Research On Chinese Health Risks Model And Risk Appraisal

Health risk appraisal (HRA) is the basis tool, prerequisite key technologies for health management. Recently, the health risk appraisal has gradually developed into an interdisciplinary science of the epidemiology, health statistics, behavioral medicine, clinical medicine, psychology, etc., and is a hot domain of health management study. For the Chinese health risk appraisal problems in health management, in this thesis, the adult Chinese health risk appraisal questionnaire was designed; based on the questionnaire, three health risks models were established, and evaluated the validity and reliability; the statistical methods of association rule mining and item response theory were applied to health risk appraisal research. This thesis made the following tasks:1. The concept of health and self-rated, health dimensions of health measurement, evolution of individual subjective health measurement, and subjective health measurement theory and quantitative methods were briefly introduced. The health risk appraisal research status at home and abroad, scoring methods of the health risk appraisal model, validity and reliability evaluation methods were systemic reviewed. The application studies of association rule mining and item response theory in health risk appraisal research were extensive reviewed.2. According to definition of health risk appraisal and its'scope, the health risk appraisal questionnaire item pool was collected, based on 10 internationally health measurement scales,"health risk appraisal"questionnaire of Health Management Research Center of the University of Michigan, self-rated health measurement scale, etc., and standardized by"Common data elements of health record"(trail version),"Base framework and data standard of health record"(trail version) of the People's Republic, etc.. The six uniform questionnaire item selection criteria were set in the research. The Chinese health risk appraisal questionnaire V1.0 (CHRAQ V1.0) was formed after pre-survey and item amendments. The questionnaire constituted by the 64 items: personal general information (gender, age, marital status, etc.), physical measurements (height, weight, chest circumference, waistline, hip circumference, etc.), lifestyle (smoking, alcohol consumption, exercise, etc.), eating habits, personal and family health history, stress, social support, health awareness, self-reported health, and the work status of employees. According to Chinese health risk appraisal questionnaire V1.0, the basic dataset of self-rated health was established, which contains 13 categories, 98 data elements, including 33 common data elements of health record.3. Three health risk appraisal models were constructed, which were the overall health risks model, the fatty liver health risks model and the breast hyperplasia health risks model, based on the Chinese health risk appraisal questionnaire V1.0, referred to the related health risks in the literature. We evaluated the validity and reliability of these models. The results of validity analysis found: the scores of the three health risks models followed normal distribution (Kolmogorov-Smirnov Z=1.073, P=0.199; Z=1.126, P=0.158; Z=0.853, P=0.460); there was a linear relationship (r=0.774, P<0.001) between overall health risk model score and the positive rate of all the objective examination and the Spearman rank correlation coefficients between overall health risk model score and urinalysis, liver function examination, cholesterol examination, blood sugar examination, electrocardiogram, abdominal ultrasonography, chest X ray examination were significant (P<0.001); the Spearman rank correlation coefficient between the fatty liver health risks model score and fatty liver result of abdominal ultrasonography was statistically significant (r=0.497, P<0.001), and the risk score of the fatty liver group was significantly higher than that of the normal group (t= 8.310, P<0.001); the Spearman rank correlation coefficient between the breast hyperplasia health risks model score and the breast hyperplasia result of the infrared breast examination was statistically significant (r=0.524, P<0.001), and the risk score of the breast hyperplasia group was significantly higher than that of the normal group (t=5.297, P<0.001). The results of reliability analysis found: the Cronbach's alpha coefficients of the three health risks models were 0.652, 0.679, 0.791; the split-half reliability Spearman-Brown coefficients were 0.784, 0.819, 0.868; the test-retest reliability coefficients were 0.841, 0.883, 0.824 (n=88,P<0.001). Validity and reliability analysis show that: the convergent validity between the health risk score of these models and the objective examination results was acceptable; the distinguish ability of the health risk score between the different objective examination results was acceptable, too; the internal consistency, test-retest reliability and split-half reliability were high. 4. The association rule mining method was applied to the field of health risk appraisal in our research. The Apriori algorithm, which is the classic algorithm of the association rule mining, was adopted to search the association rules in the training data set (n=686) between the 28 health risks items in the Chinese health risk appraisal questionnaire V1.0 and 14 objective examinations, by SPSS Clementine 12.0 software. The association rules of 9 objective examinations were found, which were abdominal ultrasonography, liver function examination, cholesterol examination, urinalysis, fatty liver examination by abdominal ultrasonography, infrared breast examination, electrocardiogram, blood examination, chest X ray examination, and the typical association rules were explained. The association rules, the lift of which was more than 1, were used to predict the results of the 9 objective examinations in test data set (n=174), and the consistency between the predicted results and the observed results was evaluated by consistent rate, sensitivity, specificity, Kappa and Pearsonχ2 test. The results showed that: the maximum consistent rate between the predicted and actual examinations results was 0.936 (infrared breast examination), and the minimum was 0.673 (liver function examination); the maximum sensitivity was 1 (infrared breast examination), and the minimum was 0.571 (chest X ray examination); the maximum specificity was 0.894 (chest X ray examination) and the minimum was 0.400 (infrared breast examination); Kappa values were greater than 0.40 (P<0.001). The false negative rate of the association rules predicted results was low, and the false positive rate was high, which indicate that the false negative predicted results base on the health risk items by association rule mining method was less. So the Chinese health risk appraisal questionnaire V1.0 can be used as a screening tool to recommend the objective examinations. Our research confirmed that recommending objective examinations based on the health risk appraisal questionnaire was practicable and effective.5. The overall health was a latent variable and measured by 9 objective examinations. The 2-parameter dichotomous item response theory model was used to analyze the relationship between the overall health and 12 health risk factors, which was that IRT was adopted to establish an overall assessment of relationship between these objective examinations and health risk factors. The parameters of 2-parameter dichotomous IRT model were estimated by SAS/STAT PROC NLMIXED. The number of estimated parameters of 2-parameter dichotomous IRT model was less than that of Logistic regression model with random intercept, and the parameters of Logistic regression model with random intercept and the OR can be calculated by the IRT model parameters. The overall health risk factor levels were male, age (above 50 years old), widowed/divorced, retired/unemployed, BMI (≥28.0, obesity), waistline (≥85cm male,≥80cm female) , waist to hip ratio (≥0.90 male,≥0.75 female), systolic blood pressure (≥120mmHg), diastolic blood pressure (≥90mmHg), ever smoked/smoking, meat or fish based eating habits, number of symptoms in two weeks (≥2). Our research extended the application of the IRT in the field of health measurement and health risk appraisal. The structural model of IRT was focused, which characterized the relationship between the overall health (latent variable) and the health risk factors. The IRT model was used to assess the relationship between multiple health risk factors and a number of related health outcomes.