非参数分位数回归模型在医学参考区间建立中的应用

发布时间：2018-04-16 00:41

本文选题：分位数回归 + 参考区间　；参考：《中国卫生统计》2014年03期

【摘要】：目的探讨非参数分位数回归模型在医学参考区间建立中的应用。方法以血红蛋白参考区间上限(97.5%分位点)确定为实例,采用SAS软件包中的Quantreg过程拟合血红蛋白97.5%分位点的回归模型,探索影响参考区间限值分布的因素;并根据分位数回归的结果,确定各亚组特异的参考区间限值。结果性别和年龄均是影响血红蛋白参考区间上限分布的因素(P0.05);非西班牙裔黑人与其他民族相比,参考区间上限值的分布也存在统计学差异(β=-4.32,P0.001)。最终分非西班牙裔黑人和其他民族、分性别、分年龄段分别确定血红蛋白浓度参考区间上限值。结论非参数分位数回归模型对假设条件要求不高,接近医学参考区间研究的现实情况,且模型具有很强的稳健性。在考虑临床实际意义的情况下,该模型是参考区间建立过程中判定分组因素、确定稳健的参考区间限值的有效方法。
[Abstract]:Objective to explore the application of non-parametric quantile regression model in the establishment of medical reference interval.Methods taking the upper limit of the reference interval of hemoglobin (97.5B) as an example, the regression model of the 97.5 locus of hemoglobin was fitted by the Quantreg process in SAS software package, and the factors influencing the distribution of the limit value of the reference interval were explored.According to the results of quantile regression, the specific reference interval limits of each subgroup were determined.Results both sex and age were the factors influencing the upper limit distribution of hemoglobin reference interval (P0.05A), and the distribution of the upper limit value of reference interval was also significantly different between non-Hispanic blacks and other ethnic groups (尾 -4.32, P 0.001).The upper limit of the reference interval of hemoglobin concentration was determined by sex and age.Conclusion the nonparametric quantile regression model requires little hypothetical conditions, which is close to the reality of medical reference interval research, and the model has strong robustness.Considering the clinical significance, the model is an effective method to determine the grouping factors and the robust reference interval limits in the process of establishing reference interval.
【作者单位】：中国医科大学附属第一医院检验科;
【分类号】：R311

【参考文献】