Latent Class Analysis Of Risk Factors Distribution Characters In Birth Defects

Objective The distribution characteristics of birth defect generally results in heterogeneous and high-dimensional data sets with complicated structure relationship and different attribute of variable. The traditional chi-square test and logistic regression analysis for addressing this type of data can’t be used to develop overall effect of many crisscross complex factors in birth defects. To overcome these limitations, this study proposes the Latent Class Model, which can analyze the distribution characteristics of high risk factors of birth defects. This topic expect to provide decision support for birth defects prevention and treatment.Method The population and family planning commission of Shanxi Province investigated that the incidence of birth defects and risk factors of birth defects in six counties which include Pingding County, Dai County, Fenyang City, Huairen County, Zhongyang County, Jiaokou County in Shanxi Province.36716mothers who had Birth to a child from2006to2008had been investigated, including493mothers of children born with birth defects and493mothers of normal children. The children of the two groups were asked to be born in same year, and their parents had similar living conditions.Latent Class Model is used to explore the high risk distribution characteristic of two groups. First of all, exploratory latent class analysis is used to get the optimal potential categories of two group. Then confirmatory latent class analysis is used to test whether the result of exploratory latent class analysis is right. Finally the result of latent class analysis is compared and discussed.Results The results of exploratory latent class analysis:The optimal potential categories of the defect group is divided into six potential categories, which can be named the type of the mothers who passed their genetic conditions to their children(25people),the type of the mothers who lack preventive measures (102people),the type of the mothers who have more risk factors (12people),the type of the mothers who are related to other risk factors(100people),the type of the mothers who have a cold fever or has a history of medication during pregnancy (36people) and the type of the mothers who have less risk factors (218people).The optimal potential categories of the control group is divided into three potential categories, which can be named the type of the mothers who lack preventive measures (132people), the type of the mothers who have less risk factors (328people) and the type of the mothers who has a history of medication during pregnancy (33people).All the conclusions of confirmatory latent class analysis support the results of exploratory latent class analysis.Conclusion Compared with the control group, the optimal potential categories of the defect group is divided into six potential categories according to the latent class analysis while the control is three. The more potential categories which the defect group have include the mothers who pass their genetic conditions to their children, the type of the mothers who have more risk factors, the type of the mothers who are related to other risk factors. When study the risk factor distribution characteristics of the birth defects, we can see that latent class model have an advantage over other methods in dealing with a categorical variable data, which is less number of latent class can be used to explain many complicated connection between explicit classification variables.