Patients
We conducted a cross-sectional observational study using a self-administered questionnaire. First, we recruited participating clinics by mail from 141 family practices and internal medicine clinics (56 health cooperative clinics, 83 Min-Iren clinics, and 2 private clinics) across Japan. The health cooperative clinics and the Min-Iren clinics are supported by the Japanese Health and Welfare Co-operative Federation and the Japan Federation of Democratic Medical Institutions, respectively. There were no differences in the delivery of healthcare or the types of health insurance systems used at these clinics. We chose these organizations as study sites to recruit clinics located in a variety of prefectures (subnational jurisdictions in Japan) to avoid being confined to one locality. To participate in the study, primary care physicians, not diabetologists, had to care for patients with diabetes in the clinic. Among the 19 clinics that agreed to participate in the study, 2 clinics could not complete the research process. As a result, the questionnaires from 17 clinics were considered eligible (13 health cooperative clinics, 3 Min-Iren clinics, and 1 private clinic). The clinics were distributed across 12/47 prefectures.
At each clinic, a physician or a nurse explained the purpose of the study to each consecutive patient with type 2 diabetes who visited the clinic on the survey days. The physicians themselves selected the survey days, taking into account their workload and other factors on the day. Each patient met the following inclusion criteria: age ≥ 20 years and < 75 years old, had been diagnosed with type 2 diabetes ≥1 year ago, was seen regularly at the clinic (at least once every 2 months for the last year, since most patients with diabetes are seen at 4–8 week intervals in Japan [37]), and hemoglobin A1c (HbA1c) level ≥ 6.2 % within the last 3 months to confirm that the patient had required medical or lifestyle management for diabetes for at least the last few months. We excluded patients aged ≥ 75 years because the regimens and diabetes education for blood glucose control are usually less intensive for older patients than for younger patients. We also excluded patients who had infrequent contact with their physicians because we aimed to assess the impact of patient–physician communication on diabetes education. Compared with the US and other countries, the consultation time is generally shorter in Japan, as the mean office visit was 6 min long in Japan [38, 39]. Considering the shorter consultation time, at least one contact per 2 months for ≥ 1 year was considered the minimum requirement for building a patient–physician relationship in Japan. We also excluded patients with known or apparent cognitive dysfunction, such as those who had previously scored < 23 on the Mini-Mental State Examination (MMSE) [40]. We also excluded those who were unable to answer the questionnaire by themselves because the HL measurement used in this study required the patients to self-assess their ability to use health information.
The eligible patients were provided a questionnaire and were instructed to complete it by themselves at the clinic after the visit or to take it home and complete it later. By responding to the questionnaire, they were considered to have given consent to participate in the study. After completing the questionnaire, the patients mailed it to the data center. The patients were guaranteed the right to withdraw from the study at any time and that their usual care would not be affected by participating in the study. The study was conducted between March and July 2010. To increase the diversity of characteristics of the patients, up to 20 questionnaires were distributed at each clinic. The study protocol was reviewed and approved by the Tokyo Hokuto Health Cooperative Institutional Review Board.
Health literacy
We measured HL using scales developed in Japan to assess functional, communicative, and critical HL of patients with diabetes. Five, five, and four items were used to assess functional, communicative, and critical HL, respectively [25]. Each item was rated on a 4-point Likert scale, ranging from 1 (never) to 4 (often). In the functional HL subscale, the patients were asked to rate how often they needed someone’s help to read the instructions or leaflets from hospitals/pharmacies, for example. In the communicative HL subscale, the patients were asked to rate, for example, how often they had collected information from various sources since being diagnosed with diabetes. For the critical HL subscale, the patients were asked to rate, for example, how often they had considered the credibility of the information, or had checked whether the information was correct. The scores for functional HL were reversed, such that higher scores indicate higher HL. The scores for the items on each scale were summed and divided by the number of items for that scale to calculate the scale score (theoretical range 1–4). The functional, communicative, and critical HL scales were previously validated in Japanese patients with diabetes [25], and the internal consistency of each scale was adequate (Cronbach’s α = 0.85, 0.81, and 0.69, respectively). Higher scale scores indicate higher HL. There are no cut-off points for classification of adequate/inadequate HL.
Patient–physician communication
To assess how the patients perceived communication with their physician in terms of information sharing and the decision-making process, we asked the patients to rate the following five items related to communication: “I can ask the physician whatever I would like to know about my disease(s) and physical condition(s),” “the physician knows a lot about my daily life,” “the physician explains other treatment options well,” “the physician’s explanation is clear enough to me,” and “the physician respects my will and ideas at the time of decision-making.” We also assessed overall satisfaction with communication using a sixth item: “I am satisfied with the communication with the physician.” The six items were scored from 1 (I do not agree at all) to 7 (I agree very strongly). These items were not developed or validated as a scale, and the associations among each of six items were very high (r = 0.58–0.80, p < 0.001). Therefore, we selected one of the six patient-physician communication variables (“the physician’s explanation is clear enough to me”) as a representative factor in the main analysis because this item was thought to strongly reflect the perceived understandability of the physician’s communication. The other items were not included in the present analyses because they were did not necessarily reflect communication or did not always apply to daily care.
