Community-Based Screening for Diabetes in Michigan

Before the publication of ADA’s recent position statement on screening for diabetes (3), the DONs, under the guidance of the MDCH Diabetes Control Program, worked with their regional advisory councils and with local community organizations to promote, facilitate, and evaluate community-based diabetes screening. For the most part, screening was conducted by hospitals, county health departments, and home health agencies as part of community awareness and outreach programs. In a number of instances, screening was conducted by work site wellness vendors as part of comprehensive work site wellness programs. The DONs offered education in all aspects of screening, including capillary glucose testing, quality control, and infection control, and provided triplicate data collection forms for risk assessment, interpretation, feedback, and evaluation.

All subjects provided written informed consent. Each individual was screened only once. The triplicate data collection form obtained individual identifiers (name, address, city, state, zip code, and telephone number), demographics (age, sex, race, and ethnicity), medical history (history of diabetes, current pregnancy), the number of hours since the individual last ate or drank anything other than water (postprandial time), and whether the individual had a personal physician or source of primary health care. The form also incorporated the “Take the Test. Know the Score” diabetes screening instrument distributed by the ADA. The form provided spaces to record the capillary glucose screening test result and a space to indicate whether the test was whole blood or plasma calibrated. Finally, the form provided a summary assessment and recommendation and a space to indicate the results of any follow-up contact, including the date the screened individual was contacted, whether he or she received further evaluation and testing, and whether he or she was diagnosed with diabetes.

MDCH collected the forms without any individual identifiers from the DONs for all screenings conducted between 1 June 1999 and 31 December 1999. Data were collected through 30 June 2000 to permit as complete follow-up as possible. MDCH then sent the forms to the Michigan Diabetes Research and Training Center where all data were entered in duplicate into a Microsoft Excel database. Capillary whole blood glucose values were converted to plasma glucose values by multiplying by 1.14 (7). Means and SDs were assessed for continuous variables, and frequencies and proportions were assessed for categorical variables. Differences among groups and the odds ratios (ORs) were assessed with the χ² test. Screening capillary plasma glucose levels were interpreted by the DONs according to ADA and MDCH recommendations. Positive screening tests were defined as random capillary plasma glucose ≥160 mg/dl or fasting capillary plasma glucose ≥126 mg/dl (8).

To independently assess the likelihood of previously undiagnosed diabetes, a predictive equation was applied to the screening data. The equation was developed with multiple logistic regression analysis using data collected from 1,032 Egyptian subjects and was validated using data collected from 1,065 American subjects (6). The equation’s sensitivity was 65% and its specificity was 96% (6). All statistical analyses were performed using SAS software version 6.12 (SAS Institute, Cary, NC). P < 0.05 was defined as the limit of statistical significance.

All subjects provided written informed consent. In the 7 months between 1 June 1999 and 31 December 1999, the local agencies screened 3,506 individuals as a part of 127 separate screening events. The number of individuals screened in the six DON regions ranged from 57 to 1,261, and the number of individual screening events conducted by various agencies ranged from 3 to 55. Most individuals were screened at hospital sites (n = 1,202, 34%). A substantial number of individuals were also screened at health fairs (n = 429, 12%), shopping centers (n = 311, 9%), work sites (n = 249, 7%), community centers (n = 224, 7%), and other sites (n = 1,071, 31%).

Of the 3,506 individuals screened, age was missing for 444 (13%). Of those reporting age, 57 (2%) were age <20 years, 658 (21%) were ages 20–44 years, 1,033 (34%) were ages 45–64 years, and 1,314 (43%) were age ≥65 years. Sex was missing for 133 (4%). There were 2,112 (63%) women and 1,261 (37%) men. Race and ethnicity were missing for 187 (5%) subjects. There were 2,979 (90%) white subjects, 120 (4%) black subjects, 83 (2%) Native-American subjects, 78 (2%) Hispanic subjects, and 59 (2%) subjects of other races or ethnicities. Health care data were missing or unknown for 243 (7%) subjects. There were 2,955 (91%) subjects who reported that they had a personal physician or source of primary health care and 308 (9%) subjects who reported no primary care physician or source of primary health care.

