The use of insulin declines as patients live farther from their source of care: results of a survey of adults with type 2 diabetes

In spite of the widespread availability of efficacious therapies, adequate management of diabetes remains an elusive goal [1, 2]. Barriers to optimum control of diabetes in community settings include poor practice management and organization [3–5], poor adherence to medical advice by patients [6], comorbid illnesses [7], and lack of health insurance [8]. Although appropriate use of insulin can lead to better glycemic control, as well as a better self-reported quality of life [9], many patients and providers underutilize insulin for a variety of reasons, including clinical inertia [10] and medical concerns [11].

Geography has long been recognized as a determinant of health status. For example, cerebrovascular disease clusters in the "stroke belt" of the southeastern United States [12]. Death rates among European elders vary two-fold depending on place [13]. Americans are more likely to be disabled in certain parts of the country than others [14, 15].

Recently, use of Geographic Information Systems has led to insight about spatial prevalence and treatment of many conditions including diabetes [16, 17]. Geographic patterns of health care utilization have also revealed important variations in treatment patterns of various conditions among even small geographic areas [18, 19]. Mammography [20], knee joint replacement [21], treatment of colorectal cancer [22], and lower extremity amputation [23], to name a few examples, have all been shown to vary importantly by geographic location of the population independent of clinical status at the individual level.

To some degree, these variations may be due to variation among the providers (both individual and institutional) whose various practice styles can introduce patterns of care that vary at the geographic level [24]. However, in some cases there is evidence that local geography can, itself, serve as a barrier to care. Travel burden, measured as driving distance or access to transportation, has been linked to health care utilization [25–27]. In diabetes care specifically, increased travel burden, as measured by driving distance, has been associated with poorer glycemic control [28].

Modern Geographic Information Systems (GIS) make it relatively inexpensive to determine the most direct driving route between two known addresses. For rural areas, where routes are often circuitous, driving distance has been shown to be a more accurate marker of travel burden than straight-line measures [29].

In this study, we examined the relationship between insulin usage and the driving distance from a patient's home to their site of primary care. We sought to evaluate the hypothesis that insulin use is associated with driving distance and to evaluate whether any observed relationship was due to confounding by social, economic or clinical factors.

This study was part of a larger project, the Vermont Diabetes Information System (VDIS), a cluster-randomized trial of a laboratory-based diabetes decision support system in a region-wide sample of 7406 adults with diabetes from 73 primary care practices [30]. VDIS delivers improved laboratory information services to patients and providers to support and enhance their decision making about diabetes treatment and monitoring. The data presented here were collected at baseline before the VDIS intervention began. A field survey targeted at a sub-sample of subjects was designed to provide a better understanding of the non-laboratory features of the patients. Subjects were selected at random from the patients participating in the VDIS trial and invited by phone to participate in an in-home interview. Patient names were randomly sorted and patients contacted by telephone until a sample of approximately 15% of the patients from each practice agreed to an interview. If a subject could not be contacted, another name was chosen from the list. Therefore, many subjects were called only once or twice before being excluded from the study. We attempted to contact 4,209 patients and reached 1,576 (37%). Of these, 1,006 (64%) agreed to be interviewed.

Subjects who agreed were mailed a questionnaire and were scheduled for an interview by a trained field interviewer. The questionnaire covered multiple domains including demographics and medication usage. During the visit, the interviewer reviewed any missing or ambiguous questionnaire items. If necessary, the interviewer read the questions aloud for subjects and recorded their responses for them. Then the interviewer measured the subject's height, weight and blood pressure. The subjects were asked to produce every medicine they had used in the past month for the interviewer to record. The interviews took place during the baseline phase of the study before any interventions were in place. The protocol was approved by the institutional review board of the University of Vermont.

Income was recorded in seven ordered self-reported categories from less than $15,000 per year to $100,000 per year or more. For analysis, we collapsed these into two categories: above and below $30,000 per year. We collapsed self-reported education into High School Graduate or not. We collapsed self-reported race and ethnicity into two categories: Non-Hispanic white and all others. Marital status was collapsed into two categories: married (including living as married) and all others (single, widowed, divorced or separated). We recorded the presence or absence of five categories of health insurance: private (commercial indemnity or health maintenance organizations), Medicare, Medicaid, military (including active duty or Veterans benefits) and uninsured. Subjects were asked to recall the number of visits they made to their primary care provider (PCP) in the last month and whether they had seen an endocrinologist in the last year. Glycemic control was estimated by the latest glycosolated hemoglobin A1C assay performed at the local clinical laboratory.

