Table 3.

Multivariable predictors of timing of stroke

Predictors	Model for Early Stroke		Model for Delayed Stroke
	Subdistribution Hazard Ratio (95% Confidence Interval)	P value	Subdistribution Hazard Ratio (95% Confidence Interval)	P value
Age at surgery, per 5-year increase			1.22 (1.08 – 1.37)	0.0009
Female sex			1.74 (1.09 – 2.76)	0.0192
Previous cerebrovascular accident	2.74 (1.25 – 6.04)	0.0122
Intraoperative nitroprusside administration	2.94 (1.51 – 5.71)	0.0015
Intraortic balloon pump insertion			2.90 (1.65 – 5.11)	0.0002
Postoperative daily platelet count
Normal, >150×109/L	1.0		1.0
Mild thrombocytopenia, 100–150×109/L	1.17 (0.55–2.48)	0.6765	1.75 (0.95 – 3.22)	0.0736
Moderate/severe thrombocytopenia, <100×109/L	1.23 (0.51–2.99)	0.6497	2.87 (1.49 – 5.55)	0.0002

^*The final multivariable Cox regression analysis was based on n = 6,067 due to missing information (n = 63) on preoperative medication use. While in the primary stroke outcome model intraoperative nitroprusside use failed the proportional hazards assumption, for the early stroke subevent analysis there was no evidence of a non-proportional hazard (test for time effect in model p=0.46), and thus, intraoperative nitroprusside use was included as predictor of this analysis.

Early strokes (n = 35), delayed strokes (n = 75).