Diabetes and Cancer Screening Rates among Appalachian and Non-Appalachian Residents of Kentucky

Discussion

In this study we demonstrate clearly that, for Kentucky women, being up to date with breast cancer screening is negatively associated with having diabetes. Also, patients with diabetes had about half the odds of reporting “regular” mammography screening than those without diabetes. It is also noteworthy that Appalachian residents had about half the odds of reporting a colonoscopy/sigmoidoscopy within the last 10 years. Although such a cross-sectional study alone cannot demonstrate causality, our results suggest that having diabetes as a chronic illness presents a barrier to achieving guideline-concordant breast cancer mammography screening and that residents of Appalachia may face barriers to receiving colonoscopy/sigmoidoscopy screening.

We also found that proportionately less Appalachian residents than non-Appalachian residents reported receiving “regular” mammograms, and proportionately fewer were up to date in their colorectal cancer screening with colonoscopy/sigmoidoscopy. However, proportionately more Appalachian residents than non-Appalachian residents were up to date with colorectal cancer FOBT screening. The diabetic-nondiabetic screening comparisons were significant for non-Appalachian but not Appalachian residents, but we did not confirm an effect modification of Appalachian residence on the relation between having diabetes and current screening with the Breslow-May test. These results should be interpreted cautiously because of the relatively small sample size of Appalachian residents. With this sample size, we should not conclude that Appalachian residence modifies the effect of diabetes on cancer screening.

Our results are consistent with findings of Schoen et al⁵ that physicians were less likely to request a mammogram for women with a chronic illness and with the Kiefe et al⁴ research showing that women with chronic illnesses were less likely to receive either mammography or CBE. It is unclear, however, why our results would single out a relationship between diabetes and breast cancer screening while showing only a marginal relationship for cervical cancer screening (significant in bivariate but not in regression analysis) and no association for colorectal cancer screening. There are several possible explanations. First, the association between diabetes and obesity may play a role. This theory is consistent with studies by Ferrante et al²⁷ indicating that obese women are less likely to adhere to mammography recommendations. It may be that obese women are more self-conscious in the situation of exposing their breasts to multiple strangers at a radiology suite, a hypothesis consistent with the research of Wee et al²⁸ and Amy et al.²⁹ In addition, mammography may be more technically difficult, painful, or unpleasant for obese women. Some studies support the hypothesis that mammograms produce more discomfort for women with particularly large or small breasts (see, for example, Papas and Klassen³⁰). In our study, we did not know the extent to which such groups populate the diabetic and nondiabetic samples, and thus we could not infer whether or not discomfort mediated by breast size could have been a factor. However, we can speculate a smaller role for obesity to mediate rates of colorectal and cervical cancer screening because the discomfort posed by these tests is probably similar for normal-weight and moderately obese persons.

Second, a differential effect may be at play with breast cancer screening because of the level of provider involvement. It may seem counterintuitive that a recommendation for mammography, a screening generally well-accepted by women in the United States and that requires no office procedure, would be more affected by a chronic illness than those that require more office time (eg, Papanicolaou test) or extensive patient education (eg, colorectal cancer screening). However, the processes for mammography referral may be creating a different picture. Although colorectal cancer screening is initiated by the health care practitioner because of the need for counseling and referral, breast screening can be initiated and/or managed by the patient in conjunction with the screening facility. A small number of patients request their first mammogram themselves because of concerns raised through the media, public attention, or family/peer experience with breast cancer.³¹ Calls to the three major insurers in Kentucky revealed that only one of them requires a referral for screening mammograms. The other two allow for self-referral provided the patient's screening site is “in network.” For a large number of Kentuckians, therefore, as long as the patient remains without symptoms or abnormalities on examination or imaging, subsequent screening appointments can be managed by the radiology provider and the patient (Humana, United Health Care, and Anthem Customer Service in Kentucky; telephone communications; April 2010). Similarly, the Kentucky Breast and Cervical Cancer Treatment Program (BCCTP), which funds free breast examination and mammography services for women without health insurance, requires these examinations to be conducted through local health departments and often do not involve the patient's primary care provider.³² Nevertheless, the literature is clear that the single most important factor in whether or not a woman receives a mammogram is whether it was recommended by her physician or provider.³³ Though speculative, it may be that providers accustomed to the “automatic” nature of mammography and burdened by more complex visits with diabetic patients may be less likely to notice delays in mammography screening. Meanwhile, women with diabetes may be too overwhelmed by their physical condition, monitoring/medication regimen, or general stress of chronic illness to follow through reliably with mammography without their doctor's encouragement. Why the pattern for CBE generally matches that of mammography is more difficult to speculate, though the requirement to receive a CBE before mammography for those in the BCCTP population may explain some of this effect. Also, since 2002 the USPSTF has not made a recommendation for or against CBE because of “insufficient evidence” of benefit, and many providers no longer perform it routinely.²² The BCCTP program also provides Papanicolaou testing and follow-up for uninsured patients. Papanicolaou testing is less expensive and has fewer barriers than mammography, so many indigent patients continue to receive Papanicolaou tests at their primary care provider offices; however, the smaller number who presumably access all their gynecologic preventive care at the health department may also explain the relatively weaker relationship but definite trend between cervical screening and diabetes status.

