Kolmogorov-Smirnov and Shapiro-Wilks normality tests were applied to ascertain the normal distribution of a variable.
The existence of a significant relationship between the studied variables and post-TKA blood transfusion requirement were evaluated by the following methods: (1) Student
t-test for numerical data comparison and Mann-Whitney test for independent samples; (2) inferential analysis using the Spearman correlation coefficient to determine the degree of association between hemoglobin level range (ΔHb), ischemia time, and transfused volume; (3) χ
2or Fisher exact test for categorical data comparison; and (4) logistic regression analysis to assess the simultaneous influence of predictor variables. The variable-selecting process was stepwise forward at a 5% level.
A receiver operator characteristic (ROC) curve determined the accuracy of the model in predicting blood transfusion requirement. This graphical representation is built with sensitivity/specificity fractions at several cutoff points for each variable, illustrating system performance and its discrimination threshold. In addition, the ROC curve allows the identification of the best cutoff point.
Statistically significant differences were determined by
p-values < 0.05. All calculations were performed by an independent statistician, using SAS version 6.11 statistical software (SAS Institute, Inc., Cary, NC, USA).
The existence of a significant relationship between the studied variables and post-TKA blood transfusion requirement were evaluated by the following methods: (1) Student
t-test for numerical data comparison and Mann-Whitney test for independent samples; (2) inferential analysis using the Spearman correlation coefficient to determine the degree of association between hemoglobin level range (ΔHb), ischemia time, and transfused volume; (3) χ
2or Fisher exact test for categorical data comparison; and (4) logistic regression analysis to assess the simultaneous influence of predictor variables. The variable-selecting process was stepwise forward at a 5% level.
A receiver operator characteristic (ROC) curve determined the accuracy of the model in predicting blood transfusion requirement. This graphical representation is built with sensitivity/specificity fractions at several cutoff points for each variable, illustrating system performance and its discrimination threshold. In addition, the ROC curve allows the identification of the best cutoff point.
Statistically significant differences were determined by
p-values < 0.05. All calculations were performed by an independent statistician, using SAS version 6.11 statistical software (SAS Institute, Inc., Cary, NC, USA).