Measuring Functional, Communicative, and Critical Health Literacy Among Diabetic Patients

Survey participants were patients with type 2 diabetes who visited the outpatient department of general internal medicine at a university-affiliated hospital in Tokyo, Japan. The inclusion criteria were having type 2 diabetes and being under continuous care by the four attending physicians of the department who specialize in metabolic diseases. During the study period of October–December 2006, patients eligible for recruitment were randomly selected from the appointment logs of the study physicians. The selected patients were approached in the waiting room after a consultation and were given an explanation of the study purpose and procedures. The age, sex, most recently measured A1C level, and complications of each participant were obtained from a review of electronic medical records.

Among the 169 eligible patients identified, 157 patients provided written consent to participate in the study and completed the questionnaire (12 refusals [7.1%]). The major reasons for refusal were a lack of time and poor physical condition on the day of the questionnaire. For the purpose of this study, we excluded patients who had been diagnosed with diabetes within the previous year and those with an A1C level lower than 5.8% (n = 19). This resulted in a final sample of 138 patients. During the survey, a research assistant helped patients complete the questionnaire if the patient preferred assistance. Of the 138 patients, 46 (33%) preferred to receive help. The study was conducted with the approval of the ethical review committee at Teikyo University School of Medicine.

The scale items were constructed to directly reflect the previously reported definitions of functional, communicative, and critical HL (8). For functional HL, we reviewed the standard measures currently being used, although none were available in Japanese. The items were examined and adapted through discussions with coresearchers, including clinicians, public health researchers, and social science researchers working in related fields. Ease of use and comprehensibility were tested in a small sample of patients with diabetes (n = 20).

For functional HL, five items were prepared to assess the extent to which patients had experienced difficulties in reading the instructions or leaflets from hospitals and pharmacies, based on previously developed instruments (23,24). Communicative HL was evaluated with five items assessing the extent to which patients had extracted and communicated diabetes-related information since they were diagnosed with the disease. Similarly, critical HL was assessed by four items focusing on the extent to which patients had critically analyzed the information and used it to make decisions. In addition to the three subscales, the total HL scale was also formed by adding all 14 items. All scale items are presented in Table 2.

Each item was rated on a 4-point scale, ranging from 1 (never) to 4 (often). The scores for the items in a scale were summed and divided by the number of items in the scale to give a scale score (theoretical range 1–4). The scores were reversed for functional HL such that higher scores indicated higher HL.

Educational attainment, self-rated economic status, and duration of diabetes were obtained for each patient from the self-reported questionnaire. The self-rated economic status question asked how the patient perceived the financial situation of his/her family, with five options (lower, lower middle, middle, upper middle, and upper) that were then grouped into three categories by combining the lower two categories and the upper two categories in statistical analyses. Age, sex, A1C level, and presence of major diabetes complications (i.e., retinopathy, nephropathy, and neuropathy) were obtained from the review of electronic medical records. Patients were divided into three age groups (aged <65, 65–74, and ≥75 years) for the analysis.

Diabetes knowledge was measured with seven true/false questions about a patient's knowledge of diabetes and its care; e.g., “An HbA1c level of 8.0 indicates tight control of diabetes” and “In the food exchange lists, a small bowl of rice equals 2 units.” Items were developed based on existing diabetes knowledge scales (14,25,26) and patient education materials in Japan. The number of correct answers was counted as the diabetes knowledge score (theoretical range 0–7). Cronbach's α of the scale was 0.65, indicating adequate internal consistency.

The number of health information sources utilized by a patient was measured by counting the sources (i.e., physicians, other health care professionals, family/friends, other diabetic patients/patient groups, health education programs, TV/radio programs, Internet, newspapers/magazines, books, and leaflets/newsletters) through which the patient had ever sought or been provided diabetes-related information.

The self-efficacy of diabetes self-care was assessed using the four-item scale of self-care ability in the Diabetes Care Profile (27). The reliability of the scale was similar to the previously reported value (Cronbach's α = 0.83).

Cronbach's α coefficients were calculated to examine internal consistency. Exploratory factor analysis with promax rotation was performed to examine the factor structure of HL. To determine the relationships of HL scales with other measures, either Pearson or Spearman's correlation coefficient was calculated for continuous variables; either Wilcoxon's rank-sum test or Student's t test was used for categorical variables with two groups. A test for a trend across ordered groups was used for categorical variables with more than two ordered groups. Although patients were recruited from four physicians, within-physician clustering was not adjusted for because the intraclass correlations were very small. Case subjects with missing values were excluded from the respective analyses, which were conducted with Stata (version 9.2; Stata Corporation, College Station, TX).

