The Prevalence and Pattern of Complementary and Alternative Medicine Use in Individuals With Diabetes

We combined data from the household and medical conditions components of the 1996 MEPS to determine the prevalence and pattern of use of alternative care in individuals with diabetes. The household component of MEPS is a survey of the U.S. civilian, noninstitutionalized population, drawn from the National Health Interview Survey sample with oversampling of Hispanics and blacks. This survey, sponsored by the Agency for Health Care Research and Quality, provides national estimates of health care use, health conditions, health status, insurance coverage, and access to care (10). In 1996, questions on prevalence, pattern, and costs associated with CAM were incorporated into the survey. A total of ∼21,571 individuals were surveyed, and the overall response rate for 1996 was 77.7%.

For the medical conditions component of MEPS, interviewers recorded verbatim the medical conditions and procedures as reported by the respondents. Then, professional coders used the verbatim text to assign fully specified 1996 International Classification of Diseases, 9th edition, Clinical Modification (ICD-9-CM) codes, including medical conditions and V codes. The error rate for any coders did not exceed 2.5% on verification. Assigned codes were verified by contacting medical providers and pharmacies that were identified by the respondents during the interview. To protect the confidentiality of respondents, fully specified ICD-9-CM codes were collapsed to three digits. For example, the ICD-9-CM code 250 (diabetes) represented diabetes, high blood glucose, juvenile diabetes, and adult-onset diabetes or diabetes neuropathy (11).

The MEPS defined CAM as “approaches to health care that are different from those typically practiced by medical doctors in the U.S.” This definition, which is similar to the NCCAM definition of CAM, included acupuncture, nutritional advice or lifestyle diets, massage therapy, herbal remedies, biofeedback, meditation, and imagery or relaxation techniques. Other treatments included in this definition of CAM were homeopathic treatment, spiritual healing or prayer, hypnosis, and traditional therapies such as Chinese, Ayurvedic, and Native American medicine. A card with the list above was presented to respondents. Then, the interviewer read the following statement verbatim to each respondent: “In order to get as complete a picture as possible of all sources of health care, we would also like to ask about the use of other forms of health care, including treatments you may have previously told me about, such as the treatments shown on this card. Frequently, this type of care is referred to as “complementary or alternative care.” During the calendar year 1996, did you consult someone who provides these types of treatment?” Individuals who responded affirmatively were asked to indicate the specific treatment, the type of CAM providers visited, and the number of visits to such providers. Respondents were also asked to indicate the cost of CAM visits, percentage of CAM costs covered by health insurance, and out-of-pocket costs for CAM.

Variables of the study participants included age, sex, race/ethnicity, marital status, educational level, health insurance status, and income status. The total household income and the number of individuals in each household were used to compute poverty levels based on federal guidelines. Household incomes were subsequently reported as a percentage of the federal poverty level. We classified households with combined incomes <125% of the federal poverty level as “poor” and households with incomes ≥125% of federal poverty guidelines as “not poor.” Self-reported physical and mental health of respondents were classified into two categories: excellent, very good, and good were combined into one category, and fair and poor were combined to form the second category. Further details on technical and programming information are available (10,11).

The MEPS a priori designated certain conditions as priority conditions due to prevalence, expense, or relevance to policy. We selected certain chronic comorbid conditions to compare with diabetes, including cancer, chronic obstructive pulmonary disease (COPD), hypertension, and ischemic heart disease. These conditions were selected because prior studies had indicated that individuals with these conditions had higher CAM use than the general population (6,7). Consequently, we created three categories of comorbidity, depending on the number of chronic medical conditions. Individuals with diabetes only were put in a separate category. Individuals with diabetes and cancer, COPD, hypertension, or ischemic heart disease were put into a second category, whereas individuals with diabetes and two or more of these conditions were put into a third category for analyses.

