急性髓细胞白血病早期病死的预测诊断

目的建立一个数学模型，对急性髓细胞白血病（AML）早期病死（ED）作出预测诊断。方法对193例AML各因素进行Logistic多元回归分析及判别分析，建立判别函数。结果Logistic多元回归分析表明，ED的危险因素包括体温、白细胞计数、血小板计数、骨髓增生度及白血病细胞类型等；据此建立的判别函数，总的判别符合率为86．5％，ED组及非ED组判别符合率分别为78．6％和88．7％。结论本判别函数判别符合率高，错判率低，具有一定的临床应用价值。

PREDICATIVE DIAGNOSIS OF EARLY DEATH IN PATIENTS WITH ACUTE MYELOGENOUS LEUKEMIA

To establish a mathematical model to predict the early death (ED) in patients with acute myelogenous leukemia (AML). METHODS 193 cases of AML were studied using the multifactors logistic regression analysis and discriminant analysis. RESULTS Logistic regression analysis showed that temperature, leukocyte count.blood platelet count, marrow cellularity. and subtype of AML were the most important factors that influenced ED.Fisher'S linear discriminant function was set up using these 5 factors by discriminant analysis. According to thismathematical model the predicative accuracies of ED, non-ED and the whole were 78. 6%, 88. 7N and 86. 5% respectively. CONCLUSION This discriminant function had good concordance. It was a valuable model in predictingthe ED of AML.