Demographic Associations with Limited Literacy

The largest proportion of American adults with limited literacy are native-born Caucasian speakers of English. Over half of the people with NALS Level 1 skills are Caucasians, and about 57 million Caucasian Americans have limited literacy skills (NALS Levels 1 and 2) (Kirsch et al., 1993). However, many groups with higher rates of limited literacy than would be predicted from population estimates alone were identified by the NALS. Groups with lower average proficiency scores include those who are poor, members of ethnic and cultural minorities, those who live in the southern and western regions of the United States, those with less than a high school degree or GED, and those who are above the age of 65. It is important to keep in mind that the NALS was performed only in English, and that clear differentials for literacy proficiencies can be seen within each population group on the basis of nativity, education, and access to economic resources (Kirsch et al., 1993). Table 3-1 displays the percentage of persons overall and various demographic groups that have NALS Level 1 or 2 literacy skills, the lowest of five skill levels. Each of these groups is discussed briefly below in order to highlight those that could potentially benefit from interventions aimed at improving health literacy.³

TABLE 3-1

Percentage of Adult Population Groups with Literacy Skills at NALS Levels 1, 2, or 3–4.

Adults over the age of 65 have more limited literacy proficiency than younger, working adults, according to the NALS data. However, cross-tabulations indicate that scores for elders vary by education level and access to financial resources (Kirsch et al., 1993). Hispanics, African Americans, Pacific Islanders, and Native Americans are also (including Alaska Natives) over-represented in the numbers of adults with lower literacy proficiency scores, as indicated by the NALS data (Kirsch et al., 1993). These populations have increased in number since 1992, when the NALS was performed, and as a group represent a larger proportion of the U.S. population (U.S. Census Bureau, 2002). Hispanic and Asian populations in particular have increased at a greater rate than other U.S. populations. While more recent literacy data is not yet available, these population increases may represent an increase in the number of individuals, and the percentage of American adults, affected by limited health literacy.

Individuals without a high school diploma or GED have lower levels of literacy proficiency than do those with a high school diploma or education beyond high school. Nearly all adults who did not finish eighth grade scored at Level 1 or 2 on the NALS, and 77 percent of these individuals scored at Level 1. Similarly, among individuals who entered high school but did not graduate, 81 percent scored at NALS Levels 1 or 2, while only 55 percent of high school graduates scored at those levels (Kirsch et al., 1993). Since between 400,000 and 500,000 students drop out of high school each year in the United States (Young, 2002), high school dropouts constitute a large population likely living without adequate literacy skills.

Overall, more than 70 percent of immigrants tested on the NALS scored at Levels 1 or 2. The NALS was conducted in English only. Thus the finding that 25 percent of those scoring in the lowest levels of literacy proficiency were immigrants to the United States might be expected, since many of those immigrants might have just begun to learn English. In addition, 91 percent of those who did not complete a high school education in their country of origin scored at the lowest levels of proficiency (see Table 3-1). A high proportion of people entering the United States from non-English-speaking nations come from non-industrialized areas of the world in which there are limited educational opportunities. For example, more than half of adults who emigrate to the United States from Spanish-speaking countries had not finished high school in their country of origin (Greenberg et al., 2001).

Individuals who described themselves as having a physical or mental condition which prevented them from participating fully in normal activities tended to score at the lowest levels of proficiency. More than half the individuals with vision, speech, or learning disabilities performed at NALS Level 1, as did 48 percent of individuals reporting a mental or emotional condition. In contrast, 35 percent of individuals reporting a hearing difficulty performed at NALS Level 1.

Approximately 7 of 10 prisoners surveyed by the NALS demonstrated limited literacy proficiency. Individuals who did not finish school are over-represented among prison inmates, so in many cases, prisoners represent the same individuals as those with limited education. NALS investigators studied nearly 1,150 inmates in 80 federal and state prisons that had been randomly selected to represent penal institutions across the country (Haigler, 1994).

