Table 2.

Factors Contributing to Sex Difference in Long-Term Mortality Between Women and Men After Stroke Based on the Best-Fit Sex-Specific Model Within Studies

Study1 y5 y
N*Actual Confounding Factors Included in the Fully Adjusted ModelN*Actual Confounding Factors Included in the Fully Adjusted Model
Oxford1290Age3, NIHSS (2-term), prestroke mRS, and stroke type732Age3, NIHSS (2-term), and prestroke mRS
Joinville979Age3, NIHSS, and stroke type
Melbourne975Age, log NIHSS, prestroke Barthel (2-term), AF, and stroke type975Age, log NIHSS, prestroke Barthel (2-term), AF, and stroke type
Arcadia547Age, GCS, and AF
Perth183Age, NIHSS, prestroke mRS, and stroke type
Orebro377Age, log NIHSS, institutional residence, and stroke type377Age3, log NIHSS, institutional residence, and stroke type
Dijon3994Age3, LOC, AF, and smoking3094Age3, LOC, AF, smoking, and stroke type
Martinique569Age3, Barthel at onset (2-term), and history of PVD569Age3 and Barthel at onset (2-term)
Porto650Age, LOC, smoking, and prestroke mRS650Age, LOC, and prestroke mRS3
Auckland1177Age3, log GCS, prestroke dependence, AF, and stroke type1177Age3, log GCS, prestroke dependence, and AF
L’Aquila3794Age, LOC§, AF, hospital admission, stroke type, and smoking3794Age, LOC§, AF, hospital admission, and stroke type
Matão79Age§ and NIHSS
Tartu358Age and log NIHSS
Total cases14 97211 368
  • Of note, covariates were transformed, as necessary to get the best model fit. AF indicates atrial fibrillation; GCS, Glasgow Coma Scale; LOC, loss of consciousness; mRS, modified Rankin scale; NIHSS, National Institutes of Health Stroke Scale; and PVD, peripheral vascular disease.

  • * The sample size was the same among the unadjusted model and fully adjusted model.

  • Transformations were based on the powers (eg, 3rd power and 2 power terms) suggested by fractional polynomials that produced the best-fitting multivariable model.

  • Self-reported data about whether the patient lived independently before stroke.

  • § Not meeting criteria of being a confounder but remain in the fully-adjusted multivariable model.