Can This Patient Read and Understand Written Health Information?

A 67-year-old patient presents to your clinic for the first time 4 days following hospital discharge for a new diagnosis of atrial fibrillation. His heart rate is adequately controlled, and he was prescribed warfarin (5 mg daily) at discharge, with a plan for anticoagulation management. His international normalized ratio today is 1.4 (goal, 2-3), and you plan to enroll him in your nurse-run anticoagulation clinic but wonder about his ability to read and follow written instructions for managing his anticoagulation therapy.

Health literacy is “the degree to which individuals can obtain, process, and understand basic health information and services needed to make appropriate health decisions.”¹ The 2003 National Adult Assessment of Literacy estimated that 14% of adults had below basic literacy and an additional 22% had only basic literacy—resulting in more than 90 million US adults who may lack the literacy skills to effectively function in the current health care environment.² Quiz Ref IDWritten instructions are a key component in health communication, and patients with limited literacy frequently do not understand prescription medication labels³ or complex instructions like anticoagulation dosing.⁴ This limitation is most common in older patients, those with lower education levels, immigrants, and racial/ethnic minorities.^{5 - 8} Quiz Ref IDPrior research has supported the association between literacy and disease knowledge, utilization of preventive services, hospitalization, overall health status, control of chronic disease, and mortality.^{9 - 14}

Quiz Ref IDDespite the important health implications of literacy, physicians are often unaware of their patients' literacy levels^{15 - 16} and the effects on outcomes.¹⁷ Patients may not volunteer they have a problem, because nearly one-half of patients with limited literacy express shame over their inability to read.^{18 - 19} Although education attainment can be a proxy for literacy, it is often misleading, with many patients reading below their highest level of education^{20 - 23} and up to 20% of high school–educated patients having limited literacy.⁵ The implication for clinicians is that many patients will be unable to interpret prescription instructions, understand patient education materials, or use written information to prepare for clinical tests; however, clinicians often do not detect these limitations.

Effective tools for measuring literacy are regularly used in research studies (the reference standards discussed herein are summarized in the eTable). The Test of Functional Health Literacy in Adults (TOFHLA), with a Spanish version available, is considered the most comprehensive reference standard and measures comprehension of written instructions and numerical information but requires up to 22 minutes to administer.²⁴ The test consists of 3 prose passages followed by a 50-item reading comprehension section that asks patients to fill in omitted words from a passage based on multiple-choice options. The passages are taken from instructions used by patients to prepare for an upper gastrointestinal tract radiograph series, the patient “Rights and Responsibilities” section of a Medicaid application, and a standard hospital informed consent document. There is also a 17-item numerical ability test that assesses the ability to comprehend prescription labels, blood glucose test results, clinic appointment slips, and financial information. A shortened version of the TOFHLA (S-TOFHLA) has been adapted that requires only 7 to 12 minutes to administer and includes 36 items from 2 prose passages and 4 items assessing numerical ability.²⁵

A basic medical vocabulary is necessary for understanding and applying health information. In the Rapid Estimate of Adult Literacy in Medicine (REALM), the patient reads and pronounces 66 English medical words arranged in ascending order of difficulty. Points are given for correctly pronounced words, and scores (range, 0-66) are correlated with reading grade estimates, with accepted cut points of less than 61 corresponding to a less than ninth-grade reading level and less than 45 corresponding to a less than sixth-grade reading level.²⁶

Although the REALM does not measure comprehension, it is highly correlated with both the TOFHLA and S-TOFHLA, with correlation coefficients of 0.84 and 0.80, respectively.^{24 - 25} The most important differences between REALM and the S-TOFHLA are in the 25% to 75% interquartile range of the S-TOFHLA, where the correlation coefficient is only 0.62.²⁵ When compared with the S-TOFHLA as the reference standard for identifying patients with inadequate or marginal literacy, a REALM score less than 45 has a likelihood ratio (LR) of 8.3 (95% confidence interval [CI], 4.7-14), a score of 45 to 60 has an LR of 0.41 (95% CI, 0.26-0.63), and a score of 61 to 66 has an LR of 0.03 (95% CI, 0.004-0.18).²⁵ While the REALM is treated as a reference standard in this review and most of the literature, it requires only 3 minutes to administer and may be the preferred method for identifying limited literacy when time is available.

Patients who score below an accepted threshold on literacy reference standards are described variably as having low, limited, or inadequate literacy. One of the challenges in understanding the relationship between literacy and health is determining what level of literacy is truly adequate to navigate the health care system such that a patient's reading ability poses no limitations. For the purposes of this review, we use the terms inadequate, marginal, and adequate literacy to mean the specific definitions for each reference standard shown in Table 1.