Understanding of diabetes care and self-efficacy for diabetes management
We examined understanding of diabetes care and self-efficacy for diabetes management to assess the adherence to diabetes self-management. We did not use self-care behaviors because the required self-care behaviors may differ depending on the individual patient’s conditions and the goals of diabetes management. The understanding of diabetes care was assessed using an eight-item scale derived from the Diabetes Quality Improvement Project in a previous study [12]. This scale evaluates the respondent’s perceived understanding of diabetes care and self-management, how to care for feet, and what to do in the event of symptoms of low blood glucose, for example. Each item was scored on a 5-point Likert scale from 1 (I do not know at all) to 5 (I know very well). The scale score was calculated as the mean value of all eight items, with possible scores ranging from 1 to 5. Higher scores imply greater understanding of diabetes care. To use this scale, we first translated it into Japanese and confirmed linguistic validity by back-translation. Reliability and validity were examined in a pilot study of 17 patients. Principal component analysis was conducted, and the proportion of the first principal component was 59.4%, suggesting a single dimension. The test–retest reliability was adequate, with moderate agreement between the test–retest scores (r = 0.87, p < 0.001). The internal consistency of the scale was also adequate (Cronbach’s α = 0.89).
Self-efficacy for diabetes management was assessed with a four-item scale of self-care ability in the Diabetes Care Profile, as used in a previous study [25]. The scale evaluates the respondent’s confidence in blood glucose control, weight control, diabetes self-management, and coping with emotions when living with diabetes. Each item was scored from 1 (I do not think I can) to 4 (I strongly think I can). The scale score was calculated as the mean value of all four items, with possible scores ranging from 1 to 4, and higher scores implying greater self-efficacy. The internal consistency was adequate (Cronbach’s α = 0.80).
Sociodemographic and clinical variables
We obtained demographic data for each patient from the self-reported questionnaire. Clinical data were obtained from separate medical reports provided by the physicians that were matched with the patient questionnaires using an ID number. Each patient reported their height, weight, time since the diagnosis of diabetes, the number of years seen by his/her physician, educational attainment, marital status, occupational status, self-rated economic status, and access to and use of the Internet. We measured the patient’s perceived level of social support using a short-form version of the Multidimensional Scale of Perceived Social Support [41, 42]. This is a seven-item scale in which each item is scored from 1 (very strongly disagree) to 7 (very strongly agree). The Japanese version was previously validated [42] and shows adequate internal consistency (Cronbach’s α = 0.89).
As polypharmacy and complex prescription regimens are known to influence treatment adherence [43], and because the status of diabetic complications is associated with HL [25], we also obtained the following clinical information. The physicians reported the age, sex, number of prescribed glucose-lowering drugs, whether or not the patient was on insulin, the most recently measured HbA1c, the status of diabetic complications, and whether the complications were managed with or without anti-hypertensive drugs and/or anti-lipid drugs. Diabetic complications included retinopathy, nephropathy, neuropathy, cardiovascular disease, and stroke. These reports were prepared after the patients’ visits. The clinical data for the patients who did not complete the questionnaire were not used in this study.
Statistical analysis
Pearson’s correlation coefficient was first calculated to determine the relationship between HL and patient–physician communication. We next conducted bivariate analyses to examine the associations between patient–physician communication and the patient’s understanding of diabetes care and self-efficacy. Then, hierarchical linear regression analyses were performed to determine the associations between the three HL scales with the understanding of diabetes care and self-efficacy as outcome variables. Model 1 consisted of bivariate analyses. In model 2, we added sociodemographic and clinical variables that were correlated with the HL scales, the clarity of the physician’s explanation, or with the outcome variables, as possible confounders. Finally, in model 3, we added the clarity of the physician’s explanation to model 2.
To identify the sociodemographic and clinical variables to be included as covariates, we used bivariate analyses to determine their correlations with HL, the clarity of the physician’s explanation, or outcome variables. Age, educational attainment, self-rated economic status, Internet use, social support, time since the diagnosis of diabetes, insulin use, and whether or not they had diabetic complications were correlated with at least one of the three subscales of HL. Internet use and social support were also correlated with the clarity of the physician’s explanation. The number of oral glucose-lowering drugs was weakly correlated with the patient’s understanding of diabetes care (p = 0.08). We included all the variables mentioned above, and sex, as covariates in models 2 and 3.
The variables were entered as follows: age (continuous), sex (male or female), educational attainment (middle school, high school, vocational school/2-year college, or university or higher), self-rated economic status (lower, middle, or upper), Internet use (yes or no), social support (continuous), number of years since the diagnosis of diabetes (continuous), number of oral glucose-lowering drugs (one, two, or three or more), insulin use (yes or no), and diabetic complications (none or any). There was no evidence of multicollinearity among any of the included variables. All statistical analyses were performed using PASW version 18.0. Statistical significance was set at p < 0.05.