【摘要】 目的　提高胸液检测指标的综合诊断水平 ,并以量化的概率形式表示出来。方法　对 10 3 2例胸腔积液中 718例已确诊为结核或恶性胸液结果进行统计学分析 ;采用logistic多元回归模型以拟合累积诊断概率方式 ,给出发生结核性或恶性胸液两者之一的概率 (Ptb)。结果　得出诊断准确率达 98%的结核性与恶性胸液各指标的预测值 ;及诊断准确率仅达 73 %的结核性与恶性胸液各指标的临界值 ;为了提高此临界区鉴别诊断的准确率 ,建立多元logistic回归模型应用预测概率可使临界区鉴别诊断的准确率提高到 94 6%。结论　应用预测概率可使结核性与恶性胸液鉴别诊断的准确率提高到 96.3 %。更多还原

【Abstract】 Objective To improve comprehensive diagnostic level of pleural fluid measurements and to show it in its quantitative diagnostic probability .Methods Statistical analysis has been done on 718 case of finally diagnosed TB or malignant pleural fluid among 1032 case; The alternative probability of TB or malignant pleural fluid(Ptb)occurrence is obtained by adoption of multi variable logistic regression’s matching cumulate probability.Results To obtain 98% veracity predicted value of diagnosed TB or malignant pleural fluid;and To obtain only 73% veracity predicted value of diagnosis TB or malignant pleural fluid; To increase the accuracy of differential diagnosis in critical section and set up a model of multi variable logistic regression,To increase the accuracy of differential diagnosis in critical section to94.6% by applying predicted probability.Conclusion To increase the accuracy of differential diagnosis in TB and malignant pleural fluid to 96.3% by applying predicted probability.更多还原