Unrecognized diabetes is a major public health problem. Approximately one-third of all patients with diabetes are undiagnosed (1). Because treatment of diabetes improves outcomes by reducing complications (2,3,4), screening for diabetes is an attractive intervention. The American Diabetes Association therefore recommends screening all adults over 45 years of age every 3 years (and younger if at high risk) (5). One presumption underlying a recommendation for doctors to screen for diabetes is that such surveillance is not part of routine medical care. However, there is a certain level of glucose testing of patients without diabetes that occurs in medical practice, which may or may not represent intentional diabetes screening. The frequency of this ad hoc diabetes screening in medical settings is unknown, as are the factors that lead to this ad hoc screening. The objective of this study was to determine the rate and predictors of glycemic testing in a managed care population.
The study was a retrospective analysis of multiple merged administrative databases of Duke University Medical Center’s Managed Care organization. During the time of the study, Duke Managed Care was a staff-model managed care organization within Duke University Medical Center (DUMC). Administrative records were evaluated for all patients enrolled in Duke Managed Care during the study period, which was April 1996 to March 1999. Patients’ records were included if they met the following criteria: at least one billable encounter during the study period, identification of a primary care provider, and age ≥30 years. There were no exclusion criteria. The database for the study analyses was generated by merging demographic data from the managed care administrative databases; diagnostic codes were from managed care billing data, and laboratory dates, tests, and values were from the DUMC laboratory database. Approximately 96% of laboratory tests in Duke Managed Care are performed by the DUMC laboratory.
We identified patients with known diabetes by searching the billing data for diabetes diagnostic information, as determined by any one International Classification of Diseases, Ninth Revision (ICD-9), 250.xx diagnostic code. We then used the laboratory database to identify patients that had any test related to diabetes over the 3-year study period. Laboratory tests used for the study were plasma glucose (PG) tests and HbA1c measurements. The frequency of ordering of other glycemic measures (e.g., fructosamine) was so low as to be irrelevant. The databases were then merged to assess the proportion of patients with glycemic testing and the predictors of testing. Analysis was mostly descriptive. Stepwise logistic regression was performed to determine independent associations with the presence of blood glucose testing among patients without diabetes. All analyses were performed using the SAS analysis system (SAS Institute, Cary, NC).
The study population was 61% female, 74% white, and had a mean age of 51 years at the end of the study period (with 18% of the study population over the age of 65). Of the 22,169 patients in University Managed Care, 1,400 (6%) had known diabetes and were removed from further analysis. Of the remaining 20,769, 10,766 (52%) had had some sort of glycemic testing, usually PG tests; 10,003 (48%) had no testing performed. For patients aged 45 years and older, this proportion increased to 68%. We then performed stepwise logistic regression to determine the predictors of glycemic testing. For the logistic model, 3,110 patients with missing demographic information were excluded (excluded patients had a testing rate of 31%). All three demographic variables in our limited database were found to be predictors of testing in a multivariable stepwise logistic regression model for the remaining 17,659 patients. These were increasing age (odds ratio [OR] 1.8 for every 10 years of age, P < 0.0001), female sex (1.4, P < 0.0001), and white race (1.1, P = 0.001).
The data from this study show that approximately half of the patients in a managed care population are screened for diabetes during a 3-year time frame. The predictors of glycemic testing in patients without diabetes are increasing age, female sex, and white race. While age is a powerful predictor of diabetes, women and men are equally likely to have diabetes in most studies (1,6,7). Despite this, women were much more likely than men to have had glycemic testing in this study. Perhaps more important is the slight (but statistically significant) increased rate of testing among white compared with nonwhite patients in our population. This is directly in opposition to numerous studies showing increased risk for diabetes in nonwhite people (6,7,8,9,10). Conversely, whites and women do in general receive more health services (11,12). Therefore, the results from this limited database are more consistent with the idea that the “ad hoc screening” we see in managed care reflects general utilization patterns rather than an intentional approach to diabetes screening. This study cannot, however, rule out a significant undercurrent of intentional screening obscured by the large numbers of tests performed in other settings (e.g., admission to the hospital, as part of a chemistry battery, etc.).
There are several limitations to our study. We may have underestimated glycemic testing in patients without a diagnosis because we have no way of accessing fingerstick blood glucose measurements, which are not recorded in the computer. The database slightly underestimates the prevalence of diabetes due to inadequate coding; a limited chart review (410 patients) estimates the sensitivity of the database compared with medical records for the diagnosis of diabetes at 91%. This leads to a trivial overestimate of the rate of screening. Nevertheless, these data indicate that screening recommendations are only moderately adhered to, and probably by accident rather than intention. There is room for improvement in diabetes screening practices in medical centers if providers pay more attention to testing those patients who are at risk for diabetes.
Address correspondence to David Edelman, HSR&D (152), Durham VA Medical Center, 508 Fulton St., Durham, NC 27705. E-mail: email@example.com.