Health Literacy and Health Care Spending and Utilization in a Consumer-Driven Health Plan

Abstract

We examined health literacy and health care spending and utilization by linking responses of three health literacy questions to 2006 claims data of enrollees new to consumer-driven health plans (n = 4,130). Better health literacy on all four health literacy measures (three item responses and their sum) was associated with lower total health care spending, specifically, lower emergency department and inpatient admission spending (p < .05). Similarly, fewer inpatient admissions and emergency department visits were associated with higher adequate health literacy scores and better self-reports of the ability to read and learn about medical conditions (p-value <.05). Members with lower health literacy scores appear to use services more appropriate for advanced health conditions, although office visit rates were similar across the range of health literacy scores.

Health literacy, or the “degree to which individuals have the capacity to obtain, process and understand basic health information and services needed to make appropriate health care decisions” (Ratzan & Parker, 2000) is alarmingly low, with an estimated 36% of adult Americans in 2003 having only basic or below basic health literacy skills (Kutner et al., 2006). Improving health literacy is a national health priority, and there is a growing body of literature linking health literacy to health care costs, utilization, and outcomes (Nielsen-Bohlman, Panzer, & Kindig, 2004). However, most studies have examined health literacy in the context of Medicare, Medicaid, or underserved populations, and to our knowledge, none have included the commercially insured population, which accounts for two-thirds of the U.S. population (DeNavas-Walt, Proctor, & Smith, 2009). In this study, we examine whether health care spending (i.e., outpatient, inpatient, prescription), health care utilization (i.e., emergency department visits, inpatient admissions), and screening services (i.e., cholesterol tests) are correlated with health literacy scores in one commercial consumer-directed health plan.

Background

In 2010, 15% of firms offered a high-deductible health plan (HDHP), though among large firms (those with more than 1000 employees, such as the two employers in the present study), 34% offered HDHPs; enrollment in HDHPs increased to 13% in 2010 from 8% in 2009 (Kaiser Family Foundation and Health Research and Educational Trust, 2010). In a typical consumer-driven health plan (CDHP), patients pay out-of-pocket health care expenses from a tax-exempt savings account combined with a high deductible level; in 2010 the minimum deductibles allowed by the IRS were $1200 for single coverage and $2400 for family coverage (Department of the Treasury, 2006). These plans require patients to be directly involved in their health care decision-making and to manage their health care dollars. Many insurers offer their CDHP members online cost and decision tools and personalized guidance to help them navigate the health care system. Studying health literacy in this population is critical because success of these plans requires that members have a high level of health literacy in order to weigh the benefits and costs of using health care services and to distinguish between necessary and unnecessary care.

Health literacy is increasingly recognized as an important factor in health care costs and quality. Among managed Medicare enrollees, those with inadequate health literacy had significantly higher emergency room costs (a difference of $108) per year and marginally significant higher annual total ($1551) and inpatient ($1543) costs than those with adequate health literacy (Howard, Gazmararian, & Parker, 2005). A small study in a medically needy Medicaid population found a lower reading level (≤3rd grade versus ≥4th grade) to be a significant predictor of higher total and inpatient annual charges (Weiss & Palmer, 2004).

Prior studies have also examined the relationship between health literacy and healthcare utilization and outcomes. In the elderly Medicare population, inadequate health literacy has been associated with poor physical function and mental health (Wolf, Gazmararian, & Baker, 2005), poor self-reported health status (Wolf et al.,2005), increased hospitalization (Baker, Parker, Williams, & Clark, 1998; Baker et al., 2002; Cho, Lee, Arozullah, & Crittenden, 2008; Wolf et al., 2005), mortality (Baker et al., 2007), emergency department use (Cho et al., 2008), and outpatient physician visits (Baker et al., 2004). At a large urban public hospital, patients with inadequate health literacy were over twice as likely to be hospitalized as patients with adequate literacy (Baker et al., 2004). For preventive services, many studies rely on self-report for receipt of a service rather than claims data. In elderly populations, inadequate health literacy was associated with lower self-reported receipt of preventive health services such as mammogram and influenza vaccination (Scott, Gazmararian, Williams, & Baker, 2002; White, Chen, & Atchison, 2008). This association is less consistent in younger age groups, and for adults younger than age 40, low health literacy was associated with higher receipt of influenza vaccination (White, Chen, & Atchison, 2008).

