Evaluation Of Respondent Consistency’s Influence On Structural Equation Modeling Results Of Health Belief Model

ObjectiveThis thesis aims to test Health Belief Model (HBM) model fit with StructuralEquation Modeling (SEM), with the example of analysis of predictors of medicationadherence in patients with hypertension using HBM, to explore the influence ofindividual characteristics’ consistency of respondents on the Structural EquationModeling results and thus to provide references on subject selection in SEM basedstudies.MethodsOne-to-one anonymous survey was carried out in conveniently selectedCommunity Health Service (CHS) centers in Wuhan among patients withhypertension who visited the CHS centers. Database was built with EpiData3.0, anddata were entered twice. Descriptive analysis were conducted using SAS9.2;Frequency, constituent ratio were used to describe respondents’ socio-demographiccharacteristics and the answers of each item in HBM. SEM using Amos17.0wereapplied to quantitatively explore the approach and magnitude of the influence ofHBM dimensions on medication adherence behavior. Stratified analysis method wasfurther applied: Respondents were stratified by age and gender, and sub-samples of male and female as well as sub-samples of “≤70” and “＞70” were obtained. TheSEM models with sub-samples were compared with full sample SEM models toexplore the influence of age-or gender-inconsistency on the SEM results, and toexplore how the health beliefs of different age and gender of patients withhypertension influence their medication adherence behavior and the magnitude of theinfluence.Results499questionnaires were distributed and461were effective (effective rate=92.38%). Mean age of respondents were70.28(SD=8.57), and67.67%of themranged from65to80years old;58.50%of the respondents were female;37.36%ofthem had an education level of senior middle school or above, and14.13%of themwere uneducated;38.07%of the respondents had diabetes, high cholesterol, coronaryheart disease, stroke, or other chronic diseases; The average life span of takingantihypertensive drugs were11.68±10.10years.Results of full sample SEM indicated that, of all the dimensions of HBM,self-efficacy exerted the most important and the greatest impact on medicationadherence behavior, and other latent variables exert their impact on behavior throughdirect and indirect impact on self-efficacy. Self-efficacy had an direct impact of0.73on behavior; Both perceived barriers and perceived benefits exerted their impact onbehavior through self-efficacy, and the magnitude were0.190and0.241, respectively;Perceived susceptibility impacted behavior by means of perceived barriers andself-efficacy, and the magnitude was0.079; Perceived severity exerted its impact onbehavior by means of perceived benefits, and the magnitude was0.079.53%of themedication adherence behavior was explained by HBM.With the stratification by age,“≤70” sub-sample SEM model indicated littledifference compared with the overall SEM model, and self-efficacy had an directimpact of0.82on behavior;“＞70” sub-sample SEM model indicated great differencecompared with the overall SEM model: Self-efficacy impacted self-efficacy, and self-efficacy sequentially impacted behavior (magnitude=0.72). With thestratification by gender, the female sub-sample SEM model indicated little differencecompared with the overall SEM model, and self-efficacy’s impact on behavior was0.52; The male sub-sample SEM didn’t fit well possibly due to not enough samplesize.ConclusionThis study confirmed that it’s feasible to explore the medication adherencebehavior using HBM. Respondents’ age consistency impacts the SEM model fit, andthe impact of respondents’ gender consistency remains to be verified. Resultssuggested that attention should be paid to respondents’ individual characteristics inSEM based studies.InnovationWhether consistency of respondents’ individual characteristics would influenceSEM model fit was first proposed. This study tested Health Belief Model (HBM)model fit with Structural Equation Modeling (SEM) with the example of analysis ofpredictors of medication adherence in patients with hypertension, applied SEM afterstratification, and explored whether consistency of respondents’ individualcharacteristics would influence SEM model fit by comparing the sub-sample SEMmodels with the overall SEM model.