| ObjectivesThrough a cross-sectional epidemiological survey, understand sub-health population feature distribution. Using data mining techniques, combined with clinical knowledge, the multidimensional sub-health features are simplified, inductive, exploring the characteristics between the longitudinal and transverse correlation of importance of relationship relationship. This will be helpful for further understanding of sub health characteristics, refinement of health classification, for developing targeted interventions provided in the direction of train of thought.MethodsThe use of national science and technology support program sub-health status of Chinese medicine identification and classification research task group developed the" physical and mental health survey scale", in2008May to2009May in East China, North China, from Southern China, southwest9cooperative units (Guangdong Province Traditional Chinese Medical Hospital, Chinese Oriental (Beijing) health science and technology limited company, the liberation of Army General Hospital sub-health Research Institute, Affiliated Hospital of Chengdu University of Traditional Chinese Medicine, Shanghai University of Traditional Chinese Medicine affiliated Longhua hospital, China Academy of traditional Chinese medicine hospital, Guanganmen Zhejiang Province Traditional Chinese Medical Hospital, Affiliated Hospital of Nanjing University of Chinese Medicine, Chongqing Institute of traditional Chinese Medicine) medical personnel health status identification and feature epidemiological survey, establishment of sub-health state feature database. First of all, on the sub health state of multidimensional descriptive analysis, form the first stage of the multidimensional sub-health feature pool. By single factor analysis, screening out health and sub-health population difference characteristics, form the second stage of the multidimensional sub-health feature pool. By using cluster analysis, combined with clinical knowledge on the second levels of the multidimensional sub-health feature pool discomfort symptoms, lifestyle features are simplified, induction, to form the third stage of the multidimensional sub-health feature pool. Then, using a decision tree model, respectively to three different levels of the multidimensional sub-health feature pool modeling, formed three series of sub health longitudinal characteristic relation model, combined with clinical knowledge on a series of three decision tree model of the diagnostic characteristics for comparison, select a more optimized decision tree model and multidimensional feature pool. Secondly, by using association rule model, the relatively optimized feature pool were analyzed, the formation of healthy transverse feature relationship model. Finally, combined with clinical knowledge of the above two models are integrated analysis.ResultA collection, met the inclusion criteria, do not accord with the exclusion criteria of a sample of4086patients, including1716cases of healthy people (49.13%), sub-health population in1777cases (50.87%). Based on the research results, the multidimensional sub-health characteristics are summarized into six parts, respectively is unwell symptom characteristics, demographic characteristics, lifestyle features, traditional Chinese Medicine physique characteristics, psychosocial characteristics, survival quality characteristics. Were collected in six dimensions the characteristic228, wherein the unwell symptom characteristics of117, demographic characteristics28, lifestyle characteristics38, traditional Chinese Medicine physique characteristics of9, social psychological characteristics of27, survival quality characteristics of9, form the first stage of the multidimensional sub-health feature pool. By single factor analysis, screening out health and sub-health population characterized by a total of197, in which the symptoms characteristic94, demographic characteristics28, lifestyle characteristics33, traditional Chinese Medicine physique characteristics of9, social psychological characteristics of25, survival quality characteristics of8, to form the second stage of the multidimensional sub-health. Pool. By using cluster analysis, combined with clinical knowledge on the second levels of the multidimensional sub-health feature pool discomfort symptoms, lifestyle features are simplified, induction, a total that features96, wherein the discomfort symptoms characteristic of1322, demographic characteristics, lifestyle characteristics19, traditional Chinese Medicine physique characteristics of9, social psychological characteristics of25quality of life characteristics,8, to form the third stage of the multidimensional sub-health feature pool.Using decision tree models, on three different levels of the multidimensional sub-health feature modeling feature pool pool, the first decision tree model with second identical feature pool decision tree model, the diagnostic accordance rate was72.5%, for sub health state of the sensitivity was75.8%, specificity was69.1%, the positive predictive value was71.8%, the negative predictive a value of73.4%, the area under the ROC curve0.72. Third level feature pool decision tree model the diagnose accordance rate was76.2%, for sub health state of the sensitivity was77.2%, specificity was75.1%, the positive predictive value was76.2%, the negative predictive value was76.1%, the area under the ROC curve0.76. Statistical results suggest, three feature pool decision tree model relative to a reference line have diagnostic significance (P<0.05), three feature pool decision tree model of the area under the curve than the one or two feature pool decision tree model, and there was significant difference (P<0.05), the third decision tree model as a feature pool sub health characteristics of longitudinal importance relation model. Using association rule model, the optimal level third on the multidimensional sub-health feature pool were analyzed, the formation of healthy transverse feature relationship model. A total of142derived association rules, the minimum support degree is50.08%, the highest support to85.20%, the lowest confidence level to90.08%, the highest confidence95.57%, formation of sub health characteristics of horizontal relation model.ConclusionThe results show that the multidimensional characteristics of sub-health population numerous, on the basis of the literature research and epidemiological findings, can be divided into the unwell symptom characteristics, demographic characteristics, lifestyle features, traditional Chinese Medicine physique characteristics, psychosocial characteristics, quality of life, a total of six dimension. In the clinical knowledge, combined with single factor analysis, clustering analysis for the multidimensional sub-health features are simplified, inductive, and using the decision tree model, association rules model on the multidimensional sub-health characteristics of longitudinal, transverse feature relationship to explore, for clinical interventions in subhealth providing evidence and ideas:sub-health intervention direction should take the Constitution as the center to carry out, according to the constitution the biased to intervene, and biased high populations in the intervention health should also pay attention to strengthen the good diet education of spleen and stomach and nursed back to health, in addition to biased tendencies appear interpersonal sensitivity of abnormal psychology, pay attention to the timely investigation and grooming, in intervention requires a combination of population occupation characteristics. Flat and high population can Qi, yin deficiency, a good diet education for, in intervention requires a combination of crowd is the regional, gender characteristics. To the Constitution and not biased, Qi deficiency, kidney bladder syndrome features is relatively easy to clip concurrently, need attention in intervention. |
PDF Full Text Request