These previous studies examined the relationship of health literacy to health care spending, utilization, and preventive service use primarily in the Medicare and underserved populations, and none included a population of enrollees in a CDHP. Individuals enrolled in a CDHP may have different correlations between health literacy and utilization. Several studies have found that CDHP enrollees are more likely to be White and are less likely to be in poor health status compared with individuals not enrolled in CDHPs. In a study of a natural experiment, wherein employees were newly offered the choice of a consumer-directed plan or one of four PPO options, employees who chose the consumer-directed plan were younger, had higher incomes, were more likely to be White, had no one in their household with a chronic condition, and had lower health spending in the year prior to switching (Barry, Cullen, Galusha, Slade, & Busch, 2008). In a survey of employees at a large company who were also newly offered a CDHP, those who selected a CDHP were less likely to be African American or receiving care for a chronic condition and more likely to have excellent self-reported health status and report no chronic medical problem or recent physician visits. There was no difference in education level attained between those who chose a CDHP and those who did not (Fowles, Kind, Braun, & Bertko, 2004). Similarly, in an experimental study that asked participants to choose a hypothetical plan, numeric health literacy did not influence choice of a CDHP over a PPO plan (Green, Peters, Mertz, & Hibbard, 2008).

Objective

To investigate the relationship between health literacy and health care spending and utilization, responses to a set of health literacy screening questions (Chew, Bradley, & Boyko, 2004) were linked to administrative claims data for a sample of new CDHP enrollees. We hypothesized that these CDHP members with inadequate health literacy scores would have fewer screenings (screening is defined as at least one preventive office visit and a cholesterol test) and more medical and pharmacy use and costs than CDHP members with more adequate health literacy scores.

Design

Using a convenience sample from two large employers, we individually linked members’ health literacy responses to claims data to examine the relationship between inadequate health literacy scores and health care costs and utilization. By linking these data sources, we produced a data set on a 1-year (2006) cross-sectional cohort.

Participants

A web-based survey invitation was extended to newly enrolled CDHP members employed at one of two large companies. Both employers had offered a CDHP product to their employees as an option alongside a traditional plan design for the calendar year 2006. This survey invitation was placed on a secure CDHP website login page during for the first 4 months of the plan year (January 1 through May 2, 2006), the time period during which newly enrolled adult members would most likely register and login.

Two national employers allowed us to extend this invitation exclusively to their new CDHP enrollees. Employees from these two companies encompass a wide array of professions from entry level positions and skilled laborers to more executive job classifications. Both employers’ health plans started on January 1, 2006. One employer was in the banking industry and the other in the paper industry.

Survey respondents of these employers were included in the study if they were continuously enrolled for one calendar year (2006), had both medical and pharmacy coverage without coordination of benefits, were adults younger than 65 years of age, and completed all three questions on the health literacy survey. New enrollees to CDHP who fit these study inclusion criteria and participated in the study numbered 4,130.

Main Measures

Three health literacy screening questions were used to measure health literacy: “How often do you have someone help you read medical materials?” “How often do you need help filling out medical forms?” and, “How often do you have problems learning about your medical conditions because of difficulty understanding written information?” Each question's answer options ranged from 0–4 (Always = 0, Often = 1, Sometimes = 2, Occasionally = 3 and, Never = 4) with lower numbers indicating lower health literacy. These questions are slightly modified wording of a set of validated health literacy screening questions developed by Chew and colleagues. The performance of each of Chew's screening questions have been noted to differentiate between adequate and/or marginal health literacy when compared to gold standards such as the Rapid Estimate of Adult Literacy in Medicine (REALM) and the Short Test of Functional Health Literacy in Adults (S-TOFHLA; Chew et al., 2004; Chew et al., 2008; Sarkar, Schillinger, Lopez, & Sudore, 2011; Wallace et al., 2007; Wallace, Rogers, Roskos, Holiday, & Weiss, 2006).