Several other groups, not specifically studied in the NALS, are also known to have limited literacy skills. These groups include the persons who are poor and/or homeless, and military recruits. As discussed in Chapter 2, social factors affect literacy. Poverty is intertwined with many sociodemographic variables (Balsa and McGuire, 2001), which, in turn, are associated with limited literacy. Although the causal relationships are not known, individuals with limited incomes/access to resources are also more likely than those with higher incomes/access to resources to have lower literacy proficiency (Kirsch et al., 1993). Homeless individuals, who may also be affected by limited literacy and poverty, are the audience recipients of a variety of state programs developed to enhance literacy skills (Profiles of State Programs, 1990). Weiss and colleagues used Medicaid enrollment as a proxy for low income and administered the Instrument for the Diagnosis of Reading/Instrumento Para Diagnosticar Lectura (IDL) as a measure of literacy. They found that the majority of low-income individuals involved in their study had limited reading skills, with the average score at the fifth-grade level (Weiss et al., 1994). In addition, reading and mathematics skills of potential military recruits are assessed, and enlistment in the military requires passing these tests. As many recruits cannot adequately perform the high school level reading tasks they face in the military, all branches of the military operate educational programs to increase the literacy skills of their members (Hegerfeld, 1999).

Studies of Limited Health Literacy

In this section, we examine the extent of limited health literacy by examining peer-reviewed studies that provide evidence about the epidemiology, or relative rates, of health literacy skills using currently available measures in different demographic groups. The conclusions that can be drawn about health literacy from this research are limited because in most cases, the studies below identify individuals and groups in which only the print component of health literacy skills is measured (see Chapter 2).

Studies described in this and the following section represent a sample of the English-language peer-reviewed studies that measure literacy or use the REALM or TOFHLA among patients or consumers in a health context. We note that these measures are, for the most part, assessments of print literacy in the health context. While these measures all tap into some aspect of health literacy, the studies do not necessarily use the same standard for identifying people with limited skills. It is even the case that any two tests both claiming to use grade-level scores would not identify the same students as being below a selected standard because their norming samples most likely were not the same. Thus the percentages falling below a particular standard are not directly comparable, but do clearly point to where problems exist.

Studies were identified by searching the Medline (1966–2003), PsycInfo (1974–2003), ERIC⁴ (1963–2003), Sociological Abstracts (1963–2003), and CINAHL⁵ (1982–2003) databases through the OVID web gateway for the indicated years with the following terms as keyword searches: “health literacy,” “literacy and health,” and “reading and health.” To be included, the study had to define the health or health-care population cohort, the literacy measurement tool, and the result of the literacy screening assessment. Additional studies for inclusion were identified through testimony to the committee by experts in the field (see Appendix A), and the available bibliographies (Greenberg, 2001; Rudd et al., 2000b; Zobel et al., 2003).

Table 3-2 summarizes identified studies containing information on the rates of health literacy skills as currently measured among various study populations. Results are given as percentages of the study participants with (1) marginal and inadequate health literacy as measured by the TOFHLA (Parker et al., 1995) or its shorter version (S-TOFHLA; Baker et al., 1999); or (2) inadequate health literacy as indicated by a score below grade 9 on the REALM (Davis et al., 1993); or (3) what the authors reported if a standard measure was not used or if a standard measure was modified. Table 3-2 also identifies the reported demographic characteristics of the study participants that were reported to be associated with limited health literacy.

Studies noted in Table 3-2 document the prevalence of limited health-literacy skills as measured by the REALM or TOFHLA among patients in general medical and pediatric clinics; specialty care clinics including those for asthma, HIV, family planning, obstetrics, and oncology; and community-based sites including retirement homes and social service agencies. The studies show that limited health literacy skills, as measured by current assessment tools, are common, with significant variations in prevalence depending on the population sampled (Williams et al., 1998b). In many cases, however, education was not controlled for, and in some cases differences based on health literacy assessments were no longer significant when education was controlled for.

Two large multisite studies of convenience samples that provide estimates of the prevalence of limited health literacy among users of urban public hospitals showed variation in prevalence in different geographic locations. In one study, Williams and colleagues (1995) administered the TOFHLA in either English or Spanish to patients presenting for acute care at public hospitals in Atlanta and Los Angeles. The study participants included 1,892 English speakers from Atlanta or Los Angeles, and 767 Spanish-speaking individuals from Los Angeles only. Almost half (47.4 percent) of the participants in Atlanta had inadequate (34.7 percent) or marginal (12.7 percent) health literacy scores on the TOFHLA, while in Los Angeles the rates of limited and marginal health literacy scores were 12.5 percent and 9.5 percent for English speakers and 41.9 percent and 19.8 percent for Spanish speakers. Spanish-speaking participants (tested in Spanish) had higher rates of limited health literacy scores than did English-speaking participants, but these differences were not significant after controlling for years of education.

