Veterans Affairs Primary Care Organizational Characteristics Associated With Better Diabetes Control

Published Online: April 01, 2005
George L. Jackson, PhD, MHA; Elizabeth M. Yano, PhD, MSPH; David Edelman, MD, MHS; Sarah L. Krein, PhD, RN; Michel A. Ibrahim, MD, PhD; Timothy S. Carey, MD, MPH; Shoou-Yih Daniel Lee, PhD; Katherine E. Hartmann, MD, PhD; Tara K. Dudley, MStat; and Morris Weinberger, PhD

Objective: To examine organizational features of Veterans Affairs (VA) primary care programs hypothesized to be associated with better diabetes control, as indicated by hemoglobin A1c (HbA1c) levels.

Study Design: Cross-sectional cohort.

Methods: We established a cohort of 224 221 diabetic patients using the VA Diabetes Registry and Dataset and VA corporate databases. The 1999 VHA (Veterans Health Administration) Survey of Primary Care Practices results were combined with individual patient data. A 2-level hierarchical model was used to determine the relationship between organizational characteristics and HbA1c levels in 177 clinics with 82 428 cohort members.

Results: The following attributes were associated with lower (better) HbA1c and were statistically significant at P < .05: greater authority to establish or implement clinical policies (lower by 0.21%), greater staffing authority (0.28%), computerized diabetes reminders (0.17%), notifying all patients of their assigned provider (0.21%), hiring needed new staff during fiscal year 1999 (0.18%), having nurses that report only to the program (0.16%), and being a large academic practice (0.27%). Associated with higher (worse) HbA1c were programs reporting that patients almost always see their assigned provider (greater by 0.18%), having a quality improvement program involving all nurses without all physicians (0.38%), having general internal medicine physicians report only to the program (0.20%), and being located at an acute care hospital (0.20%).

Conclusion: Programs that are associated with better diabetes control simultaneously have teams that actively involve physicians in quality improvement, use electronic health information systems, have authority to respond to staffing and programmatic issues, and engage patients in care.

(Am J Manag Care. 2005;11:225-237)

The quality and outcomes of healthcare in the United States are far from optimal.1-4 The chronic care model5 posits that optimum chronic illness management requires a healthcare system that provides self-management support to patients, decision support, clinical information systems that support care, and effective connections with community resources. Such a system can improve interactions between providers and patients and, in turn, enhance patient outcomes. The model proposes a basic road map for combining these elements to achieve optimal patient outcomes.5-12

Several features of the Veterans Health Administration (VHA) make it an ideal system in which to test features of the chronic care model. First, the VHA operates the largest integrated healthcare delivery system in the United States.13 Second, in the mid-to-late 1990s, the VHA was transformed from a system with an inpatient specialty treatment focus to one that emphasizes primary care and integrated practice teams.14-16 Third, the VHA's extensive electronic medical record system facilitates care and provides the opportunity to examine the relationship between the organization of primary care and chronic disease outcomes. Therefore, study of Veterans Affairs (VA) health systems offers large-scale sources of data for understanding managed care delivery.17-19

Diabetes mellitus is an excellent disease for testing the chronic care model. Diabetes mellitus is prevalent in the United States: 12 million adults have diagnosed diabetes mellitus, and another 6 million have undiagnosed diabetes mellitus.20 Persons with diabetes mellitus are at increased risk for macrovascular and microvascular complications that compromise patients' quality of life and result in extensive use of healthcare services.20-24 Given its prevalence, morbidity, and associated health services use, diabetes mellitus is estimated to cost the United States $132 billion annually.25 The significance of diabetes mellitus in the VHA mirrors or exceeds that in the rest of the United States.26-34 For example, diabetes mellitus accounts for 25% of all pharmacy costs and 1.7 million inpatient bed days each year in the VHA.31

The morbidity and costs of diabetes mellitus can be reduced by tight control of blood sugar.35-38 However, adequate blood sugar control is often not achieved among adults with diabetes mellitus.39-41 Furthermore, the VHA External Peer Review Program and the National Committee on Quality Assurance's Health Employer Data and Information Set (HEDIS) include provision of hemoglobin A1c (HbA1c) tests and control of HbA1c among performance measures.42

The present study seeks to identify organizational characteristics of primary care programs that are associated with glycemic control and, hence, a reduced risk of long-term diabetes complications. Although not all characteristics of the chronic care model were measured, we hypothesized that those programs that more closely approximated this type of optimal system would have patients with better glycemic control.