The question, “How often do you have someone help you read hospital materials?” was most effective in detecting inadequate health literacy in samples of VA preoperative clinic and university-based vascular surgery clinic populations (Chew et al., 2004; Chew et al., 2008), whereas the question, “How confident are you filling out medical forms by yourself” performed best in VA outpatient and university-based primary care clinic populations but was not predictive in a low-income Medicaid population (Chew et al., 2008; Wallace et al., 2006). As none of the questions have been validated in a commercially-insured population, we present results for all three screening questions. Although it is possible to create a threshold in order to distinguish between adequate and inadequate health literacy, we are reluctant to impose such a distinction on the summed health literacy score in the absence of any clear gold standard upon which to rely. No study has shown that combining the three questions is more effective in detecting limited health literacy than one single question alone (Chew et al., 2004; Chew et al., 2008; Wallace et al., 2006; Wallace et al., 2007). We therefore present our results for each of the individual questions as a summed score with a possible range of 0 (lowest health literacy) to 12 (highest health literacy).

Health care claims for services in 2006 were extracted for survey respondents from a large data warehouse of CDHP members. Thomson's Decision Analyst™ software was used to generate aggregate spending and utilization parameters. Health care spending was calculated on a per-member annual basis and included total allowed amount, allowed amounts for prescriptions, medical services, and admissions. Health care utilization measures included admissions and lengths of stays, prescriptions, outpatient cholesterol screens, and emergency department visits. Further analyses of screening measures were not possible as screening criterion limit the n = size and therefore the statistical power. Member characteristics included age, gender, employer, subscriber (y/n), and residence (four U.S. Census Bureau regions: Northeast, Midwest, South, and West).

Methods

We hypothesized that a member's inability to understand health information would precipitate more health care utilization and spending and fewer health screenings. More specifically, we theorized that members with inadequate health literacy measures would have more medical and pharmacy use, higher medical and pharmacy spending, and fewer cholesterol screenings than CDHP members with higher, adequate health literacy scores. CDHP members are offered information and tools to manage their health care utilization and spending, which requires a degree of health literacy. In order to test our theories, we chose new CDHP enrollees who had no prior experience with CDHP information and tools. These new CDHP enrollees presented a robust sample from which to test our theories in a commercial health care system.

Multivariate Poisson log-linked regressions, using Version 9.1 of SAS/STAT® software, were applied to determine the effect of health literacy adjusted for member characteristics on health care spending. In each regression, a spending parameter was the dependent variable and one of four health literacy indicators (health literacy score and three individual items) was an independent variable. All regression models included a health literacy measure as the main independent variable with age, gender, employer (one of two), four U.S. Census Bureau regions of residence and 13 comorbidity variables (anxiety, osteoarthritis, bipolar disorder, asthma, congestive heart failure, chronic obstructive pulmonary disease, coronary artery disease, diabetes, hypertension, human immunodeficiency virus, low back pain, depression, and otitis media) as covariates. A total of 28 unique regressions (four health literacy variables by seven spending parameters: total, medical, pharmacy, medical and pharmacy, admissions, admissions to facilities, and emergency department spending) were performed. Covariates that could theoretically skew associations in this nonrepresentative sample were included. Age, gender, and employers could readily confound results. Members were similarly scattered among 50 states, thereby necessitating the use of a broad classification of U.S. regions. Only disorders with precise and established diagnostic & treatment guidelines were included.