A more recent multisite study used the S-TOFHLA to assess new enrollees in a Medicare managed care organization in four different geographic areas: Cleveland, Houston, Tampa, and an area of southern Florida including Fort Lauderdale and Miami (Gazmararian et al., 1999b). Participants were 304 Spanish-speaking and 2,956 English-speaking enrollees. The prevalence of inadequate and marginal health literacy scores on the S-TOFHLA among English speakers was 23.5 percent and 10.4 percent, respectively, while among Spanish speakers prevalence rates were 34.2 percent and 19.7 percent. Results varied by city, with the participants in Cleveland showing the highest rate of inadequate health literacy scores among English speakers at 34.1 percent, and those in Tampa showing the lowest rate at 16.6 percent. The geographic variation between Spanish-speaking populations was even more striking; 60 percent of those in Tampa had inadequate health literacy scores compared with 21.2 percent of those in Houston. The authors of this study indicated that the geographic variation between study sites might reflect differences in the interrelated characteristics of race, language, and socioeconomic status. However, when these variables were controlled for, participants in Cleveland continued to show the highest rates of limited health literacy scores among English speakers. Factors mediating geographic differences in skills remain unknown and warrant further investigation. Increased awareness of these variations could lead to regionally tailored health literacy interventions.

Studies shown in Table 3-2 suggest that the segments of the U.S. population that could be considered at greatest risk for limited health literacy are those that were reported to have higher rates of limited literacy in the NALS study and other sources. As with limited literacy reported in the NALS (Kirsch et al., 1993), limited health literacy as currently measured is more prevalent among the elderly (Beers et al., 2003; Benson and Forman, 2002; Gazmararian et al., 1999b; Schillinger et al., 2002; Williams et al., 1995, 1998a, b). Limited literacy and health literacy as measured by these studies in older adults is an important problem. Older individuals generate high health-care costs (Berk and Monheit, 2001). High health-care costs, in turn, are associated with limited literacy (see below for a detailed discussion). While older individuals have some of the highest rates of limited health literacy scores and limited functional literacy scores, they also have some of the most demanding health-care needs. Limited health literacy in older adults may contribute to a situation in which those most in need of health care may be those least able to access and benefit from the care.

TABLE 3-2

The Epidemiology of Health Literacy Skills Among Various Populations.

Studies with current health literacy measures have reported that demographic associations with limited health literacy include racial or ethnic minority status (Arnold et al., 2001; Beers et al., 2003; Bennett et al., 1998; Gazmararian et al., 1999b; Kalichman et al., 2000; Lindau et al., 2002; Schillinger et al., 2002), fewer years of schooling or lower education level (Beers et al., 2003; Benson and Forman, 2002; Gazmararian et al., 1999a, b; Kalichman et al., 2000; Schillinger et al., 2002), and being a Spanish-speaker (tested in Spanish; Weiss et al., 1994). Notably, these demographic groups identified by several assessments conducted in different settings are the same demographic groups identified by the NALS, discussed earlier. However, none of these studies have involved a sufficiently large random sample of adults that would allow us to extrapolate these findings to other populations.

Associations with Health Knowledge, Behavior, and Outcomes

Individuals with inadequate health literacy as currently measured report less knowledge about their medical conditions and treatment, worse health status, less understanding and use of preventive services, and a higher rate of hospitalization than those with marginal or adequate health literacy (for review, see Parker et al., 2003). Table 3-3 displays a sampling of studies that provide evidence of health-related associations with health literacy as measured by currently available assessments of print literacy in the health context. These studies were identified using the same methodology as described in the previous section.

TABLE 3-3

Health-Related Associations of Health Literacy Skills.

