HbA_1c Levels Among American Indian/Alaska Native Adults

The IHS Diabetes Care and Outcomes Audit is conducted yearly through medical chart reviews. Methods for the Diabetes Care and Outcomes Audit have been described previously (11–13). Briefly, IHS and tribal clinic facilities are encouraged to maintain diabetes registries of all individuals with diabetes. Selected clinical variables and interventions are measured annually using a systematic random sample of charts at each facility. The methods for randomization and sample size collection have been reported elsewhere (13). In short, a sample size is chosen for each facility that is sufficient to provide an estimate within ±10% of the true rate of adherence for each facility, with a confidence level of ≥90% (14). The regional diabetes coordinators and professional staff members, trained by the coordinators, perform the Diabetes Care and Outcomes Audit. A uniform set of definitions is used throughout the U.S. The abstracted data are then entered into a general-purpose microcomputer-based software program. Data from participating sites were combined regionally and then aggregated to determine national rates. Data were collected following standardized protocols, using fixed definitions and data collection forms (13). Although no validation studies of the audit have been conducted, some areas have performed additional audits using different reviewers to both audit the same charts and assure accuracy in data entry procedures. Also, a comparison study has shown good observed agreement between the manual and electronic audits (0.78–1.00%) at a particular site.

We assessed glucose control by HbA_1c or by using the mean of the last three blood glucose readings. Because different laboratories are used throughout the nation, results were converted into a standardized HbA_1c value (16–19). If no HbA_1c was available, the mean of the last three blood glucose readings was used to calculate an “estimated HbA_1c value.” The following formula was used: HbA_1c = (60.16 + mean glucose)/30.9 (19).

Demographic, clinical, and standard of care variables were collected through the Diabetes Care and Outcomes Audit and included sex, age, height, weight, duration of diabetes, treatment type, and proteinuria status. Age was calculated as the date of audit minus the date of birth. BMI was calculated from recorded height and weight as follows: (703 × weight in pounds)/(height in inches × height in inches). Duration of diabetes was recorded as time since diagnosis in years. Treatment type was recorded as diet alone (no medication), oral agent, insulin, and oral agent plus insulin (combined with the insulin group for purposes of analyses). Proteinuria was defined as having ≥1+ (≥30 mg/dl) protein in a urine dipstick in the past year and did not include microalbuminuria.

An index for quality of care was developed using six of the standard of care variables from the Diabetes Care and Outcomes Audit: foot examination, eye examination, diabetes diet instruction, exercise instruction, general diabetes education, and influenza vaccination. One point was assigned for each of the six items that were recorded as completed within the year before the audit. For example, if all the standards of care measures were documented as having been completed within the year before the Diabetes Care and Outcomes Audit, the diabetes care health care index was 6, whereas if no measurements of the variables were documented as having been completed, the health care index was 0. If an activity was missing, it was recorded as “not done” and assigned a score of 0.

The analysis of the primary study end point, HbA_1c, was conducted for all people with measured and/or calculated HbA_1c. To test whether there were differences in HbA_1c by age, we used ANCOVA. ANOVA was used to test for differences in HbA_1c levels by age and treatment and for HbA_1c levels by diabetes health care index by age. The covariates for adjustment included BMI, duration of diabetes, treatment type, and the diabetes health care index. All statistical analyses were performed on SAS software (Cary, NC) (20).

Of the 11,419 potential participants aged ≥18 years who were identified from the 1998 National Indian Health Care and Outcomes Audit, 9,622 had measured HbA_1c and 1,158 had calculated HbA_1c levels. Data on duration of diabetes or BMI were missing for 1,920 participants, and these participants were excluded from the analyses. A total of 8,860 participants, 78% of the total sample, was used for analysis purposes.

Table 1 summarizes the demographic and selected clinical and measurements of care data. Consistent with sex differences in the prevalence of diabetes in American Indian/Alaska Native adults, more women than men were included in the sample. The mean age was 54.9 years, the mean duration of diabetes was 8.8 years, and the mean BMI was 32.9. The overall mean HbA_1c level was 8.6%. Of the patients, 25% had proteinuria, and most participants were treated with oral agents alone (57%). Diabetes health care factors ranged from 56% for influenza vaccine to 70% for general diabetes education.

HbA_1c was significantly different by age-group (Table 2). The youngest adults had the highest HbA_1c levels (P < 0.0001). In addition, BMI was highest in the youngest age-group, and it decreased with increasing age (P < 0.0001, data not shown). For the 1,920 patients who had missing data on duration of diabetes or BMI, a similar difference in HbA_1c levels by age-group was observed (P < 0.0001, data not shown). There was a statistically significant difference between age and all treatment types. The highest HbA_1c levels were found in the youngest age categories within each treatment regimen. We examined HbA_1c and BMI, using BMI categories of <25, 25–29.9, and ≥30, and we found no statistically significant association (P = 0.36).

Figure 1 shows HbA_1c levels by age-group and duration of diabetes. Within each duration category (0–4, 5–9, and ≥10 years), we observed an inverse relationship between age and HbA_1c. Table 3 presents HbA_1c level by health care index categories, with a higher index representing more health care factors. For each age-group, a higher health care index was significantly associated with improved HbA_1c levels, but it was not statistically significant for the group aged 18–39 years (P = 0.06). Within each health care index category, a statistically significant decrease in HbA_1c level was observed. Table 4 summarizes the adjusted mean HbA_1c by age-group for those with and without proteinuria and for those with unknown proteinuria status. An inverse relationship between age category and HbA_1c level for each category of proteinuria status was observed when adjusted for BMI, duration, treatment type, and health care index (P < 0.0001).