Similarly, multivariate Poisson log-linked regressions were applied to each of six utilization parameters as dependent variables with one of four health literacy indicators as an independent variable. A total of 24 utilization regressions were generated (six independent utilization variables (admissions, days admitted, prescriptions, days supply of prescription drugs, outpatient office visits, emergency department visits) with each of four dependent health literacy variables). Covariates included in these regression models were the same as for spending variables.

Together with the same covariates as the Poisson regressions above, logistic regressions were applied to two independent variables: cholesterol screening (y/n) and one or more office visits (yes/no). Item responses were collapsed into inadequate (always, often, or sometimes) and adequate (never or occasionally).

Results

For the 4,130 respondents who met study criteria, the distributions of responses to the three-item questionnaire are presented in Table 1. No respondents answered Always, indicating a lack of constant difficulty with medical language, to any of the three health literacy questions. Respondents overall reported the most trouble with reading medical materials. The least difficulty was observed in filling out medical forms.

Table 1. Responses to 3-item inadequate health literacy questionnaire (N = 4,130)

Responses to 3-Item Health Literacy Questionnaire	%
How often do you have someone help you read medical materials?
0: Always	0.0
1: Often	2.9
2: Sometimes	10.4
3: Occasionally	32.7
4: Never	54.0
How often do you need help filling out medical forms?
0: Always	0.0
1: Often	0.9
2: Sometimes	7.1
3: Occasionally	27.8
4: Never	64.3
How often do you have problems learning about your medical conditions because of difficulty understanding written information?
0: Always	0.0
1: Often	1.1
2: Sometimes	7.8
3: Occasionally	28.6
4: Never	62.5
Health Literacy Summed Score
Mean	10.5
Standard Deviation	1.7
N	4,130

Table 2 displays demographic characteristics of our sample. Half the sample is female; the age distribution reveals a wide range of both younger and older respondents; most are located in the western region of the country; the risk score suggests that as a group they are somewhat less healthy than an average, representative population (a score of 100).

Table 2. Respondent characteristics

Member characteristics	%
Female	49.9
Age Groups
18 to 25	6.3
26 to 35	19.3
36 to 45	30.3
46 to 55	29.8
56 to 65	14.4
Employer A*	35.5
Subscribers (vs. Members)	87.8
Region
Northeast	9.4
Midwest	29
South	19.5
West	42.1
Risk score
Mean	119.2
Standard deviation	204.6
Chronic conditions
Anxiety	1.5
Osteoarthritis	4.7
Bipolar	0.3
Asthma	2.3
CHF	0.2
COPD	0.6
CAD	1.8
Diabetes	5.5
Hypertension	12.9
Low back pain	9.3
Depression	3.2
Human immunodeficiency virus	0.1
Otitis media	1.7
N	4,130

Table 3 displays summary regression coefficients for a variety of different specifications with a variety of different outcome variables using generalized linear models. The rows display different annual spending measures. The columns reflect the different specifications of health literacy, the first being the sum of the three-item questionnaire responses followed by each question entered separately using indicator variables for the response level.

Table 3. Regression coefficients quantifying associations between health literacy and health care spending

Spending parameter	Summed score		Item 1*		Item 2*		Item 3*
overall mean (SD) per member in 2006 (N = 4,130)	Estimate	SE	Estimate^†	SE	Estimate^†	SE	Estimate^†	SE
Total $3,735.22 (8,269.72)	−0.05 p < .0001	0.01	−0.22 p < .0001	0.05	−0.18 p < .0001	0.07	−0.27 p < .0001	0.06
Medical $3,054.95 (11,535.20)	−0.03	0.02	−0.18 p = .0058	0.07	−0.16 p = .0419	0.08	−0.29 p < .0001	0.07
Pharmacy $883.98 (1,947.42)	−0.01	0.01	−0.13 p = .0182	0.06	−0.04	0.07	0.02	0.07
Medical and Pharmacy $3,938.93 (12,301.31)	−0.02	0.01	−0.17 p = .0032	0.06	−0.12	0.07	−0.23 p = .0006	0.07
Admissions $701.61 (6,087.35)	−0.06 p = .0035	0.02	−0.34 p = .0009	0.10	−0.25	0.13	−0.52 p < .0001	0.11
Admissions Facility $550.69 (5,321.93)	−0.06 p = .0033	0.02	−0.42 p < .0001	0.10	−0.32 p = .0115	0.13	−0.50 p < .0001	0.11
Emergency Departments $35.70 (212.77)	−0.15 p < .0001	0.02	−0.51 p < .0001	0.09	−0.34 p = .0029	0.11	−0.96 p < .0001	0.09