We included all interviewed subjects who provided complete data for all variables. Subjects were excluded from analysis as potentially having type 1 diabetes if they used insulin without any oral hypoglycemic agents and were diagnosed with diabetes before age 30. We compared the characteristics of the included subjects to all interviewed but excluded subjects using t-tests, χ² tests and Fisher exact tests.

The shortest driving distance from the patients' homes to their site of care was calculated in kilometers using ArcView 3.3 by Environmental Systems Research Institute, Inc., and a geographic data set of roads purchased from TeleAtlas, Inc. Locations were geocoded by matching addresses to the street layer, which is coded with address ranges. Driving distance was defined as the shortest distance along roads and highways. We also categorized subjects dichotomously as living less than or greater than 10 km from their PCP (10 km = 6.2 miles).

We compared the driving distances of insulin users and non-users with the Wilcoxon rank-sum test and the χ² test. We divided observations by their driving distance into four quartiles and compared the insulin use for each quartile. We tested the significance of the trend in insulin use across quartiles with a nonparametric extension of the Wilcoxon rank-sum test [31]. Because nonparametric procedures are not available, we used analysis of variance (ANOVA) to test for clustering of factors within providers' practices and expressed this phenomenon as the intra-provider (or intra-cluster) correlation coefficient (ICC) [32]. We used logistic regression to test the association between insulin use (present or absent) and each marker of travel burden (distance in kilometers and the dichotomous variable) by calculating odds ratios (OR) and their 95% confidence intervals (CI). We repeated the analysis controlling for age, sex, race, education, income, health insurance, body mass index, use of any oral hypoglycemic agents, glycemic control, and duration of diabetes in years. We adjusted the standard errors and significance values of the multivariate analyses for clustering within provider practices [33]. We used Stata 8.2 for all analyses.

The greater the driving distance for adults with diabetes to their source of primary care, the less likely they are to be using insulin. This relationship is not due to confounding by the social, demographic, or clinical factors measured here.

The mechanism for this relationship is unclear. Perhaps patients and physicians are concerned about the risks of insulin and are reluctant to use it if they feel the patient lives too far away from care for rescue in the event of hypoglycemia. We have no direct data on the attitudes of these decision makers in this regard.

Perhaps patients with more visits can be stepped through the various regimens more promptly and started on insulin sooner. In other words, travel burden might influence therapy through the frequency of medical contact. However, we found little relationship between distance and visit frequency. Furthermore, adjusting for frequency of visits (either to primary or specialty care) had little effect upon the relationship between distance and insulin use.

Possibly, providers who systematically avoid insulin in all their patients systematically attract patients from farther afield. This is unlikely given the low clustering of insulin use within providers and the minimal impact of correcting for clustering in the multivariate analyses. Although providers do systematically vary in the size of their catchment areas, they do not appear to have systematic tendencies to use or avoid insulin.

Previous findings suggest that patients who travel farther to primary care have poorer physiologic control [28]. Likewise, rural patients, who generally have higher travel burdens, tend to have poorer health status [36]. It is not known if these effects are completely explained by lower intensity of therapy. Nonetheless, it speaks to the need to reassess care for rural and exurban patients. Several possible approaches come to mind. Medical services, especially primary care, could be dispersed more widely. Although the economic and cultural barriers to placing physicians in underserved areas are formidable [37], greater use of home or worksite visits by traveling or circuit providers (physicians, nurses, physician assistants, etc.) could be attractive. Also, new technologies could replace or complement traditional face-to-face personal care. The telephone, facsimile, and the Internet have potential to alleviate the barriers associated with distance in health (as they have for business, education, government and other endeavors) [38]. However, current insurance reimbursement schemes generally discourage services that are not provided in person.

Adults with type 2 diabetes who live farther from their source of primary care are less likely to use insulin. This association is not due to confounding by age, sex, race, education, income, health insurance, body mass index, duration of diabetes, use of oral hypoglycemic agents, or frequency of primary or specialty care and may be responsible for the poorer physiologic control noted in rural patients.