三种建立肺功能正常参考值方法的比较

发布时间：2018-04-11 01:05

本文选题：正常参考值 + 分位数回归　；参考：《中国实用内科杂志》2014年S1期

【摘要】：目的比较目前常用三种建立肺功能正常参考值方法,为建立更准确的肺功能正常参考值提供方法学的参考,以利于在临床应用,为疾病的早期诊断、治疗效果及预后进行评估提供参考依据。方法以全球肺部倡议组织(Global Lungs Initiative)提供的肺功能数据为例,分别采用多重线性回归(参数回归模型)、分位数回归(非参数回归模型)和LMS方法 (半参数回归模型)三种常用方法来建立正常参考值和正常值下限,并对结果进行比较和评价。所有建模过程采用统计软件R完成。结果该数据包含了5723名正常男性1秒用力呼气容积(FEV1)、年龄和身高。三种方法建立的正常参考值存在有统计学差异(P0.05),基于传统多重回归模型计算得到的预计值与实测值的差异有统计学意义[(3.35±1.09)对(3.38±1.19),t=4.753,P0.001)],而基于LMS法和分位数回归计算得到的预计值更接近实测值[(3.38±1.10)对(3.38±1.19),t=0.033,P=0.974;(3.37±1.12)对(3.38±1.19),t=0.584,P=0.865),20岁之前和≥60岁两个年龄段,基于多重回归模型计算出来预计值低于实测值,差异有统计学意义。基于LMS法和分位数回归计算得到的预计值差异无统计学意义。结论分位数回归和LMS方法均适用于不服从正态分布的偏峰数据,拟合效果优于传统的多重线性回归,更适合用于建立全年龄段肺功能正常参考值和正常值下限,计算得到的结果更符合实际。
[Abstract]:Objective to compare the three kinds of commonly used to establish the normal reference values of pulmonary function, pulmonary function is more accurate for the establishment of normal reference value and provide methodology reference for clinical application, for early diagnosis of disease, therapeutic effect and prognosis assessment to provide a reference to the global initiative. Lung (Global Lungs Initiative) of lung the function of data as an example, using multiple linear regression (regression model), quantile regression (nonparametric regression model) and LMS method (semiparametric regression model) three methods to establish the normal reference value and the lower limit of the normal value, and evaluate the results. All the modeling process using statistical software R. Results the data contained 5723 normal men forced expiratory volume (FEV1), age and height. The establishment of normal reference values with the three methods are statistically significant (P0.05), Is calculated by traditional multiple regression model and the measured value of the value of the difference was statistically significant [(3.35 + 1.09) based on the (3.38 + 1.19), t=4.753, P0.001)], and based on the LMS method and the quantile regression to calculate the expected value is more close to the measured [(3.38 + 1.10) to (3.38 + 1.19), t=0.033, P=0.974; (3.37 + 1.12) to (3.38 + 1.19), t=0.584, P=0.865), before the age of 20 and above 60 years of age two, a multiple regression model is calculated based on the measured value is lower than the expected value, the difference was statistically significant. The LMS method and the estimated quantile regression to calculate the value of the difference no statistical significance. Conclusion based on Quantile Regression and LMS method were applied to partial peak data does not obey the normal distribution, the fitting effect is better than the traditional multiple linear regression, more suitable for all ages to establish normal lung function reference value and the normal value of lower limit, the calculated result is more realistic.

【作者单位】：广州医科大学附属第一医院广州呼吸疾病研究所呼吸疾病国家重点实验室呼吸疾病国家临床医学研究中心;
【分类号】：R56

【相似文献】