Association of Numeracy and Diabetes Control

Context Self-management of diabetes often involves interpretation of quantitative information and performing calculations. Poor numeracy skills could have an effect on diabetes management. Contribution The authors measured diabetes managementrelated numeracy skills and hemoglobin A1c in a sample of 398 diabetic patients from primary care clinics and diabetes clinics at 3 medical centers. Many could not calculate the calories in a bag of potato chips or recognize a very low blood glucose value. Hemoglobin A1c levels were slightly worse in patients with poor diabetes-related numeracy, more so if young, less literate, or with recent diagnosis. Implication Poor diabetes-related numeracy is common but is weakly related to glycemic control. The Editors Numeracy, an important component of literacy (1), can be defined as the ability to understand and use numbers in daily life (2). In 2003, the National Assessment of Adult Literacy found that approximately 90 million Americans have basic or below-basic literacy skills and more than 110 million people have basic or poor quantitative skills (3). Although there is a correlation between prose or print literacy and numeracy (4), many patients may have adequate literacy but poor quantitative skills (2) or become too anxious in situations that require application of quantitative skills (5, 6). Health care providers often present abstract or complex numerical information related to chronic illness for the patient to manipulate, interpret, and act on (7). Health-related numeracy involves not only interpreting risk (811) but also understanding measurement, estimation, time, logic, and multistep operations and identifying which math skills need to be applied to solve problems (12). Some small studies have suggested an association between numeracy skills and anticoagulation control (13), hospitalization in patients with asthma (14), and inconsistent interpretation of breast cancer risk (10, 15). In patients with diabetes, low health literacy is common and has been associated with less knowledge and understanding of diabetes self-management, greater reported risk for retinopathy, and possibly worse glycemic control (1622). In a study of Medicare patients with and without diabetes, 23% of those with adequate literacy could not determine whether a blood glucose value was within a target range (23). However, the relationship between numeracy skills and diabetes outcomes is unknown. Numeracy is important for many diabetes self-management skills, including the ability to interpret results of blood glucose self-monitoring, determine insulin requirements, count carbohydrates, and perform other daily activities of diabetes care. These require not only basic math skills, but also the ability to apply those math skills in the context of diabetes care (that is, diabetes-related numeracy). We explored the association between diabetes-related numeracy and perceived self-efficacy, self-management activities, and clinical measures in patients with diabetes. We hypothesized that poor diabetes-related numeracy would be associated with lower perceived self-efficacy, lower participation in self-management activities, and worse glycemic control. Methods Setting and Study Participants From March 2004 to November 2005, we enrolled study participants from 2 primary care clinics and 2 endocrinology clinics located at 3 medical centers. Study participants had type 1 or type 2 diabetes mellitus, were between 18 to 85 years of age, and were English speakers. Exclusion criteria included a previous diagnosis of dementia, psychosis, or blindness. In addition, we excluded patients with a corrected visual acuity of 20/50 or worse when tested with a Rosenbaum Screener (Prestige Medical, Northridge, California). Participants received $20 for participation. The Veterans Affairs Tennessee Valley Healthcare System Research and Development Committee, Nashville, Tennessee, and institutional review boards at Vanderbilt University Medical Center, Nashville, and the University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, approved this study. We obtained written consent from all participants. Measurements We collected demographic and clinical information about patients' diabetes, including self-management behaviors, use of insulin, and most recent hemoglobin A1c level, by conducting patient interviews and reviewing electronic medical records. For hemoglobin A1c levels, 96% were obtained within 6 months of the patient evaluation, and the median time between hemoglobin A1c testing and evaluation was 15 days (range, 0 to 323 days). We also obtained patient height and weight from the medical record to calculate body mass index. We assessed literacy by using the Rapid Estimate of Adult Literacy in Medicine (REALM), a validated measure of readability that correlates with reading comprehension (24, 25). If the patient scored less than