Timing and Risk Factors Associated With Postoperative Stroke in Vascular Surgery Patients Using Time-Varying Coefficients From a Cox Model

BACKGROUND: Mechanisms of postoperative stroke are poorly understood, particularly for strokes occurring after uneventful emergence from anesthesia. We sought to create a model to study retrospectively the timing and associations for stroke in a high-risk surgery population. METHODS: Using a large prospective database containing detailed information on the occurrence and timing of stroke, we identified patients undergoing procedures involving the distal vascular (DV) and the cerebral vascular (CV) to assess the association between perioperative factors and stroke. We used separate Cox regressions with time-varying coefficients, one for each cohort, to assess the association between baseline factors and the timing of postoperative stroke within the DV and CV cohorts. Using time-varying coefficients allows hazard ratios to vary over time rather than assuming that the hazard ratio remains constant with time. Propensity score matching was used to compare the timing of stroke between DV and CV groups. RESULTS: Among the 80,185 patients with qualifying procedures, there were 711 strokes (0.9%) in the first 30 days after surgery. Stroke incidence was lower for DV patients (0.5%, 306/57,553; P < .001) than CV (1.8%, 405/21,940) and the majority of strokes in the DV group were delayed, occurring between postoperative (POD) days 2 and 30 (236/306, 77%). Among the 711 patients who had a stroke, the proportion of strokes that occurred on day 0 was 8% (n = 24 of 306 strokes) in the DV group compared to 35% in the CV group (n = 140 of 405 strokes). Factors associated with stroke on POD 1 for both groups were preoperative mechanical ventilation and emergent procedures. Acute renal failure and female sex were highly associated with delayed stroke (POD 2-30). Perioperative blood transfusion was associated with an increased hazard of delayed stroke in the DV group and a hazard ratio that increased with time in the CV group. CONCLUSIONS: After adjusting for confounding, stroke was more common and occurred earlier in the CV group. Factors associated with delayed postoperative stroke include acute renal failure, emergent procedures, female sex, preoperative mechanical ventilation, and perioperative transfusion.