For total spending the regression suggests that higher levels of health literacy are associated with lower levels of total health care spending. The coefficient is quite large and significant (−0.05 (SE 0.01); p-value <.0001), suggesting that moving from occasionally needing help reading medical materials, filling out medical forms, and learning about medical conditions to never needing such assistance (a 3-point or roughly 2 standard deviation change in the health literacy score) would be associated with roughly 15% lower total health care spending. Each individual question is consistent with generally lower total spending associated with less frequent need of health literacy-related assistance (coefficients for Items 1–3: −0.22 p-value <.0001, −0.18 p-value <.01 and −0.27; all p-value <.0001).

Pharmacy spending appears to be less sensitive to the health literacy score (coefficient: −0.01 (SE: 0.01); p > .05), while inpatient spending is lower when the health literacy score is higher (coefficient: −0.06 (SE: 0.02); p < .01). Emergency department spending is strikingly lower as the health literacy score increases: the same hypothetical 3-point increase in the health literacy score is associated with nearly 60% lower emergency department spending (coefficient: −0.15 (SE: 0.02); p-value <.0001).

Table 4 displays annual utilization measures (inpatient admissions and prescriptions filled); the bottom rows indicate whether the person had at least one office visit and cholesterol screen during the year. Consistent with the finding for inpatient spending, the number of inpatient admission decreases with higher health literacy scores (coefficient: −0.08 (SE: 0.02); p-value = .001). Also consistent with the finding for pharmacy spending, the number of prescriptions is not statistically significantly related to the health literacy score (though, as with spending, the sign of the coefficient is negative: −0.01; p-value >.05). Cholesterol screening and at least one outpatient visit did not exhibit a statistically significant relationship with the health literacy score (coefficients: both 0.03; p-values >0.05).

Table 4. Regression coefficients quantifying associations between health literacy and health care use

Utilization parameter	Summed score		Item 1*		Item 2*		Item 3*
overall mean (SD) or % of members in 2006 (N = 4,130)	Estimate	SE	Estimate^†	SE	Estimate^†	SE	Estimate^†	SE
Admissions 0.049 (0.26)	−0.08 p = .001	0.02	−0.41 p < .0001	0.10	−0.32 p = .0091	0.12	−0.55 p < .0001	0.10
Days Admitted 0.15 (1.11)	−0.07 p = .0006	0.02	−0.3877 p = .0002	0.10	−0.3332 p = .0086	0.13	−0.61 p < .0001	0.11
Prescriptions 9.81 (13.43)	−0.01	0.01	−0.12 p = .0067	0.04	0.01	0.06	0.02	0.05
Days’ supply 434.46 (659.70)	<−0.01	0.01	−0.11 p = .0253	0.05	0.03	0.06	0.0991	0.06
Visits 13.12 (14.11)	−0.01	0.01	−0.06	0.04	<−0.01	0.05	−0.02	0.04
Emergency Department Visits 0.11 (0.43)	−0.08 p < .0001	0.02	−0.29 p = .0008	0.09	−0.16	0.11	−0.71 p < .0001	0.09
At least one visit (y/n) 90.9%	0.03	0.05	0.08	0.23	0.42	0.28	0.04	0.30
Cholesterol Screens 42.4%	0.03	0.02	0.02	0.10	0.20	0.13	0.20	0.12

Discussion

Recent interest in health literacy was spearheaded in 2004 by the Institute of Medicine and Healthy People 2010. The Institute of Medicine's report, Health Literacy: A Prescription to End Confusion (Nielsen-Bohlman, Panzer, & Kindig, 2004), noted in 2004 that limited information exists to support an association between limited health literacy and increased consumer, health provider, and health care system costs. In this study we found that in one sample of privately insured individuals enrolled in a CDHP, lower levels of health literacy were correlated with significantly higher levels of health care spending and a significantly greater number of inpatient admissions.

