Diabetes Care in the U.S. and Canada

The recruitment methods used in the DCCT are described elsewhere (21). In brief, the DCCT enrolled patients between 1983 and 1989 in 29 centers across North America (26 in the U.S. and 3 in Canada). The main inclusion criteria were the following: age 13–39 years, type 1 diabetes for 1–15 years, HbA_1c >6.6%, and willingness to participate in the trial (21). Exclusion criteria were the following: hypertension, hypercholesterolemia, advanced diabetes complications, other major medical conditions, previous intensive therapy, current or planned pregnancy, or any condition that would interfere with the measurement of HbA_1c (21). Recruitment strategies (physician referral, advertisements in local media, and promotion through diabetes organizations) were similar in both countries. (Fig. 1).

All DCCT centers used a standardized, uniform questionnaire to collect data, including demographics, duration of disease, history of complications, and other medical illnesses on all screened individuals. Volunteers who passed initial screening (n = 3,228) then had further eligibility testing with a second questionnaire that collected detailed demographic and clinical information concerning previous treatments, medical history, and family history. Almost all of these individuals (n = 2,849) also had a baseline HbA_1c measured using high-performance ion-exchange liquid-chromatographic methods at a central laboratory (22).

Each patient was categorized as screened in the U.S. or Canada. Baseline comparisons included all individuals who had data containing a HbA_1c measurement. In the principal analysis, mean baseline HbA_1c levels were compared between American and Canadian patients using a t test. Applying a two-sided α of 0.05, the analysis had a power of 0.90 for detecting an absolute difference of 0.4% in HbA_1c levels. The comparison was then repeated using linear regression to adjust for potential confounders by considering variables that were related to HbA_1c on univariate analysis. Using similar methods, we also compared HbA_1c levels achieved during the trial between U.S. and Canadian patients randomized to conventional therapy. To assess the generalizability of these findings, we compared the characteristics of patients who entered the trial with those who did not and used logistic regression to identify predictors of trial entry. Unit costs for out-patient physician, nurse, and dietitian services and self blood and urine glucose monitoring and hospitalization for acute complications were derived from previously published sources and expressed in 1994 USD (10).

Several differences between the screened American and Canadian volunteers with type 1 diabetes were observed (Table 1). Canadian patients were somewhat younger, less likely to be college-educated, and more often unmarried. The racial and gender composition was similar in the two groups, as were the distributions of lifestyle characteristics. Patients were generally lean and physically active. Canadian patients were more likely to have smoked than American patients, albeit fewer cigarettes per day (17 vs. 14, P = 0.007). Current or previous alcohol use was also more common among Canadians, including the average weekly consumption (6 vs. 5, P = 0.015).

Self-reported medical care was not the same for patients in the U.S. and Canada (Table 1). Canadian patients received more of their care through family practitioners (41 vs. 28%, P = 0.001) and had more visits per year to physicians (3.6 vs. 2.9, P = 0.004). A surprisingly large proportion of American patients reported no visits to a physician in the preceding year (17 vs. 10%, P = 0.006). Visits to dieticians mirrored this pattern, with American patients more likely to receive no contact (35 vs. 24%, P = 0.001) and averaging fewer visits annually (0.7 vs. 0.4, P = 0.004). Nurse visits were rare in both groups. More Canadian than U.S. patients reported that they were following a therapeutic diet, but the two groups reported similar frequencies of self-monitoring of blood glucose.

Among individuals ≥18 years of age in the U.S., college education was associated with a higher likelihood of seeing a physician in the previous year (83 vs. 78%, P = 0.02) but not in Canada (89 vs. 88%, P = 0.9). Similarly, individuals who had a white-collar occupation in the U.S. were more likely to report having seen a physician in the previous year (82 vs. 77%, P = 0.03) but not in Canada (88 vs. 93%, P = 0.4). In contrast, patients who had a blue-collar occupation were less likely to perform self-monitoring of blood glucose at least once per week than those with white-collar occupations in both countries (U.S. 40 vs. 50%, P = 0.0007 and Canada 36 vs. 49%, P = 0.2), possibly because supplies required for home glucose monitoring are not covered by either health care system.

