1、从预测模型到临床工具 流行病与卫生统计学系同济医学院公共卫生学院X对Y 有没有独立作用？独立作用的大小到底是多少？X与Y 之间是什么样的关系？X对Y有没有作用？什么因素影响X与Y 之间的关系？影响X对Y的作用？单个X与Y的关系2临床问题 哪些因素导致了人群中某些人患病或不患病？ 哪些因素决定了同一种手术方式病人的预后？ 哪些因素可以用来预测潜在患者的患病风险？ 哪些因素可以用来预测潜在患者的生存时间？ 预测模型问题3多组X与Y的关系NOX1X2XnY1X11X12X1nY12X21X22X2nY2MXm1Xm2XmnYmY为定量变量-linear regressionY 为二分变量Binary
2、 logistic regression (非个体匹配资料)Y 为生存结局COX regression (生存时间变量+结尾数据指示变量) 时间序列分析模型数据格式与模型选择数据格式与模型选择4预测模型以logistics回归为例 Y发生的概率=f(xn) =+1X1+2X2+nXn 预测模型的问题就转化为： 如何构建自变量X的集合来预测Y发生的风险5Logistics回归模型 Logit(p)=ln(p/(1-p)= +1X1+2X2+nXne=ORRR易侕分析软件中如何实现logistic回归分析实例6几个要点-1 X是分类变量时，将X转化为哑变量，在进行分析 比如血型：A B O AB
3、比如年龄：血型血型TbloodaTbloodbTbloodcTblooddA1000B0100O0010AB0001年龄分组年龄分组Age1Age2Age340=600017几个要点-2调整的OR和非调整的OR-所谓调整就是多元logistic回归分析中，分析X1与Y作用时，考虑了其他的X的影响-所谓非调整，就是单变量的logistic回归分析中，只有一个X，即X1交互作用(通常我们只考虑2个变量交互作用)-即研究的变量X1，与其他变量X2的关系-分析方法：交叉乘积项，哑变量化8几个要点-3自变量的共线性问题-所谓调整就是多元logistic回归分析中，分析X1与Y作用时，考虑了其他的X的影响
4、-所谓非调整，就是单变量的logistic回归分析中，只有一个X，即X1交互作用(通常我们只考虑2个变量交互作用)-即研究的变量X1，与其他变量X2的关系-分析方法：交叉乘积项，哑变量化9Logistic回归预测模型中初始变量的筛选 专业上的考虑： 参考文献 测量上的考虑： 共线性，相关系数0.6 经验上的考虑： 单变量分析，P3 seconds), oxygen saturation 94%, and C reactive protein.MAIN OUTCOME MEASURES:Pneumonia, other serious bacterial infections (SBIs, in
5、cluding septicaemia/meningitis, urinary tract infections, and others), and no SBIs.14经典文献赏析RESULTS:Oxygen saturation 94% and presence of tachypnoea were important predictors of pneumonia. A raised C reactive protein level predicted the presence of both pneumonia and other SBIs, whereas chest wall re
6、tractions and oxygen saturation 94% were useful to rule out the presence of other SBIs. Discriminative ability (C statistic) to predict pneumonia was 0.81 (95% confidence interval 0.73 to 0.88); for other SBIs this was even better: 0.86 (0.79 to 0.92). Risk thresholds of 10% or more were useful to i
7、dentify children with serious bacterial infections; risk thresholds less than 2.5% were useful to rule out the presence of serious bacterial infections. External validation showed good discrimination for the prediction of pneumonia (0.81, 0.69 to 0.93); discriminative ability for the prediction of o
8、ther SBIs was lower (0.69, 0.53 to 0.86).CONCLUSION:A validated prediction model, including clinical signs, symptoms, and C reactive protein level, was useful for estimating the likelihood of pneumonia and other SBIs in children with fever, such as septicaemia/meningitis and urinary tract infections
9、.15161718192021从预测模型到诊断工具-Score sheet 将预测模型的系数，转化为得分，代入方程，计算预测值 用预测值与Y的取值，进行诊断试验分析，寻找最佳cut-off point. 预测模型-Score sheet 诊断工具！22231. The performance of our scoring systems is adequate in predicting 5-year diabetic risk in Chinese old-age people. 2. The discriminative power is better than other existing risk scores in the randomly selected validation cohort.Conclusions24Questions?25