Compared to other studies, in our sample we found relatively few individuals had inadequate health literacy. If we were to follow others and categorize individuals who responded “always,” “often,” or “sometimes” to a question as having inadequate health literacy, we would find the following inadequate health literacy rates: having someone help them read medical materials, 13%; needing help filling out medical forms, 8%; and having problems learning about their medical condition, 9%. In contrast, one study of a sample of Medicare managed-care enrollees found that 24% of respondents had inadequate health literacy, although health literacy was measured by the S-TOFHLA (Howard et al., 2005). Given the younger ages of participants in our sample, these differences are not surprising.

Given reports of more homogeneous CDHP enrollees in terms of higher education, younger ages, and fewer racial and ethnic minorities, one could readily expect weaker correlations with health literacy levels. In spite of our study's lower proportion of participants with inadequate health literacy, the association between health literacy scores and health care spending and utilization remained strong.

In considering individual types of spending, consistent with other studies we find the strongest correlation between health literacy and emergency department spending. One study found that among Medicare enrollees, annual emergency department spending by those with inadequate health literacy was $108 higher than by those with adequate health literacy, although the study found no significant difference between total spending for these two health literacy groups. That we find larger differences in spending for emergency department use suggests that individuals with inadequate health literacy may be using an inefficient mix of health care services, perhaps incurring more potentially avoidable emergency department visits.

With comparable rates of at least one office visit (a fully covered health service under this study's CDHP), members with inadequate scores were still more apt to receive hospital and emergency department care. This finding leads us to believe that among CDHP members, health literacy is associated with acute health care use but has little effect on cholesterol screening and an office visit. This may point to lost opportunities to educate members before health episodes occur.

Although our study provides needed information regarding the association between health literacy and health care spending for enrollees in CDHPs, it is not without limitations. The data comes from a convenience sample of respondents from only two employers, thus findings may not be generalizable to all plan enrollees.

Also important, because data was collected at just one point in time, we cannot conclude that health literacy definitively caused increases in health care spending. Individuals with high health care spending may have more opportunities to interpret medical forms and thus are more likely to have had an experience in which they had difficulty interpreting medical information. Moreover, individuals with higher spending are more likely to have more complex health care needs, and thus to have come in contact with more complex medical information, perhaps leading to a greater number of reports of difficulty interpreting or understanding information.

Earlier reports indicated that health literacy is positively associated with socioeconomic status (Weiss, 2005; White, 2008). Factors that determine social economic status, such as occupation, income, and education were not available for our study participants. However, given that all participants in the survey were employed, it is unlikely that survey respondents’ annual incomes were below the poverty level during the study period. Nevertheless, the lack of information on socioeconomic status represents an important limitation of our work and an excellent opportunity for future research.

The U.S. region covariate added to all models is not based upon variations in health-related parameters. The health-plan members included in this study are from two national broad-based companies; the distribution of members across states was similarly spread out across 50 states.

The measures of health literacy we consider are not ideal. One research study compared measures similar to those used in this study to two more comprehensive measures of health literacy. The S-TOFHLA includes both a numeracy component and reading comprehension assessment, the REALM measures recognition of 66 common medical terms. Considering the optimal combination of three similar questions, our study found that these questions correctly identified individuals with inadequate health literacy (summed score >6) from 90.0% to 95.0% of the time. Specificity of individual items in this study ranged from .907 to .957. These measures do have the advantages of online administration and a shorter time burden for respondents, and they are less likely to cause embarrassment to patients.

