Regional and racial disparities in major amputation rates among medicare beneficiaries with diabetes: a retrospective study in the southeastern USA

Introduction

Patients with diabetes are at an increased risk of foot ulcers and non-traumatic lower extremity amputations (LEAs) due to factors including poor glycaemic control, peripheral neuropathy and peripheral arterial disease.1 Previous studies showed that black patients with diabetes are more likely to present with more advanced disease than white patients, contributing to considerably higher risks of LEAs.2–7 Patients living in rural areas have higher rates of LEA.4 8 Emerging data suggest that LEA disparities are amplified when these two social determinants of health—race and rurality—overlap.9 10 Nearly all rural counties with black majorities are located in the Southeastern USA.11 This same geographical area contains the diabetes belt, a high diabetes prevalence region identified by the Centers for Disease Control and Prevention (CDC) with higher rates of diabetes.12 While this area was identified with the data that go back to 2007 and 2008, this area is still very much relevant for health policy as evidenced by the continued attention to this area by the CDC13 14 and recent research.15–18

Overall, LEA rates have been on the decline in the USA with a slight increase more recently.19 It is unclear whether this trend has been consistent for patients of all races and for all regions, including the diabetes belt. Given that this area is home to some of the nation’s highest-risk patients, it is critical to understand how LEA rates have changed over time in this region.

Historically, Medicare has played a role in reducing racial disparities through many actions including desegregation of hospitals, data collection by race/ethnicity data to help identify disparities, and targeted programmes to reduce disparities.20 However, disparities persist. More recently, the Centers for Medicare and Medicaid Services (CMS) Office of Minority Health was formed with the mission to ‘lead the advancement and integration of health equity,’ including work to help rural Medicare patients have better access to high-quality care and reduce disparities for racial and ethnic minorities.21 In addition to Medicare efforts, the Affordable Care Act (ACA), a comprehensive federal statute in the USA that included many healthcare insurance and healthcare delivery reforms, supported the expansion of Federally Qualified Health Centers between 2011 and 2015.22 Medically underserved areas, including those in the diabetes belt, benefited from this increased access to care. However, patients who were cared for by these new health centres were less likely to be rural.23 It is unclear how minority health efforts have specifically affected rates of specific health conditions within the diabetes belt, where there is a combination of racial minority and rural populations.

Our objective is to examine how major LEA (MLEA) rates have changed in the diabetes belt and surrounding counties over the 10-year period surrounding the implementation of ACA (2006–2015) and whether and how it affected racial disparities in these regions. Further, we will explore factors that have a significant impact on MLEA rates that may be useful in identifying Medicare policies that could be implemented to further reduce MLEAs.

Methods

Results

The yearly sample size was between 898 428 and 899 846, representing a weighted population size ranging from 1 897 053 in 2006 to 2 214 129 in 2015. No patient in the samples was excluded on account of missing values such as race/ethnicity or zip code of residence. Table 1 presents the weighted population by year, the percentage of patients in the diabetes belt and surrounding counties, mean ages for each cohort, per cent female and race/ethnicity distributions.

MLEA rates were 3.9 per 1000 in the diabetes belt compared with 2.8 in the surrounding counties in 2006 and decreased to 2.3 and 1.6 per 1000 in 2015 (both p<0.001). The yearly weighted number of patients with foot ulcers and MLEAs by county type is shown in table 2 . Figure 1 presents (a) the yearly MLEA rates for patients in the diabetes belt compared with those in the surrounding counties and (b) the yearly MLEA rates by geographical area and race. Patients in the Belt experienced one full MLEA more per 1000 patients than those in the surrounding area. Black patients had 8.5 and 6.9 MLEAs per 1000 in 2006 and 4.8 and 3.5 in 2015 in the Belt and surrounding counties, respectively, while the rates were similar for white patients in the two areas (2.4 vs 2.1 in 2006 and 1.5 vs 1.3 in 2015).

