TABLE 2.

Multivariate Model: BSI Rate Ratio^a,b

Variable	All BSIs		Gram-negative BSIs
	Rate ratio (95% CI)	P valuec	Rate ratio (95% CI)	P valuec
IV treprostinil (vs IV epoprostenol)	3.08 (2.05-4.62)	<.001	6.86 (3.60-13.07)	<.001
Age at enrollment (y)
≤18 (vs >18 to <41)	0.46 (0.20-1.04)	.06	0.25 (0.06-1.07)	.06
≥41 to <65 (vs >18 to <41)	0.62 (0.41-0.93)	.02	0.53 (0.28-1.00)	.05
≥65 (vs >18 to <41)	0.28 (0.12-0.66)	.004	0.37 (0.14-1.02)	.06
CTD-APAH (vs all other PAH subgroups)	0.52 (0.28-0.96)	.04	0.54 (0.26-1.14)	.10
Year of infection
2008 (vs 2007)	1.17 (0.76-1.80)	.49	0.66 (0.33-1.32)	.24
2009 (vs 2007)	0.78 (0.49-1.25)	.30	0.57 (0.29-1.14)	.12
2010 (vs 2007)	0.56 (0.30-1.06)	.08	0.29 (0.11-0.76)	.01
Year of infection by medication interaction	NA	.87d	NA	.86d

^aBSI = bloodstream infection; CI = confidence interval; CTD-APAH = connective tissue disease–associated PAH; GEE = generalized estimating equations; IV = intravenous; NA = not available; PAH = pulmonary arterial hypertension.

^bAll numbers were reported from the main effects–only multivariate Poisson regression GEE model. The GEE model was used because patients may contribute multiple risk periods to the analysis, which may not be independent.¹⁵

^cP values were obtained using the Wald χ² statistic assessing model beta coefficients.

^dP values were obtained from a supplementary GEE model that was fit including an interaction term. The interaction was not included in the final model because there was no evidence that year was associated with the size of the medication effect.