Ironically, as the field of health literacy has expanded in scope and depth, the term “health literacy” itself has come to mean different things to various audiences and has become a source of confusion and debate. In 1999, the American Medical Association's Ad Hoc Committee on Health Literacy defined health literacy as “the constellation of skills, including the ability to perform basic reading and numerical tasks required to function in the health care environment,” including “the ability to read and comprehend prescription bottles, appointment slips, and other essential health-related materials.”² The definitions used by Healthy People 2010⁵ and the Institute of Medicine (IOM)³ were similar: “The degree to which individuals have the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions.” These definitions present health literacy as a set of individual capacities that allow the person to acquire and use new information. These capacities are relatively stable over time, although they may improve with educational programs or decline with aging or pathologic processes that impair cognitive function.⁶

Others have argued that if health literacy is the ability to function in the health care environment, it must depend upon characteristics of both the individual and the health care system. From this perspective, health literacy is a dynamic state of an individual during a health care encounter. An individual's health literacy may vary depending upon the medical problem being treated, the health care provider, and the system providing the care.

Some also view health knowledge as part of health literacy. For example, the IOM expert panel divided the domain of “health literacy” into (1) cultural and conceptual knowledge, (2) oral literacy, including speaking and listening skills, (3) print literacy, including writing and reading skills, and (4) numeracy. The American College of Physicians Foundation recently advertised a set of informational cards that physicians could give to their patients to “raise their level of health literacy.”⁷ From this perspective, health literacy is an achieved level of knowledge or proficiency that depends upon an individual's capacity (and motivation to learn) and the resources provided by the health care system.

All these perspectives represent reasoned ideas based on different orientations to the problem. Nevertheless, the lack of shared meaning for the central term in a field is obviously problematic. For example, there has been confusion and disagreement between the authors of research articles or grants and reviewers. It is not surprising that experts disagree about how health literacy should be measured as it may be that they are not really talking about the same underlying construct. If health literacy is a capacity of a person, measures of an individual's reading ability and vocabulary are appropriate. In contrast, if health literacy depends on the relationship between individual communication capacities, the health care system, and the broader society, measures at the individual level are clearly inadequate. If knowledge is part of the definition of health literacy, this too must be measured.

To address this problem, I will first present a conceptual model of the domains of health literacy and the relationship of health literacy with health outcomes, and suggest possible terms that may be used to describe these domains. This model is designed to supplement the models of health literacy presented within the IOM report to allow more specific and precise discussion of measures of health literacy. I hope this will serve as a first step in a longer process of achieving a shared terminology for researchers and other experts in the field, as recommended in the IOM report. I will then review available measures of these domains and discuss which measures may be most useful in research and clinical practice.

Reading fluency is the ability to mentally process written materials and form new knowledge. The first National Adult Literacy Study (NALS) divided reading fluency into 3 skill sets: (1) the ability to read and understand text (prose literacy), (2) the ability to locate and use information in documents (document literacy), and (3) the ability to apply arithmetic operations and use numerical information in printed materials (quantitative literacy).⁸ Other investigators have used the term “numeracy” for the latter skill.^9–11 In the Test of Functional Health Literacy in Adults (TOFHLA), the “numeracy” items actually assess the NALS subdomains of quantitative and document literacy.^12,13 Other “numeracy” tests focus on an individual's ability to understand probabilities and percentages.⁹

Prior knowledge (i.e., an individual's knowledge at the time before reading health-related materials or speaking to a health care professional) is composed of vocabulary (knowing what individual words mean) and conceptual knowledge (understanding aspects of the world, e.g., how different parts of the body work or what cancer is and how it injures the body). Vocabulary is distinct from reading fluency, although the 2 are very highly correlated because people acquire much of their vocabulary through reading. Although the more expansive definitions of health literacy discussed above (e.g., the IOM report) include conceptual knowledge as part of health literacy, this model views conceptual knowledge as a resource that a person has that facilitates health literacy but does not in itself constitute health literacy.

Reading fluency allows an individual to expand one's vocabulary and gain conceptual knowledge. The converse is also true. Vocabulary and background knowledge of the general topics covered in written materials improve an individual's comprehension of these materials, as shown by the double-headed arrow between the 2 individual capacity boxes in Figure 1. In other words, it is easier to read and comprehend materials that contain familiar vocabulary and concepts. Two individuals with similar general reading fluency may have different abilities to read and understand health-related material as a result of differences in their baseline knowledge of health vocabulary and concepts.

The notion that individuals will understand written and spoken communication better if they are familiar with the words and concepts presented makes sense. However, we do not know the average difference between an individual's (1) general reading fluency, vocabulary, and knowledge and (2) the person's health-related reading fluency, vocabulary, and knowledge. Nevertheless, the distinction between general reading fluency and health-related reading fluency is important for research because a measure of an individual's ability to read and understand health-related materials is likely to be more closely related to health outcomes than a measure of general literacy. Thus, a study that uses a measure of health-related reading fluency as a predictor variable will have greater power to detect associations with health outcomes than a study that uses a measure of general reading fluency.