The prevalence of limited literacy varies substantially,⁵ and there may be little added value from testing in clinic settings with a known high prevalence. A systematic review that pooled data from more than 300 studies estimated that 26% of patients had inadequate literacy and an additional 20% had marginal literacy, with a range between 0% and 68%.⁸ These estimates reflect oversampling from health care settings serving patients with lower socioeconomic status and education and may therefore overestimate the prevalence nationwide. Among seniors enrolling in Medicare managed care plans, 34% of English-speaking and 53% of Spanish-speaking enrollees had inadequate or marginal literacy.⁶

The proper use of tools to identify limited literacy is controversial, with potential application as a screening tool to identify individuals who may need more assistance or for case finding to support or refute an increased suspicion for limited literacy in an individual patient. Some experts have expressed reservations about routine screening for limited literacy and point out that nearly half of patients with limited literacy are ashamed of their inability to read, and screening could potentially cause harm.^{18 - 19} Those who oppose routine screening argue that conclusive evidence for specific interventions is lacking²⁷ and that recommended communication strategies are likely to benefit patients of all literacy levels and to harm none.²⁸ Providing information written at a low literacy level and communicating without medical jargon should be accomplished for all patients, not just those who screen positive for limited literacy. In particular, ensuring patients' understanding by having them “teach back” the material would provide universal precautions that ensure comprehension regardless of a person's literacy level.²⁹

While nearly all physicians would agree with these recommendations, they may not effectively adhere to them owing to habit or the time limitations of a busy practice. There is evidence from a single randomized trial suggesting that physicians who were informed of their patients' limited literacy were 3 times more likely to use recommended communication strategies for patients with limited literacy but were less likely to feel satisfied and effective in their visits.³⁰ Sixty-four percent of physicians and 96% of patients felt that literacy screening was worthwhile by the end of the trial. In addition to improved patient-physician communication, there is also some evidence that patients with limited literacy may benefit more from multifaceted interventions.^{31 - 32} While the most appropriate role for literacy testing is still controversial, some physicians may reasonably choose to evaluate their patients' literacy and use this information to tailor their care. Thus, we identified and appraised brief instruments for identifying patients with limited literacy in clinical settings.

The literature review was conducted on articles published from 1969 through February 2010 and included any studies indexed in PubMed (including MEDLINE) and Psychinfo (eFigure). The key word search in PubMed included literacy, numeracy, reading ability, reading skill, WRAT (Wide Range Achievement Test), wide range achievement, TOFHLA, test of functional health, REALM, and rapid estimate of adult. For Psychinfo, we used only the term health literacy to limit search results. Our search yielded 2360 PubMed abstracts and 162 Psychinfo abstracts, and a query of experts and bibliographies yielded 2 additional abstracts not identified in the original search. All abstracts were independently reviewed by 2 members of the study team, and we selected only articles reporting on studies with participants 18 years or older that included an acceptable reference standard for literacy. We contacted the corresponding authors of 7 studies requesting additional information not included in the original manuscript^{33 - 39} ; 4 provided additional data included in this systematic review.^{35 ,37 - 39}

For the reference standard, we accepted the REALM, Wide Range Achievement Test, TOFHLA, and S-TOFHLA. When multiple reference standards were used in a single study sample, we only report the results using the TOFHLA or S-TOFHLA. Based on these criteria, we identified 26 articles for full-text review and data abstraction. Each of these articles underwent independent full-text review by 2 reviewers. Studies were included if they reported original research on adult participants and included measures of diagnostic test characteristics or the data required to calculate this information. Studies were excluded if they did not meet these criteria; had 10 or fewer patients; were written in a language other than English; reported a screening test only for special populations, such as those with learning disabilities or cognitive impairment that directly affect reading ability; evaluated literacy within a specific content area (eg, genetics literacy); or only presented psychometric properties such as Cronbach α without providing diagnostic test characteristics.

We abstracted information describing each study, including the health care setting, patient population, and sampling strategy. We collected information on diagnostic test characteristics including sensitivity, specificity, and LRs. When feasible, we separated results for tests performed in Spanish and English and also omitted data for patients with visual impairment when calculating diagnostic test characteristics. The abstraction from each reviewer was synthesized into a common data table, and disagreements were settled by a third reviewer.

Many of the questions asked patients to report their responses on a graded scale (eg, ranging from “always,” “often,” “sometimes,” “occasionally,” “never”), resulting in multiple potential cut points for each question. We present multilevel LRs, in which each response option is calculated as a positive LR compared with all other response options.⁴⁰ Some studies classified patients with marginal literacy as adequate, while other studies categorized marginal literacy as inadequate. Separate results for each categorization are presented when possible. We provide summary measures only for marginal literacy categorized together with inadequate literacy to focus on the broadest definition of limited literacy. While this values sensitivity over specificity, we believe this is most appropriate, given the higher risk of falsely reassuring (ie, false-negative) results.

