Background Over the last two decades, pharmaceutical intervention for the treatment of type 2 diabetes has expanded. Studies over this same time demonstrated the benefits of tight glycemic control. Unfortunately, despite the availability of novel therapies, glycemic control remains problematic. Nonpharmacologic interventions need to be explored, including patient empowerment. Improving patient knowledge of diabetes may ultimately improve glycemic control. To test this hypothesis, we compared patients' diabetes knowledge with their glycemic control.
Methods The Michigan Diabetes Knowledge Test, designed by the University of Michigan, was administered to patients with type 2 diabetes at three University of New Mexico primary care clinics. Patient records were reviewed. The most recent hemoglobin A1c (HbA1c) value was recorded. The data were analyzed using linear regression analysis.
Results Seventy-seven patients completed surveys and had HbA1c values available. Only questions 1 to 14 of the 23-question survey were used because they pertained specifically to type 2 diabetes. HbA1c was inversely correlated with the number of questions answered correctly on the test (r = -.337, p < .003). Using “all subsets” regression, a correct response to questions 1, 3, and 9 specifically correlated with a lower HbA1c (p < .0001).
Conclusions These results demonstrate that an inverse linear relationship exists between performance on this diabetes test and HbA1c values. Improvement in patient knowledge of diabetes and the importance of treatment may indeed improve glycemic control and ultimately decrease complications. Studies aimed at empowering patients with disease knowledge may help control the ramifications of the growing diabetes epidemic.
- diabetes mellitus
- glycemic control
- patient education
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Type 2 diabetes has reached epidemic proportions, with over 17 million people in the United States alone being affected. Type 2 diabetes is the underlying or contributing cause of approximately 323,000 deaths in the United States annually.1The percentage of health care dollars used to treat this disorder is staggering. Approximately 6% of the population has diabetes, but over 15% of total US health care expenditures go to treat diabetes.2The complications drive up the cost of treating diabetes, including vascular disease, end-stage renal disease, blindness, and amputations.2Development and/or progression of diabetic complications can be decreased with improved glycemic control.3-5
Unfortunately, the data generated from clinical trials confirming the benefits of improved control have not been effectively translated to routine patient care.6,7Despite having an increasing array of medications to treat diabetes, glycemic control has been difficult to attain.8There are several plausible explanations to explain the barriers to improved glycemic control. Primary health care providers may not be aware of the research findings demonstrating the benefits of improved glucose control, or they may not have the time and tools to use the information generated from these studies. Patients may not be compliant with treatment plans for a variety of reasons, including cost, complicated regimens, and side effects. Additionally, patients may not understand the long-term ramifications of poor glycemic control.8,9,10To test this latter hypothesis, we performed the following study to determine if patients' knowledge of diabetes had any relationship to their overall glycemic control.
Subjects with type 2 diabetes from three University of New Mexico-affiliated primary care practices were evaluated. Diabetes knowledge was measured using the Michigan Diabetes Knowledge Test (J. T. Fitzgerald, University of Michigan, Ann Arbor, MI). This test, developed at the University of Michigan, has been validated.11It consists of 23 questions. Only questions 1 to 14 were examined because the remainder of the questions pertain to type 1 diabetes. Potential participants with type 2 diabetes were asked by the triage staff at each of these practice sites to participate in the study. If they agreed, they completed the Diabetes Knowledge Test. The nurse or physician then recorded the subject's most recent hemoglobin A1c (HbA1c) value at the top of the survey. Surveys were discarded if they did not include the HbA1c value. Any question that was left unanswered was marked as incorrect for the purpose of data analysis.
A regression analysis of HbA1c values was performed with the total number of correct answers of the 14 questions as the predictor variable. Stepwise regression was performed to determine individual contributions of the questions, and the resulting “best” model was verified using “all subsets” multiple regression. A subscale was defined as the sum of correct answers for the subset of questions selected by the stepwise regression and compared with the “best” model in terms of the multiple correlation R 2.
