Blood Management and Risk Assessment for Transfusion in Pediatric Spinal Deformity Surgery

Table 6

Logistic regression for variables which affect the occurrence of a transfusion.

		B	SE	Wald	df	Sig.	Exp (B)
Step 1^a	ZHgp	−7.23	0.259	7.776	1	0.005	0.485
	ZCobbM	1.038	0.308	11.326	1	0.001	2.824
	DIAG_BI	1.480	0.467	10.037	1	0.002	4.392
	Anest	−0.999	0.485	4.237	1	0.040	0.368
	AcTrax	−2.332	0.507	21.112	1	0.000	0.097
	Constant	1.870	0.437	18.307	1	0.000	6.487

^aVariable(s) entered on step 1: ZHgp, ZCobbM, DIAG_BI, Anest, and AcTrax. Calculations were performed with the standardized value for continuous variables. Categorical variables were defined as dummies with 1 for “yes” and 0 for “no.” The advantages of using standardized variables are related to the better understanding of relative ODDS in the presence of continuous and categorical variables as predictors. To use this regression as a predictive value, the following should be considered: the “Cobb” variable has a normal distribution with mean = 74.716 and SD = 22.621; the variable “HgP” is also normally distributed with mean = 13.192 and SD = 1.4329. To use the provided coefficients, the users should follow the formula: logit = 1.870–0.723 [(HgP-13.192)/1.4329] + 1.038[(Cobb-74.716)/22.621] − 0.99(remifentanil) − 2.332(tranexamic acid) + 1.480 (neuromuscular diagnosis) and entering the HgP value and Cobb angle of their patient. To transform the logit into a more familiar concept of probability (P), the formula to be used is  = EXP (logit)/[EXP (logit) + 1].