The most widely used measures are the Rapid Estimate of Adult Literacy in Medicine (REALM)¹⁴ and the Test of Functional Health Literacy in Adults (TOFHLA).^12,13 Neither test is a comprehensive assessment of an individual's capacities. Rather, the tests measure selected domains that are thought to be markers for an individual's overall capacity. The REALM is a 66-item word recognition and pronunciation test that measures the domain of vocabulary. The TOFHLA measures reading fluency. It consists of a reading comprehension section (a 50-item test using the modified Cloze procedure) to measure prose literacy and a “numeracy” section with 17 items assessing individuals' capacity to read and understand actual hospital documents and labeled prescription vials. Although the 2 tests measure different capacities, the tests are highly correlated (Spearman's correlation coefficient 0.80).¹⁵ Both tests are also highly correlated with general vocabulary tests (i.e., the Wide Range Achievement Test, Revised).¹⁶

Most recently, Weiss et al.³¹ developed the Newest Vital Sign (NVS), consisting of a nutrition label for ice cream with 6 questions about the information contained in the label. The NVS takes approximately 3 minutes to complete. The first 4 questions require document and quantitative skills, including the ability to calculate percentages. The sensitivity of a score of <2 for detecting patients with inadequate or marginal health literacy was 72% for the English version (NVS-E) and 77% for the Spanish version (NVS-S); the specificity was 87% and 57%, respectively. The ability of the NVS-E to identify patients with low literacy was markedly higher than for education and age alone (area under the receiver operating characteristic (ROC) curves of 0.88, 0.72, 0.71, respectively); data for the NVS-S were not reported. Thus, the NVS is fairly short, moderately accurate, and appears better than demographics alone for identifying individuals with limited reading capacities.

The second approach is to measure self-assessed health literacy based upon individuals' self-reported difficulty in understanding health care professionals and the written materials they encounter in their health care setting. This directly assesses the mismatch between an individual's capacities and her communication demands and provides important information from the patient's perspective. Williams et al.³² found that 3 questions about the ability to read the newspaper, the ability to read forms, and other written materials from hospitals, and the use of a “surrogate reader” to help understand health information each had low sensitivity for detecting patients with “inadequate functional health literacy” as measured by the TOFHLA. Chew et al.³³ assessed 16 questions with 5-point Likert-scale response options for their ability to identify patients with “inadequate health literacy” according to the S-TOFHLA. They found that 3 questions had a reasonably high predictive value for identifying 15 of 332 patients with inadequate literacy: (1) “How often do you have someone help you read hospital materials?” (2) “How confident are you filling out medical forms by yourself?” and (3) “How often do you have problems learning about your medical condition because of difficulty understanding written information?” (area under the ROC curves 0.87, 0.80, and 0.76, respectively). However, this study did not determine whether the predictive value of these questions was better than demographics alone (i.e., education, age, race/ethnicity) and whether the findings were replicable in a different population.

The study by Wallace et al.³⁴ in this issue of JGIM advances our understanding of the value of these screening questions. The 3 questions described above were tested in 305 adults in a university-based primary care clinic. In contrast to Chew's findings, the question “How confident are you filling out medical forms by yourself?” had the best predictive value (area under the ROC curve of 0.82) for identifying individuals with a REALM score of ≤44 (≤6th grade). This was substantially better than the ROC curve for a model that included age (<65 vs ≥65), race/ethnicity (white vs nonwhite), and education (<high school vs ≥high school).

All studies of screening questions or short tests have compared an instrument with a “gold standard,” such as the TOFHLA or the REALM. However, the more important question is whether the screening test predicts individual capacities above and beyond the powerful predictors of age, race/ethnicity, and years of school completed (as was done in the studies by Weiss and Wallace as described above). For example, in a study of Medicare managed care enrollees, demographics predicted inadequate or marginal literacy (according to the S-TOFHLA) with an area under the receiver-operator curve of 0.81 (unpublished data).³⁵ Studies should always assess the change in the c-statistic when the screening question is added to a predictive model that already contains age, race/ethnicity, and education and determine whether the change is clinically important (vs statistically significant). Models should enter age and education as precisely as possible, either as continuous or polytomous variables (e.g., 0 to 8, 9 to 11, 12, and >12 years of school) instead of the more weakly predictive dichotomous variables used in the study by Wallace.

There is no established test to measure comprehension of spoken health-related information. Like health-related print literacy, health-related oral literacy depends on vocabulary and prior conceptual knowledge of the topic being discussed. While health-related oral literacy does not depend upon reading fluency, the cognitive processes necessary for understanding the spoken word and the printed word are deeply intertwined. For all of these reasons, individuals with limited health-related print literacy are likely to also have limited health-related oral literacy, and the 2 may really be part of a single latent variable. Thus, while health-related oral literacy seems to be an appealing, distinct concept, it may be hard to measure independently from print literacy.