A Study Based On Multi-level Logistic Model: Trend And Influence Factors Of Self-evaluated Health Of Corps Residents

Objective:TO analyze resident self-evaluated health and trends of Xinjiang Production and Construction Corps in 1998, 2004, 2010, through a multi-level statistical model to explore whether the aggregation of division, mission field(divisional) population level of Corps resident self-evaluated health exists, and analyze the influencing factors, thus provide evidence for improving the Corps resident self-evaluated health. Methods:Data from Health Services Survey in Xinjiang Production and Construction Corps in 1998, 2004, 2010, the three data were taking a multi-stage stratified cluster random sampling method to obtain. Measurement data was described by mean ± SD, and enumeration data was described by percentage, self-evaluated health of different demographic characteristics used chi-square test for univariate analysis, used MLwi N2.20 fitting multi-level model to analyze the influencing factors. Results:1. 60.80% of the corps resident who evaluated their health status as good in 1998, while 43.61% and 73.83% evaluated their health status as good in 2004 and 2010 after standardized with data in 1998 as the standard age-specific population constitute.2.Variance in the mission field(divisional) level of corps resident self-evaluated health has statistically significant in 1998(χ2 = 9.877, P <0.05), 2004(χ2 = 9.288, P = 0.002), and 2010(χ2 = 7.367, P = 0.007).3.Data fitting multi-level models in 1998 shows: two weeks discomfort(OR = 2.522), half of chronic disease(OR = 4.039), annual per capita household income(OR = 0.842), 15 to 39 years old as a reference,40 to 64 years(OR = 1.774), 65 years and older(OR = 1.575), unmarried as a reference, other(divorced, widowed)(OR = 1.273), with the industry as a reference, retirees(OR = 1.223), the illiterate and primary school as a reference, junior high school(OR = 0.814), senior middle school and higher(OR = 0.761), within one year of hospitalization(OR = 1.478), exercise(OR = 2.768), ethnic(OR = 0.676),these factors affect resident self-evaluated health(P <0.05).4. Data fitting multi-level models in 2004 shows: two weeks discomfort(OR = 1.623), half of chronic disease(OR = 2.387), physical and mental condition(OR = 1.891), education(OR = 0.945), ethnic(OR = 0.640),15 to 39 years old as a reference, 40 to 64 years(OR = 1.194), 65 years and older(OR = 1.125), annual per capita household income to the poor level of the reference, medium(OR = 0.864), higher(OR = 0.725), within one year of hospitalization(OR = 1.655), sex(OR =1.130),exercise(OR = 0.611), these factors affect resident self-evaluated health(P <0.05).5. Data fitting multi-level models in 2010 shows: two weeks discomfort(OR = 1.756),half of chronic disease(OR = 2.901), health index(OR = 3.463), the illiterate and primary school as a reference, junior high school(OR = 0.882), senior middle school and higher(OR = 0.776), annual per capita household income(OR = 0.882), within one year of hospitalization(OR = 1.632), sex(OR =1.246), exercise(OR = 0.795), with the industry as a reference, retirees(OR = 1.240), property or unemployed(OR = 1.232), students(OR = 0.446), age(OR =1.636)these factors affect resident self-evaluated health(P <0.05).6.The comparison between multi-level logistic regression analysis and traditional logistic regression analysis shows: Multi-level logistic regression analysis meaningful variables had statistical significance in traditional logistic regression analysis, significant variables in traditional logistic regression analysis, some variables did not enter the multi-level logistic regression analysis,-2 times the log-likelihood values in multi-level logistic regression analysis decrease. Conclusion:1.Self-evaluated health of Corps residents has significant difference in 1998, 2004, 2010. Self-evaluated health of residents in 2004, 2010 was below the national average. Self-evaluated health status in 1998 closes to the average level of the country.2.Self-evaluated health of Corps residents in 1998, 2004 and 2010 appears high levels of aggregation and is suitable for analysis by multi-level statistical models.3.Age, education, annual per capita household income, exercise, half chronic disease, two weeks discomfort, within one year of hospitalization, physical and mental condition, the EQ-5D index score affect Corps resident self-evaluated health.4. There was an agreement of self-evaluated health and other health indicators, self-evaluated health can be used as a valuable health measuring tool reflects the health of the residents.