Limitations of this study include the relatively small number of Appalachian residents that participated in this study. A larger sample might have provided the power to show other statistically significant relationships. Also, the inability to contact subjects without a landline telephone likely resulted in an undersampling of patients who are both younger and of lower socioeconomic status (SES).³⁴ That the high school graduation rate of 87% in our study was significantly higher than the approximately 72% state average and the 60% to 65% average in Appalachian counties³⁵ lends credence to the theory that low SES persons may have been undersampled. Although low SES is certainly associated with not having any telephone access, anecdotally, since the early 2000s there has been a movement away from landlines toward the use of only cell phones in both of the above demographic groups, which would raise the absolute effect of undersampling present as compared with older telephone-based studies. Even by 2006, more than 25% of adults younger than age 30 and 17% of adults in households earning 200% of the Federal poverty level or less were accessible only by cell phone. Households without a landline are also less likely to have health insurance or a usual source of care.³⁴ Because all the screening tests in our study except for Papanicolaou tests concerned people older than 40 years of age, missing data from young cell phone-only users probably did not affect our results significantly. Unintentionally undersampling people with low SES means that fewer people with health insurance and a usual source of care would have responded. It is also important to note that Appalachian residents have disproportionately low SES.³⁶ It is possible that the low number of Appalachian participants is partially due to our landline sampling technique and that some of the differences between Appalachian and non-Appalachian residents and between Appalachian diabetics and nondiabetics would have reached significance had we been able to sample cell phone-only households. It should be noted that both of these limitations actually bias the results to the null, so the significant findings in our study would likely be even stronger with more complete sampling. This could explain why differences between Appalachian and non-Appalachian populations were found in the Hall et al⁹ study using data from 1996 to 1998, before any significant cell phone penetration, whereas our data from 2007 to 2008 was not able to distinguish these differences.

Another limitation—and opportunity for future research—is that this survey did not enable us to assess the severity of diabetes in the sampled population or to assess multiple other chronic conditions. (The only other chronic condition surveyed was asthma, which will require an additional manuscript to address sufficiently.) Comorbid conditions such as hypertension, asthma, lipid disease, obesity, and other heart disease could have biased the results to the extent that these conditions either increase or decrease the likelihood of screening. The 2005 Fleming et al² study showed that some conditions, such as diabetes, increased the risk of late-stage breast cancer, whereas others, such as cardiovascular disease, actually decreased the risk of late-stage disease. Thus, it is unclear whether multiple chronic conditions in the same person would cancel each other out (with regard to risk) or exacerbate the risk. Because it is also unclear how much chronic disease burden is present among diabetics compared with nondiabetics as a group, a reasoned assessment of the level and direction of bias cannot be done. Because colorectal cancer screening was relevant to both women and men, but breast cancer screening was measured only among women, we questioned whether the differences were because of different screening practices between men and women rather than between diabetics and nondiabetics. However, when men were excluded from the analysis, there was no difference in colorectal cancer screening between women with and without diabetes, just as there had been no difference with men and women combined.

It is worth noting that some of the recommendations for screening have changed since the initiation of this study. Papanicolaou tests, for instance, are not started before age 21 and are recommended at 3-year intervals after several normal Papanicolaou tests.^23,24 The USPSTF has recently recommended against routine mammography for women younger than 50 years old.^21,22 However, our results reflect the state of the guidelines at the time the survey was conducted.

Our results raise several issues for future study, chiefly the reasons for the difference between mammography and colorectal cancer screening with regard to effects of diabetes. Interventions focusing on provider prompts and reminders for screening may not be effective in cases in which screenings have become more automatic and less provider-directed. On the other hand, making providers aware of their role in encouraging screening for their diabetic patients could be an effective directed intervention if diabetics do indeed need assistance in keeping up with screenings. Regardless, it is important to better understand where to focus interventions, and a qualitative approach may shed light on the modern mammography experience for diabetics. Other possible research includes an analysis assessing other chronic conditions' and/or multiple comorbid conditions' effects on appropriate screening achievement, as well as which systems-based interventions could improve screening rates in patients with chronic illnesses.