a sixth-grade reading level, then the remainder of the instruments was administered orally to ensure that the patient understood the content of the survey questions. All participants were given the option of oral administration if desired. We measured general numeracy with the math section of the Wide Range Achievement Test, 3rd edition (WRAT-3), a validated instrument that evaluates calculation skills (26). We measured diabetes-related numeracy skills with the Diabetes Numeracy Test (DNT) (available at www.mc.vanderbilt.edu/diabetes/drtc/preventionandcontrol/tools.php) (27). Experts in diabetes, health literacy, and numeracy developed the 43-item DNT. It tests a wide range of applied quantitative skills fundamental to the daily care of diabetes. In contrast to the WRAT-3, which primarily measures calculation skills, the DNT is composed of word problems assessing calculation; interpretation of tables, graphs, or figures; and selection of necessary math functions to solve diabetes-specific problems. The DNT is valid and internally reliable, with a KruderRichardson coefficient (28) of 0.95 (27). We provided participants with basic-function calculators with a large number screen and large buttons, and participants had no time limit to complete the DNT. We measured diabetes self-management activities by patient self-report and with the Summary of Diabetes Self-Care Activities scale (29). We measured patient knowledge of diabetes with the Diabetes Knowledge Test (30) and perceived self-efficacy of diabetes self-management behaviors with the Perceived Diabetes Self-Management Scale (31). The estimated time to complete all surveys was 30 to 90 minutes. Statistical Analysis We performed analyses by using STATA, version 8.0 (Stata, College Park, Texas), and adjusted models by using R, version 2.6.0 (www.r-project.org). Patients' characteristics and responses were described using medians and interquartile ranges (IQRs) for continuous variables and percentages for categorical variables. Patient performance on the DNT was calculated as a percentage of questions answered correctly (score range, 0% to 100%). Missing values were considered to be incorrect responses on the DNT. We characterized patients by quartile of DNT score. We analyzed the associations between patient characteristics, including hemoglobin A1c level, and DNT score by using the Cuzick nonparametric test for trend across the DNT score quartiles (32). We examined literacy and general numeracy, measured by REALM and WRAT-3, as continuous measurements and also categorized them a priori as representing either less than ninth-grade literacy or math level or greater than or equal to ninth-grade literacy or math level (25, 26). We determined the total DNT score and proportion of correct responses to individual questions of the DNT stratified by patient literacy and numeracy levels. We performed comparisons of the proportion of correct responses between the stratified groups by using chi-square or Wilcoxon rank-sum tests, as appropriate. To describe the adjusted relationship between DNT score and hemoglobin A1c level, we applied generalized least-squares methods with restricted maximum likelihood procedures (33) and included in the model variables that we believed a priori to have an association with DNT score and glycemic control. Covariates were age, sex, race, annual income (<$20000, $20000 to $40000, $40000 to $60000, or >$60000), type of diabetes, years since diabetes diagnosis, and clinic site. We performed log transformation of hemoglobin A1c level because of its nonnormal distribution, but presented the hemoglobin A1c back-transformed for ease of interpretation. We performed regression diagnostics to ensure normality of residuals. We excluded patients with missing values from the analysis. However, an analysis that included imputations for missing values yielded similar results to those presented. We chose not to include education, literacy, and general numeracy skill level in our models because of high colinearity between these variables and DNT score and because of the potential for overadjustment (34). We evaluated for interaction between DNT score and covariates, and this was not statistically significant. Role of the Funding Source This study was funded with support from the American Diabetes Association, Pfizer, and the Vanderbilt Diabetes Research and Training Center. The funding sources had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript. Results Participant Characteristics From March 2004 to November 2005, we evaluated 615 patients for possible enrollment. Of these, 191 declined to participate and 18 were excluded because of poor vision (n= 7), age younger than 18 years or older than 85 years (n= 4), lack of English fluency (n= 2), or other exclusion criteria (n= 5). Of 406 patients who consented, 398 (98%) completed the study. Table 1 lists the characteristics of the study sample. The median a