Table 2.

Effect of Patient Characteristics^a on Warfarin Use by Multilevel Model

Characteristic	OR (95% CI)
Age in 2008, years
66–69	Reference
70–74	1.05 (0.95–1.15)
75–79	1.05 (0.96–1.15)
80–84	0.95 (0.87–1.04)
≥85	0.64 (0.59–0.70)
Race
non-Hispanic white	Reference
black	0.86 (0.76–0.97)
Hispanic	0.96 (0.84–1.11)
other/unknown	0.81 (0.69–0.95)
Sex
male	1.04 (0.99–1.09)
female	Reference
Census division
New England	1.49 (1.26–1.76)
Middle Atlantic	1.39 (1.22–1.59)
East North Central	1.33 (1.17–1.50)
West North Central	1.62 (1.40–1.86)
South Atlantic	1.13 (1.01–1.28)
East South Central	1.09 (0.93–1.26)
West South Central	Reference
Mountain	1.27 (1.07–1.50)
Pacific	1.08 (0.94–1.23)
Cardiologist visit in 2007
no	Reference
yes	1.23 (1.17–1.29)
CHADS2 scoreb in 2007
0	Reference
1	0.92 (0.86–0.99)
≥2	1.09 (1.01–1.17)
Medicaid eligibility in 2008
no	Reference
yes	0.89 (0.84–0.95)
Elixhauser comorbidity scorec
0	Reference
1–2	0.88 (0.84–0.93)
3–4	0.67 (0.63–0.72)
≥5	0.56 (0.51–0.63)

CHADS₂ = congestive heart failure, hypertension, age, diabetes, and stroke or transient ischemic attack; PCP = primary care physician.

^aPatients with contraindications were excluded.

^bThe age criteria from the CHADS₂ score were removed so that the effect of age alone could be studied.

^cExcluding conditions associated with the CHADS₂ score (congestive heart failure, paralysis, diabetes, and hypertension) to prevent collinearity.