Quality scores were assigned using the recommended approach for Rational Clinical Examination articles.⁴¹ We considered studies to be of high quality (level 1) if they had all of the following: (1) sample sizes of at least 100 participants; (2) applied a reference standard regardless of the screening result; (3) independently interpreted the reference standard and screening instrument; and (4) evaluated the screening questions in a sample that represented a wide spectrum of adult community or primary care patients. Level 2 studies did not meet at least 1 of these criteria but were not believed to be so flawed that the results were invalid. Examples of level 2 studies included studies conducted only in a sample with a particular disease (eg, diabetes). Level 3 studies had 1 or more flaws thought to likely invalidate the results. We retained only the level 1 and 2 studies for data synthesis.

When studies did not directly report the LRs, we calculated LRs based on the available information to reconstruct the contingency table. For articles not reporting 95% CIs for the LRs, CIs were calculated using methods reported previously.⁴² The summary LRs are the result of meta-analyzed data calculated in Comprehensive Meta-analysis version 2.2.048 (Biostat Inc, Englewood, New Jersey) using univariate random effects for the individual studies.⁴³ We measured heterogeneity using the I² statistic; however, even when statistical heterogeneity was present, the studies were similar enough in design that we chose to combine them into summary measures.

We identified and included 10 unique studies that met all inclusion and exclusion criteria and provide data on 6 unique tests for limited literacy. All studies were conducted in the United States. Results are presented in Table 2, Table 3, and Table 4.

Newest Vital Sign. The Newest Vital Sign test evaluates literacy based on reading and applying information included on an English or Spanish nutritional label from a pint of ice cream (2-6 minutes to administer).⁴⁴ Patients are asked 6 questions about serving size, nutrition information, and ingredients. Three separate levels have been suggested by the creators of this test and were available for data synthesis: 0 through 1 correct answers on the English version had a summary LR of 3.2 (95% CI, 1.9-5.4) for inadequate or marginal literacy^{44 - 45} ; 2 through 3 correct answers were indeterminate, with a summary LR of 0.77 (95% CI, 0.34-1.8); 4 through 6 correct answers effectively ruled out inadequate or marginal literacy, with a summary LR of 0.08 (95% CI, 0.02-0.45). The Spanish version did not perform as well for identifying Spanish speakers with literacy problems (Table 2).

Medical Term Recognition Test. The Medical Term Recognition Test (METER) is a self-administered word recognition test patterned after the REALM. From a list of 40 medical words mixed with 40 nonwords, the patient marks only those items recognized as actual words (2 minutes to administer). In a single study of 150 mostly white patients awaiting a cardiac stress test, the METER was highly correlated with REALM scores, and 3 separate literacy categories were suggested based on METER scores of 0 through 20, 21 through 34, or 35 through 40. The METER had an LR of 22 (95% CI, 2.6-179) for inadequate and marginal literacy when only 0 through 20 correct words were identified; an LR of 7.7 (95% CI: 3.8-15.7) for 21 through 34 words; and an LR of 0.27 (95% CI, 0.14-0.52) for 35-40 words (Table 2).³⁹ Given the significant overlap in word recognition with the REALM, these LRs likely inflate the accuracy of this test; however, no studies have compared the METER with the TOFHLA.

Use of a Surrogate Reader. Quiz Ref IDFive studies evaluated the use of surrogate readers as a marker for limited literacy.^{5 ,35 - 38} This question was first asked directly by Williams et al as, “Do you usually ask someone to help you read materials you receive from the hospital?” (yes/no response).⁵ Subsequent versions instead ask, “How often do you need to have someone help you when you read instructions, pamphlets, or other written material from your doctor or pharmacy?”^{35 ,38} or “How often do you have someone help you read hospital materials?”^{36 - 37} Response options were “always,” “often,” “sometimes,” “rarely,” or “never.”

When patients respond “yes” to the question, “Do you usually ask someone to help you read materials you receive from the hospital?” the LR is 4.5 (95% CI, 3.8-5.1) for inadequate literacy, while a negative response has an LR of 0.55 (0.51-0.59).⁵ When patients report the frequency of using a surrogate reader, a response of getting help “sometimes” or more frequently had a summary LR of 2.9 (95% CI, 2.3-3.7) for inadequate or marginal literacy; “rarely” had a summary LR of 1.0 (95% CI, 0.80-1.3); and “never” had a summary LR of 0.53 (95% CI, 0.38-0.74) (Table 3).

