Utilization and outcomes of HbA_1c testing: a population-based study

Improved glycemic control through blood glucose monitoring and pharmacotherapy is recommended to reduce the risk of microvascular complications in people with type 1 and type 2 diabetes mellitus.1^,2 Measurement of glycosylated hemoglobin (HbA_1c) has been advocated as a means of determining the adequacy of glycemic control.3 Guidelines developed in at least 13 countries recommend one or more HbA_1c tests annually and target levels under 8%.4

Using population-based administrative health databases, we examined HbA_1c testing frequency and outcomes between September 1, 1999, and September 1, 2000, for all individuals with diabetes aged 20 years or more who lived in eastern Ontario. In 2000, the adult population of eastern Ontario (defined as all areas east of a line connecting Arnprior in the north and Trenton in the south) was 1.09 million; the county-level prevalence of diabetes in 1999 ranged from 4.9 to 6.4 cases per 100 population.5^,6 Patients who died or moved out of the area during the study year were excluded to ensure an entire year of observation.

People with diabetes mellitus were identified using a validated algorithm that combined diagnostic information from physician service claims and hospital discharge abstracts.7 Women with gestational diabetes were excluded. The Canadian Institute for Health Information Discharge Abstract Database was used to count hospital admissions owing to hyperglycemia, which were based on a most-responsible ICD-9 diagnosis (International Classification of Diseases, 9th revision) coded 250.1 through 250.3. The Ontario Health Insurance Plan (OHIP) database was used to count visits to general or family practitioners, specialists (i.e., endocrinologists and internists) and hospital emergency departments by people with a diagnosis of diabetes. The Database of Laboratory Tests in Eastern Ontario was used to determine the numbers and results of all HbA_1c tests performed in eastern Ontario laboratories during the study year, with the exception of tests conducted at hospitals in the towns of Cornwall and Hawkesbury (about 1.5% of all tests in eastern Ontario).6 Each person in the 3 databases had a unique anonymous identifier that allowed accurate data linkage across the data sets.

We summarized HbA_1c testing frequency and outcomes and used logistic regression analysis to investigate the association of frequency and outcomes with age, sex, rural residence and utilization of the health care system. Because results could be clustered by physician offices or laboratories, which could increase the likelihood of apparent statistical significance, we used a conservative p value of 0.001 to determine significance. We categorized test outcomes as ideal or optimal according to the Canadian Diabetes Association guidelines8 by comparing each test result with the laboratory's upper limit of normal for the test.

We identified 63 699 people in eastern Ontario with a diagnosis of diabetes. Of these, only 58% had HbA_1c tests during the study year (Table 1). Fewer than half of those who had one or more tests had results that were ideal or optimal. Testing was most common among patients 50–74 years old; the youngest adults (aged 20–34 years) were the least likely to have had tests. The likelihood of having had an HbA_1c test also increased with the number of physician visits. Younger individuals had poorer test outcomes, as did those who visited specialists more often and underwent more tests.

Our results are similar to those of a smaller 1995 Canadian study of physician charts,9 which found that only 53% of patients with type 2 diabetes had HbA_1c testing during 1 year and that 87% of those had optimal or acceptable blood glucose levels (although only 57% of the test results would be categorized as ideal or optimal if current guidelines8 were used). The association between number of visits to specialists and suboptimal test results may well reflect selective referral patterns to specialists for patients with more refractory hyperglycemia.

Our study had limitations, although we expect their effect on the findings was small. First, the results might not apply to the rest of the province. Second, some patients could have had tests performed by laboratories outside of the study area; because of the large study area and referral patterns within it, however, this would affect only a small minority of patients. Third, validation studies suggest that up to 14% of people with diabetes would not have been captured by our administrative databased algorithm because they use physician services infrequently. Exclusion of these patients would likely contribute to an overestimation of testing rates, since their pattern of physician use also suggests that they are unlikely to have undergone HbA_1c testing.7 Fourth, most specialist physicians in the Kingston area are not paid on a fee-for-service basis and are not included in OHIP billing data. Nevertheless, excluding the Kingston area from our analyses made no substantial difference to our findings and did not affect our conclusions. Fifth, it is not possible to determine if a lack of testing is attributable to a physician's failure to order a test or the patient's failure to follow through with recommended testing.

Our finding that the youngest adults were tested least often and had the poorest test outcomes is particularly worrisome. These patients have the longest time to live with diabetes and accordingly the most to gain from interventions directed at secondary prevention. Other studies have also reported lower testing rates among younger people and improved glycemic control with age.10^,11^,12^,13^,14 Further research is required to identify how patient or physician factors, service availability and lack of continuity of primary care contribute to this gap. In the meantime, educational interventions for patients and providers should be considered, as well as policy initiatives to ensure equitable access to care.