We used SAS statistical software (SAS Institute, Cary, NC) (12) and SUDAAN statistical software (Research Triangle Institute, Research Triangle Park, NC) (13) for statistical analyses to generate variance estimates that accounted for the complex sampling design of the MEPS. In addition, SUDAAN was used to generate population estimates to make them applicable to the civilian noninstitutionalized adult population with diabetes in the U.S. We performed three statistical analyses. First, estimates of CAM use were compared in individuals with and without diabetes, cancer, COPD, hypertension, and ischemic heart disease in the general U.S. population. Then, the estimates of CAM use in individuals with diabetes were compared with those in individuals with cancer, COPD, hypertension, and ischemic heart disease using χ² statistics. Second, we used χ² and Student’s t test statistics to compare the type of CAM used, the type of CAM provider seen, the pattern of use, and the costs associated with CAM use. Third, multiple logistic regression was used to determine factors that were independently associated with CAM use in the general population and in individuals with diabetes. In the multiple logistic regression models for the general population, whether CAM was used in 1996 was entered as a dichotomous dependent variable; independent variables included age, sex, education, race/ethnicity, marital status, and census region. Other independent variables included perceived physical health, perceived mental health, health insurance coverage, and household income as percentage of federal poverty level, employment, and number of comorbid conditions. Similarly, in the model for individuals with diabetes, the use of CAM was entered as the dependent variable. However, due to the small sample size of CAM users, we restricted the independent variables to age, sex, race, education, household income, and number of comorbid conditions. The Institutional Review Board of our institution approved the study.

Individuals with diabetes (8 vs. 5%, P = 0.0067), cancer (9 vs. 5%, P = 0.0007), and hypertension (6 vs. 5%, P = 0.0277) were more likely to use CAM than their counterparts with no chronic medical conditions in the general population. Patients with ischemic heart disease and COPD did not differ significantly in their use of CAM when compared with individuals without both chronic conditions. The estimates of CAM users among individuals with diabetes did not differ significantly from that in individuals with cancer, hypertension, ischemic heart disease, or COPD. Extrapolating to the U.S. population in 1996 (∼265,639,034 civilian noninstitutionalized individuals), 12,963,185 individuals reported use of CAM. Similarly, among an estimated 9,652,897 individuals with diabetes in 1996, 759,683 reported use of CAM.

The sample characteristics of individuals with diabetes are presented in Table 1. Approximately 44% were aged ≥65 years, 55% were women, 72% were neither black nor Hispanic, and 59% were married. Of these individuals, 61% had at least a high school education, 35% were employed during the year of study, and only 8% reported not having health insurance coverage. A total of 64% of individuals had diabetes in addition to at least one other chronic medical condition. Of these individuals with diabetes, 46% had hypertension, 28% had ischemic heart disease, 10% had some type of cancer, and 7% had COPD.

The pattern of use of CAM and the costs associated with CAM use in individuals with diabetes are presented in Table 2. The five most commonly used CAM therapies in individuals with diabetes, in order of importance, were nutritional advice and lifestyle diets, spiritual healing, herbal remedies, massage therapy, and meditation training. On the contrary, the most commonly used CAM therapies in individuals without diabetes, in order of importance, were massage therapy, herbal remedies, spiritual healing, nutritional advice and lifestyle diets, and meditation training.

Both groups received treatment from a variety of CAM providers, including physicians and nurses. Individuals with diabetes were more likely to see a nurse for CAM than individuals without diabetes (13 vs. 2%, P = 0.0275). In contrast, individuals with diabetes were less likely to see a massage therapist (16 vs. 32%, P = 0.0099) and an herbalist (6 vs. 14%, P = 0.0165) than individuals without diabetes. Individuals with diabetes were more likely to report the use of CAM for a specific health problem (81 vs. 63%, P = 0.0075). Similarly, individuals with diabetes were more likely to report discussing the use of CAM with their regular physician (57 vs. 29%, P = 0.0024) and being referred by their physician to a CAM provider (43 vs. 10%, P = 0.001).

The mean number of visits to CAM providers were not significantly different in individuals with and without diabetes (9 ± 2 vs. 13 ± 2, P = 0.1781). Similarly, the mean amount in dollars spent per person for CAM was not different in both groups ($414 ± 269 vs. $236 ± 26, P = 0.5106). The percentage of CAM costs covered by health insurance was different between both groups, but this difference did not achieve statistical significance (27 vs. 12%, P = 0.0549).

