As far as we are aware, this is the first assessment of age of diagnosis of diabetes in Appalachia. The unadjusted average ages, while statistically significantly different, do not differ substantially. However, an unadjusted comparison confounds risk profile with county development level; because risk profiles differ dramatically among levels of county development (e.g., high poverty and low educational attainment in the distressed counties, low poverty and high educational attainment in the achievement counties [6
]), this is not the best comparison. The adjusted comparison gives a better picture of the true differences.
The regression model indicates that residents of distressed and at-risk counties had diabetes diagnosed at younger ages than people with diabetes with similar risk profiles living in non-Appalachian counties. The coefficients for Appalachian counties other than distressed and at-risk, in both the primary and the sensitivity analysis, were near zero. This suggests that, for Appalachian counties other than the distressed and at-risk counties, most if not all of the differences in age at diagnosis is attributable to risk profile; however, in distressed and at-risk counties, people had their disease diagnosed at younger ages than risk profiles alone explain. Similarly, the sensitivity analysis suggests that, for competitive and achievement Appalachian counties, most if not all of the differences in age at diagnosis between these counties and the transitional counties are attributable to risk profile differences; however, in distressed counties, people had their disease diagnosed about three years younger than risk profiles alone explain. Putting this in perspective, the effect on age at diagnosis of residence in a distressed county is roughly comparable to that of obesity.
We have no data concerning the time between developing diabetes and having it diagnosed. However, residents in the distressed and at-risk counties have limited access to care [7
], which makes it unlikely that the time between developing diabetes and having it diagnosed in these counties is shorter than it is in the more affluent counties. In addition to cost and transportation, other issues make rapid diagnosis unlikely. Coyne [8
] reports that cultural attitudes in Appalachia can be a barrier to obtaining care, including the practice of seeking medical care only as a "last resort" and a distrust of health care providers. A relatively large number of health care providers working in Appalachia are foreign born, and Coyne [8
] reports that cultural differences with foreign-born providers and high turnover are other barriers to seeking care. Thus, it is likely that people in distressed and at-risk counties developed diabetes younger than their non-Appalachian counterparts, instead of having developed diabetes at the same age and receiving earlier diagnoses.
People who develop diabetes at younger ages can spend more time with undiagnosed, and therefore untreated, diabetes. At the national level, estimates of the percentage of all diabetes cases that are undiagnosed range from 27% [9
] to 32% [10
], representing an added burden of disease in distressed and at-risk counties. The prevalence of undiagnosed diabetes in Appalachia could be substantially different from the national level. We are aware of no estimates of the prevalence of undiagnosed diabetes in Appalachia, and so cannot assess how large the additional burden is. Untreated diabetes is associated with greater incidence of complications later in life, such as vision loss [11
] and renal and cardiovascular damage [12
]. Therefore, residents of the distressed and at-risk counties might be at greater risk of eventually developing complications.
Those with lower incomes and those with only a high school education, on average, tended to have diabetes diagnosed later. People in these conditions are less likely to have access to care [13
], which could delay diagnosis. They also might be less aware of health issues and therefore less likely to seek care [14
]. People with incomes ≥ $50,000 had their diabetes diagnosed sooner, probably because of greater ease in accessing care.
The US Surgeon General has indicated that tobacco use increases the risk of diabetes [15
]. We found that people who reported ever smoking had diabetes diagnosed about two years later than comparable nonsmokers. The reason is unknown. Since the message that smoking is harmful to health is pervasive, one might speculate that people who elect to smoke might be less concerned about health. If so, smokers might be less likely to seek care, possibly resulting in later diagnoses.
Obese people had diabetes diagnosed about three years earlier than the nonobese. Obesity is a well-known risk factor for type 2 diabetes. Therefore, obese people might be screened for diabetes more often than nonobese people. Obesity was measured at the time of interview, and the respondent might or might not have been obese at the time of diagnosis; however, obesity, once developed, often persists [16
]. Respondents who have been obese since youth are at risk for developing diabetes at a younger age [16
The lack of access to full-service grocery stores, which contributes to poor diet, contributes to the high rates of obesity in distressed counties. One study of convenience stores in an Appalachian county found that none carried fresh or frozen green vegetables, low-fat milk, or low-fat cheese [17
]. Another study found that Appalachian youth knew what healthy foods were, but ate packaged foods because healthy alternatives were unavailable [18
People who reported no leisure-time physical activity reported diagnosis of diabetes about a year later than those who did. We are not certain why, because physical activity is a well-known way to prevent or delay type 2 diabetes. One possible reason is that the BRFSS measures leisure-time physical activity. Nonleisure-time physical activity, which was unmeasured, might be substantial in Appalachia, due to agricultural work or employment in the mining industry. The failure to account for nonleisure-time physical activity could bias the results. Also, the BRFSS measures physical activity at the time of the survey. Respondents included in this analysis all had diagnosed diabetes. Their prediagnosis activity status is unknown
We found that racial and ethnic minorities had diabetes diagnosed at earlier ages. This is not consistent with Koopman et al. [19
], who found no significant difference among racial and ethnic groups in a 1999-2000 national sample. This could be due to Koopman et al's use of a national (versus subnational) sample, use of 1999-2000 (versus 2006-2008) data, or failure to adjust for income and education. Similarly, a study conducted in Norway found that mean age at the time diagnosis for minority groups was eight to 15 years younger than for other Norwegians [20
]. This study's authors speculated that the pathophysiological processes for diabetes started or accelerated earlier in minority groups.
We found that people with insurance coverage at the time of the survey had their diabetes diagnosed about six years later than those who did not, and people who had received medical care in the year preceding the survey had their diabetes diagnosed about four years later than those who did not. These findings are consistent with the notion that access to medical care helps to delay the development of diabetes, and people without access to medical care are likely to develop the disease at younger ages.
Our analysis is subject to several limitations. Data from the BRFSS are self-reported and subject to nonresponse bias, social desirability bias, bias from exclusion of households without land-line telephones, and recall bias. Recall bias concerning the age of diagnosis of diabetes is of particular concern. We can think of no reason that people in the poorer counties would be more likely to incorrectly recall their diabetes as being diagnosed at younger ages than people in other counties, but neither can we totally rule out this possibility. Our measures of access to care were as of the time of the survey. Respondents' insurance status and receipt of medical care might have been different when they developed diabetes or had the condition diagnosed than it was at the time of the survey. We were unable to distinguish type 1 from type 2 diabetes. While physical activity and weight loss can prevent or delay type 2 diabetes, no way of preventing type 1 diabetes is known. However, nationally, type 2 diabetes accounts for between 90% and 95% of all cases of diabetes [7
]. Finally, ARC county classifications can change, and we used the classifications as of 2007.
It would have been desirable to provide state, rather than regional, estimates. The number of respondents with diabetes stratified simultaneously by development level and state was too small to support state-level estimates. Thus, our analysis could only be done on a regional level.