. Author manuscript; available in PMC: 2013 Sep 1.

Published in final edited form as: J Ment Health Policy Econ. 2012 Sep;15(3):105–118.

Table 2.

Characteristics of Beneficiaries with Depression and Heart Failure Who Entered the Coverage Gap in 2007

A. Before Propensity Score Weighting
Variables	No Coverage (n=2948)		Generic-only (n=1224)		LIS (n=9809)

	N	%	N	%	N	%	P-value
Female sex	2123	72	879	72	7696	78	<.001
Race
Non-Hispanic White	2809	95	1185	97	7401	75	<.001
African American	61	2	9	1	1037	11	<.001
Hispanic	51	2	21	2	1059	11	<.001
Asian	13	0	3	0	189	2	<.001
Age
65–74	696	24	363	30	3148	32	<.001
75–84	1186	40	497	41	3799	39	0.2
≥ 85	1066	36	364	30	2862	29	<.001

	Mean	SD	Mean	SD	Mean	SD

Zip Code Completion of College	0.26	0.16	0.26	0.16	0.21	0.13	<.001
Zip Code log of median household income	10.69	0.35	10.68	0.34	10.5	0.34	<.001
Prescription drug risk score	1.23	0.34	1.27	0.35	1.28	0.36	<.001

B. After Propensity Score Weighting^†
Variables	No Coverage (n=2948)		Generic-only (n=1224)		LIS (n=9809)

	N	%	N	%	N	%	P-value
Female sex	2123	76	879	77	7696	77	0.15
Race
Non-Hispanic White	2809	82	1185	82	7401	82	0.88
African American	61	8	9	7	1037	8	0.09
Hispanic	51	7	21	8	1059	8	0.06
Asian	13	2	3	2	189	1	0.25
Age
65–74	696	30	363	30	3148	30	0.96
75–84	1186	39	497	39	3799	39	0.84
≥ 85	1066	30	364	30	2862	31	0.68

	Mean	SD	Mean	SD	Mean	SD

Zip Code Completion of College	0.22	0.30	0.22	0.48	0.22	0.17	0.61
Zip Code log of median household income	10.59	0.75	10.57	1.14	10.58	0.42	0.08
Prescription drug risk score	1.26	0.77	1.27	1.22	1.27	0.43	0.07

Abbreviations: LIS = low-income-subsidies

^†

All numbers in panel B are adjusted using the inverse propensity score weights. Propensity scores were calculated using logistic regression models that predict the probability of being in a study group, controlling for beneficiary-level age, sex, race, and prescription drug hierarchical condition, and Zip Code level education and log of median household income.

To assess differences in baseline socio-demographic characteristics among beneficiaries in three different coverage groups, descriptive statistical analyses were performed using two-sided Chi-square test for categorical variables and Fisher’s two-sided F test for continuous variables.

We excluded those with end stage renal disease (n = 389) and outliers (n = 269) defined as data point more than 4 interquartile ranges (IQRs) below the 1^st quartile or above the 3^rd quartile.

Bold denotes statistically significant at α = 0.05.