Of the 3,506 individuals screened, 475 (14%) were inappropriately screened based on published guidelines for screening. Rates of inappropriate screening did not differ by DON. Those inappropriately screened included 57 subjects (2% of total population screened) who were age <20 years, 51 subjects (1%) who were currently pregnant, and 367 subjects (11%) who reported a medical history of established diabetes. Individuals screened at work sites (25%), community centers (16%), and hospitals (13%) were more likely to report previously diagnosed diabetes than those screened at shopping malls (9%), health fairs (4%), and other sites (8%).

Table 1 describes the characteristics of the 3,301 subjects age ≥20 years who were not pregnant and did not have a history of diabetes who were appropriately screened. In general, subjects who were appropriately screened tended to be female and white and have a personal physician. The characteristics of appropriately screened subjects did not differ by DON or screening site. The mean diabetes risk score (as determined by the ADA screening instrument, “Take the test. Know the score”) was 10 and >57% of the subjects had a high risk score of ≥10. Only 11% of subjects were fasting (postprandial time ≥8 h).

Only 153 (5%) of the appropriately screened subjects had fasting capillary plasma glucose levels ≥126 mg/dl or random levels ≥160 mg/dl. Individuals who screened positive were significantly more likely to be nonwhite (OR 1.73, CI 1.07–2.79) and were more likely to report having no personal physician or source of primary health care (OR 1.36, CI 0.82–2.27). Follow-up was attempted for 106 (70%) of those who screened positive and was successful for 53 (50%). The yield of diagnosed diabetes was, however, very low. The proportion successfully followed was 53 (2%) and the proportion diagnosed with diabetes was 14 (0.5%). In general, rates of successful follow-up and diagnosis did not differ by DON or screening site (data not shown).

To explain the low prevalence of positive screening tests, we applied the predictive equation to the screening program data. The estimated prevalence of undiagnosed diabetes as determined by the predictive equation was 11%. This prevalence was consistent with the results of the Third National Health and Nutrition Examination Survey (9).

Screening for diabetes has been a popular activity among local community organizations. Hospitals, home health agencies, and county health departments frequently conduct diabetes screening to increase awareness and provide education about diabetes and to promote and advertise their services and programs. Early identification and treatment of individuals with previously undiagnosed diabetes is only one of many objectives.

Our evaluation of diabetes screening conducted by local community organizations with guidance from the DONs and MDCH demonstrated that most individuals screened were at high risk based on questionnaire screening scores and age. At the same time, however, most individuals were at low risk based on race and ethnicity (90% white) and based on the fact that most (91%) reported having a primary care physician or access to routine medical care. In addition, it is apparent that criteria for screening were often not rigorously applied: ∼2% of those screened were age <20 years, 1% were pregnant, and 11% reported previously diagnosed diabetes. Although the latter is a relative contraindication to “screening,” it could certainly be argued that additional monitoring of glycemic control and education are a health benefit.

The ADA screening instrument, “Take the Test. Know the Score,” identified 57% of individuals as being at higher risk for previously undiagnosed diabetes, suggesting a high false-positive rate. The prevalence of positive capillary plasma glucose screening tests was only 5%. With appropriate training and support from the DONs and MDCH, the screening conducted by voluntary community organizations demonstrated good rates of attempted and successful follow-up. Unfortunately, the yield (the proportion of appropriately screened subjects found to have previously undiagnosed diabetes) was poor: <1%. Other community-based diabetes screening programs using both the ADA screening instrument (10) and measured glucose levels have also demonstrated a low yield of individuals with undiagnosed diabetes (2,11,12).

To further investigate the low prevalence of positive screening tests and low yield, we applied a multivariate logistic regression equation to estimate the prevalence of undiagnosed diabetes in the population. Compared to the ADA criteria, the equation identified more subjects with undiagnosed diabetes (11 vs. 5%). This suggests that the ADA plasma glucose criteria are set too high and miss a substantial portion of nonfasting individuals with undiagnosed diabetes.

Based on these results and the ADA’s updated position statement on screening in diabetes, MDCH has elected not to support community screening for diabetes. Improved definition of a positive test, relying on biochemical testing such as random plasma glucose, targeting racial and ethnic minority groups, and targeting medically underserved individuals with no primary care physician, no insurance, or no access to routine health care could substantially improve the yield of community-based diabetes screening. Further studies are needed to demonstrate these potential benefits before wide-spread community screening can be recommended.

This project was supported by the Michigan Diabetes Research and Training Center under National Institutes of Health Grant DK-20572.