LASSO及其拓展方法在回归分析变量筛选中的应用

多重共线性是影响回归分析结果的一个重要问题, 近年来发展的LASSO方法对于筛选解释性较高的变量、处理高维数据和解决多重共线性问题具有强大的优势。该方法是在模型估计中增加了惩罚项, 能将一些不必要变量的回归系数压缩为零进而从模型中剔除, 达到变量筛选的目的。本文将重点介绍LASSO这一方法, 并与最优子集、岭回归、自适应LASSO与弹性网络的结果进行比较, 结果显示LASSO与自适应LASSO在解决自变量多重共线性问题以及增强模型解释性、预测精度方面均有较好的表现。     

通过极长链脂肪酸检测从无症状的肾上腺脑白质营养不良患者中筛查肾上腺功能减退者

目的:血浆极长链脂肪酸检测使得在大范围的危险人群中筛查无症状的X-连锁肾上腺脑白质营养不良(X-ALD)患者成为可能。本研究评估了49例无症状患者(年龄4.5±3.5岁)因未被诊断出的肾上腺功能减退而引起的负担。研究从测定血清促肾上腺皮质激素(ACTH)和进行标准剂量ACTH刺激实验开始直到采用肾上腺替代疗法(随访时间2±1.7年)。结果:研究之初有39例(80%)患者的肾上腺功能减退,34例(69%)患者血清ACTH水平升高,ACTH刺激实验在21例(43%)患者中表现为不正常。血清ACTH和年龄在实验之初具有一定的相关性(r=0.32,P=0.05)。随访结束时,86%…     

断点回归设计在流行病学研究中的应用

断点回归设计是一种严格的类实验方法,基于样本在临界值周围分布的随机性,使用两侧数据分别进行回归,从而避免潜在混杂因素的干扰,为因果关联的推断提供强有力的证据。本研究对断点回归方法以及其研究进展进行了介绍,并结合案例对其应用方法进行了详细阐述,以期为断点回归在流行病学研究中的使用提供参考。     

医学英汉双语角在七年制组织学与胚胎学PBL教学中的应用

我国临床医学七年制的教育目标是培养高层次临床医学人才,要求学生掌握宽厚的医学知识和技能,具备一定的科研能力、终身学习能力和国际交流能力.通过教学实践我们认为以问题为基础的教学(problem based learning,PBL教学)在强化基础理论,锻炼临床思维等方面确有优势,然而单纯的PBL教学在提高专业外语能力方面的优势并不突出. 医学英汉双语角(medical English-Chinese corner,MECC)[1]是以提高七年制学生和来华留学生医学外语水平并有效促进其临床能力为目的,以医学知识水平近似的中外医学生为会话双方,以英汉两种语言为交流媒介,以医学知识为会话内容的,规定话题的,阶梯性语言训练模式.实践证明该模式在提高医学外语交流能力方面效果突出[2].因此,我们大胆将MECC与PBL结合起来在七年制组织学与胚胎学教学中进行实践,取得了较好的效果与反响.     

中国成年人高血压患病区域聚集性及危险因素的多水平模型分析

Objective To analyze the region cluster and risk factors of hypertension in the Chinese adult population and to explore the application of multilevel regression model in the risk factors of hypertension. Methods Multi-stage random sampling technique was used to choose 15 540 individuals aged 35-74 years from 10 regions in China. Two-level logistic regression models were fitted under MLwiN 2.02 software. Results The region cluster of hypertension existed and variance portion coefficient was 3.1%. After adjusting for the age and gender, overall obese people (BMI≥28 kg/m2) were 4.50(95%CI: 4.00-5.06) times, overweight people (BMI=24-27.9 kg/m2) were 2.26 (95%CI: 2.07-2.46) times more likely to be hypertensive as compared with those of normal BMI (18.5-23.9 kg/m2), and those centrally obesive people (Waist circumference≥85 cm in male or 80 cm in female) were 2.62 (95%CI: 2.42-2.83) times more likely to be hypertensive as compared with those of normal WC. The age-and gender-adjusted odds ratios (Ors) of triglyceride (TG), serum total cholesterol (TC), glucose, low-density lipoprotein cholesterol (LDL-C) , high-density lipoprotein cholesterol (HDL-C) and drinking alcohol were 2.10 (95% CI: 1.89-2.33) , 2.08 (95% CI: 1.84-2.35) , 1.85 (95% CI: 1.60-2.14) , 1.58 (95% CI: 1.38-1.81), 1.49(95%CI: 1.32-1.69) and 1.15(95%CI: 1.05-1.27), respectively. Conclusion The prevalence of hypertension was not only affected by individual risk factors, such as obesity, drinking alcohol, abnormal glucose and serum lipids profile, but also affected by the geographic environment where people resided in. Population-and risk factors targeted strategies, proved a promising way to reduce individual risk of hypertension in the primary prevention of hypertension.     