Acute complications occurred with similar frequencies in the two volunteer populations. American and Canadian patients reported similar rates of hospitalization for diabetic ketoacidosis (7 vs. 8 episodes per 100 patient years, P > 0.20). Canadians tended to report severe hypoglycemia more often, including episodes that required assistance (9 per 100/year vs. 4 per 100/year, P = 0.06) or episodes that led to hospitalization (4 per 100/year vs. 2 per 100/year, P = 0.10). Mild hypoglycemia was an equally common occurrence for patients in both countries (4 vs. 4 per week, P = 0.8).

Several demographic and lifestyle characteristics were independently related to baseline HbA_1c on multivariate analysis (Table 2). Lower HbA_1c levels were associated with male sex, Caucasian race, higher education, white-collar profession, and being married. Patients who had never smoked had lower HbA_1c levels. Paradoxically, patients reporting no regular alcohol use had higher HbA_1c levels, even when the comparison was limited to those consuming one drink per day or less versus nonusers (9.11 vs. 8.75, P = 0.0002). Greater weight was associated with lower HbA_1c levels, even after adjustment for age and sex. Level of exercise was not significantly related to glycemic control in this group of patients.

Many clinical characteristics were also related to the adequacy of glycemic control. Patients with longer disease duration had higher HbA_1c levels. Conversely, lower HbA_1c levels were associated with provision of care by a specialist, more frequent visits to a nurse or dietitian, more frequent blood glucose monitoring, and adherence to a regimented diet. The frequency of physician visits was unrelated to the patient’s HbA_1c.

There were no significant differences in glycemic control between the two populations. Specifically, the mean baseline HbA_1c level was 8.9 (range 4.7–15.5) for Canadian patients and 9.0 (4.5–17.3) for American patients (difference −0.10, 95% CI −0.34 to + 0.15). Adjustment for demographic, lifestyle, and clinical predictors yielded similar results; specifically, the adjusted mean HbA_1c concentration was slightly but nonsignificantly lower for Canadian than American patients (9.0 vs. 9.2, P = 0.15). The proportion of patients with excellent glycemic control (HbA_1c ≤7.0%) was similar in the U.S. and Canada (13 vs. 14%, P > 0.20), as was the percentage of patients with suboptimal levels (HbA_1c ≥8.5%: 44 vs. 43%, P > 0.20) and very poor levels (HbA_1c ≥ 10%: 28 vs. 26%, P = 0.5).

Annual expenditures for diabetes care tended to be ∼17% lower in the U.S. than in Canada ($650 vs. $780, P > 0.1). This difference was largely due to lower expenditures for out-patient physician ($59 vs. $72 per person, P = 0.004) and dietitian ($4 vs. $7 per person, P = 0.004) services in the U.S. Less frequent hospitalizations for hypoglycemia also contributed to lower expenditures in the U.S. ($31 vs. $93 per person, P = 0.1). Controlling for physician specialty did not alter these results. All of these estimates assumed equivalent fees in the two countries.

Several characteristics were predictive of trial entry: Caucasian race, higher education, white-collar profession, marriage, provision of diabetes care by a specialist, dietary adherence, more frequent nurse visits, and more frequent urine testing (Table 3). Patients who were recruited into the trial also exhibited healthier lifestyle behaviors and were relatively more compliant with self-care. For example, patients entering the trial were less likely to have ever smoked cigarettes (36 vs. 41%, P = 0.008) or to have drank alcohol (39 vs. 42%, P = 0.10) and were more likely to take part in vigorous exercise (1.9 vs. 1.4 h per week, P = 0.01). A similar percentage of American and Canadian patients who were screened ultimately entered the DCCT (19 vs. 21%, P = 0.20).

During the DCCT, American and Canadian patients randomized to conventional therapy achieved similar mean HbA_1c levels (9.1 [95% CI 5.6–13.4] for American patients vs. 9.2 [95% CI 5.9–12.5] for Canadian patients, P = 0.5). For conventional patients, baseline HbA_1c levels were the largest determinant of glycemic control during the trial; however, level of education and duration of diabetes had lesser but significant effects. Even after adjusting for these factors, American and Canadian patients had similar mean HbA_1c levels during conventional therapy (9.0 vs. 8.9, P > 0.20).

We studied ∼3,000 patients with type 1 diabetes in North America. The principal finding was that glycemic control was similar for those in the U.S. and Canada. After adjustment for baseline demographic, lifestyle, and clinical factors, HbA_1c levels tended to be slightly lower (not statistically significant) in Canadian patients. Whereas outcomes of other diseases have been shown to vary substantially between health care systems, no differences in glycemic control were observed for patients with type 1 diabetes who were screened for the DCCT in the U.S. and Canada.

