Table 2.

Results from Logistic Regression Models for Blood Culture Contamination

	Arm 1	Arm 2	Arm 3	Overall trial
	OR (95% CI)	OR (95% CI)	OR (95% CI)	P
As randomized, unadjusted	0.74 (0.58–0.94)	0.94 (0.72–1.23)	0.56 (0.44–0.71)	.02
As randomized, adjusted by no. of draws	0.74 (0.58–0.95)	0.93 (0.71–1.23)	0.56 (0.44–0.71)	.02
As randomized, adjusted by no. of draws, age, sex, race, payer, surgery, and comorbidities	0.73 (0.57–0.94)	0.93 (0.7–1.22)	0.55 (0.43–0.71)	.02
As treated, unadjusted	0.74 (0.58–0.94)	0.93 (0.71–1.22)	0.56 (0.44–0.71)	.02
Randomization to all 3 arms, unadjusted	0.75 (0.58–0.96)	0.9 (0.68–1.19)	0.56 (0.44–0.71)	.03
Accounting for randomization strata, unadjusted	0.73 (0.58–0.94)	0.94 (0.72–1.23)	0.56 (0.44–0.71)	.02
Weekday draws (entire set collected on weekday)	0.77 (0.57–1.05)	0.82 (0.6–1.13)	0.56 (0.41–0.76)	.18
Weekend draws (any in set collected on weekend)	0.66 (0.44–0.99)	1.2 (0.72–1.98)	0.56 (0.39–0.83)	.06
Day draws (entire set collected 5 am–4:49 pm)	0.71 (0.5–1.02)	0.97 (0.68–1.39)	0.64 (0.46–0.9)	.24
Night draws (any in set collected 5 pm–4:59 am)	0.79 (0.56–1.11)	0.9 (0.59–1.37)	0.44 (0.31–0.62)	.01

NOTE. P values in the pairwise analysis were as follows: P = .19 for the comparison of arm 2 with arm 1; P = .11 for the comparison of arm 3 with arm 1; and P = .005 for the comparison of arm 3 with arm 2.