Request permission

Although amputation rates declined rapidly in both the Belt and the surrounding areas, black patients consistently had rates more than three times higher than white patients during the 10-year period in the Belt. Differences in rates between black and white patients were larger among patients in the diabetes belt than in the surrounding counties. The gap in rates between black and white patients narrowed over the study period. The trends did not show any noticeable changes in the MLEAs before and after the ACA.

Patients in the diabetes belt had a 44% higher rate of MLEA in 2015 compared with the surrounding areas. Table 3 presents the full model for a weighted logistic regression for MLEA controlling for covariates. We found that black patients had 22% higher odds (OR 1.216; 95% CI 1.057 to 1.400) of experiencing an MLEA compared with surrounding areas, adjusting for patient demographics, socioeconomic status, diabetic complications, comorbidities, antidiabetic and antilipemic medications, and access to care factors. When race/ethnicity was interacted with geographical regions, we found that blacks had 52% (OR 1.515; 95% CI 1.235 to 1.860) and 69% (OR 1.692; 95% CI 1.443 to 1.985) higher odds of having an MLEA compared with whites in the same region, while blacks in the diabetes belt had almost twice higher odds (OR 1.986; 95% CI 1.653 to 2.387) compared with whites in the surrounding areas (online supplemental table 2). In the same region, the odds of experiencing an MLEA were not significantly different between whites and Hispanic or other/unknown groups. Figure 2 displays the adjusted ORs for geographical region by race/ethnicity groups.

Request permission

Another model that included the number of short-term general hospitals, federally qualified health centres in county, and nurses and physician assistants in the county was estimated to examine how disparities in MLEA rates between geographical regions were after accounting for these additional factors for health services availability. The OR of patients in the diabetes belt having an MLEA was 1.188 (95% CI 1.031 to 1.369) compared with those in the surrounding areas after accounting for services availability. A cross-model test shows that the ORs for the diabetes belt compared with the surrounding areas beforeand after (online supplemental table 3) adjusting for services availability were not significantly different (ORs=1.216 vs 1.188; p=0.292). Other notable findings were that the odds of MLEA were higher for males, black patients (compared with white patients) and those with PCIP qualifying doctor visits, while they were lower for those with Medicaid buy-in and residents in states with expanded Medicaid (table 3). The number of short-term general hospitals and the physician assistants were significantly associated with reduced MLEA risk but the number of FQHC’s in the county was not (online supplemental table 3). The number of nurses in county who can provide healthcare services to patients with diabetes (advanced practice registered nurses, nurse practitioners, clinical nurse specialists, certified registered nurse anaesthetists and physician assistants) was associated with increased risk of MLEAs, while the number of physician assistants in the county was associated with decreased risk.

Discussion

Our data show persistent disparities in MLEA rates between the diabetes belt and surrounding areas. Over the 10-year study period, we consistently observed higher MLEA rates in the diabetes belt compared with the surrounding areas with rate ratios ranging from 1.37 to 1.56 over the study period (table 2). Both sets of counties exhibit large racial disparities in MLEA rates between white and black patients with larger disparities in the diabetes belt. Geographical differences may be attributed, in part, to a higher proportion of rural, black patients living in the diabetes belt than surrounding counties. This suggests racial minorities in the diabetes belt are doubly jeopardised for MLEA.

Our study was not designed to identify why black patients living the diabetes belt had higher rates of MLEA than their neighbouring counterparts. However, our findings linking rural residence, state Medicaid buy-in and limited care access to higher MLEA risks suggest an eroding healthcare infrastructure may be a factor.26 Over 80% of rural counties with majority Black populations are health professional shortage areas; most of these overlap with the diabetes belt.26 Furthermore, our group has shown that diabetes quality care metrics between the diabetes belt and surrounding counties are nearly equivalent after adjusting for provider shortages.27 When considering how efficient counties are in terms of providing diabetes preventive care, too few foot exams and eye exams were often identified as sources of inefficiencies.27 Although we did see a slowly narrowing gap in MLEA rates between regions and races over the horizon, there does not seem to be any clear evidence that this trend may have been affected by the ACA. A potential reason for this may be that expansion of federally qualified health centres had limited reach to rural patients, such as those that served in the diabetes belt.23

