 years old) + 0.439 × sex + 1.641 × + 0.633 × family history of hypertension are overweight or obese + 0.198 × smoking + 0.350 × heavy drinking + 0.362 × triglyceride abnormalities. The area under the ROC curve of the logistic regression model and Cox regression models were 0.715 and 0.814, respectively, which showed that Cox regression model had higher predict the performance(P <0.05). Conclusions:1. Different chronic disease has different risk factors spectrum. Smoking, drinking, dyslipidemia, obesity, intake of fruits and vegetables, sleep problems, lack of physical exercise are common risk factors for hypertension,diabetes and other chronic diseases. The same risk factors have different effects on different chronic diseases.2. The conceptual framework of chronic disease risk assessment designed in this study can be used as a reference for other chronic risk assessment questionnaire. Chronic disease risk assessment basis questionnaire and special post-disease risk assessment questionnaires provided a common set for common chronic disease risk assessment. The standardized questionnaires developed data standards, data elements meaning and acceptable data expression form, which can provide reference standard and basis for building health examination database.3. The basic data set and data standards designed for chronic disease risk assessment in this study can ensure the contents of chronic disease risk assessment unified and consistent and can be information exchange standard among different health examination institutions.4. Harvard Cancer Index provide help and foundation for risk factors measurement and risk level assessment of common chronic diseases, especially for cancer.5. Synthesis analysis can provide a methodological basis for chronic disease risk assessment studies based on cross-sectional health examination data. It can also provide a practical opportunity and means for the effective use of cross-sectional data.6. Joint model can be effectively applied to construct chronic disease risk assessment tool, which can reduce measurement bias and improve the quality of disease risk assessment.7. Different statistical modeling methods have their own advantages and disadvantages. In the actual construction of disease risk prediction model, the combination of multiple modeling methods can help build effective disease risk assessment model and improve the quality of chronic disease risk assessment. |
PDF Full Text Request