The sociodemographic and clinical characteristics of the participants are displayed in Table 1. The mean age of the patients was 65 years, and the mean duration since the diagnosis of diabetes was 11.5 years. Based on the most current test results for A1C, 25% of the patients were in tight control (≤6.5%), 49.3% were in fair control (6.6–7.9%), and 25.7% were in poor control (≥8%). The patients had seen the physician at this outpatient service for mean ± SD 6.7 ± 5.7 years (not shown in Table 1).

The scores for each item of the functional, communicative, and critical HL scales are presented in Table 2. The scores were higher for functional HL items and relatively lower for critical HL items.

The internal consistency values of the functional, communicative, and critical HL scales were adequate (α = 0.84, α = 0.77, and α = 0.65, respectively). The item-total correlations were all positive and ranged 0.60–0.83 for functional HL, 0.60–0.85 for communicative HL, and 0.63–0.83 for critical HL. Because the deletion of any item would not result in an increase in Cronbach's α more than 0.02, it was decided to retain all items in each scale. Cronbach's α of the total HL scale was 0.78.

The HL scale was designed to measure three different HL levels. Exploratory factor analysis was performed to assess whether three underlying factors and each item's load on the factors could be identified as designed. The initial factor solution showed four factors with eigenvalues >1 (3.7, 3.1, 1.4, and 1.2), accounting for 26.6, 22, 9.8, and 8.6% of the variance, respectively. Although the eigenvalues declined sharply between the second and third factors, a forced factor solution with three promax-rotated factors was examined to assess whether the items clustered into three factors of HL as designed. The promax rotation showed that all five functional HL items loaded on the first component and all five communicative HL items loaded on the second component (Table 2). One critical HL item (i.e., patients “considered whether the information was applicable to your situation”) loaded on both the second and third components, with a slightly higher factor loading on the second. The analysis of reliability suggested that Cronbach's α of the critical HL scale would increase slightly by excluding this item (α = 0.67), while that of total scale would decrease slightly (α = 0.76). Because the item was considered an important component of the theoretical definition of HL and the difference in the loading and compromise in the internal consistency was relatively small, we decided to include this item in the critical HL scale. Although we ran the subsequent analyses using a scale excluding this item, the results did not change significantly.

Among the three HL scores, the association of functional HL with communicative and critical HL was not statistically significant (r = 0.09, P = 0.291 and r = 0.05, P = 0.548, respectively), while communicative and critical HL scores did have a significant positive association (r = 0.52, P < 0.001).

Bivariate relationships between HL scores and other parameters are shown in Table 3. The results were generally consistent with previous findings and with the theoretical basis of the HL concept. Diabetes knowledge score showed a strong positive association with all HL scores, whereas the number of information sources and self-efficacy were not significantly associated with functional HL scores. Functional and critical HL scores were significantly lower among older patients, but there was no significant age-related difference in communicative HL scores. There were no significant differences in any of the HL scores between sexes. HL scores were generally higher for those with higher educational attainment and economic status, although the relationship between self-rated economic status and critical HL was less clear. With regard to clinical characteristics, a higher communicative HL score was related to a lower A1C level, and the functional HL score was higher for patients with no diabetes complications.

Unlike previous studies that focused solely on functional HL, this study aimed to measure a broader concept of HL, one including the ability to extract, understand, and use health-related information. The results of this study support the reliability and validity of our HL scales in patients with type 2 diabetes.

The internal consistency of each HL scale was adequate. An examination of factor structure suggested three different levels of HL, although communicative and critical HL scales were closely related to each other. Indeed, a correlation analysis suggested that associations among HL scales were all in the positive direction but were statistically significant only between the communicative and critical HL scales. The score on each item was generally higher for functional HL items compared with communicative and critical HL items, consistent with the theoretical supposition that communicative and critical HL involve more advanced skills.

The relationships between HL scores and other measures were generally in the expected directions, supporting the validity of the scales. At the same time, correlations among scales were only moderate, indicating that each HL scale reflects a unique construct. Limited functional HL is related to poor knowledge of diabetes, as previously reported (11–14), and all three HL scales were positively associated with diabetes knowledge. The number of information sources used by patients to obtain diabetes information was not correlated with functional HL. In contrast, those with higher communicative and critical HL scores consulted a greater number of information sources, which is easily interpretable given the definitions of communicative and critical HL. Furthermore, as has been suggested by a previous study (17), self-efficacy of diabetes self-care did not significantly differ with functional HL score. On the other hand, communicative and critical HL scores had significant positive correlations with self-efficacy, suggesting that advanced levels of HL may be more closely related to higher self-efficacy in the management of diabetes.