This study had sufficient power to detect a difference in HbA_1c on the order of 0.4%, but a smaller difference between the two countries may have been missed. However, the clinical relevance of the observed difference of 0.2% is questionable (23). Selection bias may jeopardize the generalizability of this study because trial enrollment was based on the patient’s degree of motivation and other characteristics that could influence compliance. Although DCCT patients were highly educated, their baseline HbA_1c levels were not substantially lower than those observed in other population-based cohorts of patients with type 1 diabetes in the U.S. (24, 25). Stringent selection criteria could minimize, but should not eliminate, differences between the two countries, given that the same recruitment strategies were used in all centers. Furthermore, crude and adjusted analyses yielded similar results.

A strict research protocol can mask differences between health care systems, but this is unlikely to explain our findings. If care of conventionally treated patients had been identical across centers during the trial, then the distribution of HbA_1c levels would have been quite narrow, contrary to what was observed. Many studies including the DCCT have noted substantial variation in the delivery of care across clinical centers (26–28). Furthermore, glycemic control before the DCCT was the most important predictor of glucose response within the conventional group (23), contrary to the supposition that clinical protocols control all major determinants of patient outcome.

Despite equivalent mean HbA_1c levels, the frequency and pattern of care delivery differed between the two health care systems. Even among DCCT patients, Canadians had more visits per year to physicians, thereby mirroring patterns of use in the general population (8). More physician use in Canada might result from greater health needs; however, our data do not support this premise, as patients at recruitment were free of chronic complications that would demand special medical attention. Moreover, nationality had little impact on the frequency of ketoacidosis, although there was a suggestion that severe hypoglycemia was more common among Canadian patients. In general, hospitalization rates for other medical conditions are higher in Canada compared with the U.S. (9,29), possibly because of differences in thresholds for admission. However, Canadians were more likely to report having any hypoglycemic episode that required assistance for reasons that remain unclear.

Surprisingly, baseline costs of diabetes care were lower for American than Canadian patients. This finding is in sharp contrast to aggregate and disease-specific data demonstrating higher health care expenditures per capita in the U.S. than in Canada. (29, 30). One explanation is that our analysis assumed similar costs of services in the two countries and did not take costs of equipment, medical therapy, laboratory testing, or ophthalmologic consultation into consideration. Regardless, higher expenditures in Canada did not result in better HbA_1c levels. However, small differences in spending (such as $130 USD) may have little impact on glycemic control, whereas large differences ($3,000 USD), as observed between treatment arms in the DCCT, have a tremendous impact.

There are other limitations to our analysis. Subsets of the population not well represented in this study may benefit to a greater degree in one health care system than another. For example, economically disadvantaged groups may receive more care under a universally administered system. Information on the insurance status of patients screened in the U.S. was not available; thus, we were unable to assess whether individuals who lacked adequate health care coverage had worse outcomes. Although subgroup analyses suggesting more frequent physician visits among those with lower earning potential in Canada should be viewed with caution, the differences merit further study.

A final limitation is that this study compared glycemic control achieved in American and Canadian patients recruited over a decade ago, when standards of care were significantly different than they are today. This time lag is unfortunate but typical of large epidemiological studies. However, our findings may still be valid as health care systems evolve slowly, and health care professionals are slow to adopt even the best evidence for care of their patients. Moreover, widespread incorporation of new management strategies into clinical practice should occur in both countries at a similar pace.

In summary, mean HbA_1c concentrations were uninfluenced by different patterns of health care delivery in the U.S. and Canada. Furthermore, factors that are beyond the control of care providers, such as patients’ perceptions and attitudes, seem to play a pivotal role in disease management. Yet, adjustment for differences in sociodemographic factors did not alter these results. Moreover, we found that the care of patients with type 1 diabetes before enrollment into the DCCT was far from ideal in both countries.

G.L.B. was supported by a fellowship award from the Canadian Institutes of Health Research, B.Z. was supported by the Sam and Judy Pencer Family Chair in Diabetes, and D.A.R. was supported by an Ontario Ministry of Health Career Scientist Award and the de Souza Chair of Clinical Trauma Research at the University of Toronto.

We acknowledge the DCCT/Epidemiology of Diabetes Interventions and Complications (EDIC) study group for providing baseline data for this study.