Our results are generally consistent with the declining trends in LEA rates in recent years that have amply been documented in the literature. Goldberg et al studied Medicare patients with diabetes over the period 1999–2006 and reported higher rates of amputation among Black patients.6 Margolis et al found a decreasing rate of incidence of LEA over 2006–2008 with higher rates clustered in some of the hospital referral regions in states overlapping with the diabetes belt, including Mississippi and Louisiana.28 29 Harding et al19 and Gregg et al30 presented amputation trends for patients 65 and older, with the general trend showing reduced rates of amputation during the early 2000s followed by flattening out or slight increase in amputation rates by around 2015. The findings of Harding et al19 are consistent with our work in that the rates for black patients are consistently higher than those for White patients. Our results go beyond what has previously been studied by providing rates for the diabetes belt specifically and highlighting the racial disparities in MLEA rates within the Belt and between the Belt and the surrounding counties.

Similar to the findings from Harding et al19 and Gregg et al,30 the rates among white patients in the diabetes belt and surrounding counties seem to be levelling off toward the end of the study period, potentially suggesting that we might have reached a floor for how low the MLEA rates can go under the current practice. Additional, targeted policies are needed to further reduce MLEA rates among disparity populations. Specifically, policies to prevent MLEAs among black patients in general and those in the diabetes belt are needed to eliminate the large racial disparities. Policies that specifically address patients in the diabetes belt should be prioritised given the disparity in MLEA rates between the diabetes belt and surrounding counties. These could include investing in healthcare infrastructure in rural counties with predominantly Black populations, which would potentially reduce both MLEA rates as well as disparities in premature death and other health outcomes.11 A policy specifically aimed at providing timely ambulatory care for Black patients, who are experiencing increasing difficulties accessing medical care independent of cost barriers,31 should be carefully considered.

Our results did not show any significant decreases in MLEAs after the ACA. While provisions of the ACA have reduced racial and rural disparities, further progress is needed.32 The CMS Office of Minority Health was one of six minority health offices established through the ACA to help highlight and address disparities. The geographical and racial MLEA disparities highlight a need for targeted interventions and strategies from the Office of Minority Health and other federal and local agencies to eliminate these disparities, particularly for the doubly disadvantaged black patients living in the diabetes belt.

As suggested by Eichner and Vladeck,20 Medicare can be a catalyst for reducing racial disparities in MLEA rates. CMS has several options to use for reducing disparities that include adjusting benefit and cost-sharing structures to reduce or eliminate out-of-pocket costs for low-income patients and implementing evidence-based guidelines for care to ensure all patients are treated with quality care despite physician bias. Further, CMS Quality Improvement Organisations can be used to engage patients at a community level to reduce disparities, an approach that would likely be helpful for reducing the regional and racial disparities we see in the diabetes belt.

Other research has highlighted potential approaches for reducing MLEA rates that may help reduce racial and geographical disparities. A pay-for-performance programme for diabetes providers that would provide incentives for reducing the rate of MLEAs may also have a significant impact on MLEA disparities.33 Multidisciplinary care teams including podiatrists, endocrinologists, wound care nurses and surgeons to manage diabetic foot ulcers have also become a recommended standard of care.34 These approaches are in line with Eichner and Vladeck’s suggestions for actions Medicare could take to address racial and ethnic health disparities, namely adjusting payment approaches and improving quality of care to reduce health disparities.

Conclusion

Our findings show persistent disparities in the major amputation rates between the diabetes belt and surrounding counties, with racial disparities much larger in the Belt than in the surrounding areas. Targeted policies to prevent MLEAs among black patients are needed to reduce the persistent racial disparities, particularly for those black patients in the diabetes belt who are ‘doubly disadvantaged.’