A number of researchers examined relationships between patients' scores on the REALM or TOFHLA, knowledge of illnesses, and chronic condition management. Studies included in Table 3-3 demonstrate that hypertension, diabetes, asthma, and HIV/AIDS patients with lower scores have less knowledge of their chronic illness and its management than those with higher scores (Kalichman et al., 2000; Schillinger et al., 2002; Williams et al., 1998a, b). Examination of health literacy scores and health-care management found that patients with limited health literacy as determined by current measures have a decreased ability to share in decision making about prostate cancer treatment (Kim et al., 2001), lower adherence to anticoagulation therapy (Lasater, 2003; Win and Schillinger, 2003), and worse glycemic control (Arnold et al., 2001; Kalichman and Rompa, 2000; Schillinger et al., 2002; Williams et al., 1998a, b).

As noted in Table 3-3, several studies have demonstrated a relationship between hospitalization rates and limited health literacy as measured by the TOFHLA or REALM (e.g., Baker et al., 1997, 2002a). For example, Gordon and colleagues (2002) demonstrated that although rheumatoid arthritis patients of differing literacy levels had similar levels of rheumatoid arthritis-related dysfunction, those scoring lower on the REALM had significantly more hospital visits in the previous 12 months than age- and sex-matched controls with higher scores. Arozullah and colleagues (2002) investigated the relationships between health literacy level, the preventability of hospital admissions, and the causes of preventable hospital admissions. Patients scoring at or below the third-grade level on the REALM had an increased risk (odds ratio (OR) = 8.5) of preventable admission as compared to those scoring at the ninth-grade level and above, while patients with fourth- to eighth-grade literacy had an OR for preventability of 2.5. Furthermore, preventable admissions among patients with literacy levels at or below sixth grade were more likely than preventable admissions among those with above sixth-grade literacy to be related to system-level factors such as the availability of outpatient diagnostic care.

Health literacy level has also been linked to self-reported health status. Baker and colleagues found that among both individuals at a public hospital (1997) and in a Medicare managed care health plan (2002a), those with inadequate health literacy as determined by the TOFHLA were significantly more likely than patients with adequate health literacy scores to report their health as poor.

Health literacy levels may influence health status by affecting care-seeking behavior. A study by Bennett and colleagues (1998) indicated that individuals with lower health literacy scores on the REALM may enter the health-care system when they are sicker than those with higher health literacy scores; among men with prostate cancer, those scoring lower on the REALM were more likely to be initially diagnosed at a more advanced stage of disease than those with higher health literacy. However, this difference was not statistically significant after adjustments for race, age, and location or care. Several studies have showed that individuals with lower health literacy scores are less likely than those with higher health literacy scores to make use of preventive health-care services. Scott and colleagues (2002) found that, among Medicare managed care enrollees, those with inadequate health literacy, as measured by the S-TOFHLA, used preventive health services (including influenza and pneumococcal vaccinations, mammogram, and Pap smear) less than enrollees with higher health literacy. Similarly, Fortenberry and colleagues (2001) administered the REALM test to 1,035 people in four sites across the country, and found that higher REALM scores were independently associated with gonorrhea testing in the previous year. In contrast, Moon and colleagues (1998) found no relationship between parental literacy levels, as measured by the REALM, and use of preventive health services for pediatric patients.

Finding 3-3 Adults with limited health literacy, as measured by reading and numeracy skills, have less knowledge of disease management and of health-promoting behaviors, report poorer health status, and are less likely to use preventive services.

While research in developing countries rarely examines health literacy, a number of studies have examined the association between literacy or reading levels and health outcomes. A Bolivian study found that children of individuals participating in health, literacy, or small business financing programs offered through an international non-governmental organization were at less risk of becoming malnourished than children from comparison communities (Gonzales et al., 1999). A World Bank study of the association between maternal educational attainment and child health in Morocco found that the mother's health knowledge was associated with improved child health and nutrition (Glewwe, 1999). In contrast, a 10-year study of factors influencing infant and child health in Pakistan showed notable improvements in indicators of health, including a 69 percent increase in vacci nation rates, although the numbers of literate women did not change and there was a 12 percent increase in the number of literate men. This study also suggests that health behaviors may improve as a result of changes in health-care delivery without improvements in individual literacy (Northrop-Clewes et al., 1998).