The independent predictors of CAM use in both the general population and in individuals with diabetes are shown in Table 3. In the general population, women, those with high school education or more, those in poor physical health, those who were employed, and those with diabetes or diabetes in combination with other chronic medical conditions were most likely to use CAM. In addition, Hispanics and blacks were less likely to use CAM than whites and individuals of other races or ethnicities. Individuals who lived in the West were more likely to use CAM than those living in the Northeast, Midwest, or South. Finally, individuals with private health insurance were less likely to use CAM than those who were uninsured.

Among individuals with diabetes, those aged ≥65 years were three times more likely to use CAM than those aged <65 years. Individuals with high school education and higher were 2.4 times more likely to use CAM than those who had not completed high school. Sex, race/ethnicity, household income, and comorbidity were not significant predictors of CAM use in individuals with diabetes.

This study has shown that although individuals with diabetes are more likely to use CAM than individuals without diabetes, estimates of CAM use in individuals with diabetes are comparable to those in individuals with other common chronic medical conditions. In addition, using nationally representative data, this study has provided preliminary estimates on pattern of use, associated costs, and factors associated with CAM use in individuals with diabetes.

The estimates of CAM use in this study differ considerably from those of Eisenberg et al. (4,6) but closely approximate those from two other studies using data from the MEPS (7,9). There are three possible reasons for these differences. One reason is the heterogeneity of CAM practices (5,14), which means that estimates of CAM use will change depending on what is included or excluded in the definition of CAM. The second reason is the oversampling of Hispanics and blacks and the use of direct household interview rather than telephone interviews in MEPS. This approach may have increased the proportion of minorities and individuals of lower socioeconomic status, who have been shown to have lower usage of CAM (6, 7). The third reason is the difference in target population and study hypothesis across studies. This is particularly obvious in the MEPS sample. When Bausell et al. (7) focused on adults aged ≥18 years, they found that 9% of U.S. adults visited CAM providers. However, when Druss and Rosenheck (9) used the same data set with similar age cutoffs, but tested different hypotheses, they found that 6.5% of the U.S. population used both conventional and unconventional therapies. Both studies differ from our study, in which no age category was excluded and the emphasis was on CAM use as opposed to visits to CAM providers.

Patients with diabetes and other common chronic medical conditions were more likely to use CAM than individuals in the general population. This is not surprising because earlier studies have linked increased CAM use with the presence of chronic as opposed to acute or life-threatening medical conditions (4,8). However, the finding that diabetes is an independent predictor of CAM use in the general population is surprising, particularly because the presence of multiple chronic comorbid conditions did not explain the association between diabetes and CAM use. Adjusting for covariates, individuals with only diabetes were two times more likely to use CAM, whereas individuals with diabetes and additional chronic conditions were 1.8 times more likely to use CAM than the general population without chronic medical conditions. This observation will need to be explored in future studies.

Nutritional advice and lifestyle diet, spiritual healing, herbal remedies, massage, and meditation were the most frequently used CAM treatments among individuals with diabetes; this pattern approaches the pattern of CAM use in the general population (4,6,15,16). Although nutritional counseling and lifestyle modification are essential components of routine diabetes care, it is important to recognize that in this study, such advice/diets were obtained from CAM providers. The MEPS definition of CAM, which is similar to the NCCAM definition (5), implies that nutritional advice and lifestyle diets provided by CAM practitioners differ from conventional nutritional/dietary recommendations endorsed by diabetes educators or physicians.

Examples of nutritional advice and lifestyle diets offered by CAM practitioners include Ayuverdic diets, naturopathic or homeopathic nutrition/diets, and orthomolecular therapies such as magnesium, melatonin, or megadoses of vitamins (5). In addition, special diets, such as those proposed by Drs. Atkins, Ornish, Pritikin, and Weil, also qualify as CAM lifestyle nutrition/diets (5). It is unlikely that nutritional advice and lifestyle diets offered by CAM practitioners are consistent with the American Diabetes Association guidelines for dietary management of diabetes (17). In addition, it is currently unknown whether additional nutritional advice and lifestyle diets by CAM practitioners complement and reinforce American Diabetes Association guidelines or conflict with conventional dietary recommendations.