Confidence With Filling Out Medical Forms. In 3 studies patients were asked, “How confident are you filling out medical forms by yourself?”^{36 - 37 ,46} Responses were “extremely,” “quite a bit,” “somewhat,” “a little bit,” or “not at all.” In the 2 least-confident groups, patients had a summary LR of 5.0 (95% CI, 3.8-6.4) for inadequate or marginal literacy; those “somewhat confident” had a summary LR of 2.2 (95% CI, 1.5-3.3); and those expressing confidence were less likely to have a problem, with a summary LR of 0.44 (95% CI, 0.24-0.82) (Table 3).

Self-rated Reading Ability. Not surprisingly, patients who acknowledge trouble reading forms and written hospital materials have an LR of 28.6 (95% CI, 16.3-52.1) for inadequate literacy.⁵ However, the LR of only 0.81 (95% CI, 0.78-0.84) suggests that many patients who deny they have a problem either overestimate their reading ability or are reluctant to disclose their illiteracy.

Self-rated reading ability was also evaluated by asking, “How would you rate your ability to read?” with responses of “excellent,” “very good,” “good,” “okay,” “poor,” “very poor,” and “terrible.”³⁵ A response of “okay” or worse had an LR of 5.1 (95% CI, 3.2-8.3) for inadequate or marginal literacy; “good” had an LR of 1.0 (95% CI, 0.61-1.8), and “very good” or “excellent” had an LR of 0.16 (95% CI, 0.05-0.46) (Table 4).

Difficulty Learning About Health. In 2 separate studies, the question, “How often do you have problems learning about your medical condition because of difficulty understanding written information?” was assessed as an indicator of limited literacy.^{36 - 37} Response options were “always,” “often,” “sometimes,” “occasionally,” and “never.” A response of “sometimes” or more frequently had an LR of 2.4 (95% CI, 1.9-3.0); “occasionally” had an LR of 0.88 (95% CI, 0.64-1.2); and “never” had an LR of 0.65 (95% CI, 0.54-0.78) (Table 4).

Four separate studies evaluated combinations of single-item questions along with educational attainment to see if a multi-item tool performed better than single questions.^{35 - 37 ,46} No combination of questions or patient education level was able to significantly improve on the test characteristics of the best single-item test for each study.

You estimate your patient's pretest probability of inadequate or marginal literacy at 34%, based on the estimated prevalence in the Medicare population. In discussing the use of warfarin for prevention of stroke, you ask, “How confident are you filling out medical forms by yourself?” He responds, “a little bit [LR, 5.0], but my wife helps take care of that stuff.” The patient's posttest probability is more than 70%, and you recognize that additional time and effort may be needed to ensure adequate patient understanding. With his permission, you ask his wife to come in from the waiting room and join the discussion. You start by asking what they know about this new medication and why he is taking it. Being careful to avoid medical jargon, you carefully add to their explanation, including why it has to be checked so frequently. Before they go, you give him a handout, written below a sixth-grade level, with his medication instructions and follow-up appointment and ask him to teach back how he is to take this medicine and follow up. While these communication strategies are beneficial for all patients, you and your nurse are more cognizant of your patient's potential limitations and careful to take the time needed for clear health communication in all future interactions.

Quiz Ref IDLiteracy can be measured accurately in health care settings with tests that require several minutes to administer; however, when time is limited, several single-item questions are effective. Based on the available data, questions about patients' confidence with medical forms or whether they use a surrogate reader both performed moderately well for identifying patients with inadequate and marginal literacy and have been evaluated in several studies. There was limited evidence about the evaluation of literacy in Spanish-speaking patients, and it is not known whether the diagnostic test performance of these questions differs when asked by practicing clinicians.

Limited literacy is common and associated with health outcomes, but physicians are often unaware of patients' reading abilities.¹⁵ To maximize patient safety and quality of care, it is essential that clinicians emphasize clear communication for all patients, whether using written, spoken, or other methods for delivering information. In some situations, clinicians may choose to screen and identify patients whose literacy poses a significant barrier to effective communication. When several minutes are available for testing, the REALM (3 minutes), S-TOFHLA (7-12 minutes), and possibly the METER (2 minutes) are the most accurate tools for identifying patients with limited literacy. However, for rapid testing we recommend asking patients how confident they are filling out medical forms, how often they have someone help them read health information, or to rate their own reading ability. Patients who test positive with these questions are at higher risk for poor health outcomes, and clinicians should regularly assess adequate recall and comprehension of information to promote high quality and safe delivery of health care.