Table 2.

Multinomial Logistic Regression Model for the Association of Participant Characteristics With Medication Usage, Among Those Aged 65 and Older (2009–2010)

	1–4 Medications		≥5 Medications
	OR (95% CI)	p Value	OR (95% CI)	p Value
Age (y)	1.05 (1.01, 1.09)	.017	1.10 (1.05, 1.15)	.001
Sex: female	1.14 (0.76, 1.73)	.499	1.02 (0.57, 1.84)	.942
Race*,†
Non-Hispanic black	1.12 (0.53, 2.36)	.749	1.14 (0.57, 2.29)	.691
Other	1.25 (0.75, 2.06)	.366	2.00 (1.15, 3.48)	.017
Income-poverty ratio	1.12 (0.96, 1.31)	.144	1.21 (1.00, 1.46)	.052
Smoking status*,‡
Former	1.45 (0.87, 2.43)	.144	2.17 (1.23, 3.81)	.011
Current	0.89 (0.43, 1.83)	.730	1.05 (0.43, 2.53)	.911
Body mass index (kg/m2)	1.05 (0.99, 1.12)	.121	1.10 (1.03, 1.18)	.010
Usual health care source(s)	2.74 (0.74, 10.22)	.123	4.87 (0.74, 32.12)	.094
No. of health care visits, past year*,§
4–9	1.42 (1.03, 1.96)	.034	2.50 (1.64, 3.83)	<.001
≥10	6.92 (1.82, 26.38)	.007	17.69 (4.45, 70.26)	<.001
No. of chronic diseases	1.96 (1.40, 2.74)	.001	3.66 (2.56, 5.23)	<.001

Notes: Backward model selection was conducted, forcing in age, sex, and race. Other variables considered for inclusion were marital status, education, income-poverty ratio, smoking status, alcohol use, body mass index, sedentary activity, usual health care source(s), private health insurance coverage, no. of chronic conditions, no. of health care visits in the past year, functional limitation, limitation in activities of daily living, and confusion/memory problems. The threshold for inclusion was p ≤ .10. CI = confidence interval; OR = odds ratio.

*Wald test for significance of race, p = .033; smoking, p = .146; and no. of health care visits, p = .001.

^†Reference = Non-Hispanic white.

^‡Reference = never.

^§Reference = 0–3.