Table 3.

Comparing analytical methods to adjust for confounding by indication in a simulation study with six-month survival as binary outcome

Approach	Scenario 1* OR (95% CI)	Scenario 2* OR (95% CI)	Scenario 3* OR (95% CI)	Scenario 4* OR (95% CI)
Unadjusted model	1.02 (1.00–1.04)	0.69 (0.68–0.71)	0.96 (0.93–0.98)	0.72 (0.70–0.74)
Covariate adjustment	1.67 (1.63–1.71)	0.99 (0.97–1.02)	1.52 (1.47–1.56)	1.03 (1.00–1.06)
Propensity score matching	1.46 (1.43–1.50)	0.90 (0.88–0.92)	1.46 (1.41–1.50)	0.94 (0.91–0.97)
Hospital-level approach**	NA	NA	1.05 (1.04–1.07)	1.04 (1.02–1.05)

Notes:

^*Scenario 1 = observed confounders, no hospital variation; scenario 2 = observed and unobserved confounders, no hospital variation; scenario 3 = observed confounders, hospital variation (17%–58%); scenario 4 = observed and unobserved confounders, hospital variation (17%–58%).

^**Per 10% change.

Abbreviation: NA, not applicable.