We studied the effect of primary care organization during fiscal year (FY) 1999, which covers October 1, 1998 through September 30, 1999. In FY 1999, the VHA included 172 hospitals, more than 600 outpatient clinics, 132 nursing homes, 40 domiciliaries, 73 comprehensive home care programs, and 206 counseling centers.16 That year, 3.7 million people received care through the VHA.43 The institutional review board of the Durham Veterans Affairs Medical Center approved the study protocol.

Data Sources

1999 VHA Survey of Primary Care Practices. Information on the organization of VHA primary care clinics was obtained from the 1999 VHA Survey of Primary Care Practices.44,45 Between October and December 1999, the survey was sent to clinical leaders of primary care programs at individual VHA facilities with at least 4000 unique patients and at least 20 000 outpatient visits in FY 1998. This included 170 VA medical centers and 65 community-based outpatient clinics. The objective was to evaluate the organizational features of primary care programs so as to elucidate the relationships among organizational features and clinical and economic outcomes. There were 219 responses (160 medical centers and 59 community-based outpatient clinics), a response rate of 93%.44,45

VA Diabetes Registry and Dataset. The VA Diabetes Registry and Dataset was used to establish the cohort of diabetic patients and obtain outcome data.46 During FY 1999, 503 371 patients were in the registry.47 The registry contains 3 files. The pharmacy file has information on patients who have ever had prescriptions filled at a VA pharmacy for insulin, oral hypoglycemic agents, or blood glucose monitoring supplies. The laboratory file contains test results directly related to diabetes care. The vital signs file has information on patient height, weight, body mass index, blood pressure, and receipt of influenza and pneumonia vaccines.46

Veterans Affairs Corporate Databases. Data on patient demographics, healthcare use, comorbidities, and vital status were obtained from multiple VA corporate databases housed at the VA Austin Automation Center, Austin, Tex. Most databases maintained at the VA Austin Automation Center result from downloads from individual Veterans Information System Technologies Architecture health information systems located at VHA facilities.48,49 The VA Austin Automation Center has data on all VHA inpatient and outpatient encounters since 1980.50,51 Vital status of veterans is recorded in the Beneficiary Identification and Record Locator System.52 Information on patient comorbidities used for risk adjustment is located in The Johns Hopkins Adjusted Clinical Groups Case-Mix System files.53


Using the VA Diabetes Registry and Dataset, we identified patients who met the following inclusion criteria during FY 1999: (1) had at least 2 non-mental health outpatient visits with an associated diagnosis of diabetes mellitus, or any non-mental health inpatient discharge with an associated diagnosis of diabetes mellitus; (2) had a prescription filled for insulin, an oral hypoglycemic agent, or blood glucose monitoring supplies (from the VA Diabetes Registry and Dataset pharmacy file); and (3) had at least 1 outpatient visit to a VA primary care clinic. Exclusion criteria were age younger than 18 years, switching primary VA facilities during the study period, any VA endocrinology visit during FY 1999 through FY 2001, and pregnancy.

Encounter data were obtained from the inpatient bed section and outpatient event files. Specific International Classification of Diseases, Ninth Revision, Clinical Modification codes used to identify patients with diabetes mellitus were obtained from the 2003 version of the Clinical Classifications Software from the Agency for Healthcare Research and Quality.54

Individuals were considered primary care patients of the VA facility where they made the greatest number of primary care visits during FY 1999. In case of a tie, the patient was randomly assigned to one of the tied locations. Once patients' primary care clinics were determined, a random sample of 800 diabetic patients meeting the inclusion criteria was drawn from each VA facility operating in FY 1999. If a location did not have 800 patients meeting the criteria, all eligible patients were retained. The Figure illustrates the process of developing the diabetic cohort.



Our outcome was HbA1c, which measures mean glucose control during approximately 120 days.55 Results from facilities that measured total glycosylated hemoglobin (an older measure of glycemic control) were converted to HbA1c using the laboratory equipment manufacturer's conversion formula. During the study period, the VA considered HbA1c of 7.0% or higher to be elevated to a concerning level; levels of 9.0% or higher were considered markedly elevated.27 Specifically, the study modeled the associations among organization-level and individual-level independent variables and the last HbA1c value obtained during the study period of FY 2000 through FY 2001 (October 1, 1999, through September 30, 2001).

Organizational-level Variables

PDF is available on the last page.
Diabetes Compendium
Oncology Compendium