An inverse relationship of HbA_1c level and age was found among American Indian/Alaska Native adults nationally after adjusting for BMI, diabetes duration, treatment type, proteinuria, and health care index. Although these findings are cross-sectional and are limited by our measure of health care index, they lend support to prior published findings that younger American Indian/Alaska Native people with type 2 diabetes have worse diabetes control (8,9). The factors underlying the elevated HbA_1c levels among younger American Indian/Alaska Native adults are not fully understood, but they do not appear to be explained by BMI, diabetes duration, treatment type, proteinuria, or diabetes health care index. We cannot directly measure whether these results are caused by a cohort effect (different age-groups represent different birth cohorts, and the same differences will continue to exist as the cohorts age) or represent developmental effects (as the cohorts age, their glycemic control will improve). However, because glycemic control generally deteriorates with time (7), it is likely a cohort effect. A longitudinal follow-up study would help to assess this question. There is an urgent need to understand the factors related to the elevated HbA_1c levels because the prevalence of diabetes among young American Indian/Alaska Native people is on the rise (5,6), and it is therefore likely that we may observe an increase in diabetes-related morbidity and mortality unless more effective and efficient interventions are developed to improve glycemic control in young American Indian/Alaska Native adults and thereby prevent complications of diabetes.

The higher HbA_1c levels seen in the younger age-groups may be related to rapid changes in lifestyle that have occurred in many American Indian/Alaska Native communities. In a study in New Mexico, HbA_1c levels were highest for those who reported consuming the most fat and sugar (8). Moreover, high consumption of fat and sugar was most closely associated with elevations in HbA_1c levels among those aged <55 years (8).

In this study, we found that BMI increased with younger age, which is consistent with national BMI trends (21). Studies have shown that American Indian/Alaska Native people are becoming more overweight and obese, an observation that may be associated with increased insulin resistance, and this increase in obesity may influence the ability to effectively treat diabetes because greater obesity and insulin resistance have been associated with poorer glucose control (22–27). However, higher BMI values in the younger age-groups did not account for the high HbA_1c levels observed among younger American Indian/Alaska Native adults. We lack data on insulin resistance in this study to assess the role of obesity and insulin resistance in the relationship between age and HbA_1c.

Environmental and socioeconomic factors play a role in diabetes outcomes, including glucose control, morbidity, and mortality (28). Access to care and medications may also influence health outcomes; however, these factors are unlikely to play a major role in the current study because health care and medications are generally provided free of charge for American Indian/Alaska Native people through IHS and tribal health programs, although in some situations, patients are charged a co-pay, and not all clinics carry all medications. However, younger adults may have difficulty attending clinic appointments and assuming self-care activities, given their work and child/family care activities. Prior studies have noted improved compliance with treatment type and clinic appointments with increasing age (10,29). In the current study, the diabetes care health care index was used to assess whether a poor health care index explained poorer glycemic control in young adults. Overall, each increase in the index was associated with an improvement in HbA_1c; however, higher HbA_1c values with decreasing age persisted within each health care index category. These results do not support the hypothesis that worse diabetes control in the younger age-group is caused by poorer health care indices for this population (9).

There are several limitations of this study. First, these data were collected as secondary data. Although the audit criteria are clearly set to include random and adequate sampling, standard definitions, and standard reporting protocol, it is possible that all reviewers did not always follow these standards. The large dataset and the many years of audit implementation should decrease the influence of deviation from the protocol by a minority of reviewers. No validation studies of the audit have been conducted. However, some areas have performed additional audits using different reviewers to both audit the same charts and assure accuracy in data entry procedures. A comparison study has shown good observed agreement between the manual and electronic audits (0.78–1.00%) at a particular site.

Cross-sectional studies are subject to survival bias. This is unlikely to have biased the current results, given the persistence of the relationship within each diabetes duration category. Selection bias is a possibility because not all Indian health care facilities participated in the audit. A larger percentage of women than men were included in the data, which is consistent with prior reports (11). The results do not appear to be biased by the presence of a higher percentage of elders with renal disease (with higher circulating insulin levels and therefore better glycemic control), given the persistence of the results when stratified by proteinuria status.

We cannot assume the health care index is reflective of patient self-care activities alone because it contains items that are influenced by both system as well as patient self-care factors. It would be helpful to have additional items reflective of system- and patient-centered factors in the future to address this issue.

The current study brings a major health challenge to health professionals in the Indian health care system: to understand the physiological, socioeconomic, and lifestyle factors influencing diabetes control in younger American Indian/Alaska Native adults and to develop and implement effective and culturally appropriate interventions in this high-risk group. Unless we are successful in these endeavors, the morbidity and mortality associated with diabetes in young American Indian/Alaska Native people can be expected to rise in the next decade.

This study was supported by the New Mexico Veterans Affairs Health Care System, Albuquerque, New Mexico, and by National Institutes of Health-National Institute of Diabetes and Digestive and Kidney Diseases Grant RO1-DK-47096-06.

This manuscript represents the results of a study that Dr. Janette S. Carter was actively involved in at the time of her untimely death. Dr. Carter dedicated her career to improving care of people with diabetes, especially in American Indian/Alaska Native communities. The diabetes community has suffered a huge loss with her passing.