基于GPLM的40～65岁绝经后骨质疏松症风险判别模型分析  

Establishment of a Discriminant Model for 40～65 Years Old Postmenopausal Osteoporosis Women Based on Generalized Partial Linear Model

作　　者：谢雁鸣[1] 蔡博婧[2] 田峰[3] 易丹辉[2] 虞鲲 康澍[5] 李建鹏[1] 崔庆荣[1] 

机构地区：[1]中国中医科学院中医临床基础医学研究所,100700 [2]中国人民大学统计学院,100872 [3]中国中医科学院博士后流动站,100700 [4]上海市大华医院,200237 [5]北京中医药大学东直门医院,100700

出　　处：《中国卫生统计》2012年第6期837-839,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金面上项目"基于非齐性Markov model建立病证结合的绝经后骨质疏松证早期风险评估模型"(No:30873339)

摘　　要：目的建立基于广义偏线性模型(generalized partial linear model,GPLM)的,包括危险因素和中医证候要素内容的绝经后骨质疏松症(postmenopausal osteoporosis,PMOP)风险判别模型。方法在获取1740例社区PMOP高危人群危险因素及证候问卷调查数据基础上,筛选出与PMOP发病相关的重要危险因素和中医症状为协变量,以骨密度定性诊断为结局变量,建立基于GPLM的PMOP判别模型。结果 GPLM模型线性部分参数估计提示:是否绝经、体重指数、下肢抽筋、下肢骨痛、绝经年限(线性效应)具有统计意义(P<0.05);模型非线性部分参数估计提示:绝经年限(非线性效应)具有统计意义(P<0.05)。与logistic回归模型相比,拟合GPLM模型时加入了"绝经年限"的非线性效应,其AUC值为0.7971,具有统计学意义(χ2=21.9162,P<0.001)。结论绝经年限与PMOP发病之间存在非线性效应。将西医危险因素和中医症状相结合,建立基于GPLM的PMOP判别模型,反映病证结合特点,与logistic回归模型相比,具有更好的判别准确性。Objective To establish a discriminant model for PMOP with risk factors and TCM symdrome essential factors based on generalized partial linear model（GPLM）.Methods Questionnaire survey data was collected on the risk factors and TCM syndromes of 1740 community population with high risk of PMOP.PMOP risk factors and TCM symptoms were selected out as the covariates,the BMD qualitative diagnosis as outcome variable,and a PMOP discriminant model was established applying the GPLM method.Results The linear parameter estimation of the GPLM model indicated that whether menopause,body mass index,lower limb cramps,lower limb pains and duration of menopause（linear effect） had statistical significance（P0.05）,and the nonlinear parameter estimation of the model indicated that duration of menopause（nonlinear effect） had statistical difference（P0.05）.The AUC value of GPLM was 0.7971 with the nonlinear effect of duration of menopause.The value was 21.9162,P0.001.Conclusion The nonlinear effect between duration of menopause and PMOP was found.GPLM had more accurate discriminative characteristics,being compared with linear logistic regression models.

关 键 词：绝经后骨质疏松症 危险因素 中医证候 广义偏线性模型 判别模型 

分 类 号：R580[医药卫生—内分泌]