All data were analyzed using SAS software, version 6.1 (SAS institute, Cary, NC).
Seventy-seven patients completed the diabetes knowledge tests and had HbA1c data available. The average number of questions answered correctly was 8.5 ± 2.3 of 14. The mean HbA1c value was 8.05 ± 1.6 (reference range 4-5.4%). Regression analysis demonstrated an inverse correlation between the number of correct responses on the test and HbA1c values (r = -.337, p = .003) (Figure 1). There were no significant differences between the three clinics in performance on the test or in HbA1c values. Additionally, there were no significant differences by gender.
We further examined if a correct response to specific questions on the test correlated with a lower HbA1c. Using stepwise regression, a correct response to questions 1, 3, and 9 did correlate with a lower HbA1c (Table 1). This model was verified by all subsets regression. Forming a subscale, these three questions performed as well as the “best multivariate model” and slightly better than the univariate model using all 14 questions (see Table 1).
We evaluated patient knowledge of type 2 diabetes using the Michigan Diabetes Knowledge Test, a validated and reliable test.11Our results suggest that an inverse correlation exists between patients' diabetes knowledge and their HbA1c. Unfortunately, previous studies have found that patients with diabetes have poor insight into their disease, regardless of socioeconomic or education status.12,13Fortunately, patient education tools have been developed that do improve self-care behaviors.13Additionally, studies have demonstrated that a patient educational intervention results in a decrease in HbA1c. Deichmann and colleagues found that the median HbA1c value fell from 8.5 to 7.8% (p < .006) following a patient education intervention.14These studies corroborate our findings that glycemic control is improved with improved patient knowledge of diabetes.
These results can be used in several ways. Primary and specialty providers have educational tools in their practices, and patient education should be part of all diabetes treatment plans. Counseling should be reiterated any time that escalation of medication is being discussed to reinforce the role of medications and to revisit a patient's diabetes knowledge. Patient empowerment is a strong tool and may result in significant improvements in glycemic control. Patient empowerment is attained through education and patients' improved knowledge of their disease processes. Additionally, improving glycemic control through patient education may decrease the economic impact of diabetes.15
Second, knowledge tests such as the Michigan Diabetes Knowledge Test could be used as a focal point for education. The diabetes team could look at the questions answered incorrectly and use those as the education focal point.
Interestingly, when we analyzed specific questions that better correlated with a lower HbA1c, questions 1, 3, and 9 were identified as the best multivariate model by stepwise regression, with nearly equal regression coefficients as the full 14-question test. Comparing this subscale with the univariate model using all 14 questions, it appears that a shorter questionnaire may be as good as the longer test in assessing diabetes knowledge. Questions 1, 3, and 9 asked about diabetes knowledge of diet and exercise. These are indeed the areas in which patients can intervene.
There were several limitations to our study. Selection bias may have occurred because there was no randomization process. Patients chose whether to participate. If patients feel that they might not do well on the test, they might be reluctant to participate. This, however, does not appear to be the case in our study because the range of incorrect responses was wide, from 1 to 12 of 14, with a mean of 5.5. Because only patients who came to the clinics were surveyed, these results may not be applicable to the general population, particularly rural, underserved minority patients with diabetes. The Michigan Diabetes Knowledge test was provided only in English. As our overall population consists of approximately 10% Spanish-only speakers, we may have introduced further bias. Additionally, we did not collect demographic data. Owing to the diversity of our population, the data could have been skewed by age, sex, and socioeconomic factors. However, this study consists of three heterogeneous clinic populations, and no difference was seen between them, suggesting that the results were consistent across various ethnic and socioeconomic groups.
In conclusion, we have demonstrated that diabetes knowledge, as ascertained by the Michigan Diabetes Knowledge Test, is associated with better glycemic control. Empowering diabetic patients with knowledge of their disease process may help combat this financially draining epidemic.
We would like to thank the Michigan Diabetes Research and Training Center for use of the Michigan Diabetes Knowledge Test.