Finally, the study design was not optimal to examine whether enrollment in a CDHP (compared with enrollment in a more traditional plan type) results in greater correlation between health literacy and spending. A more ideal study design would compare the correlation between health literacy and spending for a sample of CDHP enrollees and a sample of non-CDHP enrollees, in which individuals would be employed by two similar firms and employees would have no plan choice. Unfortunately, this study design was not available to us.

The results of this study should encourage commercial health plans and their customers (employers) to create understandable and actionable communications for their members. It is hoped that plain language interventions, designed to address inadequate health literacy in the commercial insurance market, could result in decreased health care costs and could potentially improve the health of enrollees.

Conclusion

Adequate health literacy is associated with reduced health care spending among members of a commercial health care plan. This association is strongest for inpatient health care and emergency department spending. Hospital admissions and emergency department use correlate similarly with health literacy measures. Cholesterol screening and one office visit, (a totally covered health benefit) did not correlate with any of our health literacy indicators. With comparable numbers of office visits, CDHP members with lower, inadequate health literacy scores still used services more appropriate for advanced health conditions (hospitalization and emergency department use), leading us to believe that health literacy may be associated with acute health care use. Employers and insurers should be alerted to the influence of health literacy on health care costs and utilization for enrollees within a CDHP and should consider health literacy interventions for individuals in these plans.

Acknowledgments

The authors wish to thank Steven Rush, Kari Duma-White, and Mona Shah for their encouragement and diligence in reviewing this paper. Special thanks also go to John Azzolini at Thomson Reuters for his assistance with this project.

No outside funding was used for this research project. Two authors are employed by and own stock in UnitedHealthcare, the for-profit company where the study was conducted. The findings and conclusions in this manuscript are those of the authors, who are responsible for its content, and do not necessarily represent the views of any third parties.

Notes

*One of two employers.

Note. All regressions included four U.S. regions, gender, employer (1 of 2), age, and 13 comorbidity variables (y/n) as covariates. Northeast, female, employer 1 and comorbidity (yes) were references. Age was entered as a continuous variable. The U.S. regions covariate included four regions: Northeast, Midwest, West and South. Regions were classified by state of participant's residence as follows: CT, ME, MA, NH, RI, VT, NJ, NY and PA: Northeast region; IN, IL, MI, OH, WI, IA, KS, MN, MO, NE, ND and SD: Midwest region; DE, DC, FL, GA, MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK and TX: South region and AZ, CO, ID, NM, MT, UT, NV, WY, AK, CA, HI, OR and WA: West region.

*Item 1: How often do you have someone help you read medical materials?

Item 2: How often do you need help filling out medical forms?

Item 3: How often do you have problems learning about your medical conditions because of difficulty understanding written information?

†Adequate versus inadequate health literacy; adequate health literacy includes responses of never or occasionally; reference includes always, often, or sometimes.

Note. All regressions included four U.S. regions, gender, employer (1 of 2), age, and 13 comorbidity variables (y/n) as covariates. Northeast, female, employer 1 and comorbidity (yes) were references. Age was entered as a continuous variable. The U.S. regions covariate included four regions: Northeast, Midwest, West and South. Regions were classified by state of participant's residence as follows: CT, ME, MA, NH, RI, VT, NJ, NY and PA: Northeast region; IN, IL, MI, OH, WI, IA, KS, MN, MO, NE, ND and SD: Midwest region; DE, DC, FL, GA, MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK and TX: South region and AZ, CO, ID, NM, MT, UT, NV, WY, AK, CA, HI, OR and WA: West region.

*Item 1: How often do you have someone help you read medical materials?

Item 2: How often do you need help filling out medical forms?

Item 3: How often do you have problems learning about your medical conditions because of difficulty understanding written information?

†Adequate versus inadequate health literacy; adequate health literacy includes responses of never or occasionally; reference includes always, often, or sometimes.