中国成年人高血压患病区域聚集性及危险因素的多水平模型分析

Objective To analyze the region cluster and risk factors of hypertension in the Chinese adult population and to explore the application of multilevel regression model in the risk factors of hypertension. Methods Multi-stage random sampling technique was used to choose 15 540 individuals aged 35-74 years from 10 regions in China. Two-level logistic regression models were fitted under MLwiN 2.02 software. Results The region cluster of hypertension existed and variance portion coefficient was 3.1%. After adjusting for the age and gender, overall obese people (BMI≥28 kg/m2) were 4.50(95%CI: 4.00-5.06) times, overweight people (BMI=24-27.9 kg/m2) were 2.26 (95%CI: 2.07-2.46) times more likely to be hypertensive as compared with those of normal BMI (18.5-23.9 kg/m2), and those centrally obesive people (Waist circumference≥85 cm in male or 80 cm in female) were 2.62 (95%CI: 2.42-2.83) times more likely to be hypertensive as compared with those of normal WC. The age-and gender-adjusted odds ratios (Ors) of triglyceride (TG), serum total cholesterol (TC), glucose, low-density lipoprotein cholesterol (LDL-C) , high-density lipoprotein cholesterol (HDL-C) and drinking alcohol were 2.10 (95% CI: 1.89-2.33) , 2.08 (95% CI: 1.84-2.35) , 1.85 (95% CI: 1.60-2.14) , 1.58 (95% CI: 1.38-1.81), 1.49(95%CI: 1.32-1.69) and 1.15(95%CI: 1.05-1.27), respectively. Conclusion The prevalence of hypertension was not only affected by individual risk factors, such as obesity, drinking alcohol, abnormal glucose and serum lipids profile, but also affected by the geographic environment where people resided in. Population-and risk factors targeted strategies, proved a promising way to reduce individual risk of hypertension in the primary prevention of hypertension.     

河南省不同肥胖类型农村居民多代谢异常分析

目的:探讨河南省农村居民不同肥胖类型人群代谢异常的差别。方法:采用随机整群抽样的方法抽取河南省某农村18岁以上常住居民1158人进行横断面调查。根据体质指数(BMI)和腰围(WC)将其分为正常组、全身性肥胖组、中心性肥胖组和复合型肥胖组。对年龄、血压、血糖、总胆固醇、甘油三酯、高密度脂蛋白及低密度脂蛋白进行分析,对多代谢异常进行聚集性分析。结果:男性的高血压、高总胆固醇、低高密度脂蛋白的患病率均低于女性(P<0.05),男性的血脂异常率(58.6%)低于女性(65.9%)(χ2=6.642,P=0.010)。男性人群中,只有1种代谢异常时中心性肥胖组高血脂患病率最高(χ2=10.669,P=0.014),复合型肥胖组合并高血压、高血脂的患病率和合并高血糖、高血脂的患病率均较高(χ2=12.601,P=0.006;χ2=25.922,P<0.001),合并3种代谢异常时全身性肥胖组的患病率较高(χ2=27.431,P<0.001)。女性人群中,复合型肥胖组合并高血压、高血脂的患病率较高(χ2=35.637,P<0.001),合并3种代谢异常时复合型肥胖组的患病率较高(χ2=34.477,P<0.001)。多重线性回归分析显示WC标准偏回归系数多大于BMI。结论:河南省农村居民肥胖人群多代谢异常患病率较高,多代谢异常个体聚集现象严重;人群多代谢异常与中心性肥胖关系更密切。     

患和不患支气管肺发育不良的超低出生体重的学龄期儿童的心血管异常

背景:随着围生期和新生儿科新的治疗方法的引入,支气管肺发育不良(BPD)的临床征象似乎也发生了变化。本研究的内容就是“新BPD”的结局。目的:评价患和不患BPD的学龄期超低体重儿(VLBW,出生体重<1500g)使用表面活性剂期间的心血管方面的异常改变。方法:34例BPD的V LBW患儿,年龄7～8岁,对其进行血压(BP)、心电图(ECG)、二维多普勒和M型超声心动描记图、肺通气量和全身容积描记图等检查。按年龄和性别匹配的原则,设置了有34例不患BPD的VLBW儿童的非BPD组和包括34例足月正常儿的足月产组。结果:非BPD组的舒张压x(s)显著高于BP…     

医学统计学教学改革的若干问题探索

医学统计学是一门重要的基础课程,也是医学生较难掌握的一门课程.随着医学科学的不断发展,医学统计学的思维与理念已经渗透到医学科学研究的每一个角落.     

中部某综合性大学医学生睡眠状况动态调查及影响因素研究

目的 动态评估医学生睡眠质量及其影响因素变化,为改善医学生睡眠质量提供依据。方法 采用分层整群抽样,利用自编《医学生睡眠影响因素调查表》和匹兹堡睡眠质量指数（Pittsburgh sleep quality index, PSQI）量表对中国中部某综合性大学医学院各年级学生于学期初、中、末进行调查,采用线性回归分析探究影响医学生PSQI评分的因素,并按性别、年级进行分层分析,采用广义估计方程评价学期不同时段影响睡眠的因素。结果 最终1 631人次纳入研究。其中男生932人次,女生699人次。学期初、中、末PSQI评分为（4.79±2.24）、（4.65±2.21）和（4.90±2.34）,平均为（4.78±2.26）,总体睡眠障碍（PSQI≥7）率为19.0%,三次调查显示男生睡眠状况优于女生（P值均＜0.05）。广义回归方程分析显示医学生睡眠质量主要与午睡情况、压力、睡前电子产品使用及认知、情绪及寝室睡眠环境有关。结论 医学生睡眠障碍发生率高,影响医学生睡眠质量的因素主要为午睡情况、压力、睡前电子产品使用及认知、情绪及寝室睡眠环境。