Financial Associations of Low Health Literacy

The limited information available suggests that limited health literacy may be associated with increased consumer, health provider, and the healthcare system costs. Weiss and colleagues (1994) initially found no association between literacy skills, as measured by grade-equivalent reading level, and total 1-year health-care charges in 402 randomly selected Arizona Medicaid enrollees. The relationship between costs incurred and reading level as assessed by the IDL in the 72 patients in the original sample who were not seeking prenatal care was reanalyzed since 330 patients in this study were receiving prenatal care that had a fixed cost. The health-care costs of the remaining 72 patients showed greater variation. Consistent with REALM scoring, which categorizes the lowest reading group as third grade and under, the authors divided the sample at grade 3. Patients with a reading level at or below third grade had mean Medicaid charges of $10,688, while patients who read above the third-grade level had mean charges of $2,891 (p = 0.025) (Weiss and Palmer, 2004).

Friedland (1998) presented an analysis of the association of literacy and health-care utilization using data from the Health Care Financing Administration,⁶ the NALS, and the Survey of Income and Program Participation. This estimate was derived from predicted levels of functional literacy and estimates of health-care use. Differences in health-care spending were estimated by comparing health-care utilization by people with a lower probability of having functional literacy skills to health-care utilization by people those with a higher probability of having functional literacy skills. Friedland suggested that the additional health-care resource attributable to inadequate health literacy (NALS Level 1) in 1996 was $29 billion if inadequate health literacy was equivalent to inadequate literacy, and would have grown to $69 billion if even half of the marginally literate (NALS Level 2) were also considered not health literate (Friedland, 1998). This estimate has been cited elsewhere and, while illustrating the depth of the problem of limited health literacy, does not directly address the issue of cost.⁷

Associations of patient literacy with health-care utilization were examined by Baker and colleagues (1998, 2002a). Two analyses showed that limited literacy patients have higher rates of hospitalization that may be associated with greater resource use. Of 979 patients seen in the emergency department of Grady Memorial Hospital (Atlanta) in 1994–1995, logistic regression showed that those with inadequate health literacy, as measured by the TOFHLA, were more likely to be hospitalized (31.5 percent) than patients with adequate health literacy (14.9 percent). The OR after adjusting for confounding variables was 1.69. The adjusted relative risk increased to 3.15 among patients with inadequate literacy who had been hospitalized in the previous year (Baker et al., 1998). More recently, Baker and colleagues (2002a) studied a prospective cohort of 3,260 Medicare managed care enrollees. Of the 29 percent of these patients hospitalized over a 2-year period, the risk of admission was inversely related to health literacy as measured by the S-TOFHLA. The adjusted relative risk of admission was 1.29 for those with inadequate literacy and 1.21 for those with marginal literacy; however, it should be noted that education was not controlled for in this study.

For the purpose of this Institute of Medicine report, the committee commissioned an examination of the expenditure data collected in association with the Baker et al. (2002a) study from David Howard, a health economist at Emory University.⁸ Using econometric regression techniques, Howard found that predicted inpatient spending for persons with inadequate health literacy, as measured by the S-TOFHLA, was $993 higher than that of persons with adequate health literacy. This difference fell to $450 after controlling for health status. It is not clear whether this control is appropriate as there may be a bidirectional relationship between literacy and health status. Emergency care costs incurred by individuals with inadequate health literacy scores were higher than in patients with adequate literacy, while pharmacy expenses were similar in both groups and outpatient expenditures were lower. Although this Medicare managed care sample is not representative of the U.S. population as a whole, the results are consistent with previous reports that limited-literacy individuals make greater use of services designed to treat complications of disease and fewer services designed to prevent complications (Baker et al., 1998, 2002; Gordon et al., 2002; Scott et al., 2002).

These data suggest that patients with limited literacy may interact with a complex health-care system in ways that interfere with ideal utilization patterns and therefore could be more expensive. However, since the causal relationships between literacy and health-care utilization and cost have not been discovered, it is not possible to establish a valid cost figure for the impact of limited health literacy. If the magnitudes suggested by Howard and Friedland approach the actual costs, they clearly underscore the importance of addressing this risk factor from a financial perspective, in addition to health outcome implications. More research is needed to expand these limited results and move to a clearer estimation of these effects.

Finding 3-4 Two recent studies demonstrate a higher rate of hospitalization and use of emergency services among patients with limited literacy. This higher utilization has been associated with higher health-care costs.