With regard to sociodemographic factors, advanced age, lower educational attainment, and lower economic status tended to be associated with lower HL, which was generally consistent with a recent meta-analysis study (28). However, the significance of associations with each of these factors was different among the three HL scales, indicating that each HL scale reflects a unique construct. Using these scales in further investigations may allow identification of patient characteristics that are likely to affect each type of HL.

As in previous studies (12–14,19,20), our study did not find a significant relationship between functional HL and A1C, whereas a higher communicative HL score was related to a lower A1C. This suggests that the skills of extracting, communicating, and applying information might be more important in managing diabetes than the skills necessary to read the information. On the other hand, a lower functional HL was significantly associated with higher rates of diabetes complications. These bivariate relationships between HL and health outcomes should be carefully examined in future studies with larger samples, considering the potential confounders. It is possible that patients with lower functional HL were older and more likely to have complications. However, our post hoc analysis suggested that these findings remained significant even after adjusting for patient sociodemographic characteristics (i.e., age, sex, education, and self-reported economic status) and duration of diabetes (data not shown). Our findings suggest that HL may have a direct relationship with clinical outcome in patients with diabetes and that each HL level might act in a different way. The ways in which each HL level influences patient self-care behaviors and health outcomes should be further explored.

Several limitations should be noted in interpreting the findings of this study. First, this study was conducted at a single university hospital in a metropolitan area and with a sample of patients who had an established relationship with their physician. It is possible that the patients in this study were better educated and less likely to be illiterate than those at local clinics or in rural areas. Additionally, the findings may be specific to the Japanese context. This may limit the generalizability of our findings, but it is noteworthy that a difference in the degree of HL was observed in this relatively homogenous sample of patients. Although sociocultural context and practice style may differ between Japanese and Western settings, the associations of HL with patient characteristics and with other measures in our study are generally similar to those reported in previous Western studies.

A second possible limitation of our study is that HL was measured based on a self-reported questionnaire. It has been indicated that individuals with reading problems are often ashamed and hide their inability to read (29), which might have led to an overestimation of the HL levels in this study. Moreover, our scale items may not cover the whole concept of the three levels of HL as defined by Nutbeam (8). Furthermore, because standard measures of functional HL such as Short Test of Functional Health Literacy in Adults (S-TOFHLA) or Rapid Estimate of Adult Literacy in Medicine (REALM-R) were not available in Japanese, we were unable to examine the relationship between our HL scale and these measures, an issue that should be explored in a future study with an English-speaking population. Further revisions and validation of our HL scales, along with an examination of their predictive validity and responsiveness to change, might enhance their utility.

Despite these limitations, this study suggests that HL is a measurable and important concept in considering education for patients with chronic diseases such as diabetes. Beyond previous measures focusing solely on functional HL, our scales included three levels of HL, each of which might have different effects on patient outcomes. Our scales also proved to be easy to administer in a clinical setting.

There has been a major concern that much of the diabetes-related health information requires literacy skills that are well above the population average, rendering them of little value to patients with limited literacy (30). Previous studies reported a greater benefit of face-to-face disease management programs for patients with limited literacy (13,20), which might suggest that those patients were less capable of seeking relevant information from other sources and applying it to the management of their disease compared with patients having higher HL. Physicians ought to be aware of these types of differences among patients and adopt flexible strategies of patient education that identify and accommodate their patients’ needs and abilities in managing disease. Exploring the functional, communicative, and critical HL levels of patients may help physicians better understand their patients’ potential barriers to engaging in self-management of disease and health-promoting behaviors. Furthermore, because the characteristics of the health system significantly contribute to poor communication, especially for patients with limited HL (22), it is necessary to consider changing the systems for providing health care and health information in a way that could reduce HL demands for such patients.

This study was supported by a Grant-in-Aid for Young Scientists from the Japanese Ministry of Education, Culture, Sports, Science, and Technology.

We thank Drs. Tamio Teramoto, Shin Fujimori, Makoto Kinoshita, and Toshikazu Yamanouchi at Teikyo University School of Medicine for their cooperation in conducting the survey and Mayuko Yoshikawa for helping with the data collection.