Spiritual healing in MEPS refers to healing by someone other than self, such as the clergy or a spiritualist, and differs from self-prayer. This differentiation is important because an earlier study (6) reported that although only 7% of the U.S. population reported spiritual healing by others, up to 35% of the U.S. population used self-prayer as a form of treatment. Spiritual healing was the second most common type of CAM used by individuals with diabetes (21%) and the third most frequently used CAM in individuals without diabetes (24%). This finding suggests that substantial percentages of the U.S. population believe and seek spiritual healing as a form of treatment.

It seems that the search for spiritual healing may be well founded, based on the results of a recent study (18). This systematic review of 23 trials and 2,774 patients found that prayer and distant healing yielded statistically significant treatment effects in 13 patients (57%), no effect over control interventions in 9 patients (39%), and a negative effect in 1 patient (4%). In contrast, the use of herbal remedies, which was reported by 20% of individuals with diabetes, has not been shown to improve glucose control and may even be harmful in individuals with diabetes (19,20).

There are some limitations to this study. A major limitation is the small number of individuals with diabetes that used CAM in the 1996 MEPS household sample. This limited the type of analysis that could be performed and the number of independent predictors of CAM entered into the multiple logistic regression models. However, the MEPS is the first nationally representative survey that provides detailed information about the use of CAM and has data on medical conditions. The MEPS offers an exceptional opportunity to provide baseline data on CAM use in individuals with diabetes in the U.S. noninstitutionalized civilian population.

A second limitation of this study is that MEPS based CAM use on visits to a practitioner and excluded treatments obtained by individuals without consultation with a CAM practitioner. This may decrease the estimates of CAM use in both comparison groups. Recall bias is another potential limitation. Studies have shown that self-reports are reliable for the diagnosis of diabetes (21,22), but no studies validate recall of visits to CAM practitioners. If recall was low, it was likely similarly low across the sample because there are no obvious reasons to expect differential reporting of CAM use. The fourth limitation is the ambiguous definition of CAM (5,14), which allows for estimates that are dependent on the inclusion and exclusion criteria used to define CAM. This limitation may interfere with the ability of researchers to compare findings across studies.

The major implication of this study is that individuals with diabetes seem to use CAM as a complement rather than as an alternative to conventional treatment. In this study, 57% of individuals with diabetes who used CAM discussed it with their regular physician and 43% were actually referred to CAM users by a physician. This is reassuring because it means that patients with diabetes are not abandoning conventional treatments, which have been rigorously tested, for unconventional treatments, which lack properly designed efficacy trials. On the other hand, it means that health care providers will need to acknowledge CAM use, learn to discuss CAM use with their patients, and be able to do so candidly and without prejudice.

In an editorial published in the Journal of the American Medical Association, Wayne Jonas said “alternative medicine is here to stay” and that the challenge of the health care community is to “separate the pearls from the mud” (23). The increase in prevalence of diabetes (1), the search by patients for holistic medicine, and the skepticism some patients have about the efficacy of modern medicine (24) make the appropriateness of this statement even more evident. Differentiating efficacious CAM treatments from bogus treatments will remain a daunting task, and the absence of properly designed and conducted efficacy trials for many treatments further complicates the problem (25). Nonetheless, to foster collaboration with patients, health care providers must respect patients’ wishes to use CAM and provide unbiased advice about CAM treatments to patients. In addition, health care providers will need to understand the benefits and limitations of currently available alternative treatments.

There are three important areas for future research. First, there is a need to replicate the findings from the 1996 MEPS in current surveys. Second, there is an urgent need to use rigorous research designs to establish the efficacy of several complementary and alternative treatments that are currently being used by individuals with diabetes. Third, future studies must determine the effectiveness of CAM use in typical clinical situations and the effect of CAM on the quality of life in individuals with diabetes.

This study was funded by grants (1P01HS1087-01 and 1K08HS11418-01 to L.E.E.) from the Agency for Health Care Research and Quality, Rockville, MD, and the Centers for Disease Control and Prevension (Grant U501CC417281-02 to L.E.E. and D.Z.).

The concents of this article reflect the personal opinions of the authors and do not represent the official opinion of either the Agency for Health Care Research and Quality or the Centers for Disease Control and Prevention.