Latent Variables Analysis Of Health-promoting Behaviors And Health Risk Factors

Healthy lifestyle behaviors which include health-promoting behaviors andhealth-protecting behaviors can prevent or delay the occurrence of various typesof chronic diseases. Health-promoting behaviors are more proactive action toimprove personal health, and health protective behaviors which are action tomaintain one’s current health through reducing risk factors. Health behaviors arecharacterized by multiple dimensions and multiple variables and cannot bemeasured directly. Therefore, we used the latent variable analysis for dataintegration and information extraction.Objective:1. To extract potential dimensions of health-promoting behaviors.2.To compare the differences among latent class models constructed for healthrisk factors and health status between different gender and age groups.3. Toevaluate the effects of controllable health behaviors on health behaviors.4. Toidentify the major risk factors which contribute to chronic diseases and build astructural equation model for predicting chronic diseases. Materials:1．A total of1034questionnaires of community-based survey datawere used to detect the potential dimensions of health-promoting behaviors.2． In order to compare the differences among latent class models constructedfor health risk factors and health status between different gender and age groups,a total of5677personal health examination data were collected in a healthmanagement center in Beijing from January to December,2010.3. Foranalyzing the effects of the controllable health behaviors on the health status anddevelope the structural equation model for chronic disease measurement, wecollected medical checkup data continuously in a health management center inBeijing from January in2008to December in2010.Methods:1. Ordinal factor analysis is used to detect the potential dimensions ofhealth-promoting behaviors.2. Latent class analysis (LCA) method was used toexplore the latent classes of health risk factors and health status.3. To evaluatethe effects of controllable health behaviors on the health status, multivariatemultilevel analysis method was used.4. Some data mining methods, such asgenetic algorithms, artificial neural network (ANN), multivariate adaptiveregression splines (MARS), etc., were used to identify major risk factorscontributed to chronic diseases. Structural equation model (SEM) method wasalso used to evaluate the performance of health-promoting behaviors andlifestyle-based risk factors on chronic diseases incidence.Results:1. Take the Health-promoting lifestyle profile for Chinese elderly(HPLP-CE) as an example, we detected6potentional dimensions (stressmanagement, self-actualization, health responsibility, vitality, interpersonalsupport and nutrition behavior) of health-promoting behaviors. The split-halfreliability of the scale was0.857, the test-retest reliability was0.68andCronbach’s coefficients were0.912, which were all better than that of HPLP-C.Correlation between HPLP-CE scores and SF-36scale scores was significantly positive. The subjects with chronic diseases scored lower than non-chronicallyelderly for HPLP-CE scale in4demensions of stress management,self-actualization, vitality and interpersonal support.2. According to the WHO report on major health risk factors amongdeveloping countries, we develope latent class models for full sample,age-group and gender-group samples. LCA models of health risk factors of allsamples were divided into three classes and were respectively named healthygroup (23.26%), unhealthy group (34.84%) and lack of physical exercise group(41.89%) according to their characteristics. LCA models with different genderwere developed, and the same pattern of health behavior as in all samples werefound in female group, while high alcohol consumption plus smoking class(15.96%) was more prominent instead of lack of physical exercise class in malegroup. For male, in high alcohol consumption plus smoking class, more than80%subjects have smoking and drinking habits. LCA models among differentage groups indicated that youth group (<30) and old-age group (≥60) weredifferent from that of all samples. We found that2-class LCA models of thesetwo age groups fit the data best, which were respectively named healthy groupand the unhealthy group according to the conditional probability. In addition, wedeveloped LCA models of health status which were measured by body massindex, age, diastolic blood pressure, systolic blood pressure, fasting glucose andtriglycerides, and found that3-class model fit the data best, which were namednormal group (37.15%), high-risk group (16.5%) and critical high-risk group(43.65%).3. We evaluate the effects of controllable health risk behaviors on healthstatus. With the increasing of BMI, triglyceride TC, total cholesterol TG, lowdensity lipoprotein LDL were significantly increased, while high densitylipoprotein HDL significantly reduced. For male, with average BMI increased 1Kg/m2, the TC increased by10.14mg/dl, the TG increased by1.72mg/dl, LDLincreased by1.91mg/dl, and HDL decreased by1.01mg/dl; for female, as theaverage BMI increased by1Kg/m2, TC increased by8.50mg/dl and TG increaseby1mg/dl, LDL increased by1.40mg/dl, and HDL decreased by1.27mg/dl.People who participate in physical exercise always have significantly lower TCTG and LDL. For male, regular smoking and drinking can significantly increaseTC, TG and LDL, and people deficient of vegetables, fruit intake showed higherTC and TG levels.4. Physical examination indicators such as age, fasting blood glucose, bodymass index, systolic blood pressure and total cholesterol were identified forpredicting chronic diseases incidence. Then we construct a SEM to evaluate theperformance of health-promoting behavior and health-related risk factors onchronic disease incidence. The final SEM model fit well with data after beenmodified, and CFI=0.92, TLI=0.90and RMSEA=0.05<0.08. The directeffect of health promotion behaviors that affect chronic diseases prediction was0.26, while that of lifestyle-based health risk factors was0.30.Conclusions:1. The potential dimensions of HPLP-CE were detected, in whichmental and inner self-reflection planes were the most salient factors forassessing health-promoting behaviors of seniors.2. The LCA results showedthat the latent class model differed for different gender-group and age-group,among which lack of physical exercise group is more prominent. Tailoringhealth education messages using latent class analysis may be a promising newapproach to address multiple behavior change more effectively.3. Age, fastingblood glucose, body mass index, systolic blood pressure and total cholesterolwere identified for predicting chronic diseases incidence.4. Health-promotingbehaviors and health risk factors have similar direct effects on the occurrence ofchronic diseases. The contributions of this study are as follows:①Take the HPLP-CE as anexample, we detected6potentional dimensions (stress management,self-actualization, health responsibility, vitality, interpersonal support andnutrition behavior) of health-promoting behaviors.②Latent variables analysistechnology were used to explore the latent classes of health risk factors andhealth status, which provided analysis strategy for the effectively use of healthexamation data.③Genetic algorithms, artificial neural networks and multivariateadaptive regression splines and other data mining methods were used fordetecting risk factors of chronic disease with muti-phase.④Risk factors ofchronic diseases can be regarded as measurement indicators for predicting theincidence of chronic diseases, SEM was used to analyze the effects ofhealth-promoting behaviors and health risk factors on chronic diseases.In summary, latent variables analysis approaches were used to detectpotential dimensions of health-promoting behaviors and health risk factors, andSEM for chronic diseases measurement were builded. Latent variables analysisof health-promoting behaviors and health risk factors can provide scientificbasis for effective interventions for chronic diseases, and to make interventionsmore targeted and accuracy. In this study, we provided analysis strategy for theeffectively use of health examation data. The analysis results may provideevidence for health management and then effectively promote health status ofChinese residents through adoption of evidence-based selective interventions.