Table 2. Marginal Estimates of the Adjusted Prevalence of Making No HSA Contributions in the Last Year Among US Adults in HDHPs Who Had an HSA.

Characteristic	Sample size	No. of respondents	Weighted % (95% CI)a	P valueb
Educational level
High school or less	134	88	62.6 (52.4-72.8)	NAc
Some college	198	116	59.1 (51.3-66.9)	.59
Bachelor’s degree	316	147	49.5 (43.2-55.7)	.04
Master’s degree or higher	216	94	46.1 (38.3-53.9)	.02
Race/ethnicity
White	716	360	52.3 (48.2-56.4)	NAc
Black	42	25	66.6 (51.2-81.9)	.10
Hispanic	56	29	44.9 (30.7-59.0)	.33
Other	50	31	66.7 (49.9-83.5)	.13
Source of health insurance
Employer without plan choice	187	108	61.0 (53.2-68.7)	NAc
Employer with choice of plans	629	320	52.2 (47.8-56.6)	.06
Insurance exchanged	21	6	30.9 (6.9-54.9)	.03
Other source	27	11	52.6 (33.5-71.6)	.41
Health status
Excellent	117	64	57.1 (46.7-67.5)	NAc
Very good	390	192	54.4 (48.6-64.0)	.65
Good	286	147	50.7 (44.1-57.4)	.33
Fair	59	31	51.6 (35.8-67.4)	.58
Poor	12	11	75.1 (35.8-115.5)	.46
Chronic condition
Yes	409	218	53.2 (46.8-59.5)	.81
No	455	227	54.2 (49.2-59.3)	NAc
Level of health insurance literacye
Lowest tertile	266	155	58.6 (51.5-65.6)	NAc
Middle tertile	276	146	56.9 (50.5-63.3)	.73
Highest tertile	322	144	47.3 (40.7-54.0)	.03
Level of financial literacyf
Lowest tertile	140	89	58.1 (47.8-68.3)	NAc
Middle tertile	206	119	57.4 (49.6-65.3)	.92
Highest tertile	518	237	51.1 (46.0-56.2)	.26

Abbreviations: HDHPs, high-deductible health plans; HSA, health savings account; NA, not applicable.

^aBased on marginal effects from a logistic regression model in which the dependent variable was $0 in savings in an HSA in the past 12 months, if a respondent reported having an HSA and responded to questions about savings. Savings level of $0 was defined as either reporting not saving any money in the last 12 months for health care or not saving any money for health care through their HSA. Sample was anyone who reported having an HSA and had nonmissing values for covariates in model (n = 864). Survey weights were based on the full sample of respondents with nonmissing covariates (n = 1564). Prevalences are adjusted for age, sex, income, region, and level of consumer engagement.¹⁸ Estimates of coefficients for each of the predictor variables in the model can be found in eTable 5 in the Supplement.

^bCalculated using logistic regression models.

^cIndicates reference category.

^dRespondents answered they had “health insurance that you bought through a state or federal individual marketplace/exchange.”

^eIndicates tertiles of participants’ scores for the Health Insurance Literacy Measure.²⁰

^fIndicates tertiles of participants’ sum of scores for 3 measures of financial literacy developed by Lusardi and Mitchell.¹⁹