Over the past 2 decades, numerous studies have clearly demonstrated that both racial and ethnic disparities exist in healthcare delivery and outcomes in the United States.^1-3 In 2002, the Institute of Medicine defined disparities as “racial or ethnic differences in the quality of healthcare that are not caused by accessrelated factors or clinical needs, preferences, and appropriateness of intervention.”¹ The report also highlighted the uncertainty in understanding the root causes of this disparity. As part of “Healthy People 2010,” the Department of Health and Human Services called for an elimination of disparities in healthcare.² In order to meet this demand we must better understand the basis of the problem.

Previous data show that African Americans with diabetes are less likely to have recommended process of care measures such as glycated hemoglobin (A1C) and lipid measurements performed.^3,4 There is also evidence to suggest that African Americans with diabetes in particular have disparate results on outcome measures such as blood glucose control, blood pressure control, A1C values, cholesterol values, and incidences of retinopathy, chronic kidney disease, and lower-extremity amputations.^5-7 While it is clear that disparities in both process and outcomes exist, there continues to be a lack of understanding as to how these disparities are related and if they are consistently related to these opportunity gaps described above.⁸ This connection is critical to help guide parity in healthcare outcomes for all patients.

Physicians’ understanding of the significance and magnitude of the effects of racial disparities on healthcare appear to be lower than one might expect given the previously cited publications. In a study of cardiologists, for example, only one-third felt that racial health disparities existed.9 In the same study, only 1 in 8 felt that disparities existed within their own practice. In a review of cardiothoracic surgeons, while some surgeons acknowledged that disparities existed, most attributed these disparities to patient characteristics.¹⁰ These examples are not unique to 1 specialty or even to the care of chronic diseases in general. Rather, they are endemic to the US healthcare system.

Change in healthcare requires a multidisciplinary effort that includes significant physician, patient, and system changes. Physician-led, multidisciplinary teams need accurate, timely data to drive change.¹¹ The evidence is variable that quality improvement practices can lessen disparities in diabetes care.¹² One method is a systematic data collection and review with patient registries. Patient registries can provide a useful tool to aid in data collection and analysis.¹³ In order to understand the effect of disparities on processes and outcomes of care and maximize the benefit of these registries, it is imperative that race be included as a variable.¹⁴ We are using registries to understand quality opportunities and to judge performance at the healthcare provider level. Risk adjustment is important to isolate the effect of interventions or the contribution by the provider.

Using diabetes as a sample population, the authors explored a national database from osteopathic residency programs— the American Osteopathic Association Clinical Assessment Program (AOA-CAP). The authors evaluated the delivery of diabetes care to determine if there were differences in the process and outcomes of care associated with 2 racial categories.

METHODS

This study entailed retrospective analysis of national registry data collected from osteopathic family medicine and internal medicine residency programs. The AOA-CAP is a web-based registry providing a standard method of sampling patients and collecting information from medical records on key processes and outcomes of diabetes care. The diabetes module has been operational since 2003. Data are collected for the purposes of quality improvement and satisfaction of core competencies of systems-based practice and practice-based learning within osteopathic graduate medical education. The process and outcomes measures for the diabetes care database were developed from guidelines for diabetes care promulgated by the American Diabetes Association and the American Association of Clinical Endocrinologists.^15,16 Measure development is overseen by the AOA-CAP Steering Committee with representation from internal medicine and family medicine program directors and leadership from the AOA. Clinical indicator definitions and constructs are “harmonized” with national organizations, including the National Committee on Quality Assurance, the American Medical Association Physician Consortium for Performance Improvement, and the National Quality Forum^17-19 Table 1 lists the numerator and denominator of each of the process of care measures and the intermediate outcome measures used in this analysis. Materials for AOA-CAP include documents standardizing patient selection, sampling, and data collection.

Residents imported the data on their own patients as part of a separate educational activity. There were no resident incentives or penalties based on how they performed on the data set for their patients. Information was abstracted from patient medical records using the most recent visit as a starting point and review of care prior to the most recent visit during a specified time period for each measure. Case selection for diabetes mellitus was based on any International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis of diabetes. Patients included in the database had to be older than 18 years and have had diabetes for at least 1 year, as well as have had at least 2 visits to the clinic. Patients treated by diet and lifestyle alone were excluded, as they were often monitored less closely. A sample size of 40 patients per residency program was recommended in order to provide a meaningful sample. However, cases from all programs were used regardless of the number of cases contributed by the program. Analysis excluded data from non-Caucasian and non– African American patients. Information regarding patient demographics and treatment were collected from each medical record. Data elements were entered in a Health Insurance Portability and Accountability Act–compliant manner into a web-based data collection form along with program identification.